Salty Matters

The Blog is written by me, John Warren. Once every three or four weeks or so I will post an article or two on an evaporite topic that has piqued my interest. On the Saltwork Publications webpage (under "the Works") there is a growing library of pdfs and epubs based on these blogs. These articles on the website have much higher resolution extractable graphics in than in the blog. There is also a link to this set of pdfs and epubs on the home page (

Salt Dissolution (3 of 5): Natural Geohazards

John Warren - Tuesday, October 31, 2017


Surface constructions and other anthropogenic activities atop or within evaporite karst terranes is more problematic than in subcopping carbonate terranes due to inherently higher rates of dissolution and stoping (Yilmaz et al., 2011; Cooper and Gutiérrez, 2013; Gutiérrez et al., 2014). Overburden collapse into nearsurface gypsum caves can create stoping chimneys, which break out at the surface as steep-sided dolines, often surrounded by broader subsidence hollows. Such swallow-holes, up to 20 m deep and 40 m wide, continue to appear suddenly and naturally in gypsum areas throughout the world.

Unlike the relatively slow formation of limestone karst, gypsum/halite karst develops on a human/engineering time-scale and can be enhanced by human activities (Warren, 2016, 2017). For example, in 2006, the Nanjing Gypsum mine in China broke into a phreatic cavity in a region of gypsum karst, driving complete flooding of the mine in some three days. Associated groundwater drainage caused a sharp drop in the local piezometric level of up to 90 m in a well in nearby Huashu village. Resultant ground subsidence severely damaged nearby roads and buildings (Wang et al., 2008). In Ukraine, dewatering of gypsum karst to facilitate sulphur mining substantially increased the rate of gypsum dissolution and favoured the expansion of sinkholes within an area affected by the associated cones of water-table depression (Sprynskyy et al., 2009). Natural evaporite karst enhanced by intrastructure focusing of drainage creates the various scales of problem across the Gypsum Plain of West Texas and New Mexico (Stafford et al., 2017).

Although halite is even more susceptible to rapid dissolution than gypsum, it typically is not a major urban engineering problem; large numbers of people simply do not like to live in a climate that allows halite to make it to the surface. However, in the Dead Sea region, the ongoing lowering of the water level encouraged karstic collapse in newly exposed mudflats and has damaged roads and other man-made structures (Frumkin et al. 2011; Shviro et al., 2017). Catastrophic doline collapse atop poorly managed halite/potash mines and solution brine fields is an additional anthropogenically-induced or enhanced geohazard in developed regions is discussed in detail in Warren, 2016 (Figure 1).

Gypsum karst is a documented natural hazard in many parts of Europe (Figure 2), and similar areas of shallow subcropping gypsum are common in much of the rest of the world (Table 1). For example, areas surrounding the city of Zaragoza in northern Spain are affected, as is the town of Calatayud (Gutiérrez and Cooper, 2002; Gutiérrez, 2014). Gypsum dissolution is responsible for subsidence and collapse in many urban areas around northern Paris, France (Toulemont, 1984), in urban areas in and around Stuttgart and other towns peripheral to the Harz Mountains in Germany (Garleff et al., 1997), in Pasvalys and Birzai in Lithuania (Paukstys et al., 1999), in the Muttenz-Pratteln area in northwestern Switzerland (Zechner et al., 2011), in the Perm area of Russia (Trzhtsinsky, 2002), in the Sivaz region of Turkey (Karacan and Yilmez, 1997), in the region centred on the city of Mosul in northern Iraq (Jassim et al., 1997) and in a number of areas of rapid urban development in eastern Saudi Arabia (Amin and Bankher, 1997a, b). Large subsidence depressions caused by gypsum dissolution in China have opened up in the Taiyuan and Yangquan regions of Shanxi Coalfield and the adjacent Hebei Coalfield.

Variation in the watertable level, induced by groundwater pumping or uncontrolled brine extraction, can be an anthropogenic trigger for dolines surfacing. As the watertable declines it causes a loss of buoyant support to the ground, it also increases the flow gradient and water velocity, which facilitates higher rates of crossflow and deeper aquifer recharge in subsequent floods and so reduces the geomechanical strength of the cover and washes away roof span support (Figures 1, 3). Dolines can also be associated with groundwater quality issues. Collapse dolines or sinkholes are frequently used as areas or sumps for uncontrolled dumping industrial and domestic waste. Because of the direct connection between them and the regional aquifer, uncontrolled dumping can cause rapid dispersion of chemical and bacterial pollutants in the groundwater. In the case of Riyadh region Saudi Arabia, a lake of near-raw sewage has appeared in Hit Dahl (cave) and is likely related to the increased utilisation of desalinated water for sanitation and agriculture (Warren, 2016). In the Birzai region of Lithuania numerous sinkholes developed in Devonian gypsum subcrop are in direct connection across the regional hydrology. Accordingly, the amount of agricultural fertilizer use is limited to help protect groundwater quality.

One of the problems associated with rapid surfacing of evaporite collapse features is that any assignment of sinkhole cause will typically lead to an assignment of blame, particulary when anthropogenic infrastructure has been damaged or destroyed by the collapse, or lives may have been lost. Areas of natural evaporite karst are typically areas of relatively shallow evaporites. Shallow evaporites make such regions suitable for extraction via conventional or solution mining. When a collapse does occur in a mined area, one group (generally the miners) has a vested interest in arguing for natural collapse, the others, generally the lawyers and their litigants, will argue for an anthropogenic cause. The reality is usually a combination of natural process enhanced to varying degrees by human endeavours. In the examples in this section, much of the driving process for the collapse is natural, while the cause of any unexpected karst-related disaster is typically geological ignorance combined with political/community intransigence. See Chapter 13 for a further discussion of karst and stope examples that include collapses and explosions where the anthropogenic drivers can dominate.

Problems in the Ripon area, Yorkshire, UK

The town of Ripon, North Yorkshire, and town’s surrounds experiences the worst ongoing gypsum-karst related subsidence in England (Figures 3, 4; Cooper and Waltham, 1999). Some 43 events of subsidence or collapse in the caprock over the Ripon gypsum have occurred over the last 160 years, within an area of 7 km2 (Figures 4). This gives a mean rate of one new sinkhole every 26 years in each square kilometre. Worldwide, the highest documented event rate occurs in Ukraine, in an area of thin and weak clay caprocks above interstratal gypsum karst, where new sinkholes appear at a rate of 0.01 to 3.0 per year per km2 (Waltham et al., 2005). In the Ripon area, numerous sags and small collapses also typify surrounding farmlands. Subsidence features are typically 10-30m in diameter, reach up to 20m in depth and can appear at the surface in a matter of hours to days (Figure 3). To the east of the town, one collapse sinkhole in the Sherwood Sandstone is 80 m in diameter and 30 m deep, perhaps reflecting the stronger roof beam capacity of the Sherwood Sandstone.

When a chimney breaks through, the associated surface collapse is very rapid (Figure 3 b-e). For example, one such subsidence crater, which opened up in front of a house on Ure Bank terrace on 23rd and 24th April, 1997, is documented by Cooper (1998.) as follows (Figure 3b).

“...The hole grew in size and migrated towards the house, to measure 10m in diameter and 5.5m deep by the end of Thursday. Four garages have been destroyed by the subsidence. This collapse was the largest of one of a series that have affected this site for more than 30 years; an earlier collapse had demolished two garages on the same site, and a 1856 Ordnance Survey map shows a pond on the same site. The hole is cylindrical but will ultimately fail to become a larger, but conical, depression. As it does so, it may cause collapse of the house, which is already damaged, and the adjacent road. The house and several nearby properties have been evacuated and the nearby road has been closed. The gas and other services, which run close to the hole, have also been disconnected in case of further collapse.”

Cooper (1998) found the sites of most severe subsidence in the Ripon area (including the house at Ure Terrace and in the vicinity of Magdelen's Road) are located at the sides of the buried Ure Valley, an area where the significant volumes of water seeps from the gypsum karst levels into the river gravels (Figure 4). In 1999 the Ure Terrace sinkhole was filled using a long conveyor belt that was cantilevered over the hole so that no trucked needed to back up close to the sinkhole opening. The hole was surcharged to a height of 0.5m. The hole remains unstable, but the collapse of the fill is monitored to document fill performance and the fill is periodically topped up. After the sinkhole was filled, the road adjacent to the sinkhole was re-opened and the site of the sinkhole fenced. The severely damaged Field View house remains standing next to the sinkhole. The nearby Victorian Ure lodge was not directly damaged by the 1997 sinkhole, but its western corner fell within the council-designated damage zone, and was left unoccupied. It fell into disrepair and was subsequently demolished (Figure 3b). A similar fate befell houses damaged by the surfacing of collapse sinkholes in and around Magdelen's Road, which is located a few hundred metres from Ure Terrace (Figure 3c-e). Shallow subcropping Zechstein gypsum (rehydrated anhydrite) occurs in two subcropping bedded units in this area, one is in the Permian Edlington and the other is in the Roxby Formation (Figure 4b). Together they form a subcrop belt about a kilometre wide, bound to the west by the base of the lowest gypsum unit (at the bottom of the Edlington Formation) and to the east by a downdip transition from gypsum to anhydrite in the upper gypsum-bearing unit of the Roxby Formation. The spatial distribution of subsidence features within this belt relates to joint azimuths in the Permian bedrock, with gypsum maze caves and subsidence patterns following the joint trends (Cooper, 1986). Most of the subcropping gypsum is alabastrine in the area around Ripon, while farther to the east, where the unit is thicker and deeper, the calcium sulphate phase is still anhydrite.

Fluctuations in the watertable level tied to heavy rain or long drought are thought to be the most common triggering mechanism for subsidence transitioning to sinkhole collapse. Many of the more catastrophic collapses occur after river flooding and periods of prolonged rain, which tend to wash away cavern roof span support. Subsidence is also aggravated by groundwater pumping; first, it lowers the watertable and second, it induces considerable crossflow of water in enlarged joints in the gypsum. When recharged by a later flood, the replacement water is undersaturated with respect to gypsum.

Thomson et al. (1996) recognised four hydrogeological flow units driving karst collapse in the Ripon area (Figure 4):

1) Quaternary gravels in the buried valley of the proto-River Ure

2) Sherwood Sandstone Group

3) Magnesian Limestone of the Brotherton Fm. and the overlying/adjacent gypsum of the Roxby Fm.

4) Magnesian limestone of the Cadeby Fm. plus the overlying/adjacent gypsum of the Edlington Fm.

Local hydrological base level within this stratigraphy is controlled by the River Ure, especially where the buried Pleistocene valley (proto-Ure) is filled by permeable sands and gravels, as these unconsolidated sediments, when located atop a breached roof beam, are susceptible to catastrophic stoping to base level (Figure 4). In the area around Ripon the palaeovalley cuts down more than 30 m, reaching levels well into the Cadeby Formation, so providing the seepage connections or pathways between waters in all four units wherever they intersect the palaeovalley. There is considerable groundwater outflow along this route with artesian sulphate-rich springs issuing from Permian strata in contact with Quaternary gravels of the buried valley (Cooper, 1986, 1995, 1998).

The potentiometric head comes from precipitation falling on the high ground of the Cadeby formation to the west and the Sherwood Sandstone to the east. Groundwater becomes largely confined beneath glacial till as it seeps toward the Ure Valley depression, but ultimately finds an exit into the modern river via the deeply incised sand and gravel-filled palaeovalley of the proto-Ure. Waters recharging the Ure depression pass through and enlarge joints and caverns in the gypsum units of the Edlington and Roxby Formations, so the highest density of subsidence features are found atop the sides of the palaeovalley. This region has the greatest volume of artesian discharge from aquifers immediately beneath the dissolving gypsum bed. Although created as an active karst valley, the apparent density of subsidence hollows is lower on the present Ure River floodplain than the surrounding lands as floodplain depressions are constantly filled by overbank sediments (Figure 4b).

Cooper (1998) defined 16 sinkhole variations in the gypsum subsidence belt at Ripon, all are types of entrenched, subjacent and mantled karst. Changes in karst style are caused by; the type of gypsum, the nature and thickness of the overlying deposits, presence or absence of consolidated layers overlying the gypsum and the size of voids/caverns within the gypsum.

To the west of Ripon, the gypsum of the Edlington Formation lies directly beneath glacial drift. These unconsolidated drift deposits and the loose residual marl atop the dissolving gypsum gradually subside into a pinnacle or suffusion (mantled) karst. But between Ripon town and the River Ure, the limestone of the Brotherton Formation overlies the Edlington Formation. There the karst develops as large open caverns beneath strong roof spans (entrenched karst). Ultimate collapse of the roof span creates rapid upward-stoping caverns in loosely consolidated sediment. Stopes break though to the surface as steep-sided collapse dolines or chimneys with sometimes catastrophic results. A similar entrenched situation is found east of the Ure River but there karstified gypsum units of both the Edlington and the Roxby formations are involved.

There are also thick beds of gypsum in the Permian Zechstein sequence that forms the bedrock in the Darlington area. In this area, subsidence features attributed to gypsum dissolution are typically broad shallow depressions up to 100 m in diameter, and the ponds, known as Hell Kettles, are the only recognized examples of steep-sided subsidence hollows around Darlington (Figure 5). Historical records suggest that one of the ponds formed in dramatic fashion in AD 1179 (Cooper 1995). The southern pond appears to be the most likely one to have formed at that time because it is many metres deep and is fed from below by calcareous spring water that is rich in both carbonate and sulphate. The 2D profiles have revealed evidence of foundering in the limestone of the Seaham Formation at depths of c. 50 m (Figure 5; Sargent and Goulty, 2009). The foundering is interpreted to have resulted from dissolution of gypsum in the Hartlepool Anhydrite Formation at ≈ 70 m depth. The reflection images of the gypsum itself are discontinuous, suggesting that its top surface has karstic topography. The 3D survey also acquired and interpreted by Sargent and Goulty (2009) reveals subcircular hollows in the Seaham Formation up to 20 m across, which are again attributed to foundering caused by gypsum dissolution.

Problems with Miocene gypsum, Spain

Karstification has led to problems in areas of subcropping Miocene gypsum in the Ebro and Calatayud basins, northern Spain (Figure 6). Cliff sections and road cuts indicate the widespread nature of karstification in the gypsum outcrops and subcrops in Spain (Figure 7b) Areas affected are defined by subsidence or collapse in Quaternary alluvial overburden and include; urban areas, communication routes, roads, railways, irrigation channels and agricultural fields (Figure 7a; Soriano and Simon, 1995; Elorza and Santolalla, 1998; Guerrero et al., 2013; Gutiérrez et al., 2014). In the region there can be a reciprocal interaction between anthropic activities and sinkhole generation, whereby the ground disturbance engendered by human activity accelerates, enlarges and triggers the creation of new sinkholes. Subsidence is particularly harmful to linear constructions and buildings and numerous roads, motorways and railways have been damaged (Figure 7a, b). Catastrophic collapse and rapid karst chimneying into roads and buildings can have potentially fatal consequences. For example, several buildings have been damaged around the towns of Casetas and Utebo. In the Portazgo industrial estate some factories had to be pulled down due to collapse-induced instability (Castañeda et al., 2009). A nearby gas explosion was attributed to the breakage of a gas pipe caused by subsidence. The local water supply is also disrupted by subsidence and pipe breakage so that 20,000 inhabitants periodically lose their water supply. The most striking example of subsidence affecting development comes from the village of Puilatos, in the Gallego Valley. In the 1970's this town was severely damaged by subsidence and abandoned before it could be occupied (Cooper 1996).

Collapse affects irrigation channels in the countryside with substantial economic losses (Elorza and Santolalla, 1998). In 1996 a doline collapse surfaced and cut the important Canal Imperial at Gallur village. New dolines often form near unlined irrigation canals. The ongoing supply of fresh irrigation waters to field crops can also encourage sinkhole generation in the fields. Though not directly visible, natural sinkholes also form in the submerged beds of river channels cutting regions of subcropping gypsum.

On December 19th, 1971, a bus fell from a bridge into the Ebro River at Zaragoza, near where the ‘San Lazaro well’ (a submerged gypsum sinkhole) is located (Figure 8a). Ten people lost their lives in this accident , while the remainder of the passengers were rescued, after being stranded on the bus roof in the flowing river for some hours (Figure 8b). After survivors were rescued, river waters washed the bus from the foot of the bridge supports into the nearby 'San Lazaro well (collapse sinkhole) in the water-covered floor of the river. Nine of the ten bodies in the bus were never found, although the bus was later recovered from the sinkhole. Locals suggested that bodies were carried deeper into the various interconnect phreatic sinkhole caverns fed by this losing stream.

Karstification in the Zaragoza region is characterised by the preferential intrastatal dissolution of glauberite bed, which are more soluble than the gypsum interbeds, this leads to collapse and rotation of gypsum blocks and river capture (Guerrero et al., 2013).

Sometimes even well-intentioned attempts to remediate culturally significant buildings under threat of evaporite karst collapse can exacerbate collapse problems. Gutiérrez and Cooper (2002) cite examples from the city of Calatayud, Spain. Subsidence-induced differential loading across doline edges drives the tilting of the 25-metre high tower (mudéjar) of the San Pedro de Los Francos church, which leans towards and overhangs the street by about 1.5 metres. (Figure 9) In places, the brickwork of the church indents the pre-existing tower fabric, which probably dates from the 11th Century or the beginning of the 12th Century. This indentation and the non-alignment of the church and the tower walls indicates that most of the tower tilting occurred prior to the construction of the church. In 1840, the upper 5m of the tower was removed and the lower part buttressed for the safety of the Royal family, who visited the town and stayed in the palace opposite. On 3rd June 1931, San Pedro de Los Francos church was declared a “Monument of Historical and Artistic value.” Due to its ruinous condition, the church was closed to worship in 1979. Micropiling to improve the foundation was started in 1994, but this corrective measure was interrupted when only half of the building was underpinned. Very rapid differential settlement of the building took place in the following year, causing extensive damage and aggravating the subsidence problem.

Colegiata de Santa María la Mayor was constructed between the 13th and 18th centuries, it has an outstanding Mudéjar (a 72 m high tower) and numerous Renaissance features; it is considered the foremost monument in the city of Cataluyud. As with the San Pedro de los Francos Church, recent micropiling work, applied to only one part of the cloister, has been followed by alarming differential movements that have drastically accelerated the deterioration of the building. Large blocks have fallen from the vault of the “Capitular Hall” and cracks up to 150 mm wide have opened in the brickwork of the back (NW) elevation, which has now been shored up for safety. The dated plaster tell-tales placed in these cracks to monitor the displacement demonstrate the high speed of the deformation produced by subsidence in recent years. On the afternoon of 10 September 1996, the fracture of a water supply pipe flooded the cloisters and the church with 100 mm of muddy water. Ten years earlier a similar breakage and flood had occurred. These breaks in the water pipes are most likely related to karst-induced subsidence. Once they occur, the massive input of water to the subsurface may trigger further destruction via enhanced dissolution, piping and hydrocollapse (Gutiérrez and Cooper, 2002).

Gypsum karst in Mosul, Iraq

A similar quandary of multiple areas of structural damage from gypsum-induced subsidence affects large parts or the historic section of the city of Mosul in northern Iraq (Jassim et al., 1997). The main part of its old quarter is over a century old and some buildings are a few hundred years old. Mosul lies on the northeastern flank of the Abu Saif anticline and near to its northern plunge (Figure 10a). It was built on the western bank of the Tigris River on a dip slope of Middle Miocene Fatha limestone that is directly underlain by bedded gypsum and green marl (equivalent to Lower Fars Formation). Houses in the old city were built on what seemed to be at the time a very sound rock foundation.

Water distribution in the city was done on mule back in the early part of last century and the estimated water consumption did not exceed 10 litres per person per day (Jassim et al., 1997). Discharge from households was partly to surface drainage and partly to shallow and small septic tanks. The modern piped system of water distribution did not start until the 1940s, resulting in a sudden increase in water consumption (presently around 200 litres per person per day) and it was not associated with a complementary sewer system. Increased water consumption meant larger and deeper septic tanks were dug at the perimeter of buildings (which never seemed to fill) resulting in a dramatic increase in water percolating downwards, water that was also more corrosive than previously due to the increased use of detergents and chlorination. This water passes through the permeable and fractured limestone to the underlying gypsum. On its way through the limestone it enlarges and creates new dissolution cavities, but eventually finds its way into the older gypsum karst maze, which is then further widened as water drains back into the Tigris (Figure 10b). Caverns in the gypsum enlarge until the roof span collapses. Since the 1970s more and more buildings in the old city have fractured and many are subject to sudden collapse. The problem is further intensified due to the expansion of the city in the up-dip direction (west and southwest) including the construction of industrial, water-dependent centres with integrated drainage. Water seeping/draining from these newly developed up-dip areas eventually passes under the old city before discharging in the Tigris river. The process was slightly arrested in the 1980s by the completion of a drainage system for the city, but the degradation of the old city continues.

Coping: man-made structures atop salts

The towns of Ripon in the UK and Pasvales and Birzai in Lithuania house some 45,000 people, who currently live under the ongoing threat of catastrophic subsidence, caused by natural gypsum dissolution (Paukstys et al., 1999). Special measures for construction of houses, roads, bridges and railways are needed in these areas and should include: incorporating several layers of high tensile heavy duty reinforced plastic mesh geotextile into road embankments and car parks; using sacrificial supports on bridges so that the loss of support of any one upright will not cause the deck to collapse; extending the foundations of bridge piers laterally to an amount that could span the normal size of collapses; and using ground monitoring systems to predict areas of imminent collapse (Cooper 1995, 1998).

