We would like to thank Dr. Clement Flaux and Dr. Doriane Delanghe, for help at the sedimentology laboratory, CEREGE, Aix-Marseille University, France. The authors thank ANR GEOMAR and PHC IMHOTEP Campus France for co-financing the research.
1In the Western Desert of Egypt, on the margins of the hyper-arid Sahara, the mean annual precipitation is around 0.5 mm/year (GOUDIE, 2013). However, the role of water flow in drylands is known to be an important landscape-forming process (WAINWRIGHT and BRACKEN, 2011) and episodic runoff can have significant geomorphic effects (THOMAS, 2011). Ephemeral surface runoff dominates deserts. It occurs in the form of flash-flood events. For example, a continued eight hours of rainfall caused intensive flash floods in Wadi Urana, Makkah (Saudi Arabia), the estimated peak discharge rate was 223 m/s and the total flow volume was approximately 7.8 million m3 (El-BASTAWESY et al., 2013). In Upper Egypt (January 28, 2013), a torrential rainfall storm struck Wadi Qena, the estimated total run off volume was 66 million m3 (MOAWAD et al., 2016).
2Climate, topography, vegetation and geology are variables that control surface runoff in deserts. Indeed, surface runoff is attributed to weather anomalies; spotty rainy storms show both temporal and spatial variability (GOUDIE, 2013). The extent of periods of aridity leads to distinctively patchy vegetation characteristics which often provide extensive bare surface areas that enhance runoff production (WAINWRIGHT and BRACKEN, 2011). BANNARI et al., (2017) have demonstrated that storm rainfall and morphology (elevation, slope, curvature, and water catchment) are the major contributing factors in driving flash‑flood erosion.
3Geomorphologically, flash floods acts through water erosion and inundation. The geomorphologic features produced by flood runoff have been observed in arid and semi-arid regions. These landforms include, for example, rill-gullies (ZHAO et al., 2017), boulder berms (KEHEW et al., 2010) and ephemeral lake-pans (BRYANT and RAINEY, 2002).
4Gully erosion processes often occur in narrow channels and, over short periods, remove the soil to considerable depth (POESEN et al., 2002). Permanent versus ephemeral gullies are defined in terms of their dimensions and survival. Ephemeral gullies can easily be altered by repeated runoff (POESEN et al., 2002). Gullying processes represent major source of sediment production in semi-arid and arid regions (MARZOLFF and POESEN, 2009) and typically form flood fans at local base level.
5In deserts, saline ephemeral lakes occupy local basins. Their hydrology is related to water flow-accumulation of the endogenic drainage system, groundwater sources, evaporation rates and human impacts. Ephemeral lakes are characterized by prominent changes of surface water area. In hot deserts, the lack of water resources coupled with high evaporation rates encourage the development of playas after ephemeral lakes. Changes in playa water balance can occur at various time scales, from short (day–month) to long (decade–century) timescales (SHAW and BRYANT, 2011).
6The current study highlights geomorphological imprints of episodic runoff in the context of the 4th November 2015 flash flood event that hit Wadi An-Natrun depression, Western Desert, Egypt.
7Wadi An-Natrun is one of the five largest mega-depressions dominating the northern part of the Western Desert of Egypt. It is located relatively close to the Nile Delta. It is a totally land-locked depression, and has an elongated shape that trends in a SE-NW direction. The surface area of Wadi An-Natrun depression is 1090 km2, including the depression floor and sloping escarpments. Around a third of this area (the depression floor) lies below sea level. The lowest point in the depression floor attains 23 meters below sea level (Figure 1).
8Some isolated hills are found on the depression floor, such as Qaret Paromeos (+28 meters relative to the surrounding depression floor). These hills exhibit conical to flat peaks, and have very steep upslope sides of up to 90°. The Wadi An-Natrun depression is drained by thirteen drainage basins concentrated on the west-northwest border. The mean areas of these basins equal 45.3 km2. The largest basin is Wadi Raqabt El-Haytah which covers 205 km2 and has 169 meters of elevation range (Figure 1).
Figure 1 ‑ Study area
9Ephemeral lakes (ponds) are characteristic geomorphic landforms of An-Natrun depression. These lakes are located on the depression floor, namely Al-Fasdah, Umm-Rishah, Ruzita, Al-Hamrah, Al-Bidah, and Jaar (Figure 1). Groundwater represents the main source of An-Natrun saline lakes. Some lakes receive wastewater from irrigation and sewage. The water levels of An-Natrun’s lakes change seasonally. For example, shallow saline lakes shrink in volume by > 60% in summer, which produces mud flats and micro features, such as salt crusts, microbial mats and cracks (TAHER, 1999).
