Ground fissures are linear or curvilinear cracks that are common in the Main Ethiopian Rift, which is part of the East African Rift System. They are a unique type of geohazard affecting roads and railway in many parts of the rift valley. The vicinity of Lake Ziway is underlain by lacustrine, pyroclastic, and/or volcaniclastic sediments that are cut by numerous ground fissures, which run for 2–3 km with up to 1–3-m-wide openings. Plausible causes of ground fissure events, as described in previous studies, include extensional movement due to active rift tectonics, hydro-compaction, and piping. However, the exact mechanism of their evolution is poorly understood. Based on field observations and sediment characterization, we propose a strong case for the role played by internal erosion or piping of pumice deposits in the development of ground fissures. Internal erosion causes the formation of subsurface conduits, which grow in size and eventually collapse, forming ground fissures. Typical sediment-erosion–related geomorphologic features such as disappearing streams, sinkholes, blind gullies, and piping mounds have been observed. These features indicate the presence of a network of pipes that connect the groundwater with surface water in a manner very similar to karst hydrology. Pumice deposits in the study area are found to be the most susceptible to internal erosion leading to ground fissuring due to their bimodal grain size distribution and ultralow density (average specific gravity = 0.6).

The Main Ethiopian Rift (MER) is a linear belt of extensional tectonic structures marked by normal fault systems. The MER is also riddled with ground fissures, which are commonly 1–3-m-wide cracks that may extend for 2 to 3 km (Figure 1). There have been very few reports of their impact on infrastructure in Ethiopia since the 1950s (Gouin and Mohr, 1967). Asfaw (2000) argued that ground fissures should be considered as extensional geologic structures similar to normal faults. Mohr (1967), Asfaw (1998), and Yirgu et al. (1997) proposed a direct relationship between ground fissures and earthquakes. Most previous research has linked ground fissures to rift tectonics based on their orientation being parallel with the NE-SW rift axis (Asfaw, 1998, 2000; Yirgu et al., 1997). Despite their widespread occurrence in the Ethiopian rift valley and their geohazard potential, ground fissures have not received enough attention in terms of geological and engineering investigations.

Ground fissures have also been reported and studied in many other parts of the world. The earliest report was from the Picacho Mountains in Arizona, southwestern United States, where a 15 km fissure opened following a heavy rainfall around the year 1927 (Leonard, 1929). More ground fissure events and subsequent studies were undertaken mainly in Arizona and other southwestern U.S. states (Carpenter, 1993; Galloway, 1999). Most of the fissures that form in the southwestern United States are long, linear features or parallel anastomosing cracks that show mainly horizontal movement with a minor vertical component (Carpenter, 1993). There are also several reports of ground fissuring from China’s Shanxi graben system and the North China Plain, where formation of ground fissures is attributed to pre-existing faults coupled with excessive groundwater withdrawal (Lee et al., 1996; Peng et al., 2016, 2018, 2020). A recent giant ground fissure (approximately 50 ft [15 m] deep and 65 ft [20 m] wide) from the Kenyan part of the rift valley occurred in 2018, cutting the Mai Mahiu–Narok road. This ground fissure received worldwide attention by the press and media because it was wrongly thought to be a sign of splitting of the African continent. Prior to the incident, several other ground fissures had been reported from the Kenyan segment of the East African Rift System (EARS), specifically in the Lake Nakuru area (Ngecu and Nyambok, 2000). In western Saudi Arabia, a rise in the number of ground fissures associated with differential compaction of loose sediments due to excessive groundwater withdrawal has been reported (Bankher and Al-Harthi, 1999).