Dams to store urban water supplies are costly structures and failure can lead to disaster, large scale mortality and financial liability (for example, Cooper and Gutiérrez, 2013). For example, at two and a half minutes before midnight on March 12, 1928, the St. Francis Dam (California) failed catastrophically and the resulting flood killed more than 400 people (Figure 11). The collapse of the St. Francis Dam is considered to be one of the worst American civil engineering disasters of the 20th century and remains the second-greatest loss of life in California’s history, after the 1906 San Francisco earthquake and fire. The collapse was partly attributed to dissolution of gypsum veins beneath the dam foundations. The Quail Creek Dam, Utah, constructed in 1984 failed in 1989, the underlying cause being an unappreciated existence of, and consequent enlargement of, cavities in the gypsum strata beneath its foundations.

Unexpected water leakage from reservoirs, via ponors, sinkholes and karst conduits, leads to costly inefficiency, or even project abandonment. Unnaturally high hydraulic gradients, induced by newly impounded water, may flush out of the sediment that previously blocked karst conduits. It can also produce rapid dissolutional enlargement of discontinuities, which can quickly reach break-through dimensions with turbulent flow. These processes may significantly increase the hydraulic permeability in the region of the dam foundation, on an engineering time scale.

Accordingly, numerous dams in regions of the USA underlain by shallow evaporites either have gypsum karst problems, or have encountered gypsum-related difficulties during construction (Johnson, 2008). Examples include; the San Fernando, Dry Canyon, Buena Vista, Olive Hills and Castaic dams in California; the Hondo, Macmillan and Avalon dams in New Mexico; Sandford Dam in Texas; Red Rock Dam in Iowa; Fontanelle Dam in Oklahoma; Horsetooth Dam and Carter Dam in Colorado and the Moses Saunders Tower Dam in New York State. Up to 13,000 tonnes of mainly gypsum and anhydrite were dissolved from beneath a dam in Iraq in only six months causing concerns about the dam stability (Figure 13). In China, leaking dams and reservoirs on gypsum include the Huoshipo Dam and others in the same area. The Bratsk Dam in eastern Siberia is leaking, and in Tajikistan the dam for the Nizhne-Kafirnigansk hydroelectric scheme was designed to cope with active gypsum dissolution occurring below the grout curtain. Gypsum karst in the foundation trenches of the Casa de Piedra Dam, Argentina and El Isiro Dam in Venezuela, caused difficult construction conditions and required design modifications.

Another illustration of the problems associated with water retaining structures and the ineptitude, or lack of oversight, by some city planners comes from the town of Spearfish, South Dakota (Davis and Rahn, 1997 ). As discussed earlier in this chapter, the Triassic Spearfish Formation contains numerous gypsum beds in which evaporite-focused karst landforms are widely documented across its extent in the Black Hills of South Dakota (Figure 12). The evaporite karst in the Spearfish Fm. has caused severe engineering problems for foundations and water retention facilities, including wastewater stabilization sites. One dramatic example of problems in water retention atop gypsum karst comes from the construction in the 1970s of now-abandoned sewage lagoons for the City of Spearfish.

Despite warnings from local ranchers, the Spearfish sewage lagoons were built in 1972 by city authorities on alluvium atop thick gypsum layers of Spearfish Formation. Ironically, at one point during lagoon construction, a scraper became stuck in a sinkhole and required four bulldozers to pull it out. Once filled with sewage, within a year the lagoons started leaking badly; the southern lagoon was abandoned after four years because of ongoing uncontrollable leaks, and the northern lagoon did not completely drain, but could not provide adequate retention time for effective sewage treatment. Attempts at repairs, including a bentonite liner, were ineffective, and poorly treated sewage discharged beneath the lagoon’s berm into a nearby surface drainage. The lagoons were abandoned completely in 1980. This was after a US $27-million lawsuit was filled in 1979 by ranchers whose land and homes were affected by leaking wastewater. A mechanical wastewater treatment plant was constructed nearby on an outcrop of the non-evaporitic Sundance Formation. The engineering firm that designed the facility without completing a knowledgeable geological site survey was reorganised following the lawsuit.

Likewise, the development of Chamshir Dam atop Gascharan Formation outcrop and subcrop in Iran is likely to create ongoing infrastructure cost and water storage problems (Torabi-Kaveh et al., 2012). The site is located in southwest of Iran, on Zuhreh River, 20 km southeast of Gachsaran city. The area is partially covered by evaporite formations of the Fars Group, especially the Gachsaran Formation. The dam axis is located on limestone beds of Mishan Formation, but nearly two-thirds of the dam reservoir is in direct contact with the evaporitic Gachsaran Formation. Strata in the vicinity of the reservoir and dam site have been brecciated and intersected by several faults, such as the Dezh Soleyman thrust and the Chamshir fault zone, which all act in concert to create karst entryways, including local zones of suffusion karst. A wide variety of karstic features typify the region surrounding the dam site and include; karrens, dissolution dolines, karstic springs and cavities. These karst features will compromise the ability of Chamshir Dam to store water, and possibly even cause breaching of the dam, via solution channels and cavities which could allow significant water flow downstream of the dam reservoir. As possible and likely partial short term solutions, Torabi-Kaveh et al. (2012) recommend the construction of a cutoff wall and/or a clay blanket floor to the reservoir

Difficulties in building hydraulic structures on soluble rocks are many, and dealing with them greatly increases project and maintenance costs. Gypsum dissolution at the Hessigheim Dam on the River Neckar in Germany has caused settlement problems in sinkholes nearby. Site investigation showed cavities up to several meters high and remedial grouting from 1986 to 1994 used 10,600 tonnes of cement. The expected life of the dam is only 30-40 years, with continuing grouting required to keep it serviceable.

Grouting costs in zones of evaporite karst can be very high and may approach 15 or 20% of the dam cost, currently reaching US$ 100 million in some cases. In karstified limestones grouting is difficult, yet in gypsum it is even more difficult due to the rapid dissolution rate of the gypsum. Karst expansion in limestone occurs on the scale of hundreds of years, in gypsum it can be on the order of a decade or less. Grouting may also alter the underground flow routes, so translating and focusing the problems to other nearby areas. In the Perm area of Russia, gypsum karst beneath the Karm hydroelectric power station dam has perhaps been successfully grouted, a least in the short term, using an oxaloaluminosilicate gel that hardens the grout, but also coats the gypsum, so slowing its dissolution. The Mont Cenis Dam, in the French Alps, is not itself affected by the dissolution of gypsum. However, the reservoir storage zone is leaking and photogrammetric study of the reservoir slopes showed ongoing doline activity over gypsum and subsidence in the adjacent land.

Probably the worst example tied to and evaporite karst hazard is the significant dam disaster waiting to happen that is the Mosul Dam in Iraq (Figure 13; Kelley et al., 2007; Sissakian and Knutsson, 2014; Milillo et al., 2016). It is ranked as the fourth largest dam in the Middle East, as measured by reserve capacity, capturing snowmelt from Turkey, some 70 miles (110 km) north. Built under the despotic regime of Saddam Hussein, completed in 1984 the Mosul Dam (formerly known as Saddam Dam) is located on the Tigris river, some 50 km NW of Mosul.

The design of the dam was done by a consortium of European consultants (Sissakian and Knutsson, 2014), namely, Swiss Consultants group, comprising: Motor Columbus; Electrowatt; Suiselectra; Societe Generale pour l’Industrie. The construction was carried out by a German-Italian consortium of international contractors, GIMOD joint venture, comprising: Hochtief; Impregilo; Zublin; Tropp; Italstrade; Cogefar. The consultants for project design and construction supervision comprised a joint venture of the above listed Swiss Consultants Group and Energo-Projekt of Yugoslavia, known as MODACON.

As originally constructed the dam is 113 m in height, 3.4 km in length, 10 m wide in its crest and has a storage capacity of 11.1 billion cubic meters (Figure 13b). It is an earth fill dam, constructed on evaporitic bedrock atop a karstified high created by an evaporite cored anticline in the Fat’ha Formation, which consists of gypsum beds alternating with marl and limestone (Figure 13a, 14). To the south, this is same formation with the same evaporite cored anticlinal association that created all the stability problems in the city of Mosul (Figures 10). The inappropriate nature of the Fat’ha Formation as a foundation for any significant engineering structure had been known for more than a half a century. Then again, absolute rulers do not need to heed scientific advice or knowledge. Or perhaps he didn’t get it from a well-paid group of Swiss-based engineering consultants. As Kelley et al. (2007) put it so succinctly....“The site was chosen for reasons other than geologic or engineering merit.”

The likely catastrophic failure of Mosul Dam will drive the following scenario (Sissakian and Knutsson, 2014); “... (dam) failure would produce a flood wave crest about 20 m deep in the City of Mosul. It is estimated that the leading edge of the failure flood wave would arrive in Mosul about 3 hours after failure of the dam, and the crest of the flood wave would arrive in Mosul about 9 hours after failure of the dam. The total population of the City of Mosul is about 3 million, and it is estimated that about 2 million people are in locations within the city that would be inundated by a 20 m deep flood wave. The City of Baghdad is located about 350 km downstream of Mosul Dam, and the dam failure flood wave will arrive after 72 hours in Baghdad and (by then) would be about 4 m deep.”

The heavily karsted Fat’ha Formation is up to 352 m thick at the dam and has an upper and lower member. The lower member is dominated by carbonate in its lower part (locally called “chalky series”) and is in turn underlain by an anhydrite bed known as the GBo. Gypsum beds typify its upper part,and the evaporite interval is capped by a limestone marker bed. The upper member, crops out as green and red claystone with gypsum relicts, around the Butmah Anticline. Thickness of individual gypsum beds below the dam foundations can attain 18 m; these upper member units are intensely karstified, even in foundation rocks, with cavities meters across documented during construction of the dam (Figure 14). Gypsum breccia layers are widespread within the Fatha Formation and have proven to be the most problematic rocks in the dam’s foundation zone. The main breccia body contains fragments or clasts of limestone, dolomite, or larger pieces of insoluble rocks of collapsed material. The upper portion of the accumulation grades upward from rubble to crackle mosaic breccia and then a virtually unaffected competent overburden. Breccia also may form without the intermediate step of an open cavity, by partial dissolution and direct formation of rubble. As groundwater moves through the rubble, soluble minerals are carried away, leaving insoluble residues of chert fragments, quartz grains, silt, and clay in a mineral matrix. These processes result in geologic layers with lateral and vertical heterogeneity on scales of micro-meters to meters.

High permeability zones in actively karsting gypsum regions can form rapidly, days to weeks, and quickly become transtratal. So predicting or controlling breakout zones via grouting and infill can be problematic (Kelley et al., 2007; Sissakian and Knutsson, 2014). For example, four sinkholes formed between 1992 and 1998 approximately 800 m downstream in the maintenance area of the dam (Figure 13a). The sinkholes appeared in a linear arrangement, approximately parallel to the dam axis. Another large sinkhole developed in February 2003, east of the emergency spillway when the pool elevation was at 325 m. The Mosul Dam staff filled the sinkhole the next day, with 1200 m3 of soil. Another sinkhole developed in July 2005 to the east of the saddle dam. Six borings were completed around the sinkhole and indicated that the sinkhole developed beneath overburden deposits and within layers of the Upper Marl Series. Another cause for concern at Mosul Dam in recent years is a potential slide area reported upstream of the dam on the west bank. The slide is most likely related to the movement of beds of the Chalky Series over the underlying GBo (anhydritic) layer.

To “cope” with ongoing active karst growth beneath and around the Mosul Dam, a continuous grouting programme was planned, even during dam construction, and continues today, on a six days per week basis. It pumps tens of thousands of tons of concrete into expanding karst features each year (Sissakian and Knutsson, 2014; Milillo et al., 2016). The dam was completed in June 1984, with a postulated operational life of 80 years. Due to insufficient grouting and sealing in and below the dam foundation, numerous karst features, as noted above, continue to enlarge in size and quantity, so causing serious problems for the ongoing stability of the dam. The increase in hydraulic gradient created by a wall of water behind the dam has accelerated the rate of karstification in the past 40 years.

Since the late 1980s, the status of the dam and its projected collapse sometime within the next few decades has created ongoing nervousness for the people of Mosul city and near surroundings. All reports on the dam since the mid 1980s have underlined the need for ongoing grouting and monitoring and effective planning of the broadcasting of a situation where collapse is imminent. For “Saddam’s dam” the question is not if, but when, the dam will collapse. To alleviate the effects of the dam collapse, Iraqi authorities have started to build another “Badush Dam” south of Mosul Dam so that it can stop or reduce the effects of the first flood wave. However this new dam has a projected cost in excess of US$ ten billion and so lies beyond the financial reach of the current Iraqi government. Problems related to the dam increased with the takeover of the region by the forces of ISIL.

Today, the Mosul dam is subsiding at a linear rate of ~15 mm/year compared to 12.5 mm/year subsidence rate in 2004–2010 (Milillo et al., 2016). Increased subsidence restarted at the end of 2013 after re-grouting operations slowed and at times stopped. The causes of the observed linear subsidence process of the dam wall can be found in the human activities that have promoted the evaporite–subsidence development, primarily in gypsum deposits and may enable, in case of continuous regrouting stop, unsaturated water to flow through or against evaporites deposits, allowing the development of small to large dissolution cavities.

Large vertical movements that typified the dam wall have resulted from the dissolution of extensive gypsum strata previously mapped beneath the Mosul dam. Increased subsidence rate over the past five years has been due to periods when there was little or no regrouting underlying the dam basement. Dam subsidence currently seems to follow a linear behavior but on can not exclude a future acceleration due to increased gypsum dissolution speed and associated catastrophic collapse of the dam (Milillo et al., 2016).

Given the existing geologic knowledge base in the 1980s, in my opinion, one must question the seeming lack of understanding in a group of well-paid consultant engineering firms as to the outcome of building such a major structure, atop what was known to be an active karstifying gypsum succession, sited in a location where failure will threaten multimillion populations in the downstream cities. The same formation that constituted the base to the Mosul dam was known at the time to be associated with ground stability problems atop similar gypsum-cored anticlines in the city of Mosul to the south. Even more concerning to the project rationale should have been the large karst cavities in highly soluble gypsum that were encountered a number of times during feasibility and construction of the dam foundations (Figure 14). Or, perhaps, as Lao Tzu observed many centuries ago, “ ...So the unwanting soul sees what’s hidden, and the ever-wanting soul sees only what it wants.”

Canals, like dams, that leak in gypsum karst areas can trigger subsidence, which can be severe enough to cause retainment failure. In Spain, the Imperial Canal in the Ebro valley, and several canals in the Cinca and Noguera Ribagorzana valleys, which irrigate parts of the Ebro basin, have on numerous occasions failed in this way. Similarly, canals in Syria have suffered from gypsum dissolution and collapse of soils into karstic cavities. Canals excavated in such ground may also alter the local groundwater flow (equivalent to losing streams) and so accelerate internal erosion, or the dissolution processes and associated collapse of cover materials. In the Lesina Lagoon, Italy, a canal was excavated to improve the water exchange between the sea and the lagoon. It was cut through loose sandy deposits and highly cavernous gypsum bedrock, but this created a new base level, so distorting the local groundwater flow. The canal has caused the rapid downward migration of the cover material into pre-existing groundwater conduits, producing a large number of sinkholes that now threaten an adjacent residential area.

Pipelines constructed across karst areas are potential pollution sources and some may pose possible explosion hazards. The utilization of geomorphological maps depicting the karst and subsidence features allied with GIS and karst databases help with the grouting and management of these structures. In some circumstances below-ground leakage {Zechner, 2011 #26} from water supply pipelines can trigger severe karstic collapse events. Where such hazards are identified, such as where a major oil and gas pipeline crosses the Sivas gypsum karst in Turkey, the maximum size of an anticipated collapse can be determined and the pipeline strength increased to cope with the possible problems.

Solving the problem?

Throughout the world, be it in the US, Canada, the UK, Spain, eastern Europe, or the Middle East, it is a fact that weathering of shallow gypsum forms rapidly expanding and stoping caverns, especially in areas of high water crossflow, unsupported roof beams, and unconsolidated overburden and in areas of artificially confined fresh water. Rapid karst formative processes and mechanism will always be commonplace and widespread (Table 2). Resultant karst-associated problems can be both natural and anthropogenically induced or enhanced. It is fact that natural solution in regions of subcropping evaporites is always rapid, and even more so in areas where it is encouraged by human activities, especially increased cycling of water via damming, groundwater pumping, burst pipes, septic systems, agricultural enhancement and uncontrolled storm and waste water runoffs to aquifers.

Typically, the best way to deal with a region of an evaporite karst hazard is to map the regional extent of the shallow evaporite solution front and avoid it (Table 3). In established areas with a karst problem the engineering solutions will need to be designed around hazards that will typically be characterised by short-term onsets, often tied to rapid ground stoping/subsidence events and quickly followed by ground collapse. If man-made buildings of historical significance are to be restored and stabilized in such settings, perhaps it is better to wait until funds are sufficient to complete the job rather than attempt partial stabilization of the worst-affected portions of the feature. Significant infrastructure (including roads, canals and dams) should be designed to avoid such areas when possible or engineered to cope with and/or survive episodes of ground collapse.

A piecemeal approach to dealing with evaporite karst can intensify and focus water crossflows rather than alleviate them. In the words of Nobel prizewinner, Shimon Peres; “If a problem has no solution, it may not be a problem, but a fact - not to be solved, but to be coped with over time.”


Alberto, W., M. Giardino, G. Martinotti, and D. Tiranti, 2008, Geomorphological hazards related to deep dissolution phenomena in the Western Italian Alps: Distribution, assessment and interaction with human activities: Engineering Geology, v. 99, p. 147-159.

Amin, A., and K. Bankher, 1997b, Causes of land subsidence in the Kingdom of Saudi Arabia: Natural Hazards, v. 16, p. 57-63.

Amin, A. A., and K. A. Bankher, 1997a, Karst hazard assessment of eastern Saudi Arabia: Natural Hazards, v. 15, p. 21-30.

Biddle, P. G., 1983, Patterns of drying and moisture deficit in the vicinity of trees on clay soils: Geotechnique, v. 33, p. 107-126.

Castañeda, C., F. Gutiérrez, M. Manunta, and J. P. Galve, 2009, DInSAR measurements of ground deformation by sinkholes, mining subsidence, and landslides, Ebro River, Spain: Earth Surface Processes and Landforms, v. 34, p. 1562-1574.

Cooper, A. H., 1986, Subsidence and foundering of strata caused by the dissolution of Permian gypsum in the Ripon and Bedale areas, North Yorkshire: Harwood, Gill M., Smith, Denys B. The English Zechstein and related topics. Univ. Newcastle upon Tyne, Newcastle upon Tyne, United Kingdom. Geological Society Special Publications, v. 22, p. 127-139.

Cooper, A. H., 1995, Subsidence hazards due to the dissolution of Permian gypsum in England: Investigation and remediation, in B. F. Beck, ed., Karst Geohazards - Engineering and Environmental Problems in Karst Terrane. Proceedings of the fifth multidisciplinary conference on sinkholes and the environmental impacts of karst, Gatlinburg, Tennessee: Rotterdam, A.A. Balkema, p. 23-29.

Cooper, A. H., 1998, Subsidence hazards caused by the dissolution of Permian gypsum in England: geology, investigation and remediation, in J. G. Maund, and M. Eddleston, eds., Geohazards in Engineering Geology, v. 15: London, Geological Society, London, p. 265-275.

Cooper, A. H., and F. Gutiérrez, 2013, Dealing with gypsum karst problems: hazards, environmental issues, and planning, in J. F. Shroder, ed., Treatise on geomorphology, Elsevier, p. 451-462.

Cooper, A. H., and J. M. Saunders, 2002, Road and bridge construction across gypsum karst in England: Engineering Geology, v. 65, p. 217-233.

Cooper, A. H., and A. C. Waltham, 1999, Subsidence caused by gypsum dissolution at Ripon, North Yorkshire: Quarterly Journal of Engineering Geology, v. 32, p. 305-310.

Dahm, T., S. Heimann, and W. Bialowons, 2011, A seismological study of shallow weak micro-earthquakes in the urban area of Hamburg city, Germany, and its possible relation to salt dissolution: Natural Hazards, v. 58, p. 1111-1134.

Davis, A., and P. Rahn, 1997, Karstic gypsum problems at wastewater stabilization sites in the Black Hills of South Dakota: Carbonates and Evaporites, v. 12, p. 73-80.

Driscoll, R., 1983, The influence of vegetation on the swelling and shrinking of clay soils in Britain: Geotechnique, v. 33, p. 93-105.

Elorza, M. G., and F. G. Santolalla, 1998, Geomorphology of the Tertiary gypsum formations in the Ebro Depression (Spain): Geoderma, v. 87, p. 1-29.

Ford, D. C., 1997, Principal features of evaporite karst in Canada: Carbonates and Evaporites, v. 12, p. 15-23.

Frumkin, A., M. Ezersky, A. Al-Zoubi, E. Akkawi, and A.-R. Abueladas, 2011, The Dead Sea sinkhole hazard: Geophysical assessment of salt dissolution and collapse: Geomorphology, v. 134, p. 102-117.

Galve, J. P., F. Gutierrez, P. Lucha, J. Bonachea, J. Remondo, A. Cendrero, M. Gutierrez, M. J. Gimeno, G. Pardo, and J. A. Sanchez, 2009, Sinkholes in the salt-bearing evaporite karst of the Ebro River valley upstream of Zaragoza city (NE Spain) Geomorphological mapping and analysis as a basis for risk management: Geomorphology, v. 108, p. 145-158.