10The sedimentary rocks in Wadi An-Natrun depression are of Pliocene to Pleistocene age. Wadi An-Natrun Upper-Pliocene formation dominates the depression floor. This is composed of two members (ABOU-KHADRAH, 1973; GAD, 2012), which are from the bottom of El Solymanya member (light green sandstone alternating with shelly limestone, appears on the NE corner of Wadi An-Natrun) overlain by El-Muluk member (dark gray to green clays, with sand beds, limestone and gypsum layers). The Pleistocene deposits are of fluvial, lacustrine and aeolian origin. The latter form mega sand dunes (vegetated dunes up to 15 meters in height) are found west of the Jaar Lake.
11The Pliocene groundwater aquifer is the only source of water in Wadi An-Natrun depression (MASOUD and ATWIA, 2011). Water logging naturally occurs in the depression. It is also attributed to infiltration of irrigation water because the flood irrigation method is frequently used. The vegetation cover in the depression comprises vascular plants, such as trees-shrubs, perennial herbs and annual herbs. They are categorized into four habitats including croplands, orchards, wastelands and lakes (ABD El‑GHANI et al., 2015).
12The study area lies in an arid climate zone. Based on data from the Egyptian metrological authority records (An-Natrun ground station 1990-2014), the average monthly temperatures range between 12.6°C in January and 29.8°C in August. The area is characterized by irregular rainfall. The mean annual rainfall is 42 millimeters (1975-2005), with nine rainy months during the year (the summer months are totally dry). In Wadi An‑Natrun, for example the long-term rainfall data (1975-2005) show that the maximum amount of rainfall in one day was 79 mm (November 17, 1980). During the same period (1975-2005), only eleven days had rainfall of more than 10 mm per day. The weather satellite observation showed that by the early morning on hour (3: 00 -3: 59 UTC) on November 4, 2015 intensive rainfall (25 to 30 mm/h) hit the study area. The rainfall continued for seventeen hours and caused flash flooding. This led to causalities and caused significant damage to Wadi An-Natrun and its surroundings (SALEEM, 2017).
13The current study applies numerous geographical methods including field activities and laboratory analyses. Four field studies were carried out from December 2015 to April 2016. The field studies allowed us: (1) to check the satellite image observations, (2) to measure the flood landforms using traditional methods (such as cross sections, and measuring blocks dimensions), (3) to apply the close-range photogrammetry method, and (4) to collect sedimentary samples.
14Traditionally, photogrammetry is related to conventional aerial photography. Close range photogrammetry (CRP) refers to photos taken with an object-to-camera distance of less than 300 meters (MATTHEWS, 2008). The CRP tool has many advantages. It is a fast surveying tool and it is inexpensive in comparison to other tools such as terrestrial laser scanner and LIDAR. It allows small morphologies to be measured with centimeter-scale dimensions. Furthermore, it offers an instantaneous snapshot of temporal features. This tool has been used in geomorphological investigations, especially for quantifying gully erosion (CASTILLO et al., 2015; STOCKER et al., 2015).
15The use of CRP for topographic surveying passes through successive steps. After determining the object, a Ground Control Points (GCPs) should be positioned, their coordinates (X, Y and Z) are known in a reference coordinate system using a field surveying tool. GCPs are calibration marks that are distributed across the target area and appear on images. The basic concept for image acquisition is to capture the area of interest from different angles. The number of required images depends on the total area of interest.
16The current study utilizes terrestrial CRP to record micro- to meso-scale landforms such as gulley networks and mud polygons. The professional Agisoft photoscan® software (http://www.agisoft.com/) is used for digital image processing. This software offers the complete processing workflow including photo alignments, defining GCPs, building point clouds and generating an image-based 3D model, georeferenced ortho-mosaic and Digital Elevation Model (DEM). These products are used for further processing within the software package ArcGIS-ArcScene®, including digitizing geomorphological features on vector layers and extracting measurements.
17Five short cores were used to sample subsurface sediments. The depth of cores varies between 30 cm and 90 cm. We lightly hammered PVC tubes into the ground. In the lab we split these tubes horizontally using a mechanical saw. This allowed us to describe the facies including color, grain size and texture. Conventional wet and dry sieving methods were applied for 22 samples. Also, the Beckman laser particle size analyzer was used. This machine precisely measures the granulometry of the finer grain sizes, from 0.017 μm – 2000 μm. The sedimentary analyses were performed in order to probe the erosional-depositional processes.
18We used seven satellite images captured by the MSI sensor onboard Sentinel-2 European satellite (Table l). This dataset covers the two-year period between summer 2015 and summer 2017 (Table 2).