The study area (bounded by 7°56′10.12″N and 8°03′11.95″N latitude and 38°38′22.49″E and 38°45′03.64″E longitude) is in the vicinity of Lake Ziway; this area is underlain by unconsolidated pyroclastic/volcaniclastic and lacustrine sediments, which are riddled with several ground fissures that threaten the stability of the newly built Meki-Ziway freeway. The objective of this study was to investigate the mechanisms (surficial or tectonic processes) that promote the formation and development of ground fissures based on data collected from the environs of Lake Ziway (Figure 2). Lake Ziway is one of four major lakes, also including Shala, Abaya, and Langano, that were connected to form an ∼2,960 km2 paleolake in the late Quaternary occupying the MER (Gillespie, 1983). The MER is part of EARS, which is major tectonic feature marking a zone of extension along the boundary between the Nubian and Somali Plates. The extensional tectonics in the MER has produced two main sets of NE-SW–trending boundary faults and NNE-SSW–trending faults, which are sub-orthogonal to the E-W extensional faults (Corti et al., 2013). Lake Ziway is fed by two major streams (Ketar and Meki) and drained by the Bulbula River. The average rainfall is 837 mm (Alemayehu, 2010), and the lake receives an approximate groundwater inflow of 220,700 m3/d and outflow of 40,047 m3/d (Ayenew, 2001). Groundwater recharge is mainly from the highlands to the east and west of the lake. Aquifers are represented by lacustrine/alluvial/volcaniclastic sediments and fractured volcanic units (Fentaw and Mihret, 2011).

Possible Genetic Models for Ground Fissure Development

Ground fissures affecting sediments can be the result of surficial processes such as internal erosion, decay of organic matter, hydro-compaction (rapid settlement of loose soil upon addition of water), fault movements, and/or horizontal seepage stress (Ayalew, 2004). However, most previous works in Ethiopia (Gouin and Mohr, 1967; Asfaw, 1982, 1998; and Ayalew, 2004) report that the orientation of ground fissures is subparallel to the NE-SW rift axis, suggesting a tectonic control on their genesis. From cursory observations in the Ziway area, one can notice a significant number of ground fissures that do not conform with the NE-SW orientation of the rift axis. The ubiquity of ground fissures and the rarity of earthquake activity in the Ziway area make it difficult to directly link tectonics with the ground fissures. Ayalew (2004), based on a more comprehensive study in the Muleti area (∼100 km south of Lake Ziway) in the central part of the Ethiopian rift valley, strongly disagreed with the connection between ground fissures and earthquake activity and proposed two mechanisms as probable causes: (1) the possible presence of unrelieved stress that could spread to the surface upon being subjected to heavy rainfall, which would reduce tensile strength, leading to the formation of fissures, and (2) rainfall-induced hydro-compaction of loosely packed alluvial soils, which could densify and build tensile stress that would eventually produce cracks. The effect of groundwater withdrawal has been extensively investigated by Sheng et al. (2003), who suggested differential aquifer deformation due to groundwater extraction causing extensional stress along planes of weakness such as fault planes and buried bedrock edges. In the Ziway area, however, there is little groundwater extraction that could account for ground fissure development.

An alternative model that may explain ground fissure development in the Ziway area is the geomorphologic expressions of internal erosion, also known as piping of sediments. According to Bernatek-Jakiel and Poesen (2018), internal erosion of soils can produce surface features such as sinkholes, blind gullies, and piping mounds (Figure 3). These features, as well as disappearing streams, are also well evident in Google Earth imagery in the Ziway area, suggesting the influence of internal erosion (Figure 4). Internal erosion of soils/sediments, also known as suffusion of internally unstable soils/unconsolidated sediments, leads to the formation of open conduits or pipes that serve as passageways for turbulent groundwater flow. Due to seepage forces, internally unstable sediments lose finer grain sizes through pore spaces created by larger grains. Continued internal erosion forms subsurface conduits that can collapse, initially forming sinkholes, which then coalesce to form ground fissures. Over time, ground fissures widen as walls collapse to become blind gullies (Figure 5). Internal instability depends on the grain size distribution and the magnitude of seepage forces. It is therefore necessary to characterize the piping potential of the mainly unconsolidated pyroclastic rocks underlying the Ziway area based on the high hydraulic gradient and bimodal grain size distribution.