Garleff, K., H. Kugler, A. V. Poschinger, H. Sterr, H. Strunk, and G. Villwock, 1997, Germany, in C. Embleton, and C. Embleton, eds., Geomorphological hazards of Europe, Vol. 5. Developments in Earth Surface Processes, v. 5, p. 147-177.

Guerrero, J., F. Gutiérrez, and J. P. Galve, 2013, Large depressions, thickened terraces, and gravitational deformation in the Ebro River valley (Zaragoza area, NE Spain): Evidence of glauberite and halite interstratal karstification: Geomorphology, v. 196, p. 162-176.

Gutierrez, F., 2010, Hazards associated to karst (Chapter 13), in I. Alcántara-Ayala, and A. S. Goudie, eds., Geomorphological Hazards and Disaster Prevention, Cambridge University Press, p. 161-176.

Gutiérrez, F., 1996, Gypsum karstification induced subsidence - effects on alluvial systems and derived geohazards (Calatayud Graben, Iberian Range, Spain): Geomorphology, v. 16, p. 277-293.

Gutiérrez, F., 2014, Evaporite Karst in Calatayud, Iberian Chain, in F. Gutiérrez, and M. Gutiérrez, eds., Landscapes and Landforms of Spain: World Geomorphological Landscapes, Springer Netherlands, p. 111-125.

Gutiérrez, F., A. Cooper, and K. Johnson, 2008, Identification, prediction, and mitigation of sinkhole hazards in evaporite karst areas: Environmental Geology, v. 53, p. 1007-1022.

Gutiérrez, F., and A. H. Cooper, 2002, Evaporite dissolution subsidence in the historical city of Calatayud, Spain: Damage appraisal and prevention: Natural Hazards, v. 25, p. 259-288.

Gutiérrez, F., M. Parise, J. De Waele, and H. Jourde, 2014, A review on natural and human-induced geohazards and impacts in karst: Earth-Science Reviews, v. 138, p. 61-88.

Jassim, S. Z., A. S. Jibril, and N. M. S. Numan, 1997, Gypsum karstification in the Middle Miocene Fatha Formation, Mosul area, Northern Iraq: Geomorphology, v. 18, p. 137-149.

Johnson, K., 2008, Gypsum-karst problems in constructing dams in the USA: Environmental Geology, v. 53, p. 945-950.

Jones, C. J. F. P., and A. H. Cooper, 2005, Road construction over voids caused by active gypsum dissolution, with an example from Ripon, North Yorkshire, England: Environmental Geology, v. 48, p. 384-394.

Karacan, E., and I. Yilmaz, 1997, Collapse dolines in Miocene gypsum - An example from SW Sivas (Turkey): Environmental Geology, v. 29, p. 263-266.

Kelley, J. R., L. D. Wakeley, S. W. Broadfoot, M. L. Pearson, C. J. McGrath, T. E. McGill, J. D. Jorgeson, and C. A. Talbot, 2007, Geologic Setting of Mosul Dam and Its Engineering Implications: US Army Corps of Engineers; Engineer Research and Development Center Report ERDC TR-07-10.

Martinez, J. D., and R. Boehner, 1997, Sinkholes in glacial drift underlain by gypsum in Nova Scotia, Canada: Carbonates and Evaporites, v. 12, p. 84-90.

Milillo, P., R. Bürgmann, P. Lundgren, J. Salzer, D. Perissin, E. Fielding, F. Biondi, and G. Milillo, 2016, Space geodetic monitoring of engineered structures: The ongoing destabilization of the Mosul dam, Iraq: Nature Open Reports, v. 6, p. 37408.

Paukstys, B., A. H. Cooper, and J. Arustiene, 1999, Planning for gypsum geohazards in Lithuania and England: Engineering Geology, v. 52, p. 93-103.

Sargent, C., and N. R. Goulty, 2009, Seismic reflection survey for investigation of gypsum dissolution and subsidence at Hell Kettles, Darlington, UK: Quarterly Journal of Engineering Geology and Hydrogeology, v. 42, p. 31-38.

Shviro, M., I. Haviv, and G. Baer, 2017, High-resolution InSAR constraints on flood-related subsidence and evaporite dissolution along the Dead Sea shores: Interplay between hydrology and rheology: Geomorphology, v. 293, p. 53-68.

Sissakian, V., N. Al-Ansari, and S. Knutsson, 2014, Karstification Effect on the Stability of Mosul Dam and Its Assessment, North Iraq: Engineering and Mining Journal, v. 6, p. 84-92.

Sissakian, V. K., V. K. Al-Ansari, and S. Knutsson, 2015, Karst Forms in Iraq Journal of Earth Sciences and Geotechnical Engineering, v. 5, p. 1-26.

Soriano, M. A., and J. Simon, 1995, Alluvial dolines in the central Ebro basin, Spain: a spatial and developmental hazard analysis: Geomorphology, v. 11, p. 295-309.

Sprynskyy, M., M. Lebedynets, and A. Sadurski, 2009, Gypsum karst intensification as a consequence of sulphur mining activity (Jaziv field, Western Ukraine): Environmental Geology, v. 57, p. 173-181.

Stafford, K. W., W. A. Brown, T. Ehrhart. Jon, A. F. Majzoub, and J. D. Woodard, 2017, Evaporite karst geohazards in the Delaware Basin, Texas: review of traditional karst studies coupled with geophysical and remote sensing characterization: International Journal of Speleology, v. 46, p. 169-180.

Thierry, P., A. Prunier-Leparmentier, C. Lembezat, E. Vanoudheusden, and J. Vernoux, 2009, 3D geological modelling at urban scale and mapping of ground movement susceptibility from gypsum dissolution: The Paris example (France): Engineering Geology, v. 105, p. 51-64.

Tolmachev, V., A. Ilyin, B. Gantov, M. Leonenko, V. Khomenko, and I. A. Savarensky, 2003, The main results of engineering karstology research conducted in Dzerzhinsk, Russia (1952-2002), in B. Beck, ed., Sinkholes and the engineering and environmental impacts of karst: proceedings of the ninth multidisciplinary conference, September 6-10, 2003, Huntsville, Alabama, American Society of Civil Engineers, p. 502-516.

Tolmachev, V., and M. Leonenko, 2011, Experience in Collapse Risk Assessment of Building on Covered Karst Landscapes in Russia, in P. E. van Beynen, ed., Karst Management, Springer Netherlands, p. 75-102.

Torabi-Kaveh, M., M. Heidari, and M. Miri, 2012, Karstic features in gypsum of Gachsaran Formation (case study; Chamshir Dam reservoir, Iran): Carbonates and Evaporites, v. 27, p. 291-297.

Toulemont, M., 1984, Le karst gypseux du Lutetien superieur de la region parisienne; caracteristiques et impact sur le milieu urbain: Revue de Geologie Dynamique et de Geographie Physique, v. 25, p. 213-228.

Trzhtsinsky, Y., 2002, Human-induced activation of gypsum karst in the southern Priangaria (East Siberia, Russia): Carbonates and Evaporites, v. 17, p. 154-158.

Waltham, T., F. Bell, and M. Culshaw, 2005, Sinkholes and Subsidence: Karst and Cavernous Rocks in Engineering and Construction: Berlin Heidelberg, Springer Praxis Books, 382 p.

Wang, G., G. You, and Y. Xu, 2008, Investigation on the Nanjing Gypsum Mine Flooding, in H. Liu, A. Deng, and J. Chu, eds., Geotechnical Engineering for Disaster Mitigation and Rehabilitation: Proceedings of the 2nd International Conference GEDMAR08, Nanjing, China 30 May – 2 June, 2008: Berlin, Heidelberg, Springer Berlin Heidelberg, p. 920-930.

Warren, J. K., 2016, Evaporites: A compendium (ISBN 978-3-319-13511-3): Berlin, Springer, 1854 p.

Warren, J. K., 2017, Salt usually seals, but sometimes leaks: Implications for mine and cavern stabilities in the short and long term: Earth-Science Reviews, v. 165, p. 302-341.

Yaoru, L., and A. H. Cooper, 1997, Gypsum karst geohazards in China, in B. F. Beck, and J. B. Stephenson, eds., Engineering Geology and hydrogeology of Karst Terrains: Proceedings of the Sixth Multidisciplinary Conference on Sinkholes and the Engineering and Environmental Impacts of Karst Springfield, Missouri, 6-9 April 1997, Balkema, Rotterdam, p. 117-126.

Yilmaz, I., M. Marschalko, and M. Bednarik, 2011, Gypsum collapse hazards and importance of hazard mapping: Carbonates and Evaporites, v. 26, p. 193-209.

Zechner, E., M. Konz, A. Younes, and P. Huggenberger, 2011, Effects of tectonic structures, salt solution mining, and density-driven groundwater hydraulics on evaporite dissolution (Switzerland): Hydrogeology Journal, v. 19, p. 1323-1334.


Dissolving salt (1 of 5): Landforms

John Warren - Thursday, August 31, 2017



This series of five articles discuss effects of nearsurface evaporite dissolution at various scales and times, in terms of; 1) landform expression, 2) evaporite-hosted caves and speleothems 3) natural evaporite karst geohazards 4)Anthropogenically enhanced karst features, 5) economic associations tied to evaporite paleokarst (hydrocarbon and metal).

As salt dissolves in the subsurface, it creates void spaces of various sizes and shapes into which overlying strata can drape or brecciate, so creating characteristic landforms at both local (m to km) and regional (km to tens of km) scales (Figure 1). Water sources producing these features can be shallow unconfined meteoric or marine, confined shallow or deeply circulating meteoric, or hypogene, basinal and hydrothermal. Solution-related evaporitic landforms can be either active karst or palaeokarst features. Paleokarst refers to ancient karstic features that are no longer active and tied to ancient basin flushing hydrologies and now buried land surfaces. Active karst is responding to the modern ambient hydrology and typifies landscape elements discussed in the following section.

Evaporite solution karst landforms overprint nearby carbonate or evaporite strata, with many descriptive features and terms are common to textures and geometries across both host lithologies. Evaporite-related karst includes many varieties of karren, sinkholes (dolines and caves), blind valleys, poljes and subsidence basins (Ford and Williams, 2007; Warren, 2016). Bedrock exposures of gypsum or salt at a finer scale are sculpted into irregular curved and grooved surfaces called rillenkarren or other varieties of gravity-oriented solution flutes (Macaluso and Sauro, 1996; Stenson and Ford, 1993).

Local-scale karst geometries that define evaporite-related subsidence range in lateral extent from the metre-scale cones of suffosion karst to collapse cones to kilometre-scale doline depressions (Figure 1). In regions of soil-mantled outcrops of bedded evaporites, these features tend to occur at higher densities, within or near broad subsidence valleys. In the context of the lateral extents of evaporite-related landforms, it is also useful to separate landforms associated with the dissolution of bedded evaporite versus dissolution landforms and namakier residues that occur above variably active salt diapir crests. Features related to dissolving diapirs and namakiers are discussed in an earlier article (Salty Matters, March 10, 2015).

The geometry of the karst set-up hydrology is distinct in sediments above diapiric versus bedded evaporite substrate, and this article focuses on landforms related to a dissolving substrate made up of shallow dipping salt beds. The next section outlines characteristics feature of evaporite karst at the metre to kilometre scale, as defined in Figure 1. Then we will compartmentalise these features using the classification of Gutiérrez et al., 2008, and finally look at three of the more distinctive regions responding to the dissolution of subcropping evaporite beds.

Evaporite solution dolines

Dolines are closed circular to elliptical hollows or depressions, often funnel shaped, with diameters ranging from a few metres to a few kilometres and depths from a metre or so, to hundreds of metres (Figure 1; Ford and Williams, 2007). They indicate subsidence and collapse in underlying salt or carbonate units. Valley sides in larger subsidence dolines can be steep and expose karstified gypsum, or can be gentler slopes covered by soil.

Larger dolines may enclose one or more smaller sinkholes and can be further subdivided into, suffusion, subsidence sag, and collapse dolines (Figure 1). A collapse sinkhole is initiated by collapse of a roof span into an underlying solution cavity. A subsidence sag doline indicates a more diffuse broader and gentler lowering of the ceiling above the dissolving bed. Suffosion dolines are small-width high-density karst features (1-2m diameter) covered and filled with soil and debris that has washed or fallen into closely-spaced fissures cutting into the evaporite bed. They typically indicate that a dissolving salt mass is very close to the landsurface.

Dissolution, collapse and suffosion processes are more active, more rapid, more frequent and more noticeable in regions of shallow evaporite units, compared to carbonate terranes at the same depths. For shallow weathering evaporites versus carbonates, at roughly the same depths in the Perm and Bashkir regions of Russia, Gorbunova (1979) reported doline densities of 32 and 10/km2 respectively.

Suffosion dolines

Suffosion in a karst terrane describes the downward migration of unconsolidated cover deposits through voids. High levels of dispersed impurities (mostly clays and muds, often dolomitic and other dissolution residues) in a rapidly dissolving evaporite mass, means a soil or carapace of insoluble residues quickly covers the subcropping top and edges of a dissolving salt mass (Figures 1a, 2). Downwash transport into growing fissures and voids in the shallow underlying evaporite is via downwashing of fine particles carried by percolating waters, cohesionless granular flows, viscous sediment gravity flows (non-Newtonian), freefall of particles, and sediment-laden water. The carapace is continually undermined by ongoing rapid solution, as residual debris is continually and rapidly washed into the doline crevices into the growing cavities in the evaporite. This creates a unique soil-covered dimpled landscape, which is typified by a high-density doline terrain where the dissolving evaporite is only a metre or so below the landsurface.

Densities of up to 1000 suffosion dolines/km2 occur in many evaporite-cored fold axes or at subcrop contacts of evaporites with other outcropping lithologies. Densities of 1100-1500/km2 have been documented in evaporite karst in the Italian Alps and in regions to the west of Sivas, Turkey (Figure 3; Belloni et al., 1972; Kacaroglu et al., 1997). The inherently high solubility of evaporite salts explains similar densely-packed schlotten depressions and large karren shafts seen in gypsum karst of Antigonish County, Nova Scotia, where extrapolated doline densities in the latter case range up to 10,000/km2 (Martinez and Boeher, 1997). Extrapolated, because such dense networks do not extend over any more than a square kilometre or two and are typically found near retreating escarpment edges underlain by shallow subcropping salts. Such high densities tend to occur in more humid rather than arid areas, with high hydraulic gradients so that the resulting suffosion dolines have diameters around 5 m or less.

Subsidence Dolines

Solution (subsidence) dolines, as described in much of the literature tend to be larger down-warped doline craters or bowl-like subsidence depressions (Figure 1c). In contrast to the numerous small steep-sided suffosion and collapse dolines formed atop shallow subcropping shallow evaporites, these larger, bowl-like dolines can have diameters of 100-500 m and depression depths of 10-20 m or more. Most subsidence sag dolines have a well-developed soil cover and a thick sediment fill, with dissolving salt units found depths measured in tens to hundreds of metres below the land surface (Figure 4). Compared to collapse and suffusion dolines, subsidence or sag dolines have lower angle to near flat slopes into the deeper parts of the doline hollow, the doline walls at the outcrop level are usually hosted in non-evaporites, with the dissolving salt bed lying some distance below the landsurface. Compared to suffusion karst, subsidence dolines occur in regions of more deeply buried evaporites (tens of metres) including; Italy (Belloni et at. 1972; Burri 1986; Ferrarase et al., 2002), Spain (Gutiérrez, 1996) and the Pecos Valley of West Texas and New Mexico (Gustavson et al., 1982; Davies, 1984a, b; Quinlan et al., 1986). At the larger end of the scale of subsidence doline development, a subsidence doline merges into a subsidence basin (Figure 1d). The latter is a large solution hollow that had created enough accommodation space to be considered a small sedimentary basin, generally filling with varying combinations of fluvial lacustrine and other continental sediments.

Collapse dolines

Collapse dolines are steep-sided sinkholes, often defined by cave entrances that contain large blocks of roof material (Figure 1c). They form when solution of an underlying evaporite bed creates a roof span that can no longer be supported by the overlying lithology. Collapse doline walls are frequently asymmetrical; one wall is steep, and the other one is gentle (Figure 5). Soil covered doline floors, when not water covered, tend to display either concave-up or flat geometries. Apart from blocks of the collapsed roof span, active doline floors can be veneered by thin collapse breccias in a matrix of insoluble residues. The high density of dolines in areas of evaporite subcrop and the ubiquity of the associated breccias indicates the inherently higher solubility of the evaporite salts compared with interbedded and overlying carbonates. Active collapse sinkholes atop shallow bedded evaporites typify terminations of dry arroyos in many deserts and may also line up along active or former river courses, as in Bottomless Lake State Park, New Mexico (Figure 6).

Sinkhole Classification

Following the definitions of Waltham et al. (2005), Gutiérrez et al. (2008) constructed a nongenetic classification of subsidence dolines or sinkholes (Figure 7). It is based on two observations that refer to the material affected by downward gravitational movements (cover, bedrock or caprock) and the primary type of process involved (collapse, suffosion or sagging). The classification also applies to both evaporite and carbonate karst. The term “cover” refers to allogenic unconsolidated deposits or residual soil material, bedrock to karst rocks and caprock to overlying non-karst rocks. Collapse indicates the brittle deformation of soil or rock material either by brecciation or the downward migration deposits through conduits and its progressive settling, while sagging is the ductile flexure (bending) of sediments caused by the lack of basal support. In practice, more than one material type and several processes can be involved in the generation of sinkholes. These more complex sinkholes can be described using combinations of the proposed terms with the dominant material and/or process followed by the secondary one (e.g. cover and bedrock collapse sinkhole, bedrock sagging and collapse sinkhole).

The cover material may be affected by any of three subsidence mechanisms. The progressive corrosional lowering of the rockhead may cause the gradual settling of the overlying deposits by sagging, producing cover sagging sinkholes. An important applied aspect is that the generation of these sinkholes does not require the formation of cavities These depressions arc commonly shallow, have poorly defined edges and may reach several hundred metres across.

Cover deposits may migrate downward into fissures and conduits developed in the rockhead by action of a wide range of processes collectively designated as suffusion: downwashing of particles by percolating waters, cohesion-less granular flows, viscous gravity flows (non-Newtonian), fall of particles, and sediment-laden water flows. The downward transport of the cover material through pipes and fissures may produce two main types of sinkholes depending on the rheological behaviour of the mantling deposits. Where the cover behaves as a ductile or loose granular material, it may settle gradually as undermining by suffosion progresses. This creates sags and slumps in the overburden materials. Where cover behaves in a more brittle manner, collapse breccias form.

Sinkholes that result from the combination of several subsidence processes and affect more than one type of material are described by combinations of the different terms with the dominant material or process followed by the secondary one (e.g. bedrock sagging and collapse sinkhole). The mechanism of collapse includes any brittle gravitational deformation of cover and bedrock material, such as upward stoping of cavities by roof failure, development of well-defined failure planes and rock brecciation. They define suffosion as the downward migration of cover deposits through dissolutional conduits accompanied by ductile settling. Sagging is the ductile flexure of sediments caused by differential corrosional lowering of the rockhead or interstratal karstification of the soluble bedrock. Sag plays a major role in the generation of sinkholes across broad areas underlain by shallow dissolving evaporites (not included in previous genetic classifications mostly based on carbonate karst). Likewise, collapse processes are more significant in extent and rate in areas underlain by evaporites than in carbonate karst, primarily due to the order of magnitude greater solubility of the evaporites and the lower mechanical strength and ductile rheology of gypsum and salt rocks (Warren, 2016).

Broader-scale evaporite landforms

There are some unique aspects to regional landform and doline density associations above dissolving evaporites because; 1) halite, and to a lesser extent anhydrite, units are many times more soluble than carbonate or siliciclastic strata, and 2) matrix in most salt units away from its retreating edges tends to remain impervious even as salt masses approach the landsurface (Figure 8; melting block of ice analogy, see Warren, 2016). Such features form at the basin-scale in regions where an ancient salt bed is moving from the mesogenetic to the telogenetic realm, especially in regions where mountains are growing adjacent to the uplifted evaporite basin. The broader-scale landscape features not seen in carbonate karst in similar regional settings are; 1) laterally migrating subsidence basins (Figure 8a) and 2) regions of breccia chimneys (Figure 8b).

Subsidence basins

A subsidence basin is a large (tens of kilometres width) solution hollow that creates enough accommodation space to be considered a small sedimentary basin with widths measured in tens of kilometres (Figure 8a). In a continental setting, the basin is filling with varying combinations of fluvial and lacustrine sediments.

Subsidence troughs are large-scale elongate depositional depressions created by interstratal solution along the dissolving edge of shallow dipping ancient uplifted salt bodies. The largest solution-induced depositional basins tend to occur along the margins of the great interstratal halite deposits, creating a solution form that may be represented by a shallow retreating salt slope at the surface, with a laterally migrating monoclinal drape of beds in the vicinity of a salt scarp, which is defined the dissolving edge of the underlying salt bed or beds. Subsidence basins are filled or partly filled by clastic sediments (Olive, 1957; Quinlan, 1978; Simpson, 1988). If the subsidence zones atop a retreating salt edge lack a significant volume of sediment fill, it called is a subsidence trough, as seen in the vicinity of the outcropping edge of the Jurassic Hith Anhydrite in Saudi Arabia (see part 2).