Table 1 ‑ The Sentinel-2 MSI spectral bands
S2-MSI Bands
|
Central Wavelength (nm)
|
Spatial Resolution (m)
|
Coastal-Aerosol ()B1)
|
443
|
60
|
Blue-visible )B2(*
|
490
|
10
|
Green-visible (B3)*
|
560
|
10
|
Red-visible (B4)*
|
665
|
10
|
Red-Edge (B5)
|
705
|
20
|
Red-Edge (B6)
|
740
|
20
|
Red-Edge (B7)
|
783
|
20
|
NIR (B8)*
|
842
|
10
|
NIR (B8a)
|
865
|
20
|
Water vapor (B9)
|
945
|
60
|
Cirrus (B10)
|
1375
|
60
|
SWIR (B11) *
|
1610
|
20
|
SWIR (B12) *
|
2190
|
20
|
* Bands used in the current study
Source: [https://earth.esa.int/web/sentinel/user-guides/sentinel-2-msi/resolutions/spatial].
Table 2 ‑ The Sentinel 2 Satellite images used in the current study
Image (ID)
|
Notes
|
T36RTU_2015816
|
Two months before flood event ( summer 2015)
|
T36RTU_20151114
|
Ten days after the flood event (Autumn 2015)
|
T35RQP_20151114
|
Ten days after the flood event (Autumn 2015)
|
T36RTU_201642
|
Five months after the flood event (winter 2016)
|
T35RQP_2016042
|
Five months after the flood event (winter 2016)
|
T36RTU_2016820
|
Ten months after the flood event (summer 2016)
|
T36RTU_201717
|
Fourteen months after the flood event (winter 2017)
|
T36RTU_2017825
|
Twenty two months after the flood event (summer 2017)
|
19The comparison of two images (T36RTU 2015816 and T36RTU 20151114) shows the flash-flood abrupt impacts. The other images are utilized to monitor long-term geomorphological changes of the ephemeral lakes and salt pans. In the current study we have applied the satellite image-based spectral indices, as follows:
|
(Green - Near Infrared) / (Green + Near Infrared)
|
In MSI (Sentinel-2):
|
(band3 – band8) / (band3 + band8)
|
The Soil Salinity and Sodicity Indices (SSSI-1 and SSSI-2) have been used to map slight to moderate salinity on agriculture plots (BANNARI et al., 2016). This method uses two SWIR bands (specifically band numbers 9 and 10 of the Advanced Land Imager, EO-1). The current study implements the MSI’s equivalent SWIR bands according to the following equation:
|
(SWIR1 × SWIR2 – SWIR2 × SWIR2) / SWIR1
|
(BANNARI et al., 2016)
|
In MSI (Sentinel‑2):
|
(band11 × band12 – band12 × band12) / band11
|
|
20Over two years, the Sentinel-2 datasets, combined with field observations show the following changes in the ephemeral lakes and playas on the depression floor. According to the topographic maps (scale 1/50.000) six ephemeral lakes occupy local topographic basins. All the lakes are found below the contour line -10 meters (below sea level). The Jaar Lake is found in the most north-western part of the depression floor at about -20 meters below sea level, and it represents the main sink of Wadi Raqabt El-Haytah (Figure 1).
21On August 16th 2015 (two months before the flash flood), only Al-Hamrah and Al-Bidah basins contained water bodies. The four other basins (El-Fasdah, Umm-Rishah, Ruzita and Jaar) were mostly dry and they exhibited mud playas (pans) covered by white and reddish salt crusts (Figure 2a)
Figure 2 ‑ Flood impacts on Jaar and Al-Bidah Lakes

22On November 14th 2015 (ten days after the flash flood), we observed an abrupt increase in the surface area of water bodies. For example, the waterbody of Al-Bidah Lake doubled in area. El-Fasdah, Umm-Rishah, Ruzita and Jaar also witnessed inundation (Figure 3). The water accumulation around Jaar-Lake partially drowned the sand dunes (Fig 2c).
Figure 3 ‑ Water bodies changes in two years (August 2015 – August 2017)
23By April 2nd 2016 we noted a slight increase in the surface area of the main lakes, expect Al-Bidah Lake (Figure 3). This could be attributed to discharge waste irrigation water during the winter. By August 20th 2016, these lakes lost wide areas of their water bodies (Figure 3). In ten months (November 2015-August 2016), the water body of Jaar Lake shrank by about 60% and It yielded salt pans on the lake margins, specifically in the inter-dune areas west of Jaar Lake (Figure 2e). By the following summer 2017 (around two years after the flash flood) the water body of Jaar Lake gradually decreased, consequently the Jaar playa developed. This playa has a high salinity relative to the water body (Figure 2f).