Soil/Sediment Susceptibility to Internal Erosion

Internal erosion is triggered when groundwater seepage forces exceed in situ soil/sediment effective stress, leading to the formation subsurface pipes. Seepage forces are a function of hydraulic gradient and saturated density of sediments/soils. The critical hydraulic gradient value at which effective stress of soil/sediment becomes zero and piping occurs is ∼1 for most soils and siliciclastic sediments with specific gravity of ∼2.65. At vertical hydraulic gradients larger than 1, soil particles would lose strength and begin to flow and internally erode. The critical hydraulic gradient (ic) is calculated by the equation below (Terzaghi, 1939):
formula
where ϒsat is the saturated unit weight of soil/sediment, and ϒw is the unit weight of water.

Internal erosion can also occur at hydraulic gradient values less than 1, particularly for soils/sediments that lack certain grain size fractions, exhibiting a bimodal grain size distribution (Skempton and Brogan, 1994). Such soils are known as gap graded because they consist of large grain sizes without intermediate sizes, allowing finer grains to be washed through the large pore spaces created by the coarser fractions. Well-graded soils are therefore more stable as pore spaces are progressively filled by finer fractions.

The degree of gap gradation can be evaluated by visual inspection of cumulative grain size distribution curves to identify if a given soil/sediment is missing certain grain size ranges. Cumulative grain size distribution curves for gap-graded soils typically have flat section(s) on the curve where grain sizes are missing (Figure 6), as opposed to the uniform gentle slopes of well-graded soils. To evaluate the presence of flat sections indicating gap-graded distribution, Kenney and Lau (1985) proposed the use of the H/F value, where H is the difference in percent of grains passing between arbitrary grain sizes of one diameter (1D) and 4D, and F is the percent of grains corresponding to grain size D (Figure 6). Several H/F values can be calculated, but the minimum is used to identify gap-graded distributions. Soils/sediments with H/F <1 are considered as gap graded and therefore susceptible to internal erosion. Other graphically determined quantitative measures of piping susceptibility include Terzaghi’s filter criteria rule (Terzaghi, 1939). Terzaghi’s filter criteria originally was designed to select soils that would act as filters blocking the transport of fines from earth dams. The criteria was modified to assess piping potential within a given soil/sediment by separating the grain size distribution curve into coarse and fine grain size fraction distribution curves. If the ratio of the grain size corresponding to 15 percent (d15) passing of coarse fraction to the grain size corresponding to 85 percent (d85) passing of the finer fraction is greater than 4, then finer grain sizes can be eroded through pore spaces created by the coarser fraction, and hence internal erosion is possible (Skempton and Brogan, 1994). Additionally, the coefficient of uniformity (CU), which measures the uniformity of the grain size distribution, is also used to characterize internal instability.
formula
where d60 is the grain size corresponding to 60 percent passing, and d10 is the grain size corresponding to 10 percent passing on the cumulative grain size distribution curve.

High CU values (>10) may indicate a well-graded grain size distribution, which is also typical for internally unstable soils/sediments (Kovacs, 1981).

To explore the presence of geomorphological evidence for internal erosion and susceptibility of sediments to the process of piping, the research methodology included Google Earth imagery interpretation, field investigation, grain size analysis, and precipitation/earthquake data interpretation.

Google Earth Imagery

Ground fissures were mapped by tracing their extent on Google Earth imagery, which provides high-resolution images that were unavailable when most of the previous works on ground fissures were published. It was also possible to constrain the time of their opening using historical Google Earth imagery, which provided high-resolution aerial photographs dating back to early 2010s. Planetscope images with 3 m resolution from www.planet.com were used to constrain timing on some ground fissures.

On Google Earth imageries, ground fissures appear as linear/curvilinear openings, which can be continuous for up to 2–3 km per fissure. In some cases, linear vegetation tracks are also common manifestations of ground fissures. Their orientation was determined using Google Earth’s “ruler” tool. The orientations of Ground fissures showed a more dominant N-S orientation as well as a significant E-W trend (Figure 7). Some fissures appeared to have polygonal geometry, and others showed branching relationships, where shorter fissures connected perpendicularly with longer, more extensive fissures. From historical Google Earth imagery, ground fissures appeared to grow over time, in some cases by as much as hundreds of meters in a few years. The historical images also showed that fissuring started out as sinkholes, which eventually connected to form linear cracks (Figure 8). It was difficult to tell if the growth was gradual or sudden just from Google Earth imagery, which does not have short temporal resolution and only dates back to the 2010s, especially for the high-resolution images. However, using the available historical imagery, the (2016–2017) time period showed widespread fissure development (Figure 7).