Salt-edge leaching of shallow dipping salt underscores a set of self-perpetuating processes. Fractures created by the collapse of overburden and intrasalt beds contribute to an expanding accommodation space generated by salt cavities and the lateral retreat of the salt scarp. The ongoing salt dissolution provides new fissures and sinks that act as additional conduits for further percolation of meteoric or upwelling of undersaturated basinal waters into the salt. This, in turn, instigates yet more salt solution in the vicinity of the collapse and typically, so creating an elongate corridor of subsidence at the surface, which can extend for many kilometres parallel to the dissolving salt edge

Hence, large solution-induced depositional basins and monoclines outline the edges of the great interstratal salt deposits of the world. Depression corridors some 5-500 km long, 5-250 km wide, with up to 100 to 500 m of subsidence induced relief and sediment fill define saline karst plains along the edges of bedded and dissolving saline giants in the Devonian evaporite subcrop of Canada (Figure 9b; De Mille et al., 1964; Tsui and Cruden 1984; Christiansen and Sauer, 2002; Tozer et al., 2014; Broughton et al., 2017), the Permian Basin of New Mexico, Oklahoma and Texas (Figure 9a; Anderson, 1981; Davies, 1984a, b; Bachman, 1984), the Perm region of Russia (Gorbunova, 1979) and the Jurassic Hith region of Saudi Arabia (Amin and Bankher, 1997a, b;). Associated regolith and sediment infill typically reduces the regional solution edge landscape to a few tens of metres of relief.

Salt Breccia Chimneys

Undersaturated waters not only influence the dissolving edges of a shallow-dipping salt bed, but can also cut up through the salt as a series of salt chimneys, located well out in the salt basin, with diameters between 20 and 250 m. Most are plumb vertical, with reported heights ranging from tens of metres to kilometers. Higher examples have usually stoped upward through one or more cover formations, which may include siliciclastics, coal measures, extrusive volcanics, etc. (Figure 9b). Breccia pipes in an evaporite mass may exhibit one of four possible states (Figure 8b): a) Active, and propagating upwards towards the surface, with fault and collapse bounded edges, but not yet expressed at the surface - a blind chimney; b) Active or inactive, expressed at the surface as a closed depression or a depression with a surface outflow channel; c) Inactive, and buried by later strata - paleokarst chimney; d) Inactive and standing up as a positive relief feature on the landsurface, because the breccia (generally cemented) is more resistant to weathering than the upper cover strata - so forming a residual pipe with positive relief. Some upstanding features are firmly cemented and resistant structure, such as the castiles above an erosional plain in the Delaware Basin of Texas (Hill, 1990)

Downward flexure in the uppermost strata atop a growing pipe tends to form a bowl of subsidence and a sag doline structure.

In some intrasalt pipes, the halite or gypsum is entirely removed, leaving only an accumulation of insolubles and collapsed breccia blocks. In others, portions of the less soluble salts can remain in the pipe at the level of the mother salt. Mature pipes are typically filled by a jumble of intrasalt and suprasalt breccia blocks. In some, a lower brecciated zone is succeeded by an upper zone in which the overlying strata failed as a coherent block, settling downwards via a cylindrical pattern of steep to vertical faults with downthrows of up to 200 m. Such variably filled and zoned salt breccia chimneys have created one type of barren zone in the potash ores of the Prairie Evaporite in Canada, and we shall discuss this in detail in part 5(economic associations).

Most breccia pipes originate via a point-source dissolution breakthrough of a halite bed (occasionally gypsum or anhydrite) and are located above a fracture junction, an anticlinal crest, or a buried reef that channels groundwater into a local area in the salt (Figure 8b). Hence, salt breccia chimneys and pipes form best where there is an artesian head to create subsalt pressures. In the Western Canada Sedimentary Basin the head comes from the adjacent uplifted Rocky Mountains, while the Delaware and Guadalupe Mountains play a similar role in the Delaware Basin. A pipe creates anomalous chemical interfaces with adjacent and overlying strata and so may be targets for the later precipitation of economic ores, including uranium and base metals (part 5).

Delaware Basin, West Texas

The Delaware Basin of West Texas and southeast New Mexico, in the southwest portion of the Permian Basin, contains bedded evaporites of the Late Permian Castile, Salado and Rustler Formations (Figure 9a). Outcrops and subcrops of these three formations constitute an area of widespread subsidence troughs, collapse sinks, dolines and breccia chimneys, all created by ongoing removal of underlying bedded Permian salts. The Delaware is an eastward-dipping basin, mainly surrounded by the Capitan Reef and its equivalents (Figure 9a). The original extent of the evaporites in the basin was much greater than today due to ongoing salt dissolution.

The area called the “Gypsum Plain” of Texas lies to the west of the Pecos River and comprises about 2,600 square kilometres of subcropping gypsum of the Castile Formation (Olive, 1957; Quinlan, 1978; Kirkland and Evans, 1980; Anderson, 1981; Stafford et al., 2008a,b; Holt and Powers, 2010). South of Carlsbad and Carlsbad Caverns, the plain exhibits many small examples of solution subsidence troughs, typically 0.7–15 km in length, 100–1500 m wide, but no more than 5–10 m deep (Figure 9a: Stafford et al., 2008a). The plain is also where the “Castile” landforms are found and indicate an overprint of bacterial and thermochemical sulphate reduction (Kirkland and Evans, 1976). Additional gypsum outcrops are present to the east in the Rustler Hills and to the north in Reeves County, Texas. Permian strata in region of the "Gypsum Plain" once contained significant volumes of halite that were dissolved well before the evaporite succession reached the landsurface.

In the centre of the Delaware Basin, the thick evaporites of the Castile and Salado Formations retain their halite beds as do the thinner evaporites of the overlying Rustler Formation. All are underlain by relatively permeable carbonates and siliciclastics, including some prolific hydrocarbon reservoirs in Permian backreef of the Central Basin Platform (Figure 10). Reef mounds, fractures and faults in these underlying sediments have provided focused conduits for upward stoping breccia chimneys through the buried evaporites as well as the subsurface formation of now-exhumed Castiles to the west. An eastward-flowing deeply-circulating regional artesian hydrology, in combination with centripetal escape of buoyant hydrocarbon-rich basinal waters, drives the formation of these chimneys, with their surface expressions occurring in areas such as the Wink Sink (chain of breached chimneys) and the Gypsum Plain (Castiles).

The upper sides of the shallow dipping salt beds are also affected by the hydrologies of the zone of active phreatic circulation. The Pecos River has migrated back and forth across the top of subcropping evaporites for much of the Tertiary. Its ancestral positions drove substantial salt dissolution, now evidenced by large sediment-filled subsidence basins and troughs in the centre of the Delaware Basin (Figures 9b, 10). The regional eastward dip of the Delaware basin sediments means first halite, and then gypsum has disappeared along the updip eastern edge of the basin. Relatively undisturbed salt remains along the more deeply buried western side of the basin that abuts and covers the Central Basin Platform. Bachman and Johnson (1973) estimate the horizontal migration rate of the dissolution front across the basin as high as 10-12 km per million years so that more than 50% of the original halite is gone. Multiple smaller examples of sediment-filled subsidence troughs occur at the edge of the gypsum plain of the Delaware Basin south of Carlsbad Caverns, where depositional troughs, 0.7 to 15 km in length, 100 to 1,500 m wide and no more than 5-10 m deep, are well documented, as are subsidence sinks within the subsidence swales, such as Bottomless Lake, which is a region where the watertable intersects a collapse chimney (Figure 10c; Quinlan et al. 1986).

The San Simon swale is a 25 km2 depression defining a residual karst feature atop the Capitan Reef on the northeastern margin of the Basin (Figures 9a, 10a, b). San Simon Sink sits atop a subsidence chimney within the San Simon swale; it is the lowest point in the depression and is some 30 m deep and 1 km2 in area. It, in turn, encloses a secondary collapse sink some a few hundred metres across and 10 metres deep (Figure 10b). During a storm in 1918, the San Simon sinkhole formed as a gaping hole about 25 metres across and 20 metres deep in the lower part of the sink. In one night, nearly 23,000 cubic metres of soil and bedrock disappeared into the collapse cavern. Annular rings that cut the surface around the San Simon sinkhole today suggest ongoing subsidence and readjustment of the sinkhole is still occurring in response to earlier collapses. The position of the San Simon sink over the Permian reef crest led Lambert (1983) and many others to suggest that the sinkhole originated as a groundwater cavity breakthrough, atop a series of stoping reef-focused breccia pipes or chimneys. Sinkhole breakouts, which can emerge in a matter of hours, continue to form across this dissolution basin, in some case aided by poorly-monitored brine extraction operations and improperly-cased water wells. But the majority of the sinkholes in the Delaware Basin are natural, not anthropogenically enhanced (Land, 2013).

Nash Draw is a southwesterly trending depression or swale, some 25 km long and 5-15 km wide, at the northern end of the Delaware Basin with its sump in a salt lake (Poker Lake) at the southern end of the draw (Figure 11). The underlying evaporitic Rustler Formation and parts of the Salado Formation have largely dissolved so that more than 100 caves, sinks, fractures, swallow-holes, and tunnels make up a complex local karst topography in the Draw, which is still active today (Figure 11b, c; Bachman, 1981, Powers et al., 2006; Goodbar and Goodbar, 2014).

An extensive drilling program conducted for the nearby WIPP site (now a low-level radioactive waste repository) showed that natural dissolution of halite in the Rustler and upper Salado formations is responsible for the subsidence and overall formation of Nash Draw (Lambert, 1983; Holt and Powers, 2010). To the immediate west of Nash Draw, the WIPP/DOE drilling program defined the formation of a solution trough in the Dewey Lake Redbed; it was created by preferential leaching of halite beds in the Rustler Formation, with interstratal anhydrite and breccia residuals (Figure 12). This dissolution occurred at a level some 400 metres above the salt-encased storage level of the WIPP waste isolation facility and so is not considered a significant risk factor in terms of longterm site stability (Holt and Powers, 2010). A heated scientific (and at times not so scientific!) debate of just how deep surface karst penetrates into the bedded halite of the Salado Formation in the vicinity of the WIPP site continues today.

Further north, near the subcropping western edge of the Northwest Permian Shelf, is Bottomless Lakes State Park, located some 20 kilometres southeast of Roswell. Encircling the lakes are the gypsum, halite and dolomitic redbeds of the Artesia Group and San Andres Formation. Away from the lakes are numerous other sinkholes and collapse dolines, most of which are circular, steep walled or vertical holes, 50-100 metres across and 30-60 metres deep, with the greatest density of features aligned along the eastern side of the Pecos River floodplain (Figure 6). Water in the various sinks that make up the Bottomless Lakes is crystal clear and brackish to saline (6,000-23,000 ppm), attesting to its passage through subsurface layers of gypsum and salt. Although the lakes are around 30 metres deep, dark-green moss and algae coat the bottoms giving the impression of great depth and hence the name of the park (Lea Lake; Figure 10d.). To the west, many of the playas in depressions near Amarillo in the High Plains of Texas have a similar genesis as solution depressions atop dissolving Permian salt, but most do not intersect the regional watertable and so do not hold permanent surface water (Paine, 1994). Further south, near Carlsbad Caverns, there are other subsidence chimneys that form lakes where they intersect the watertable and so are also locally known as “Bottomless” (Figure 9a).

Western Canadian karst

The distribution and timing of chimneys and subsidence troughs created by the subsurface dissolution of the Prairie evaporite are well known and tied to the distribution of oil sands and the quality of potash ores (part 5). Dissolution drape features are more pronounced nearer the retreating edges of the thick multilayered subsurface Devonian salt succession while breccia chimneys occur above the buried salt successions (Figure 9b). For example, the Rosetown Low and the Regina Hummingbird Trough accumulated more than 100 metres of depression trough and drape sediment during interstratal dissolution of the underlying Prairie salt (Devonian) in southern Saskatchewan, especially in Cretaceous time (Figure 13a,b; DeMille et al., 1964; Simpson, 1988). This evaporite-related Cretaceous subsidence is also the principal control on the distribution of the Athabasca oil sands in subsidence basins along the eastern side of the evaporite extent (Tozer et al., 2014). Uplift of the ancestral Rocky Mountains likely created the potentiometric head that drove much of the subsalt aquifer flow. As in the Delaware Basin, the positions of sub-salt reefs and pinnacles focused many of the upwells of deeply circulated meteoric water that ultimately created solution breccia layers and breccia chimneys (Figure 8b).

In various circum-salt subsidence troughs, now filled or partially filled with late Mesozoic and Tertiary sediment thicks, the concurrence of a supra-unconformity thick, adjacent to the sub-unconformity feather-edge of a bedded salt sequence, is at a scale that is easily recognised in seismic and constitutes one of the classic signatures of a salt collapse-induced hydrocarbon trap (Figure 8a). For now, we will focus on the influence of dissolving evaporites on the modern Canadian landscape in regions of active karst, but we will return to the topic of economic hydrocarbon associations with paleokarst in this basin in part 5.

Evaporite karst domes, and laterally extensive solution breccia units, tied to a waxing and waning cover of glacial ice and permafrost, were first documented in Canada in northern Alberta and the Northwest Territories (e.g. De Mille et al., 1964). Karst domes remain as surface features of positive relief once the surrounding evaporite mass has completely dissolved and are outlined by megabreccia with caverns. Unlike breccia chimneys, the cores of evaporitic karst domes can expose blocks from below, as well as above the original bedded and folded evaporite level. The evaporitic karst domes in western Canada are related to the stratigraphic level of former bedded salts; elsewhere in the world others are the remains of now dissolved salt thrusts, diapirs and allochthons (see diapiric breccias and rauhwacke discussion in Warren, 2016). Hence, domes and residual units are dramatic landscape features in the gypsiferous terrain of northern Alberta, Canada (Wigley et al., 1973; Tsui and Cruden, 1984). Similar features, tied to the dissolution of bedded evaporites, typify the Arkhangelsk gypsum-residue karst region inland of the Barents Sea coast of Russia (Korotkov, 1974).

Canadian karst domes range from 10 to 1000 m or more in length or diameter, and can rise to 25 m above the surrounding land surface. Many domes are highly fractured, with individual overburden blocks displaced by heaving and sliding, with the residual gypsum showing well-developed flow foliation.

At the extreme end of disturbance and dissolution range, the domes breccias are megabreccias; positive relief features made up of collapsed or even upthrust jumbles of large blocks, some the size of houses. The largest reported Canadian megabreccia example is in a steep-limbed anticline that extends along the shore of Slave Lake for a distance of 30 km. It is up to 175m in height with a brecciated crestal zone that is marked by a ‘chaotic structure and trench-like lineaments’ (Aitken and Cook 1969).

These brecciated landforms develop upon a distinctive geological association in the central Mackenzie Valley region, NWT, at a stratigraphic level equivalent to anhydrites and halites in the deeper subsurface (Hamilton and Ford, 2002 ). In widespread outcrops the breccia unit is formally named the Bear Rock Formation, when covered by consolidated strata it is termed the Fort Norman Fm (Meijer Drees, 1993; Law, 1971). It is centred in Late Silurian-Early Devonian strata and defines an outcrop and subcrop belt more than 50,000 km2 in extent. In core, the Fort Norman Formation is 250-350 m or more in thickness. It consists of a thin upper limestone (Landry Member), a central Brecciated Member and, in some cores, undisturbed lower sequences of inter­bedded dolostones and anhydrite remnants. It is conformably overlain by 90-150 m of limestones and calcareous shales (Hume Formation-Eifelian). This evaporite dissolution breccia is possibly derived from the dissolution of not one, but several subcropping Devonian salt layers. That is, although not much discussed in the literature, the mother level may not be tied to a single stratigraphic layer.

Typical till-covered collapse and subsidence karst of the Bear Lake Formation can develop through the Hume and higher formations as a consequence of interstratal dissolution of the salt layers in the Fort Norman and equivalents, where meteoric groundwater circulation and sulphate dissolution have been recorded at core depths as great as 900 m (Figure, 14, 15).

In the NWT the Bear Rock Formation is considered to host to much of this widespread solution breccia, which up to 250 m thick. If present, the Landry Member is a brecciated limestone no more than 20 m thick. The main breccia level forms a visually set of outcrops made up of chaotic, vuggy mixtures of limestone, dolomite, and dedolomite clasts, variably cemented by later calcites, with small residual clasts and secondary encrustations of gypsum. Pack breccias in the Bear Creek and its equivalents displaying rubble fabric (predominant), crackle or mosaic fabric, are common and cliff-forming (Stanton, 1966; James & Choquette, 1988). Float breccias are rarer and tend to be recessive. Meijer Drees (1985) classifies the Bear Rock Formation as a late diagenetic solution breccia created by meteoric waters. Hamilton (1995) shows that calcite, dolomite and sulphate dissolution, plus dedolomitisation with calcite precipitation, are continuing today. They are re-working older breccia fabrics, creating new ones, as well as forming a suite of surficial karstic depressions and subsidence troughs ranging from metres to several km in scale.

The Canadian example constitutes an important set of observations that also relate to many other regions with widespread, basin-scale evaporite dissolution breccias, namely; evaporite solution breccias are multistage and can encompass significant time intervals. Similar-appearance evaporite solution breccias can form at different times in a basin's burial history, from different superimposed undersaturated hydrologies. Individual breccias in any single sample are typically responses to multistage, multi-time diagenetic-fluid overprints. This is also why potash ores in the Prairie evaporite preserve evidence of multiple times of potash mobilisation and mineralogy (part 5; Warren, 2016).

Detailed evaporite karst landform studies have focused on Bear Rock Mountain (type area) and the Mackay Range, which are outlying highlands on the east and west banks respectively of the Mackenzie River, and terrain between Carcajou Canyon and Dodo Canyon in the Mackenzie Mountains (Hamilton, 1995). These sites were covered by the Laurentide Ice Sheet (Wisconsinan) but were close to its western, sluggish margin. They are at the boundary between widespread and continuous permafrost in the ground today and can display some year-round groundwater circulation via taliks. Thaw/freeze and solifluction processes compete with ongoing evaporite dissolution to mould the topography.

Regionally, the principal karst landforms hosted in and above the Bear Rock Formation are varieties of sinkholes, blind valleys, solution-subsidence troughs and fault-bounded depressions (e.g. Figures 14, 15). Sinkholes range from single colla­pse features a few metres in diameter to merged or compound dolines up to 1.5 km2 and 100 m deep. Smaller individual sinkholes and collapse dolines may retain seasonal meltwater ponds, and there are permanent lakes draining to marginal ponors in some of the larger subsidence troughs. Blind valleys have developed where modern surface streams flow for several km into the evaporite karst zone from adjoining rocks. There are many relict, wholly dry valleys that may have been created by glacial meltwaters. Subsidence troughs have developed at the surface along contacts between the Bear Rock breccias and underlying, massive dolostones that are typically gently dipping. In the centre of the Mackay Range and at Bear Rock itself are solution depressions formed where the breccias make-up hanging-wall strata on steeply inclined fault planes on the collapse or pipe edge. The Mackay example is 3.2 km long, 1.0 km wide and-160 m deep (Ford, 1998).

Where patches of the Landry Member survive above the main breccia level(s), they are often broken into large, separate slabs that tilt into adjoining depressions in sharply differing directions, creating a very distinctive topography of dissolution draping and block rotation (see also Dahl Hit in Warren, 2016). Some slabs are rotated through 80-90°. Ridges (inter-sinkhole divides) that are wholly within the main breccias often display stronger cementation, represented by pinnacles as much as 30 m high The many sinking streams pass through the permafrost via taliks and emerge as sub-permafrost springs at stratigraphic contacts or topographic low points.

According to Hamilton (1995), the variety and intensity of karst landform assemblages on the Bear Rock Formation are like no other in Canada, and he notes he had not seen or read of very similar intensive karst topographies elsewhere. He attributes their distinctiveness to repeated evaporite dissolution and brecciation, with dedolomitisation and local case hardening, throughout the Tertiary and Quaternary, with these processes occurring multiple times in mountainous terrains subject to episodes of glaciation, permafrost formation and decay, and to vigorous periglacial action.

The spectacular karst domes that typify dolines and glacially associated surfaces atop an evaporite subcrop and are particularly obvious in regions of anticlinal salt-cored structures within regions of widespread permafrost. The accumulation of ground ice in initial fractures in the evaporite layer probably contributes to the heaving, folding and other displacement of breccia fragments in the dome. Tsui and Cruden (1984) attributed the examples that they studied in the salt plains of Wood Buffalo National Park Canada (Lat. 59-60°N) to hydration processes operating on subcropping bedded gypsum during the postglacial period. Ford and Williams (2007) argued such features indicate local injection of gypsum residuals during times of rapidly changing glacial ice loading. Whether they are created by glacial unloading/reworking or are a type of gravitationally-displaced dissolution breccia in a permafrost region is debatable; that in Canada they are a widespread type of evaporite karst residue is not. They characterise those parts of the permafrost-influenced region defined by dissolution of shallow subcropping folded Devonian salts in Northern Alberta, where the evaporites are typically exposed in anticlinal crests are today still retain fractionated gypsum residuals.

And so, in addition to widespread breccias hoisting karst domes, there are numerous natural collapse dolines atop dissolving shallow salt beds. A classic example is the water filled doline some km NE of Norman Wells (Figure 15).