24Furthermore, a new 438 ephemeral lakes (ponds) were formed within the depression floor. The largest ephemeral lake is located just northwest of the Paromeos monastery (Figure 1), covering 0.35 km2. By April 2nd 2016, the Paromeos ponds were transformed into playas. The Paromeos received floodwater from some gullies which were recently incised on Qaret Paromeos.
25The current study is based on field analyses of the landforms recently formed by the November 4th 2015 flash flood. The study carried out at four sample sites (Figure 1) as follows:
26At Qaret Paromeos, many quarrying basins (pits) were left abandoned. These basins contain geomorphological evidence for recent runoff. The current case study focuses on the site named “QP inner quarry” (30° 22' 59.1" N 30° 16' 4.7" E). This site consists of an artificial quarry basin (covering 1.590 m2) and which is overlooked by a flat-summit step sided mound. The local elevation range at the study site (between the summit and the basin floor) equals 16 meters (Figure 4a).
Figure 4 ‑ Field profiles for Qaret Paromeos site
(a) Cross profile shows gully, fan and playa system, (b) Cross profile shows playa on the quarry basin
27At the QP inner quarry site, we traced flash flood landforms in detail. Water erosion has promoted gullying on the mound’s sloping side. This gully network comprises elongated rills which cut into two opposing slopes and connect to the trunk gully. The length of the trunk gully is 35 meters. The dimensions (depth-width) of the rill-trunk gully system vary according to the lithology and slope gradient. For example, the head-water rills (i.e. upstream first order tributaries) formed in a shale formation, are 20 centimeters maximum in depth. The path of the trunk gully flows down the sandstone formation. The former deepens up to 80 centimeters, forming a small waterfall, where it cuts a lithological contact from sandstone to shale (Figure 4a).
28The trunk gully shows a sharp sided wedge shape (large socket) where it ends and breaks the basin slope (Figures 4a and 5). It attains its maximum width (3.3 meters) and maximum depth (2.1 meters). This morphology is attributed to water erosion, coupled with mass sediment failure on the basin step slope. At the same elevation, some parallel incisions formed (Figure 5).
Fig 5 ‑ Aerial view (Qaret Paromeos site)
29A typical floodout fan was found where the main gully entered the basin floor and suddenly lost its flow energy and deposited most of its load. This fan covers 130 m2 and it has a maximum length of up to 12.7 meters, with an approximate volume of 30 m3 (Figure 5). On the fan, we observed the fan surface gully. This gully was probably formed by a water gullying pulse concurrent with the deposition of the flood fan. Figure 6 shows detailed morphologies of the fan surface gullying such as the main gully, three tributaries and irregular shaped mud-polygons.
Figure 6 ‑ Fan surface gully (Qaret Paromeos site)
(a) 3D orthophoto mosaic, (b) DEM, (c) Vector (micro features)
30The surface flood water accumulated on the basin floor forming an ephemeral pond for around two months after the flood event. This is attested by a mud playa covering 885 m2 (occupies about 90% of the basin area). We have traced the outer boundary of the ephemeral pond by identifying some micro-features such as mud polygons and salt evaporates as well as identifying the limit of fresh natural vegetation. The former ephemeral pond was a maximum 150 centimeters deep (Figure 4b).
31This site was occupied by ephemeral lakes located northwest of the Paromeos monastery. The field survey allowed to define the transition of the ephemeral lakes (January 2016) to playas showing large mud cracks (April 2016). Salt-tolerant vegetation densely covered the playas’ surface, specifically “Tamarix niolitica”.
32The AL site is located on the northeast escarpment of An-Natrun depression (30° 29' 24.2" N 30° 07' 22.1" E). It is located one kilometer from the An-Natrun/El-Alamein highway. The site (Figure 7a) consists of three sections such as, (1) the upper-slope, moderately sloping to the depression floor, (2) the steep slope up to 35° and (3) the lower-slope section moderately sloping to the depression floor. Consolidated sand formations dominate and are thicker than 15 meters.
Figure 7 ‑ Field profiles for El-Alamein site
(a) Cross profile shows slope sections, (b) Cross profile (playa)
33Inside the orchards (uppermost slope, Figure 7a), we observed shallow rills and sheet-flow marks. Down from the upper slope section we observed severe gullying. The latter includes parallel upslope gullies of significant dimensions (up to 1.7 meters deep and 4 meters wide). On the steep slope section, the large gullies progressed to hollows. On the inter-gully areas, we observed deep and narrow linear incisions on the initial surface (Figure 8). The gullies produced depositional flood fans at the foot of the slope. The largest fan is spatially correlated with the largest gully-hollow, while smaller gully incisions yielded small fans (Figure 8). The fans margins show salt crusts and retain subsurface moisture (March 12, 2016).