Field Investigation

Image interpretation was followed by field mapping of the study area’s geology and ground truthing of ground fissures and geomorphological features interpreted from Google Earth imagery. During field investigation, a geologic map and representative stratigraphic sections were produced (Figure 7). Ground fissures interpreted from Google Earth imagery were ground checked. Geologic contacts and geomorphologic features were mapped. Samples from the pyroclastic and lacustrine deposits were collected for grain size characterization.

Ground truthing of image-interpreted fissures showed that Google Earth imagery was a reliable tool to map fissures with openings as narrow as <1 m width. Field investigation showed that ground fissures can be 1–3 m wide with no vertical displacement component and often run for up to 2–3 km. Their depth is very difficult to estimate, but most were estimated to be at least 10–20 m deep. The geology of the Ziway area is marked by extensive unconsolidated pyroclastic sediments interlayered with lacustrine deposits (Figure 7). The lake and its environs are within fault-bounded caldera depressions, which provided accommodation space for sediment accumulation and formation of the lake. The pyroclastic deposits appear to be derived from nearby volcanoes, and the lacustrine deposits are the result of shrinking of a large paleo-lake that preceded Lake Ziway (Gillepsie et al., 1983). The pyroclastic deposits consist of unconsolidated sediments ranging from fine ash to gravel-sized pumice deposits. It is possible that some pyroclastic material may have been reworked by water action. Welded tuff with thicknesses ranging from 10 to 30 cm and lacustrine deposits were also observed. A representative stratigraphy as exposed on stream sections shows, from top to bottom, welded tuff, unconsolidated sand-sized pyroclastic deposits, unconsolidated gravel-sized pumice, silty and clayey lacustrine deposits, and unconsolidated sand-sized pyroclastic deposits.

There are no drilling data to establish the depth to bedrock, but it is possible that sediment thickness may pinch out westward to the basin boundary. The only bedrock mapped is in the southwestern corner, where rhyolite outcrops as a footwall of a normal fault (Figure 7). Field mapping also identified geomorphological features including disappearing streams, blind gullies, and sediment mounds that are indicative of a process by which surface water is connected to the subsurface via open conduits similar to karst topography (Figure 7). Most fissures lie in the middle of open fields, but they are also observed developing along the drainage ditches adjacent to foot paths (Figure 9).

Laboratory Analysis

Sediment samples were sieved in the laboratory to characterize sediments’ susceptibility to internal erosion. At the study area, pyroclastic/volcanoclastic and lacustrine sediment samples for grain size distribution analysis were collected on a stream section of Meki River (Figure 10). Samples were sieved using sieve sizes 22.4, 11.2, 8, 4.75, 4, 2, 0.85, and 0.423 mm. Saturated density and dry density were also measured.

A plot of percent by weight retained by each sieve showed that two sieve sizes (2 mm and 4.75 mm) retained high amounts for the coarse-grained pumice (SC2-1A, B, C), indicating a bimodal distribution, which was further confirmed by low H/F values (Figure 11; Table 1). The cumulative distribution curve also showed that the coarse-grained pumice layer has flat sections, indicating gap-graded distribution (Figure 11). Saturated and dry densities were also determined in the laboratory. The dry and saturated density values of the pumice layer were very low, and, correspondingly, the calculated hydraulic gradients were also very low (Table 1). The pumice layer exhibited both gap-graded distribution and extremely low critical hydraulic gradient values, making it highly susceptible to internal erosion (Table 1).