Holbrook Anticline, Arizona

Subsidence driven by natural salt solution at depth generates regional-scale drape or monoclinal folding in strata atop the retreating salt edge. This is the corollary of the formation of a subsidence trough (Figure 16a). The 70-km long dissolution front in the Permian Supai salt of Arizona is defined by more than 500 sinkholes, fissures, chimneys and subsiding depressions some 40-50m across and 20-30m deep (Neal, 1995; Johnson, 2005). Away from the main dissolution zone the northeasterly-dipping Schnebly Hill Formation (aka Supai Salt) is composed of up to 150 metres of bedded halite, with local areas of sylvite along its northern extent (Figure 16a). Atop the solution front, there are a number of topographic depressions with playa lakes that in total cover some 300 km2. Salt-dissolution induced features to include areas known as; The Sinks, Dry Lake Valley, and the McCauley Sinks (Neal et al. 1998; Martinez et al. 1998). The Sinks region includes more than 20 steep-sided caprock sinkholes, with some that are more than 100 metres across and up to 30 metres deep (Figure 16b). The McCauley Sinks are the likely surface expression of compound breccia pipes and chimneys (Neal and Johnson, 2002).

Regional expanses of Supai Salt removal produce a regional gravity depression, largely coincident with a surface topographic depression, where there is as much as 100 metres of collapse and topographic displacement. The solution front is essentially coincident with the updip end of the Holbrook Anticline, a flexure defined by dip reversal in an otherwise northeasterly dipping succession (Figure 16a).

Rather than orogenically driven, the Holbrook Anticline is a subsidence-induced monoclinal flexure created by the northeasterly migrating dissolution front (Figure 16a). It may be the largest single solution-collapse fold structure in the world (Neal, 1995). The reverse dip of the flexure directly overlies the salt-dissolution front and marks the location of two major collapse depressions known as The Sinks and Dry Lake Valley, both occur where salt is within 300 m of the surface (Peirce, 1981). Although it has periodically held surface water, reports of several hundred acre-feet of flood water in the Dry Lake Valley playa draining overnight supports the notion of active fissure and cavern formation related to salt removal at depth. Major surface drainage events took place in 1963, 1979, 1984 and 1995, with more than 50 new sinkholes forming in the valley during that period. The continuing rapid appearance of new sinkholes testifies to the ongoing nature of dissolution in the underlying evaporites. According to Neal (1995), dissolution front features began forming in the landscape in the Pliocene and continue to form today.

The caprock collapse sinkholes centre in the Coconino Sandstone bed that overlie the salt, which is located 200-300 metres below the at-surface features. These caprock sinkholes define regions of focused breccia pipe development and discreet upward cavity migration (chimney stoping). The underlying salt is bedded and there are no indications of halokinesis anywhere in the basin. Sinkhole regions lie just ahead of a dissolution front that is migrating downdip, driven by the widespread dissolution of halite.

Other karst features attributed to evaporite dissolution in the Holbrook basin are; pull-apart fissures, graben sinks (downward-dropped blocks), breccia pipes and plugs, and numerous small depressions with and without sinkholes (Neal et al., 1998, 2001). Interestingly, many of the “karst” features occur in sandstones, not limestones; such caprock collapse sinkholes can form wherever pervasive dissolution has removed the underlying salt, independent of caprock lithology. The presence of the more than 500 karst features in the Holbrook basin, some of which formed in days, evokes practical karst hazard and infrastructure concerns, even in such a sparsely populated region (Martinez et al., 1998).


Evaporite-related karst landforms are in many ways similar carbonate karst features. But, the much higher solubilities of halite and other evaporite salts compared to limestone and dolomite means there are additional features unique to regions of salt dissolution.

In the subsurface, a bedded evaporite can be composed of thick impervious halite with intrasalt beds composed of anhydrite dolomite and calcite. The edges of this halite bed can dissolve even in the deep mesogenetic (burial) realm, wherever the edge of the salt is in contact with undersaturated waters. Insoluble residues bands start to form and can take the form of a basal anhydrite or a fractionated caprock. If the rise of undersaturated water is focused, a breccia cavern can form, and once it breaches the salt, the cavity will contain collapse blocks of the less soluble intrasalt beds. The cavity can then stope to the surface, forming a breccia pipe. A transtratal breccia pipe can rise through kilometres of overburden before attaining the surface. There is no equivalent process-response in the carbonate realm.Subsidence basins, troughs and megabreccia plains are also features that owe their origins to the rapid dissolution rates inherent to salt bed-fresher water contacts

In a similar fashion, the rapid dissolution of halite in the shallower parts of a salt basin means the evaporite karst features will transition from mesogenetic or bathyphreatic cavern formation to meteoric phreatic to vadose effects. This is the emphasis of the second article in this series, which will discuss processes forming vadose and phreatic caves in evaporites.


Aitken, J. D., and D. G. Cook, 1969, Geology, Lake Belot, District of Mackenzie: Geological Survey of Canada Map 6.

Amin, A. A., and K. A. Bankher, 1997a, Karst hazard assessment of eastern Saudi Arabia: Natural Hazards, v. 15, p. 21-30.

Anderson, E. J., 1981, Deep-seated salt dissolution in the Delaware Basin, Texas and New Mexico: in Wells, S.G., Lamber, W., and Callender, J.F., eds., Environmental geology and hydrology in New Mexico: Special Publication 10, New Mexico Geological Society, p. 133-145.

Bachman, G., and R. B. Johnson, 1973, Stability of Salt in the Permian Salt Basin of Kansas, Oklahoma, Texas and New Mexico: U. S. Geol. Surv. Open-File Rept.. v. 4339-4.

Bachman, G. O., 1981, Geology of Nash Draw, Eddy County, New Mexico.: U. S. Geol. Surv. Open-File Rept. 81-3.

Bachman, G. O., 1984, Regional geology of Ochoan evaporites, northern part of Delaware Basin: New Mexico Bureau of Mines and Mineral Resources, Circular, v. 184, p. 22.

Belloni, S., B. Martins, and G. Orombelli, 1972, Karst of Italy, in M. Herak, and V. T. Springfield, eds., Karst: Important karst regions of the Northern Hemisphere: Amsterdam, Elsevier, p. 85-128.

Broughton, P. L., and D. Cotterill, 2017, Breccia pipe and sinkhole linked fluidized beds and debris flows in the Athabasca Oil Sands: dynamics of evaporite karst collapse-induced fault block collisions: Bulletin of Canadian Petroleum Geology, v. 65, p. 200.

Burri, E., 1986, Various aspects of karstic phenomena in the urbanised area of Gissi and neighbouring areas (southern Abruzzo, Italy): Le Grotte d'Italia, v. 4, p. 143-161.

Christiansen, E. A., and E. K. Sauer, 2001, Stratigraphy and structure of a Late Wisconsinan salt collapse in the Saskatoon Low, south of Saskatoon, Saskatchewan, Canada: an update: Canadian Journal of Earth Sciences, v. 38, p. 1601-1613.

Davies, P. B., 1984a, DeepSeated Dissolution and Subsidence in Bedded Salt Deposits: doctoral thesis, Stanford University.

Davies, P. B., 1984b, Structural analysis of a deep seated salt dissolution collapse chimney; implications for nuclear waste disposal: Neues Jahrbuch fuer Mineralogie, Abhandlungen, v. 149, p. 163-175.

De Mille, G., J. R. Shouldice, and H. W. Nelson, 1964, Collapse structures related to evaporites of the Prairie Formation, Saskatchewan: Geological Society America Bulletin, v. 75, p. 307-316.

Ferrarase, F., T. Macaluso, G. Madonia, P. A., and U. Sauro, 2002, Solution and recrystallisation processes and associated landforms in gypsum outcrops of Sicily: Geomorphology, v. 49, p. 25-453.

Ford, D., and P. D. Williams, 2007, Karst hydrology and Geomorphology: New York, John Wiley and Sons, Ltd, 562 p.

Ford, D. C., 1998, Principal features of evaporite karst in Canada: Suppl. Geogr. Fis. Dinam. Quat. III, Proceedings of fourth International Conference on Geomorphology, Italy 1997:Karst Geomorphology, v. 12, p. 11-19.

Goodbar, J., and A. Goodbar, 2014, A Method to Investigate Karst Groundwater Flow in Nash Draw Eddy County, New Mexico To Delineate Potential Impacts of Potash Industry Discharge and Runoff U.S. Geological Survey Karst Interest Group; Conference Paper April 2014 held at Carlsbad New Mexico.

Gorbunova, K. A., 1979, Gidratatsiya angidrita i soputstvuyushchiye yey yavleniya: Koval'chuk, A. I. Karst i gidrogeologiya Predural'ya. Akad. Nauk Sssr, Ural. Nauchn. Tsentr, Inst. Geol. Geokhim., Tr.

Gustavson, T. C., W. W. Simpkins, A. Alhades, and A. D. Hoadley, 1982, Evaporite dissolution and development of karst features on the Rolling Plains of the Texas Panhandle: Earth Surface Processes and Landforms, v. 7, p. 545-563.

Gutiérrez, F., 1996, Gypsum karstification induced subsidence - effects on alluvial systems and derived geohazards (Calatayud Graben, Iberian Range, Spain): Geomorphology, v. 16, p. 277-293.

Gutiérrez, F., J. Guerrero, and P. Lucha, 2008, A genetic classification of sinkholes illustrated from evaporite paleokarst exposures in Spain: Environmental Geology, v. 53, p. 993-1006.

Hamilton, J., and D. Ford, 2002, Karst geomorphology and hydrogeology of the Bear Rock Formation — a remarkable dolostone and gypsum megabreccia in the continuous permafrost zone of northwest Territories, Canada: Carbonates and Evaporites, v. 17, p. 114-115.

Hamilton, J. P., 1995, Karst geomorphology and hydrogeology of the northeastern Mackenzie Mountains, District of Mackenzie, N.W.T., Canada: Doctoral thesis, McMaster University, 532 p.

Hill, C. A., 1990, Sulphuric acid speleogenesis of Carlsbad Caverns and its relationship to hydrocarbons, Delaware Basin, New Mexico and Texas: American Association of Petroleum Geologists Bulletin, v. 74, p. 1685-1694.

Holt, R. M., and D. W. Powers, 2010, Evaluation of halite dissolution at a radioactive waste disposal site, Andrews County, Texas: Geological Society of America Bulletin, v. 122, p. 1989-2004.

James, N. P., and P. W. Choquette, 1988, Paleokarst: New York, Springer, 416 p.

Johnson, K. S., 2005, Subsidence hazards due to evaporite dissolution in the United States: Environmental Geology, v. 48, p. 395-409.

Kirkland, D. W., and R. Evans, 1976, Origin of limestone buttes, Gypsum Plain, Culberson County, Texas: American Association of Petroleum Geologists, Bulletin, v. 60, p. 2005-2018.

Kirkland, D. W., and R. Evans, 1980, Origin of castiles on the Gypsum Plain of Texas and New Mexico: Guidebook New Mexico Geological Society, v. 31, p. 173-178.

Korotkov, A. N., 1974, Caves of the Pinego-Severodvinskaja karst (in Russain): Geog. Soc. USSR, Leningrad.

Lambert, S. J., 1983, Dissolution of Evaporites in and Around the Delaware Basin, Southeastern New Mexico and West Texas: Sandia Nat’l. Labs. Report SAND82-0461, Albuquerque, NM, 96 pp.

Land, L., 2013, Evaporite Karst in the Permian Basin Region of West Texas and southeastern New Mexico: The Human Impact, in L. Land, D. H. Doctor, and J. B. Stephenson, eds., Sinkholes and the engineering and environmental impacts of karst; Symposium 2; Proceedings of the 13th Multidisciplinary Conference, May 6 through 10, 2013, Carlsbad, New Mexico, p. 113-121.

Law, J., 1971, Regional Devonian Geology and Oil and Gas Possibilities, Upper Mackenzie River Area: Bulletin of Canadian Petroleum Geology, v. 19, p. 437-484.

Macaluso, T., and U. Sauro, 1996, Weathering and Karren on exposed gypsum surfaces: International Journal of Speleology, v. 25, p. 115-126.

Martinez, J. D., K. S. Johnson, and J. T. Neal, 1998, Sinkholes in Evaporite Rocks: American Scientist, v. 86, p. 38.

Meijer Drees, N. C., 1985, Evaporitic deposits of western Canada: Geological Survey of Canada, v. Paper 85-20, p. 118 pp.

Meijer Drees, N. C., 1993, The Devonian succession in the subsurface of the Great Slave and Great Bear Plains, Northwest Territories: Geological Survey of Canada Bulletin, No. 393, 222 p.

Neal, J. T., 1995, Supai salt karst features: Holbrook Basin, Arizona, in B. F. Beck, ed., Karst geohazards: engineering and environmental problems in karst terrane. Proc. 5th conference, Gatlinburg 1995, Balkema, p. 53-59.

Neal, J. T., R. Colpitts, and K. S. Johnson, 1998, Evaporite Karst in the Holbrook Basin, Arizona, in J. Borchers, ed., Joseph F. Poland Symposium on Land Subsidence: Sudbury, MA, Assoc. Eng. Geologists Spec. Pub. 8, p. 373-384.

Neal, J. T., R. Colpitts, and K. S. Johnson, 2001, Evaporite Karst in the Holbrook Basin, Arizona: SMRI (Solution MIning Research Institute) Report presented at the Fall 2001 Meeting 7 - 10 October 2001 Albuquerque, New Mexico, USA.

Olive, W. W., 1957, Solution-subsidence troughs, Castile Formation of Gypsum Plain, Texas and New Mexico: Geological Society of America, v. 68, p. 351-358.

Paine, J. G., 1994, Subsidence beneath a playa basin on the Southern High Plains, U. S.A.; evidence from shallow seismic data: Geological Society of America Bulletin, v. 106, p. 233-242.

Powers, D. W., R. L. Beauheim, R. M. Holt, and D. L. Hughes, 2006, Evaporite karst features and processes at Nash Draw, Eddy County, New Mexico, in L. Land, V. Lueth, B. Raatz, P. Boston, and D. Love, eds., Caves and Karst of Southeastern New Mexico: New Mexico Geological Society, Guidebook 57, p. 253-266.

Quinlan, J. F., 1978, Types of karst, with emphasis on cover beds in their classification and development: Doctoral thesis, University of Texas at Austin, 323 p.

Quinlan, J. F., R. O. Ewers, J. A. Raty, and K. S. Johnson, 1986, Gypsum karst and salt karst of the United States of America: Le Grotte d'Italia, v. 4, p. 73-92.

Simpson, F., 1988, Solution-generated collapse (SGC) structures associated with bedded evaporites; significance to base-metal and hydrocarbon localization: Geoscience Canada, v. 15, p. 89-93.

Stafford, K. W., R. Nance, L. Rosales-Lagarde, and P. J. Boston, 2008, Epigene and Hypogene Gypsum karst manifestations of the Castile formation: Eddy County, new Mexico and Culbesron County, Texas, USA: International Journal of Speleology, v. 37, p. 83-98.

Stafford, K. W., L. Rosales-Lagarde, and P. J. Boston, 2008, Castile evaporite karst potential map of the Gypsum Plain, Eddy County, New Mexico and Culberson County, Texas: A GIS methodological comparison., v. 70, no. 1, p. 35–46.: Journal of Cave and Karst Studies, v. 70, p. 35-46.

Stanton, R. J., Jr, 1966, The solution brecciation process: Geological Society of America Bulletin, v. 77, p. 843-847.

Stenson, R. E., and D. C. Ford, 1993, Rillenkarren on gypsum in Nova Scotia: Geographie Physique et Quaternaire, v. 47, p. 239-243.

Tozer, R. S. J., A. P. Choi, J. T. Pietras, and D. J. Tanasichuk, 2014, Athabasca oil sands: mega-trap restoration and charge timing: Bulletin American Association Petroleum Geologists, v. 98, p. 429-447.

Tsui, P. C., and D. M. Cruden, 1984, Deformation associated with gypsum karst in the Salt River Escarpment, northeastern Alberta: Canadian Journal of Earth Sciences, v. 21, p. 949-959.

Waltham, T., F. Bell, and M. Culshaw, 2005, Sinkholes and Subsidence: Karst and Cavernous Rocks in Engineering and Construction: Berlin Heidelberg, Springer Praxis Books, 382 p.

Warren, J. K., 2016, Evaporites: A compendium (ISBN 978-3-319-13511-3): Berlin, Springer, 1854 p.

Wigley, T. M. L., J. J. Drake, J. F. Quinlan, and D. C. Ford, 1973, Geomorphology and geochemistry of a gypsum karst near Canal Flats, British Columbia: Canadian Journal of Earth Sciences, v. 10, p. 113-129.



Danakhil Potash; Ethiopia - Modern hydrothermal and deep meteoric KCl, Part 3 of 4

John Warren - Friday, May 01, 2015

So far we have discussed the modern salt pan geology of the Danakhil (Part 1 of 4) and the initial subaqueous setting for widespread bedded potash, now in the subsurface, mostly as a kainitite bed (Part 2 of 4). In this blog we will discuss examples of potash in the Danakhil where remobilised salts and brines are related to the circulation of hydrothermal and meteoric fluids have facilitated localised reworking of potash to the surface (part 3 of 4). These fluids are related to the thermal anomalies created by the emplacement of the Dallol mound and the chemical front created by the encroachment of the Bajada along the western margin of the saltflat. Notably, we shall see the Dallol Mound is not a volcanic cone, rather it is an anticlinal dome of uplifted and eroded bedded salt, capped and surrounded by hydrothermal crater features typified by karst pools and brine outflows. Its creation is likely related to emplacement of igneous material at depth but, as yet there, has been no breakout of volcanic rock material in the mound area. This has important economic implications for the nature of remobilised potash and the creation of potential potash ores in the Dallol Mound area, these cosiderations are separate from the regional distribution of primary potash beds (kainitite and carnallitite) that were discussed in the previous blog.

Thermal brine springs and potash occurrences near Dallol mound

Today, hot springs supply and maintain a number of hydrothermally-fed brine pools and brine filled karst lakes in various depressions both atop and near the regional anticlinal salt mound or salt dome, sometimes called Dallol Mountain (Figure 1). As it only rises some 60 metres from the surrounding surface (-81 m versus -120 m) the term mountain is a misnomer. The highly dissected and eroded slope of bedded halite that is the southwest margin of Dallol mound shows the various springs are active in a region of uplifted and eroded bedded evaporite that defines the Dallol mound (Figure 2a). For example, brine springs still supply a small carnallite deposit known as the Crescent deposit located near the uplifted black halite beds that define Black Mountain and located 1.5 km southwest of Dallol mound (Figure 2b). This potash ore is the result of hydrothermally-driven groundwater activity, likely related to the emplacement of the Dallol Mound. The uplift-related thermal hydrology has broken up the mineralogical continuity of the nearsurface evaporite beds including the equivalents to the potash-rich Houston Fm.

The Black Mountain potash deposits caught the attention of the Houston-based  Ralph M. Parsons company in 1954 where, according to Holwerda and Hutchinson, 1968, potash mining had previously already taken place at the Crescent carnallite/sylvite deposit. Earlier extraction had involved, amongst other techniques, flooding of salt pans around a continuously flowing hot spring, followed by harvesting of potash-rich salts, once natural deliquescence had flushed most of the highly soluble MgCl2 from the system. A concession was obtained Parsons linked to obligations to investigate the various potash deposits in the area, some of which were tied to actual outcrops of potash salts. The Parsons Company set up its base on Dallol Mountain at a site previously occupied by the Italian mining community, which had operated in the first few decades of last century (Figure 2a; the modification and reuse of older salt brick buildings is still evident on the ground today). As well, Parsons Co. constructed airstrips on Dallol Mountain and in the Musley area. They drilled more than 300 holes in order to better understand the the distribution of the potash beds. Drilling operations in 1959-1961 led to the delineation of the small localized "Crescent" carnallitite deposit in the vicinity of Black Mountain . This was followed by the discovery of the much larger (>80 million tonnes) "Musley" sylvite deposit near the base of the Ethiopian Highlands, some 5km W of Dallol, and extending at least 10km in a N-S orientation. A 92m vertical shaft and a total of 805m of drives were made in this deposit, but all work was stopped in 1967 after rapid influx of water into the conventional mine killed a number of workers. The political tensions in the area at the time probably also played a part in preventing mining activity in the following years.

Holwerda and Hutchinson (1968) argue that geographical location of the main "Musley" sylvite strata is directly west of Dallol Mound and at the base of the highlands. This, and the fact that sylvite is an alternation product that consistently overlays the carnallite strata and thickens (although discontinuously) along the western margin (see drill hole intersections published in Ercosplan, 2011), suggests that the potash enrichment was produced by selective leaching of MgCl2 from a carnallite precursor, driven by phreatic run-off waters sourced in the Ethiopian highlands. My own observations and plotting of enrichment fairways (using published Ercosplan 2010, 2011 data) confirms Holwerda and Hutchinson’s inferences. If diagenesis, not primary precipitation, is the prime mechanism of sylvite creation in the Musley region, then the regional sylvite control/distribution for this style of enrichment is related to a subsurface meteoric/groundwater phreatic overprint that parallels the encroaching bajada edge. It is a separate ore fairway to the more regional easterly dipping bedded kainitite/carnallitite trend.

Waters in some of the active brine-filled hydrothermal craters and dolines can locally have temperatures of more than 100°C and when waters cool they precipitate varying combinations of halite, carnallite and bischofite. The brines are so saturated with salts that if a stick is thrust into a boiling brine pool and removed it is immediately covered by layer of carnallite or bischofite and halite (Figure 2b, c). The same pools are also rich in FeCl2, sulphur and manganese, which explains the spectacular bright green, red-orange and yellow colours of many of the saline mineral assemblages precipitating in and about these active spring-formed pools. Occasional intense storm-driven sheetfloods can drive renewed activity in the various springs in vicinity of the mound, as happened in the recent floods of February 2011, when the intensity of water circulation and the areal extent of the pools greatly increased. After the same storm flood, a natural collapse doline tens of metres across formed on the western depression margin. Clearly, the local hydrothermal/karstic enhancement style of bittern enrichment is a separate process set for potash enrichment compared to the widespread earlier deposition of marine-fed subaqueous kainite. Hence, it contrasts with the much more widespread set of depositional/early diagenetic processes that laid down the bulk of the bedded potash association that is the Houston Fm. in the Danakhil Depression (as discussed in the previous Danakhil blog).

What is the Dallol Mound and what drives its uplift hydrology?

Despite the widespread misconception that the Dallol mound is a lava cone, Mount Dallol is not a volcanic-centered feature on the Danakhil landscape. A visit to the area reveals no observable volcanic products (lava, ashfall or scoria) on the surface on or near the Dallol mound. This is so even in the region of the most recent phreatic activity in 1926 where a 30 m-diameter phreatic (explosion? or daylighting hydrothermal karst) crater formed, hosted in salt beds (Figure 2b). All the rocks associated with this cavity and its formative event are not volcanic. This means the mechanism that created the Dallol Mound is unlike the magmatic events that created the world famous Erte Ale volcanic cone, with its distinctive longterm active magma lake and located some 80 km to the south of Dallol and still in the Danakhil depression. Instead, the Dallol mound crest is made up of uplifted and eroded halite and potash beds soaked in a thermal hydrology that breaks out on the lake surface as a number of hot bubbling sulphurous brine pools. This is also true of the off-mound crater that formed in 1926 near Black Mountain and still retains bubbling brines with present temperatures ~65-70 °C. Nearby “Black Mountain” is a small area of dark coloured bedded and recrystallised halite, it is not a primary volcanic feature.

As a sedimentologist visiting the area, I wondered at why the Dallol mound features had ever been called volcanic cones, hornitos, or maars (as they are widely described in the literature). To use such genetic terms in a geologically correct fashion I would like to put my hand on a piece of volcanic debris (lava, pumice, scoria or ash) in any of the craters before I call the Dallol mound a volcanic cone. And yet, many workers in the published literature dealing with the Dallol area are happy to do this. I am not saying there is no influence of magmatic heating in forming Dallol Mound, only that molten volcanic rock has yet to surface in the immediate Dallol region. Hence it is unlike the many actual volcanic cones, maars and hornitos to the south and north and this is an significant observation as it deals with mechanism of local potash enrichment. I will argue in the next section that this is because Dallol Mound is a salt uplift feature or dome capped by phreatic cone/ hydrothermal karst structures and all related to the migration of molten magma into more deeply buried salt beds, which contain hydrated salts at the level of the Houston Fm and perhaps even deeper buried hydrated salt layers (see blog 2).

Darrah et al (2013) and Detay (2011) argue that the 30m diameter 1926 crater and other nearby pools on the Dallol saltflat in the vicinity of the Dallol mound are the result of a phreatic explosions, tied to the increasing gas pressure in superficial hydrothermal reservoirs atop a deeper mass of molten rock. The mound is a landscape feature indicative of deep dyke/sill intrusion that did not surface. According to Holwerda and Hutchinson (1968) this yet-to-daylight dyke complex explains the linear orientation of the mound, its pools and other karst/erosion features on the salt flat surface in vicinity of the Dallol mound. That is, the various Dallol hot springs typically consist of 30-40m diameter circular to sub-circular ponds, initially formed by explosive vapor eruptions, to form at-surface circular features, which are widely termed maars, although I would prefer to call them "maar-like." A “maar” is defined in the AGI Glossary of Geology as “a low relief, broad volcanic crater formed by multiple shallow explosive eruptions. It is surrounded by a crater ring, and may be filled by water. Type occurrence is in the Eifel area of Germany.” Given the lack of a volcanic crater rim the Dallol Mound and adjacent brine-filled cavities are not really maars, nor are they hornitos. They will likely evolve into such features, but in their current state better considered brine-filled fumaroles or solfateras or even better, hydrothermal karst cavities that have daylighted. Once the cavities have broken out onto the salt flat surface, these circular (possibly-explosive) features can continue enlarge due to ongoing rise of undersaturated waters and so evolve into expanding hydrothermal karst pools or they can be partially to completely filled with saline precipitates (with no volcanic products derived from molten igneous rock materials).

So, instead of at-surface volcanic products such as lava and ashfall, most of the superficial precipitates/sediments observed in and around the various on- and off-structure Dallol brine pools are evaporite salts, along with some remnants of older clay-sediments. Brine fluids in various hot spring pools in the Dallol area (in the Dallol “hill” crest and the “Crescent” region near Black mountain, and in the “Boiling Lake” region south of the mound) are typically multi-coloured warm/hot ponds (Figure1, 3; Gebresilassie et al., 2011). The various pools are extremely salty (>500g/L), can be highly acidic (sometimes with a pH approaching 0.5), and gas-rich (as evidenced by steady, vigorous bubbling of gases). According to Darrah et al. (2013) the Dallol “salt dome” fluids and associated hot springs are hypothesized to result from the interaction between hot mantle fluids or basalt dyke injections with evaporite deposits at unknown depths. However, direct observations of the volumes of pool waters and the vigour of the outflow are known to increase after the occasional heavy rain event, as happened in February, 2011. Hence, it is unclear if sulfur-rich gases and the low pH brine fluids provide evidence of the interaction of hot mantle fluids with the evaporites (as inferred by Darrah et al., 2013) or the pool waters are, at least in part, related shallower ongoing hydrothermal/karst interactions with more deeply circulated meteoric waters sourced in the 1000-m high adjacent rift highlands.

Why hydrated salts are important in some salt-hosted thermal systems: a Permian Zechstein analog

Most published volcanogenic-related studies of the Dallol Mound have not considered the effects of hydrated salt layers in a situation of rising molten rock, where the country rock contains beds of hydrated evaporites such as kainite or carnallite. This situation is exposed in the dyke-intruded halite-carnallite levels in the mines of the Werra-Fulda mine district of Germany (Schoefield et al., 2013; Warren, 2015). There, the Permian Zechstein salt series contains two important potash salt horizons (2-10m thick), which are mined at a depths ≈ 800 m from within a 400m thick halite host (Figure 4a). In the later Tertiary, basaltic melts intruded these Zechstein evaporites, but only a few dykes reached the Miocene landsurface. Basaltic melt production was related to regional volcanic activity some 10 to 25 Ma. Basalts exposed in the mine walls are typically subvertical dykes, rather than sills. These basaltic intervals can crosscut the salt over zones up to several kilometres wide (Figure 4b). However, correlations of individual dyke swarms, between different mines, or between surface and subsurface outcrops, is difficult.

The basalts are phonolitic tephrites, limburgites, basanites and olivine nephelinites. Dyke margins in contact with halite are usually vitrified, forming a microlitic limburgite glass along dyke edges (Knipping, 1989). At the contact on the evaporite side of the glassy rim there is a cm-wide carapace of high temperature salts (mostly anhydrite and ferroan carbonates). Further out, the effect of the high temperature envelope is denoted by transitions to clear halite, with higher temperature fluid inclusions (Knipping 1989). All of this centimetre to metre-scale alteration is an anhydrous alteration halo, the salt did not melt (halite’s melting temperature is 804°C), rather than migrating, the fluid driving recrystallisation was largely from local movement of entrained brine inclusions. The dolerite/basalt interior of the basaltic dyke is likewise altered and salt soaked, with clear, largely inclusion-free halite typically filling vesicles in the basalt.

Worldwide, dykes intersecting salt beds tend to widen to become sills in two zones: 1) along evaporite units within the halite mass that contain hydrated salts, such as carnallite or gypsum and, 2) where rising magma has ponded and so created laccoliths at the upper or lower halite contact with the adjacent nonsalt strata or against a salt wall (Warren, 2015). The first is a response to a pulse of released water as dyke-driven heating forces the dehydration of hydrated salt layers. The second is a response to the mechanical strength contrast at the salt-nonsalt contact. The first is what is observed in the Fulda region and is also likely relevant to the formation of the Dallol Mound and its remobilised potash-precipitating brines.


In such subsurface regions, the heating of hydrated salt layers (such as carnallite or kainite), adjacent to a dyke or sill, drives off the water of crystallisation (chemical or hydration thixotropy) at a much lower temperatures than that at which anhydrous salts, such as halite or anhydrite, thermally melt (Table 1). In the Fulda region the thermally-driven release of water of crystallisation within particular Zechstein salt beds creates thixotropic or subsurface “peperite” textures in carnallitite ore layers, where heated water of crystallisation escaped from the hydrated-salt lattice. Dehydration-driven loss of mechanical strength focuses zones of magma entry into particular horizons in the salt mass, wherever hydrated salt layers were intersected (Figure 4c verses 4d). In contrast, dyke and sill margins are much sharper and narrower in zones of contact with anhydrous salt intervals (Figure 4b; Schofield et al., 2014).

Accordingly, away from immediate vicinity of the direct thermal aureole, heated and overpressured dehydration waters can enter a former Zechstein carnallite halite bed, and drive the creation of extensive soft sediment deformation and [1]peperite textures in the former hydrated layer (Figure 4d, e). Mineralogically, sylvite and coarse recrystallised halite dominate the salt fraction in the peperite intervals/beds. These are evaporite-related beds formed within a hydrated salt bed and so differ from the common notion of volcanic peperites indicating water-saturated sediment intercations with very shallow dyke or sill emplacements. Sylvite in these altered zone is a form of dehydrated carnallite, not a primary-textured salt. In the Fulda region, such altered zones and deformed units can extend along former carnallite layers to tens or even a hundred or more metres from the dyke feeder. Ultimately, the deformed potash bed passes laterally out into the unaltered bed, which retains abundant inclusion-rich primary chevron halite and carnallite (Figure 4d versus 4e). That is, nearer the basalt dyke, the carnallite is largely transformed into inclusion-poor halite and sylvite, the result of incongruent flushing of warm saline fluids mobilized from the hydrated carnallite crystal lattice as it was heated by dyke emplacement. During Miocene salt alteration/thermal metamorphism in the Fulda region, NaCl-fluids were mixed with fluids originating from thermally-mobilised crystallisation water in the carnallite, as it converted to sylvite.

In the Dallol depression I think it is highly likely that a similar set of destabilisation processes occurred when rising dyke magma reached the levels of hydrated salts (kainite and carnallite beds) in the Houston Formation of the Danakhil fill, after passing relatively passively through the Lower Rocksalt Formation (see the previous blog). Emplacement of the magma/dyke into  hydrated evaporites in the vicinity of what is now the Dallol mound would have mobilised and deformed the hydrated salt level, converting carnallite to sylvite, kainite to bischofite and lesser kieserite, as well as creating widespread cavities filled with pressured volatiles carried by MgCl and KCl brines. Once these hydrothermal cavities dissolve their way to surface, the feeder brines can cool and precipitate as prograde salts such as halite, sylvite and perhaps bischofite. Such destabilisation would have accommodated the emplacement of a basaltic sill at the level of the potash salts, in turn driving the uplift of the lake beds above this region. Mound-related uplift and hydrothermal activity then drives the formation of natural regions of ground collapse, sulphurous and acidic springs and fumaroles, along with the creation of water-filled chimneys and doline sags, filling with various hydrothermal salts, in the vicinity of the volcanic mound.

Implications for Potash distribution in the Danakhil Depression

The discussion of potash mineral-forming processes in this and the previous blog clearly underlines a trichotomy in the way potash has accumulated in halite host-beds across the Danakhil Depression. The most widespread form of potash in the Danakhil Depression is as a primary evaporite bed, composed of primary marine kainitite precipitates with a carnallite cap (Houston Formation). Across the western side of the depression this easterly dipping bed is now buried beneath 30-150 m of overburden salts. It likely precipitated as a marine seepage-fed bittern layer, at a time the Danakhil depression was hydrographically isolated from a direct surface connection with the Red Sea. Its brine hydrology was dominantly subaqueous and not unlike that of modern Lake Asal in Djibouti, although it was more saline than Asal in the subaqueous potash sump areas. Thus, the Danakhil potash bed (Houston Fm) formed sometime ago, its formative hydology is no longer present in the depression and it may be as old as Pliocene or more likely early to mid Pleistocene. There has been sufficient time for this bed to tilt toward the east. The unit is underlain by the subaqueous Lower Rocksalt Formation (LRF) and subsequently overlain by the Upper Rocksalt Formation (URF). Both these halite formations do not entrain primary potash beds. The LRF contains numerous CaSO4 layers, while the URF contains clayey laminite beds and locally hosts regions of remobilised potash salts. The URF evolves upward into the saltflat/ephemeral lake hyperarid hydrology that typifies the modern depression.

More localised forms of potential potash ore typify occurrences in the Dallol and Musley areas (Figure 2a). There potash in the Dallol Mound region is hydrothermally reworked from the uplifted equivalents of the Houston Formation. Even today this hydrology is precipitating carnallitite (associated with bischofite and minor kieserite) in various hydrothermal brine pools atop and around the Dallol Mound, such as the carnallite-dominant Crescent deposit (Figure 2b). These hydrothermal salts owes their origins to daylighting of pressurised fluid systems and cavities. They were created by the volatile products of hydrated salt layers (Houston Fm) where these salts had come into contact with thermal aureoles or actual lithologies of newly emplaced dykes that had penetrated the underlying halite section. Actual molten volcanic rock has yet to make it to the surface in the Dallol Mound region, although active volcanic mounds and flows do typify the saltflat surface tens of kilometres to the south (Erte Alle ) and north. Based on the analogy exposed within the Zechstein-hosted potash mines of the Fulda region of Germany, it is likely that as well as creating at-surface brine pools, this hydrothermal dyke-related hydrology converts any carnallitite to a sylvinite bed at the level of contact with the Houston Fm. 

Then there is the deep-meteoric alteration system that is altering the kainitite/carnallitite of Houston Fm into sylvinite, it is active along the deep meteoric alteration front located at the irregular interface between the downdip end of the Musley Fan and the updip portion of the Houston Fm. This diagenetic mechanism formed the Musley potash deposit, defined and exploited by the Parsons Company operations and documented in Holwerda and Hutchison (1968). Variations on this deep-meteoric alteration theme likely extend south and north of the Musley fan, wherever the active phreatic hydrology of the bajada located at the foot of the Ethiopian Highlands interacts and interfingers with the updip edge of the easterly dipping Houston Formation.

Once again there is no "one-size-fits-all) model for economic potash understanding (Warren, 2010, 2015). Even in what is probably the youngest known marine-fed potash system in the world, the original potash mineralogy and distribution has been altered and locally upgraded via diagenetic interactions with hydrothermal or deep-meteoric fluids. Predicting ore distributions in this, and all potash systems worldwide, requires an understanding of formative process evolution through deep time, and not just the simple application of a layer-cake primary stratigraphic model. 


Carniel, R., E. M. Jolis, and J. Jones, 2010, A geophysical multi-parametric analysis of hydrothermal activity at Dallol, Ethiopia: Journal of African Earth Sciences, v. 58, p. 812-819.

Darrah, T. H., D. Tedesco, F. Tassi, O. Vaselli, E. Cuoco, and R. J. Poreda, 2013, Gas chemistry of the Dallol region of the Danakil Depression in the Afar region of the northern-most East African Rift: Chemical Geology, v. 339, p. 16-29.

Detay, M., 2011, Le DALLOL revisité: entre explosion phréatomagmatique, rifting intra-continental, manifestations hydrothermales et halocinèse: LAVE. Liaison des amateurs de volcanologie européenne, v. 151, p. 7-19.

ERCOSPLAN, 2010, Techical report and current resource estimate: Danakhil Potash Deposit, Afar State, Ethiopia: Project Reference: EGB 08-024.

ERCOSPLAN, 2011, Preliminary Resource Assessment Study, Danakhil Potash Deposit, Afar State, Ethiopia: G & B Property: Project Reference: EGB 10-030.

Gebresilassie, S., H. Tsegab, and K. Kabeto, 2011, Preliminary study on geology, mineral potential, and characteristics of hot springs from Dallol area, Afar rift, northeastern Ethiopia: implications for natural resource exploration: Momona Ethiopian Journal of Science, v. 3, p. 17-30.

Holwerda, J. G., and R. W. Hutchinson, 1968, Potash-bearing evaporites in the Danakil area, Ethiopia: Economic Geology, v. 63, p. 124-150.

Knipping, B., 1989, Basalt intrusions in evaporites: Lecture Notes in Earth Sciences (Springer-Verlag), v. 24, p. 132 pp.

Schofield, N., I. Alsop, J. Warren, J. R. Underhill, R. Lehné, W. Beer, and V. Lukas, 2014, Mobilizing salt: Magma-salt interactions: Geology.

Warren, J. K., 2010, Evaporites through time: Tectonic, climatic and eustatic controls in marine and nonmarine deposits: Earth-Science Reviews, v. 98, p. 217-268.

Warren, J. K., 2015, Evaporites: A compendium (ISBN 978-3-319-13511-3) Released August 2015: Berlin, Springer, 1600 p.


[1] Peperite is a sedimentary rock that contains fragments of igneous material and is formed when magma comes into contact with wet water-saturated sediments. 

Danakhil Potash, Ethiopia: Beds of Kainite/Carnallite, Part 2 of 4

John Warren - Wednesday, April 29, 2015

The modern Dallol saltflat described in the previous blog defines the upper part of more than 970 metres of halite-dominated Quaternary evaporites that have accumulated beneath the present salt pan of the Northern Danakhil. The total sequence is made up of interbeds of halite, gypsum, anhydrite and shale with a potash succession separating two thick sequences of halite (Figure 1; Holwerda and Hutchison, 1968; Augustithis, 1980). At depths of more than 35-40 meters, and deepening to the east, this km-thick subcropping Quaternary halite-dominated fill contains one, and perhaps two or more, potash beds. For a more detailed description of the upper part of the fill the reader is referred to the previous blog and Chapter 11 in Warren, 2015.

Bedded Pleistocene evaporites may underlie the entire Danakil depression, but younger lava flows of the Aden Volcanic Series and alluvium washed in from the surrounding bajada obscure much of the older Pleistocene sedimentary series across much of the southern part of the depression beyond Lake Assale). Potash exploration drilling and core recovery is concentrated in the accessible parts of the northern Danakhil rift, where the saltflat facilitates vehicle access, compared with the lava-covered regions south of Lake Assale. The most recent volcanic activity affecting the known potash region was the emplacement of the Dallol Mound, which has driven local uplift of the otherwise subsurface potash section to where it approaches the surface in the immediate vicinity of the mound (Figure 2a).

Away from the Dallol volcanic mound the upper potash bed beneath the saltflat lies at a depth of 38-190 metres. A lower inferred potash bed likely occurs at depth along the eastern end of the saltflat, but this second bed is inferred from high API kicks in gamma logs run in deeper wells, no solid salt was recovered (Holwerda and Hutchison, 1968). The upper proven potash bed is now the target zone for a number of minerals companies currently exploring for potash in the region. Regionally, both potash units dip east, with the deepest indicators of the two units encountered by the drill in a single well on the eastern side of the saltflat at depths of 683 and 930 m, respectively (Figure 2: Holwerda and Hutchison, 1968). The likely Quaternary age of the potash units, the marine brine source, explains the high magnesium content of the potash bittern salts, as modern seawater contains high levels of Mg and SO4.

My study of core that intersected the potash interval and that is sandwiched between the Lower and Upper Halite units shows both the lower and the upper halite units retain pristine sedimentary textures, with features and vertical successions that indicate distinct hydrologies during their deposition (Figure 3). There is no textural evidence of halokinetic recrystallization in halites any of the studied cores and published seismic also indicates consistent dips in the evaporites . Most of the textures in the cored potash interval indicate a subaqueous density-stratified environment, with brine reworking of the upper part of primary kainitite, carnallitite units. Perennial subaqueous, density-stratified brines also typify the hydrology of the Lower Halite unit, albeit with somewhat lower salinities tan those precipitating the bitterns (Figure 3). The brine that precipitating the Upper Rocksalt Formation was shallower and more ephemeral. The following paragraphs summarise my core-based observations and interpretations that led to this interpretation of the evolving brine hydrology.

The Lower Rocksalt Formation (LRF) is dominated by bottom-growth-aligned subaqueous halite textures and lack of siliciclastic detritus, unlike the Upper Rocksalt Formation (Figure 3). Halite textures in the LRF lack porosity and dominated by coarsely crystalline beds made up of cm-scale NaCl-CaSO4 couplets dominated by upward-pointing halite chevrons and mantled by thin CaSO4 layers (Figure 3). This meromictic-holomictic textural association passes up into the upper part of the LRF with cm-scale proportions alternating of less-saline to more-saline episodes of evaporite precipitation decreasing, indicating an “on-average” increasingly shallow subaqueous depositional setting as one approaches the base of the kainitite unit. The combination of bottom-nucleated and cumulate textures in the LRF are near identical to those in the halites in the kainitite interval in the Messinian of Sicily (see later). 