Earthquake and Precipitation Data

To study the link between ground fissure development and earthquake and high rainfall events, earthquake events dating back to 1950 and rainfall data since 1980 were obtained from the U.S. Geological Survey (https://earthquake.usgs.gov/earthquakes/map/) and the Climate Research Unit Gridded Time Series (CRUTS) database (https://www.uea.ac.uk/web/groups-and-centres/climatic-research-unit/data), respectively. Table 2 shows earthquake events within 100 km from the study area since 1950. One particular earthquake, 29.5 km from Lake Ziway, appears to have been the most spatially and temporally related to ground fissures that occurred since the 2010s.

The CRUTS precipitation database is a compilation of interpolated monthly climate anomalies (Harris et al., 2020). Based on CRUTS precipitation data, the year 2016 received the largest amount of rain at 1,278.5 mm, followed by 1983 (1,215 mm), 1996 (1,207.7 mm), and 2021 (1,200.6 mm) (Figure 12).

The mechanism of ground fissuring in the central rift valley can be complicated. According to previous work, the most favorable cause is related to extensional tectonics (Asfaw 1982, 1998; Yirgu et al., 1997). Ayalew’s (2004) more extensive analysis proposed a combination of release of elastic strain due to heavy rainfall and hydro-compaction. Although plausible, it is difficult to show the presence of elastic strain in unconsolidated sediments. This study showed that the orientations of ground fissures are not in complete agreement with the general E-W–directed extension of the rift valley. An important observation made in the field was that the two sides of ground fissures do not have the same outline as one would expect from an extensional crack, and also no vertical offset typical of extensional faults was observed. Most ground fissure events do not coincide with seismic events, as ground fissuring is much more frequent than earthquake activities. Out of the earthquake events since 1950, only eight were within 100 km radius from the study area, and none was within 30 km. However, one event, from January 27, 2017, at ∼37 km from the study area, was suspected to have played a role in the development of the numerous ground fissures that formed within the 2016–2017 time frame. To verify this, satellite images with known acquisition dates were obtained from PlanetScope (www.planetscope.com). The Planetscope image (3 m resolution) shown in Figure 13 demonstrates that one of the 2016–2017 fissures existed in December 2016, which is well before the earthquake had happened on January 27, 2017.

Field investigation of geomorphologic features and sediment grain size analysis showed that piping caused by internal erosion of ultralow-density pyroclastic deposits is the most plausible mechanism. In striking similarity to areas subject to internal erosion described by Bernatek-Jakiel and Poesen (2018), geomorphologic features such as disappearing streams, blind gullies, and circular sediment mounds are well evident in the study area. Subsurface conduits can link surface water with groundwater, causing streams to disappear. The conduits could serve as subsurface continuations of surface drainage and also reappear on the surface, depositing sediments in the form of circular sediment mounds or in some cases forming circular ponds.

The presence of geomorphologic features indicative of internal erosion and the presence of ultralow-density pyroclastic sediments are proof that internal erosion can easily be triggered at very low critical hydraulic gradients that are much lower than the ideal critical gradient of 1. In addition to their low density, the gravel-sized pumice deposits show large variations in grain size, as shown by grain size distribution curves and H/F values, promoting their likelihood to internally erode and making them the most susceptible to internal erosion. Observations along the Meki River section showed open conduits that are strata-bound within the 4–5-m-thick gravelly pumice layer (Figure 10). The fact that ground fissures follow heavy rainfall is therefore due to increased groundwater flow in subsurface conduits, which increases the erosion of conduits, leading to roof collapse. Although it needs thorough investigation, the absence of ground fissures in drier areas underlain by sediments (Asfaw, 1998) is further proof that ground fissuring is erosion controlled. Along the Meki River section, collapsed pipes observed on the river section can be seen in plan view initiating ground fissure development (Figure 14). Linear ground fissures develop as subsurface groundwater conduits or pipes grow in size and eventually collapse. Continued erosion will widen ground fissures and turn them into blind gullies. A similar genetic model was proposed for the development ground fissures in the Lake Nakuru area of Kenya, where buried faults serve as sites of subsurface flow, causing erosion of overlying sediments, which subsequently collapse to form linear ground fissures (Ngecu and Nyambok, 2000). The widely publicized Kenyan fissure from 2018 followed heavy rainfall, and no preceding seismic activity was recorded. Ground fissuring activity in the 2016–2017 time period in the Ziway area, as constrained by historical Google Earth imagery, can be directly related to the year 2016, which recorded the largest amount of rainfall in over 40 years. Based on conversations with locals, the timing of several fissures in the area occurred as overnight events in May and August of 2016 following heavy rainfall. In addition to piping due to internal erosion, an additional plausible mechanism for the development of ground fissures is surface erosion, as observed along drainage ditches, causing deep erosion that eventually manifests as ground fissures.