The laminated Kainitite Member is also a subaqueous unit with layered cumulate textures (Figure 3), it was likely deposited on a pelagic bottom beneath a shallow body of marine-fed bittern waters, which never reached carnallite saturation. Above this are the variably present carnallitic Intermediate and Sylvinitite members and the overlying Halite marker beds in turn overlain by the Upper Halite unit. All retain pristine textures indicating mostly subaqueous deposition, soon followed by varying exposure and reaction with shallow phreatic brines moving across the top of Kainitite member. This shallow phreatic brine crossflow drove syndepositional mineral alteration and collapse in the upper part of the kainitite and carnallitite units.

The potash-entraining interval between the URF and LRF is called the Houston Formation has been drilled and cored extensively by explorers in the basin, showing it is consistently between 15 and 40 metres thick (Figure 1). Stratigraphically, it consists of lower Kainitite Member (4-14m thick) atop and in depositional continuity with the LRF (more than 500m thick) (Figure 3). The Kainitite Member is fine-grained, laminated, locally wavy-bedded, containing up to 50% kainite cumulates in a cumulate (non-chevron) halite background, along with small amounts of a white mineral that is likely epsomite. It is overlain by what older literature describes as the “Carnallitite Intermediate unit” (3-25 m thick). More recent potash exploration drilling has shown all the members that constitute the Intermediate Carnallitite Member is not always present within the Dallol depression. Mineralogically is at best considered as variably developed (Figure 3). Its lower part is a layered to laminated carnallite-halite mixture with some kieserite, anhydrite and epsomite. This can pass up or laterally into kainitite with sylvite. Above the Intermediate Member is the 0-10m thick Sylvinite Member containing 20-30% sylvite, along with polyhalite and anhydrite (up to 10%). Typically the sylvinite member shows primary layering disturbed by varying intensities of slumping and dissolution. Often the upper part of a carnallite unit (where present) also shows similar evidence of dissolution and reprecipitation.

Cores through the sylvinite member and parts of the upper carnallitite member sample a range of recrystallization/flow/slump textures, rather than primary horizontal-laminar textures. Beneath the sylvinite member, the variably-present upper carnallitite member contains a varied suite of non-commercial potash minerals that in addition to carnallite include, kieserite, kainite (up to 10% by volume) and polyhalite, along with minor amounts of sylvite. Minor anhydrite is common, while rinneite may occur locally, along with rust-red iron staining. Sylvite is more abundant near the top of the carnallitite member and its proportion decreases downward, perhaps reflecting its groundwater origin. Kainite is the reverse and its proportion increases downward. The sylvinite member and the carnallitite member also show an inverse thickness relationship. Bedding in the carnallitite member is commonly contorted with folded and brecciated horizons interpreted as slumps. The base of the carnallitite member is defined as the level where carnallite forms isolated patches in the kainite before disappearing entirely.

Drilling in the past few years has clearly show that in some parts of the evaporite unit, located nearer to the western side of the basin, the lower and upper carnallite units are separated by thick bischoftite intervals (Figures 2b, 3). The bischofite is layered at a mm-cm scale and with no obvious breaks related to freshening and exposure, implying it too was deposited in a perennially subaqueous or phreatic cavity setting (Pedley et al., in press).

The potash/bischofite interval passes up into a slumped and disturbed halite-dominated unit that is known as the “Marker Beds” because of the co-associated presence of clay lamina and bedded halite, along with traces of potash minerals (Figure 3). This unit then passes up into the massive Upper Rock Salt unit across an unconformity at the top of the halite “Marker Beds.” Bedded, and at times finely laminated cumulate textures in the various magnesian bittern units, are used by many to argue that the kainitite and the lower carnallitite members are primary or syndepositional precipitates.

Three types of potash-barren zones can occur within it and are possibly related to the effects of groundwaters and solution cavity cements within the carnallitite unit, perhaps precipitated before the deposition of the overlying marker halites. Barren zones in the Sylvinite member are regions where: a) the entire sylvinite bed is replaced by a relatively pure stratiform halite, along with dispersed nodules of anhydrite, b) zones up to 23 m thick and composed of pure crystalline halite (karst-fill cements?) that occur patchily within the sylvinite bed and, c) potash-depleted zones defined by coarsely crystalline halite instead of sylvinite. Bedding plane spacing and layering and some slumping styles in the halite in styles a and b are similar to that in the sylvite bed. Contact with throughflushing freshened nearsurface and at-surface waters perhaps created most of the barren zones in the sylvinite. Fluid crossflow may also have formed or reprecipitated sylvite of the upper member, via selective surface or nearsurface leaching of MgCl2 from its carnallite precursor (Holwerda and Hutchison, 1968; Warren 2015). Due to the secondary origin of much of the sylvite in the Sylvinite member, the proportion of sylvite decreases as the proportion of carnallite increases, along with secondary kieserite, polyhalite and kainite.

The kainite member is texturally distinctive and is composed of nearly pure, fine-grained, dense, relatively hard, amber-coloured kainite with ≈ 25% admixed halite (Figure 3). Core study shows the lamina style remained flat-laminar (that is, subaqueous density-stratified with periodic bottom freshening) as the mineralogy passes from the LRF up into the flat-laminated kainitite member (Figure 4: Warren, 2015; Pedley et al., in press). Throughout, the kainitite unit shows a cm-mm scale layering, with no evidence of microkarsting or any exposure of the kainitite depositional surface. That is, the Kainitite Member is a primary depositional unit, like the underlying halite and still retains pristine evidence of its dominantly subaqueous depositional hydrology. The decreased proportion of anhydrite in the Kainitite Member, compared to the underlying LRF, indicates a system that on-average was more saline than the brines that deposited the underlying halite. The preponderance of MgSO4 salts means the Kainitite unit like the underlying LRF formed by the evaporation of seep-supplied seawater.

This situation differs from the present “closed basin” hydrology of the Danakil Depression which typifies the URF and the overlying Holocene succession (Hardie, 1990; Warren, 2015).

Units atop the primary laminated textures of the kainitite, lower carnallitite and bischofite members (where present) tend to show various early-diagenetic secondary textures (Figure 4). It seems much of the sylvinite and upper carnallitite member deposition was in shallow subsurface or at-surface brine ponds subject to groundwater crossflows and floor collapse, possibly aided by seismically-induced pulses of brine crossflow. In addition, this perennial density-stratified brine hydrology was at times of holomixis subject to brine reflux and the brine-displacement backreactions that typify all evaporite deposition, past and present (Warren 2015).

The observation of early ionic mobility in potash zone brines in the Danakil depositional system is also not unusual in any modern or ancient potash deposit. It should not be considered necessarily detrimental to the possibility of an extensive economically exploitable potash zone being present in the Danakil Depression. Interestingly, all the world’s exploited potash deposits, including those in the Devonian of Canada and Belarus, the Perm of the Urals and the potash bed of west Texas, show evidence of syndepositional and shallow burial reworking of potash (Warren, 2015). Early potassium remobilization has created the ore distributions in these and other mined potash depositsTextures and mineralogies in the Upper Rocksalt Formation (URF) define a separate hydrological association to the marine-fed LRF and Houston Formation (Table 4). Compared to the LRF, the URF has much higher levels of depositional porosity, lacks high levels of CaSO4, and has high levels of detrital siliciclastics. This is especially so in its upper part, which shows textural evidence of periodic and ongoing clastic-rich sheetflooding and freshening (Figure 4). It was deposited in a hydrology that evolved up section to become very similar to that active on today’s halite pan surface. The URF contains no evidence of salinities or textures associated with a potash bittern event and is probably not a viable exploration target for solid potash salts.

Above the URF is a clastic unit with significant amounts of, and sometimes beds dominated by, lenticular gypsum and displacive halite. The unit thickens toward the margins of the depression (Figure 2). The widespread presence of diagenetic salts indicates high pore salinities as, or soon after, the saline beds that stack into the clastic unit were deposited. Some of these early diagenetic evaporite textures are spectacular, as seen in the displacive halite recovered in a core from the lower portion of the clastic overburden, some 45 m below the modern pan surface (Figure 3).

What is clear from the textures preserved in the potash-rich Houston formation and the immediately underlying and overlying halites is that they first formed in a subaqueous-dominated marine-fed hydrology (Figure 4), which evolves up section into more ephemeral saltpan hydrologies of today (see the previous blog). The potash interval encapsulated in the Houston formation has primary mineralogical associations that are derived by evaporation of Pleistocene seawater (kainitite, carnallitite). In contrast the sylvite section in the Houston tends to form when these primary mineralogies are altered diagenetically perhaps soon after deposition but, especially, when hydrothermal waters circulated through uplifted beds of the Houston Formation, as is still occurring in the vicinity of the Dallol Volcanic Mound. Or where the chemical/meteoric interface associated with the encroachment of the bajada sediment pile drove incongruent dissolution of carnallite along the updip edge of the Houston Fm (as we shall discuss in the next blog). 


Augustithis, S. S., 1980, On the textures and treatment of the sylvinite ore from the Danakili Depression, Salt Plain (Piano del Sale), Tigre, Ethiopia: Chemie der Erde, v. 39, p. 91-95.

Hardie, L. A., 1990, The roles of rifting and hydrothermal CaCl2 brines in the origin of potash evaporites: an hypothesis: American Journal of Science, v. 290, p. 43-106.

Holwerda, J. G., and R. W. Hutchinson, 1968, Potash-bearing evaporites in the Danakil area, Ethiopia: Economic Geology, v. 63, p. 124-150.

Pedley, H. M., J. Neubert, and J. K. Warren, in press, Potash deposits of Africa: African Mineral Deposits, 35TH International Geological Congress (IGC), Capetown (28 August to 4 September 2016).

Warren, J. K., 2015, Evaporites: A compendium (ISBN 978-3-319-13511-3) Released August 2015: Berlin, Springer, 1600 p. 

Danakhil Potash, Ethiopia: Is the present geology the key? Part 1 of 4

John Warren - Sunday, April 19, 2015

Geology of potash in the Danakil Depression, Ethiopia: Is the present the key?

The Danakhil region, especially in the Dallol region of Ethiopia, is world renowned for significant accumulations of potash salts (both muriates and sulphates), and is often cited as a modern example of where potash accumulates today. What is not so well known are the depositional and hydrological dichotomies that control levels of bittern salts in the Pleistocene stratigraphy that is the Danakhil fill. Geological evolution of the potash occurrences in the Dallol saltflat and surrounds highlights the limited significance of Holocene models for potash, when compared to the broader depositional and hydrological spectra preserved in ancient (Pre-Quaternary) evaporite deposits (see Warren, 2010, 2015 for a more complete analysis across a variety of evaporite salts).

Across the next four blogs, I shall discuss the geological character of the Danakhil fill and the controls on potash in the depression via four time-related discussions; A) Current continental fan - saltflat hydrology that typifies present and immediate past deposition in the depression (Danakhil Blog1). B) A time in the latest Pleistocene when there was a marine hydrographic connection exemplified by a healthy coralgal rim facies (probably ≈ 100,000 years ago, and a subsequent drawndown gypsum rim facies. Both units are discussed in this blog, (Danakhil Blog1), and C) a somewhat older Pleistocene period when widespread potash salts were deposited via a marine seepage fed hydrology (Danakhil Blog2). Then, within this depositional frame, we will consider D) the influence of Holocene volcanism and uplift driving remobilisation of the somewhat older potash-rich evaporite source beds into the Holocene hydrology (Danakhil Blog3) and finally how this relates to models of Neogene marine potash deposition (Danakhil Blog4). These observations and interpretations are based in large part on a two-week visit to the Dallol, sponsored by BHP minerals, and focused on the potash geology of the region. 

Dallol Physiography

The Danakhil Depression of Ethiopia and Eritrea is an area of intense volcanic and hydrothermal activity, with potash occurrences related to rift magmatism, marine flooding, and deep brine cycling. The region is part of the broader Afar Triple Junction and located in the axial zone of the Afar rift, near the confluence of the East African, Red Sea and Carlsberg rifts (Figure 1a; Holwerda and Hutchison, 1968; Hutchinson and Engels, 1970; Hardie, 1990). The depression defines the northern part of the Afar depression and runs SSE parallel to the Red Sea coast, but lies some 50 to 80 km inland, and is separated from the Red Sea by the Danakil Mountains. The fault-defined Danakil Depression is 185 km long, up to 70 km wide, with a floor that in the deeper parts of the depression is more than 116 meters below sea level. It widens to the south, beginning with a 10 km width in the north and widening up to 70 km in the south (Figure 1a). In the vicinity of Lake Assele, the northern portion of the Danakil is known as the Dallol Depression and has been the focus for potash exploration for more than a century and is in the deepest region of the depression with elevations ranging between 50m to 120m below sealevel (Figure 1b, c). Shallow volcano-tectonic barriers, behind Mersa Fatma, Hawakil Bay and south of the Gulf of Zula, prevent hydrographic (surface) recharge to the depression. Marine seepage is not occurring at the present time, but likely did so at the time the main potash unit was precipitated. Lake Assele (aka Lake Kurum) with a water surface at -115m msl should not be confused with Lake Asal (-155 msl), located 350 km to the southeast of the Danakil. Asal an active marine-fed hydrographically isolated lacustrine drawdown system, which today is depositing a combination of pan halite and subaqueous gypsum in the deepest part of the Asal-Ghubbat al Kharab rift (Figure 1a; Warren, 2015).

Today the halite-floored elongate saltpan, known as the Dallol saltflat, occupies the deepest part of the northern Danakil Depression, extending over an area some 40 km long and 10 km wide (Figure 1b, c). The pan’s position is asymmetric within the Danakil Depression; it lies near the depression’s western edge, some 5km from the foot of the escarpment to the Balakia Mountains and the Ethiopian Highlands, but some 50 km from the eastern margin of the depression, which is in Eritrea. The Dallol saltpan and adjacent Lake Assele today constitute the deepest continental drainage sump in the Afar depression (Figure 1b, c). The area, located east and northeast of the main modern Dallol saltpan depression, is mostly an extensive gypsum plain (Bannert et al., 1970). As we shall see, the gypsum pavement, and its narrower equivalents on the western basin flank, defines a somewhat topographically higher (still sub-sealevel) less-saline, lacustrine episode in the Dallol depression history fill. To the south of the Dallol salt pan, bedded Pleistocene evaporites may underlie the entire Danakil depression, but younger lava flows of the Aden Volcanic Series in combination with alluvium washed in from the surrounding bajada obscure much of the older Pleistocene sedimentary series in southern part of the depression beyond Lake Assele (Figures 1a).


In terms of daily and monthly temperatures, the Dallol region currently holds the official record for highest average, year-round, monthly temperatures; in winter the daily temperature on the saltflat is consistently above 34°C and in summer every day tops 40°C, with some days topping 50°C (Figure 2; Oliver, 2005). These high temperatures and a lack of rainfall, typically less than 200 mm each year, place the Dallol at the hyperarid end of the world desert spectrum and so it lies at the more arid end of the BWh Köppen climate zone (Kottek et al., 2006; Warren, 2015).

History of extraction of salt products and their transport (Table 1)

Using little-changed extraction and transport methods, salt (halite) has been quarried by local Afar tribesmen for hundreds of years. First, using axes, a crust of pan salt is chopped into large slabs (Figure 3a). Then workers fit a set of sticks into grooves made by the axes. Next, working the stick, workers lever slabs of bedded salt, which is cut into rectangular tiles of standard size and weight, called ganfur (about 4kg) or ghelao (about 8kg). Tiles are stacked, tied and prepared for transport out of the depression on the backs of camels and donkeys (Figure 3b). Around 2,000 camels and 1,000 donkeys come to the salt flat every day to transport salt tiles to Berahile, about 75 km away. Previously, salt tiles were carried via camel train to the city of Mekele, some 100 km from the Danakil. Mekele, located in the Ethiopian highlands is known as the hub of Ethiopia’s former “white gold” salt trade and still today is known as the “old” salt caravan city. Today, the salt caravans walk the extracted salt to Berahile, located some 60 km from Mekele. From there, trucks transport the salt to Mekele. Each truck can transport up to 350 camel salt loads. From the Mekele salt market, Dallol salt blocks are transported and sold to all parts of Ethiopia for use mainly as table salt or as an add-on in animal feed. The lifestyle of the miners and the camel trains is likely to change in the next few decades as sealed roads are now under construction that will link Mekele to Dallol.

Once potash (sylvite and carnallite) was discovered in the Dallol region in 1906, an Italian company by the name of Compagnia Mineraria Coloniale (CMC) established the first mining operation. In 1918 a railway was completed from the port of Mersa Fatma to a termination some 28 km from Dallol (Table 1). Rail construction took place from 1917-1918, using what was then the British and French “military-standard” 600 mm rail-gauge Decauville system. "Decauville" rail construction used ready-made sections of small-gauge track and so the trackway was rapidly assembled; <2 years to complete more than 50 km of track. Once completed, the railway transported extracted potash salt from the "Iron Point" rail terminal near Dallol, via Kululli to the port. Potash production is said to have reached some 50,000 metric tons in the 1920s, extracted from an area centred on the Crescent Deposit, which is located near the foot of uplifted lake beds on the southern flanks of Mt Dallol. However, significant salt production had ceased by the end of the 1920s, as large-scale mines in Germany, the USA, and the USSR began to supply the world market with cheaper product. Unsuccessful attempts to reopen potash production were made in the period 1920-1941. Between 1925-29 some 25,000 tons of sylvite were shipped by rail from the Dallol, with a product that averaged 70% KCl. After World War II, the British administration dismantled the railway and removed all traces of it. In 1951-1953, the Dallol Co. of Asmara sold a few tons of product from the Dallol.

The potash concession title was transferred to the American “Ralph M. Parsons Company” (Parsons) at the end of the 1950s. Parsons initiated the first systematic exploration for potash in the Danakil depression and drilled more than 250 exploration holes during their 9-year evaluation campaign. Major potash resources were confirmed a few km west of Mount Dallol, in a mineralized zone that was named the “Musley” Deposit. Following on from positive exploration results, they began an engineering study to investigate potential processing and mining methods for the Musley Deposit and subsequently in October 1965 sank a shaft into the orebody. They installed underground mine facilities and established a pilot processing plant on surface, to investigate recovery from the bulk samples collected from the underground workings. They envisaged developing the Musley Deposit as a conventional room-and-pillar operation and to this end developed six underground drifts totalling some 805 m in length. Unconfirmed reports suggest that an influx of water flooded the mine (possibly triggered by a seismic event) and after failed attempts to solve the water problem, the activities Parsons ceased activities in 1968. As of end 2014, some salt block buildings built by the Italian and other companies still partially stand as ruins, along with rusting equipment.

Based on the previous work conducted by Parsons, a German potash producer, Salzdetfurth AG (SAG), began a new exploration campaign in the Danakil Depression in 1968 and 1969. In addition to their work in the Dallol depression, SAG drilled a number of wells in a concession south of Lake Assale, and conducted a geological mapping campaign as far north as Lake Badda, on the border with Eritrea. SAG’s exploration work away from the known Dallol deposits did not prove fruitful as they drilled only one drill hole that reached the potash level. This drill hole, located approximately 25 km to the southeast of Mount Dallol, intersected a kainitite bed, with no sylvinite intersection. The SAG concession was returned to state authorities of Ethiopia. Subsequent drilling by other explorationists in this region has confirmed the deepening of the kainitite level to the southeast of Dallol and the lack of sylvinite at greater depths.

Since the dismantling of the railway, there has been no high-volume transport system to carry potash product the Red Sea coast. Currently, the Ethiopian Government is constructing all-weather roads from Dallol to Mekele and Afdera When complete this road system will facilitate transport of future potential potash product from the Dallol to Afdera, from where existing roads provide access to Serdo and from there to the seaport of Tadjoura in Djibouti (Figure 1a). This section requires an addition 30 km of all-weather road to be completed to the coast and will facilitate cost-effective transport of potash product to the large agricultural markets of India and China. The transport distance to the Eritrean coast from Dallol is much shorter, but political considerations mean such a route is not a viable option at the present time.


Surficial sediment distributions outline classic drawdown facies belts in the Dallol region, with a wadi-fed alluvial fan fringe passing down dip into sandflats (local dune fields), dry mudflats (with springs), saline mudflats and ephemeral to perennial brine pans of Lake Assele (Figure 1b). The fans, especially along the western margin of the depression are indented or locally covered by a mostly younger succession of constant-elevation marine, biochemical and evaporitic sediments fringes or “bathtub rim” facies (Figure 4).


Alluvial Fan fringe (Bajada) 

Pleistocene alluvial/fluvial beds, exposed by local uplift, deflation and ongoing watertable lowering, outcrop about updip edges to the salt-crusted parts of the northern Danakil, and form low flat-topped plateaus or mesas on the plain. These mesas define the tops of alluvial fans aprons, which are heavily dissected and eroded by occasional storm runoff and rainfall. This fan fringe contains relatively fresh water lenses in a desert setting that is one of the world’s harshest (Kebede, 2012). Most of the depositionally active fans line the western margin of the basin and many of the downdip fan edges occur slightly up dip a still-exposed gypsum pavement (Figure 5a), showing depositional equilibration largely with an earlier higher lake stage, while others, such as the Musley fan, have flowed across cut into the gypsum pavement level and now feed water and sediment directly into the edges of the saltflat that defines the lower parts of the depression (Figure 4). Watercourses of the fans that have dissected earlier wadi (bajada) deposits as well as the earlier lacustrine gypsum and limestone pavements so create excellent windows into the stratigraphy of these units. Fan avulsion is indicated by palaeosol layers exposed by downcutting of younger streams (Figure 5b, c).