A relevant geomorphological model that best explains the formation of disappearing streams, blind gullies, and ground fissures was described as pseudokarst, which produces karst-like features in soils/sediments (Kempe and Halliday, 1997). According to Haliday (2007), one of the mechanisms for pseudokarst development is piping of poorly consolidated sediments, forming piping caves, funnel-shaped sinks, and dry valleys. Piping in soils can be caused by (1) soil with vertical permeability differences that cause lateral water movement, (2) a steep slope that may cause high hydraulic gradient, and (3) the presence of dispersive soils (Wilson et al., 2017). Therefore, ground fissures in the areas surrounding Lake Ziway appear to be controlled by the presence of pumiceous layers that are highly susceptible to internal erosion and promote pseudokarst formation. Heavy rainfall usually precedes ground fissuring and acts as a triggering factor. Earthquake events can also potentially have a triggering effect. The proposed presence/pattern of subsurface groundwater conduits should be investigated using geophysical methods such as electrical resistivity and ground-penetrating radar (GPR).

Although not based on thorough investigation, areas of high ground fissure occurrence appear to be spatially linked to circular caldera basins and a nearby lake. This spatial association is true for the Ziway, Gedemsa, and Muleti areas in the central Ethiopian rift and Lake Nakuru in Kenya. This may be due to the higher hydraulic gradient within these basins and the fact that the lakes can provide large accommodation space for deposition of internally eroded sediment. The importance of pumice layers being susceptible to internal erosion is highlighted in this study. Interestingly, studies in the Muleti area of Ethiopia and Lake Nakuru of Kenya report the presence of thick deposits of pumice in areas affected by ground fissures (Ngecu and Nyambok, 2000; Ayalew, 2004). As for preferred locations/alignments of ground fissures, further research should be done too understand if there are roles played by (1) buried faults/fracture zones, which may play a passive role by serving as preferential groundwater flow conduits, promoting the development of subsurface pipes, and (2) old buried channels, which may serve as conduits of concentrated groundwater flow, promoting internal erosion. Extensional fissures in bedrock, as described in Acocella et al. (2003), may be underlying the unconsolidated sediments, providing sites of concentrated groundwater flow, and potentially explaining the ground fissures that are oriented concordant to the E-W rift extension.

The results and conclusions of this work should be limited to ground fissures forming in the rift valley underlain by low-density volcanic sediments such as pumice. It would be an over-simplification to extend the role of internal erosion to all ground fissuring observed in the MER. For infrastructure engineers, ground fissure–related studies should focus on how to predict ground fissuring and identifying engineering solutions to slow the process and protect civil structures. A model relating ground fissuring events with the presence of internally erodible sediments and rainfall amounts should be formulated. Suitable geophysical methods to map subsurface conduits should be identified to proactively design engineering solutions to stop or slow internal erosion. Possible remedies may include increasing runoff while decreasing infiltration, and designing geotechnical filter mechanisms at sites where ground fissuring is likely to affect infrastructure.

The following conclusions can be drawn:

  • (1) Ground fissures primarily affecting sediments in the Lake Ziway area do not appear to have any significant association with extensional tectonics and associated seismic activity.

  • (2) A strong argument can be made that ground fissures affecting sedimentary deposits in the Ziway area are the results of collapsing subsurface groundwater conduits. The subsurface ground conduits appear to result from internal erosion of gravelly pumice layers, which are gap graded and have ultralow density.