The Musley fan characteristics are well documented by current and previous potash explorers in the basin as these permeable gravels and sands store a reliable water source for potential solution mining/ore processing in the Musley area and so has been cored by a number of proposed water wells. Internally, the fan is composed of interfingering layers and lenses of sand, gravel and clay (paleosols), with highly porous intervals in the sand and gravels (Figure 5b, c). Depth to the watertable varies from >2m to 60m, and salinities from 760 ppm to more than 23,400 ppm. The principal source of recharge is flash flooding, originating in Musley Canyon, which drains the Western Escarpment, along with minor inflows from the adjacent uplifted volcanic block and local highly intermittent rains (Figure 1a). Of six potential water wells drilled in the fan by the Ralph M. Parsons Company in the 1960s, four returned water of good quality (<2000 mg/l), while the other two had waters with salinities in excess of 20 g/l. Pumping test data indicate average transmissivity of the water-bearing beds around 870m2/day, with salinities in the fan increasing from west to east, approaching the saltflat.

Chemical sediments outcropping in the depression

Overall, surface sediment patterns in the Danakil depression define a depositional framework of brine drawdown, related to basin isolation from an earlier hydrographic (at surface) marine connection to the Red Sea, followed by stepped evaporative drawdown. This is indicated by fringing topographically-distinct belts or rims of now inactive coralgal carbonates and gypsum evaporites (aka “bathtub ring” patterns) that cover earlier Pleistocene and Neogene clastics (Figures 3a, e, Figure 4). These “rims” of marine limestone and subsequent gypsum were followed by today’s drawdown saline-pan halite-dominant hydrology (Figures 4, 7a-c). The current hydrological package of sediments encompassing the current drawdown episode lies atop and postdates the Pleistocene potash-hosting Lower Halite Formation in the depression and is probably equivalent to the uppermost part of the clastic overburden facies, as illustrated in the drilled and cored portions of the depression stratigraphy. As we shall discuss in the next blog, only the uppermost portion of the recovered core stratigraphy has equivalents in current depression hydrology (Figure 6). 

In earlier work, some authors interpreted the fringing belts, especially the exposed coralgal reef belt, as being possibly of Pliocene or even Miocene age. However, when one looks at the stratiform nature of the outcrop trace of both the reef belt and the gypsum belt, and the carapace nature of its depositional boundaries in the field, it is immediately apparent they must be younger (Figure 5a, c; Figure 7). Both the reefal and gypsum belts track horizontal hydrological intersections with the landscape, in what is an ongoing volcanogenic and tectonically active depression. When the reefal belt image is overlain by a DEM it shows the reef belt is consistently at sea level (Figure 1c). If the outcrops of the reef belt and the gypsum pavement were older than late Pleistocene or Holocene, then ongoing episodes of tectonism and volcanism would have modified the elevations of the two outcrop belts in the landscape, as is seen in Miocene redbed outcrops. These underlying and centripetal Miocene sections clearly show the influence of ongoing tilting and tectonism and hence why the flat-lying tops to the reef and gypsum belts imply a late Pleistocene-Holocene (Figure 5d).

That is, the topographic distribution of the top of the reef facies, which lies within a metre or two of current sea level, implies that the Danakil depression had a relatively recent connection to the Red Sea. The pristine preservation of aragonitic corals and sand dollars in the adjacent marls suggest the connection was either related to the penultimate interglacial (around 104, 000 years ago) or to an early Holocene transgression into the depression. Bannert et al. (1970) assign a C14 age of 25.4-34.5 ka to this formation. However, we consider this is unlikely as DEM overlay levelling shows the reef rim, wherever it outcrops, lies within a meter of current sealevel. World eustacy clearly shows that sealevel was more than 50-60m below its present level some 25,000-30,000 years ago. A 25-35 ka determination of the reef rim would require the whole basin was subject to a single basinwide wide vertical uplift event that did not fragment or disturb the lateral elevation of the rim.

The coralgal reef terrace indicates normal marine water were once present in the Dallol depression, while the occurrences of the stratiform gypsum pavement are consistent with a former arid lake hydrology at a somewhat lower elevation than the reef rim (Figures 1c, 5a). Like the reef rim, the gypsum pavement fringe defines a consistent elevation level or surface, most clearly visible along the western margin of the present salt flat. It is the result of gypsum deposition during a period of drawdown associated with brine level stability, subsequent to the isolation of the depression from its former “at-surface” marine connection. During this time gypsum accumulated as a stack of subaqueous aligned gypsum beds, along with a series of gypsiferous tufas and rhizoconcretions in zones about the more marginward spring-fed parts of the gypsiferous lake margin (Figure 7d-f). The evolution from marine waters that deposited the reefs and adjacent echinoid-rich lagoonal marls at a higher level in the depression (the lit zone) into a more saline seepage-fed system, with no ongoing marine surface connection to the Red Sea is indicated by the diagenetic growth of large lenticular (“bird’s-beak”) gypsum crystals within the marine marls and the dominant subqueous bottom-nucleated textures in the gypsum beds. In a similar way, the now-outcropping subaqueous-gypsum drawdown rim deposits, located at higher elevations than current saline pan levels typify other drawdown saline lakes in the Afar region, such as Lake Asal in Djibouti, all such occurrences indicate an earlier, somewhat less saline, hydrological equilibrium level (Warren, 2015).

Active today is the lowest parts of the Dallol saltflat is an ephemeral saltpan hydrology indicated by bedded salt crusts dominated by megapolygonal crusts made up of aligned-chevron halite stacks separated by mm-cm thick mud layers . This current pan hydrology is associated with even greater drawdown levels compared to the former gypsum-dominant hydrology (Figure 8). Current deposits, made up a series of stacked brine-pan salt sheets,  are still quarried as a renewable resource by the local tribesmen (Figure 3). These modern brine flats accumulate pan halite whenever the Lake Assele brine edge (strandline) is periodically blown back and forth over the modern brineflat. It driven by southerly winds, which are frequent in the annual weather cycle, and can move thin sheets of brine kilometres across the pan in a few hours (Figure 1a, Figure 8). Superimposed on this southerly supply of brine is an occasional land-derived sheetflood event, driven by rare rainstorms and the deposition of silt-mud layers from water sheets sourced from the adjacent wadi belt. This ephemeral brineflat hydrology is stable with respect to the current climate (groundwater inflow ≈ outflow). It means the current brineflat of the northern Danakil low is  accumulating bedded pan salt at an even lower topographic level in the basin than the surrounding gypsum pavement, so implying today’s halite-dominant pan beds form under more arid conditions (less humid, more drawndown) than that of the gypsum pavement.

This stepped (reef to gypsum to halite) late-Pleistocene-early Holocene hydrology, captured in the modern surficial geology of the Dallol Depression, likely postdates a somewhat wetter (humid) climatic period indicated by the widespread deposition of a clastic overburden unit, atop the Upper Halite Formation (UHF; Figure 6). That is, the modern hydrology in present-day Lake Assale, and the adjacent saline mud flats of the Dallol pan, is not the same hydrology as that which precipitated the massive salt of the Upper Halite Formation (UHF). A potash-free halite unit extensively cored beneath the present clastic-dominated saline pan (to be discussed in the next blog). Texturally and hydrologically the depositional system of stacked salt crusts, which dominates the upper part of the UHF in the cored wells, is similar to today's halite-dominated passage from the salt flat into the subaqueous Lake Assale. However, as we shall see, a wetter moister period, dominated by sheetfloods and higher amounts of clastics, separates the two hydrological events in all the cored wells. Today's outcrop geology of alternating saltpan and clastic beds are a different to marine-fed seepage hydrology formed the Lower Halite Formation (LHF), with its potash bittern cap (Houston Formation). 

Most importantly there is no evidence of primary potash deposition in the modern lake/ pan hydrology of the Dallol saltflat. It is clear that the world-famous bedded potash (mostly kainitite) units of the Danakhil accumulated in a bittern hydrology that is not present in today's Dallol depositional hydrology (Blog 2). As we shall see, Holocene potash only occurs in the vicinity of the Dallol Volcanic Mound, where uplift has moved older, formerly buried, potash beds into a more active hydrothermal hydrology (Blog 3).


Bannert, D., J. Brinckmann, K. C. Käding, G. Knetsch, M. KÜrsten, and H. Mayrhofer, 1970, Zur Geologie der Danakil-Senke: Geologische Rundschau, v. 59, p. 409-443.

Ercosplan, 2011, Resource Report for the Danakhil Potash Deposit, Afar State Ethiopia, comissioned by Allana Potash. Document EGB 11-008.

Hardie, L. A., 1990, The roles of rifting and hydrothermal CaCl2 brines in the origin of potash evaporites: an hypothesis: American Journal of Science, v. 290, p. 43-106.

Holwerda, J. G., and R. W. Hutchinson, 1968, Potash-bearing evaporites in the Danakil area, Ethiopia: Economic Geology, v. 63, p. 124-150.

Hutchinson, R. W., and G. G. Engels, 1970, Tectonic significance of regional geology and evaporite lithofacies in northeastern Ethiopia: Philosophical Transactions of the Royal Society, v. A 267, p. 313-329.

Kebede, S., 2012, Groundwater in Ethiopia: Features, numbers and opportunities, Springer.

Kottek, M., J. Grieser, C. Beck, B. Rudolf, and F. Rubel, 2006, World Map of the Köppen-Geiger climate classification updated: Meteorologische Zeitschrift, v. 15, p. 259-263.

Warren, J. K., 2010, Evaporites through time: Tectonic, climatic and eustatic controls in marine and nonmarine deposits: Earth-Science Reviews, v. 98, p. 217-268.

Warren, J. K., 2015, Evaporites: A compendium (ISBN 978-3-319-13511-3) Released August 2015: Berlin, Springer, 1600 p.

Solikamsk sinkhole and potash ore

John Warren - Thursday, February 19, 2015

The Solikamsk-2 mine output currently accounts for a fifth of the Uralkali's potash capacity. The mine is one of Uralkali’s potash mines in the Kama district (western Urals) of Russia. The possible loss of the Solikamsk-2 mine in the near future may put upward pressure on the currently low world price of potash. Even if loss of production from the mine doesn't drive an increase in price, Uralkali as a company will continue to be profitable. All mines in Solikamsk basin benefit from low production costs, related to shallow ore depth. The collapse sinkhole is a classic example of what can happen if any salt or potash mine operates at a depth that is shallow enough to intersect the overlying  zone of active phreatic water crossflow.

In November 2014 the most recent example of an evaporite-related ground collapse or sinkhole, daylighted to the east of the Solikamsk-2 potash production site. It seems likely that halite-undersaturated waters, sourced from above, now flow into an area of the Solikamsk-2 workings that were no longer mined, but were still connected to the active extraction areas of the Solikamsk mine. Worldwide experience shows that flooding is difficult to control once a salt mass is breached. Increasing volumes of inflow waters into the mine workings will likely lead to the ultimate loss of the Solikamsk 2 mine, as it has in what are other now abandoned salt mines and solution brinefields across the world.

Recently the head of the Ural region's Mining Institute, Alexander Baryakh, was quoted as saying (Moscow Times Dec. 11 2014): ....."Based on our analysis and the world's experience in developing potassium mines, the risk of a negative scenario — the complete flooding on the mine — remains high. We are ready for this contingency, but we are doing everything possible to minimize related risks,"  adding that, "fortunately, the accident poses no danger to the residents of [the local town of] Solikamsk....”

When, in mid November, a stoping solution cavity above the abandoned section of the mine workings reached the surface, the resulting sinkhole diameter measured 30m by 40m. As of 6 February 2015, the at-surface sinkhole diameter had increased in size and measured 58 by 87 metres wide and was some 75 metres deep. Uralkali’s ongoing measurements show levels of brine inflow into the Solikamsk-2 mine continuously varied over this time frame. Between 11 December 2014 and 21 January 2015, the average brine inflow was around 200 cubic metres per hour. Between 22 January and 06 February 2015, the average brine inflow increased substantially, reaching approximately 820 cubic metres per hour. 

Accordingly, where underground equipment at Solikamsk 2 is not being used to mitigate the consequences of the collapse and water inflow, Uralkali has started to remove plant via the mine shaft. Three "Ural" continuous miners have already been dismantled and taken out. In line with their accident mitigation plan, Uralkali in a recent press release states: that the company continues to comprehensively monitor the situation both underground and at the surface; water inflows are monitored through brine level checks; groundwater levels are monitored by water monitoring wells and via the drilling of additional water monitoring wells currently in progress; gas levels are monitored around the sinkhole and in the mine; while the sinkhole is continuously monitored from a distance, using stationary cameras and air drones; and a seismologic network has been set up over the sinkhole area.

This latest collapse is one of a series of evaporite collapse dolines or sinkholes that have daylighted atop the mined regions in this part of Russia. Collapse cavities typically reach the surface some years after extraction operations below the collapse have ceased. For example, in 1995, a collapse sinkhole formed atop the Solikamsk-2 potash mine’s Verkhnekamsky deposit. The collapse on January 5th, 1995 resulted in a 4.7 magnitude seismic event on the Richter scale, with an associated initial 4.5 m of surface subsidence. Underground, the mine roof collapsed over an area measuring 600 m by 600 m. Across the period 1993 to 2005, several hundred earthquakes were recorded in the Berezniki-Solikamsk region with magnitudes varying from 2 to 5. These earthquakes were caused by collapsing underground tunnels of potash mines, mined out over the 70 continuous years of production. In October 2006, in order to prevent catastrophic outcomes of a sudden brine influx into the underground workings, Berezniki potash mine #1 was flooded by Uralkali. After that, three major sinkholes occurred in the region above the flooded workings.

On July 28th, 2007, a huge sinkhole appeared on the surface above the closed Berezniki mine #1, its creation likely aided by infiltration of undersaturated floodwaters water into the abandoned underground workings. At the surface this sinkhole had an initial size of 50 by 70 meters and was 15 meters deep. By November, 2008, the sinkhole had expanded into a crater measuring 437m by 323m and some 100 m deep.  The July 28th collapse released an estimated 900,000 cubic metres of gas (a mix of methane, hydrogen, carbon dioxide, carbon monoxide and other gases), which in turn led to gas explosions on the following day. Timely placement by the mine operators of a “significant volume” of backfill, prior to flooding, is credited with preventing further catastrophic collapses.

The mined potash ore level at both collapse sites (Solikamsk and Berezniki) is Middle Permian (Kungurian) in ag. The potash is halite-hosted and occurs in one of six evaporitic foreland sub-basins, extending from the Urals foreland to the Caspian basin. Sylvinite (potash) beneath the 1995 Verkhnekamsky collapse area was extracted from two to three halite-potash beds, with 10 to 16 metres of total extraction height. At that time the mine used a panel system of rooms and pillars under 200 to 400 m of overburden. Rooms were 13-16 m wide and pillars 11-14 m wide by 200 m long. Due to the relatively shallow nature of the Solikamsk and Berezniki potash ore levels, compared to other potash mines in the world, a “rule of thumb” used across the Upper Kama mining district is that surface subsidence typically reaches 50% of the subsurface excavation height around 48 months after excavation. 

The 1995 collapse event occurred 15 years after mining began, and 7 years after mining was completed in the area beneath the sinkhole. The delay before the main collapse doline surfaced implies there was a rigid bridging of overburden as a roof to the mine level. This is consistent with the uncommonly high release of seismic energy associated with the1995 collapse. The next largest collapses associated with published seismic measurements occurred in 1993 and 1997 with seismic magnitudes of 2.6 and 2.8, respectively.

An even earlier surface collapse occurred on July 25, 1986 atop a portion of the nearby 3rd Bereznki potash mine and is yet another case of a sinkhole forming atop a mine that was operating at relatively shallow depths. Potash extraction at Bereznki was active at depths of 235 and 425 m below surface. There, the targeted ore zone was overlain by a 100 m thick “salt complex” made up of halite and carnallite beds, overlain in turn by clays, carbonates, aquifers and sediments. Mining created “yield pillars,” with 5.3 m wide rooms, 3.8 m wide pillars and a 5.5 m mining height. After mining, conditions in mined-out areas were described as, “pillars crushed and roofs sagged.”

Observations of significant brine leakage into the 3rd Bereznki potash mine workings at a depth of 400m indicated a loss of hydraulic control as early as January of 1986. This was a prelude to the massive dissolution cavitation that was occurring in the 90 m interval of disturbed salt and clastics that overlay the potash level. Some 7 months later a large cavity formed in the sandstone/limestone overburden, which was nearly 200 m thick. In the mine it appeared the water inflow situation remained relatively stable, at least from January until July 1986. Failure of the mine head then occurred, the result of a cavity that had migrated vertically through more than 300m of limestones, mudstones and sandstones. 

Final cavity stoping was indicated by the near instantaneous appearance of a caprock sinkhole, which was 150 m deep and 40-80 m across and located at the top of a stoping breccia pipe or chimney of the same dimensions. Failure of this sequence began at 18:30 hours on July 25 with “clearly felt underground shocks” culminating with a final collapse at midnight, which was accompanied by an explosion with “flashes of light.” In the final stages of stoping by the rising solution pipe, before the sinkhole daylighted, it took only 12 days to migrate through the last 100 metres of mudstone. This very high rate of stoping was likely aided by structural weakness in a fracture zone along a local fold axis.

In all these cases of rapid sinkhole creation, the collapse occurred above what was formerly an active area of the mine  and took place some years after mining had ceased. In all cases, the ultimate cause of the size of the collapse was likely a combination of a significant cavity growth below what was a mechanically strong rock rock, likely a dolomite or a limestone bed. This unit had significant structural integrity and so allowed a solution cavity to expand prior to the ultimate brittle collapse of roof rock. Once collapse did occur, undersaturated groundwaters, sourced in the overburden, then reached the salt level in large volumes and further expanded the region of collapse. Likewise, once the upward stoping cavity reached the shallower unconsolidated sediment levels, the cavity's passage to the surface sped up so that it daylighted and expanded in a rapid fashion.  

Dissolving evaporites and solution dolines occur naturally in all parts of the world, wherever salt is within a few hundred metres of the landsurface, but mining of both salt and potash at depths shallower than 250-350 metres can exacerbate the speed and and intensity of what is an ongoing natural process of evaporite solution, surface collapse and sinkhole growth. While Uralkali's operations in this region continue to exploit a relatively shallow potash ore source, it will continue to supply the Company a relatively inexpensive product, but the company will have live with sinkholes breaking out above some areas of the mined region. That is, as long as Uralkali can continue to be a low cost supplier of potash, there will be likely be ongoing landsurface-stability problems. Some of the problems may not daylight until years after the extraction operation has ceased.

Recent Posts


hydrogen lithium brine Evaporite-source rock association Crescent potash Neoproterozoic Musley potash potash ore Mega-monsoon evaporite dissolution supercontinent intersalt freefight lake lithium carbonate Deep solikamsk 2 Messinian magadiite DHAL gassy salt base metal circum-Atlantic Salt Basins Dallol saltpan MOP anomalous salt zones astrakanite Hadley cell: collapse doline CO2 chert Kara bogaz gol cryogenic salt tachyhydrite Atlantis II Deep Lake Peigneur Danakhil Depression, Afar Corocoro copper saline clay kainitite SOP lithium battery CaCl2 brine Zaragoza sinkhole halite trona ancient climate Jefferson Island salt mine marine brine Weeks Island salt mine Neoproterozoic Oxygenation Event blowout authigenic silica hydrothermal potash epsomite allo-suture Stebnyk potash evaporite karst salt periphery oil gusher intrasalt salt trade eolian transport halotolerant sepiolite evaporite geohazard Five Island salt dome trend flowing salt salt karst black salt nacholite High Magadi beds NaSO4 salts silica solubility perchlorate waste storage in salt cavity crocodile skin chert climate control on salt lazurite Belle Isle salt mine dissolution collapse doline potash Quaternary climate dihedral angle Dead Sea karst collapse potash ore price bedded potash Zabuye Lake Warrawoona Group African rift valley lakes Pilbara deep meteoric potash mine stability namakier Proterozoic sedimentary copper Muriate of potash Koeppen Climate lunette recurring slope lines (RSL) snake-skin chert jadarite Paleoproterozoic Oxygenation Event natural geohazard salt leakage, dihedral angle, halite, halokinesis, salt flow, sinjarite brine evolution water in modern-day Mars HYC Pb-Zn Karabogazgol halokinetic doline dark salt evaporite-hydrocarbon association hectorite evaporite-metal association salt ablation breccia well blowout Red Sea Hyperarid hydrohalite Clayton Valley playa: Catalayud vadose zone gem lapis lazuli alkaline lake nuclear waste storage Ripon Pangaea cauliflower chert organic matter extrasalt salt tectonics mirabilite gas in salt salt seal salt mine Kalush Potash lot's wife silicified anhydrite nodules rockburst vanished evaporite source rock antarcticite Ingebright Lake Magdalen's Road bischofite Ethiopia hydrothermal karst Sumo MVT deposit knistersalz anthropogenically enhanced salt dissolution Hell Kettle auto-suture Great Salt Lake water on Mars Ure Terrace Salar de Atacama Dead Sea caves stevensite Lomagundi Event gas outburst McArthur River Pb-Zn palygorskite nitrogen Badenian venice North Pole Mesoproterozoic Precambrian evaporites mummifiction Realmonte potash Archean SedEx halite-hosted cave mass die-back Platform evaporite zeolite carnallitite meta-evaporite hydrological indicator salt suture Koppen climate methane Deep seafloor hypersaline anoxic lake sodium silicate Lake Magadi Turkmenistan Sulphate of potash H2S Stebnik Potash basinwide evaporite subsidence basin gypsum dune seal capacity capillary zone