Figure 8 ‑ Steep slope section (El-Alamein site)
34On the lower-slope section we observed a mud playa (Figure 7b). It covers an area of approximately 1700 m2 and it received water and sediment from the 280 m long main gully. At the center of the playa, mud polygons and surface cracks are observed (Figure 9).
Figure 9 ‑ CRP documentation for the playa micro features (El‑Alamein site)
(a) Orthophoto mosaic, (b) DEM, (c) Vector
35These mainly formed by the dehydration and contraction forces acting on the expansive clay regolith, with the important role of water salinity (ZHAO et al., 2014). The fraction calculation of cracks segments/intersections (S/I) defines the structure network morphology. It is 3 for a triangle, 2 for a square and 1.5 for a hexagon (VELDE, 1999). In the current case, the S/I value equals 1.77 (Figure 9c). This is consistent with a square shape, which could be related to the early development stage of the polygons.
36This site is located on the midstream area of Wadi Raqabt El-Haytah drainage basin (30° 26' 39.1" N 30° 02' 55.8" E, Figure 1). It comprises three wadis (tributaries) which drain into the local basin. This basin was found by quarrying activities. At the site WRH, the water gullied down to the bottom of the wadis. The running water triggered rocky blocks (Table 3), some blocks were transported a few meters inside the wadis (tributaries). The largest block (weight 19.5 tons) was detached and fell into the basin. Also, we observed a small flood pan on the bottom of the basin (Figure 10).
Table 3 ‑ The dimensions of blocks moved by surface runoff (WRH site)
Boulders_id
|
a axis (m)
|
b axis (m)
|
c axis (m)
|
Volume (m3)
|
Mass (tons)
|
Notes
|
WRH-1
|
0.86
|
0.7
|
0.42
|
0.25284
|
0.556248
|
Moved on the bottom of wadi
|
WRH-2
|
0.11
|
0.1
|
0.07
|
0.00077
|
0.001694
|
Moved on the bottom of wadi
|
WRH-3
|
0.15
|
0.13
|
0.004
|
0.000078
|
0.0001716
|
Moved on the bottom of wadi
|
WRH-4
|
0.12
|
0.09
|
0.06
|
0.000648
|
0.0014256
|
Moved on the bottom of wadi
|
WRH-5
|
0.14
|
0.11
|
0.07
|
0.001078
|
0.0023716
|
Moved on the bottom of wadi
|
WRH-6
|
0.09
|
0.08
|
0.06
|
0.000432
|
0.0009504
|
Moved on the bottom of wadi
|
WRH-7
|
0.09
|
0.08
|
0.05
|
0.00036
|
0.000792
|
Moved on the bottom of wadi
|
WRH-8
|
0.12
|
0.11
|
0.07
|
0.000924
|
0.0020328
|
Moved on the bottom of wadi
|
WRH-9
|
0.13
|
0.11
|
0.1
|
0.00143
|
0.003146
|
Moved on the bottom of wadi
|
WRH-10
|
0.21
|
0.18
|
0.11
|
0.004158
|
0.0091476
|
Moved on the bottom of wadi
|
WRH-11
|
0.44
|
0.39
|
0.1
|
0.01716
|
0.037752
|
Moved on the bottom of wadi
|
WRH-12
|
0.39
|
0.26
|
0.22
|
0.022308
|
0.0490776
|
Moved on the bottom of wadi
|
WRH-13
|
0.75
|
0.71
|
0.43
|
0.228975
|
0.503745
|
Moved on the bottom of wadi
|
WRH-14
|
0.46
|
0.39
|
0.15
|
0.02691
|
0.059202
|
Moved on the bottom of wadi
|
WRH-15
|
1.45
|
1.1
|
0.08
|
0.1276
|
0.28072
|
Overturned along wadi side
|
WRH-16
|
3.4
|
2.9
|
0.9
|
8.874
|
19.5228
|
Collapsed to the basin
|
Figure 10 ‑ Field profile for Wadi Raqabt El-Haytah site
37The litho-stratigraphy of the pan-playa cores (in QP, PM, WRH and AL sites) and the flood fan core (in QP site) are presented below. For each core, the subunits are coded and ordered alphabetically (Figure 11). Three main sedimentary facies have been identified.
Figure 11 ‑ Lithostratigraphy for the studied sites (Cores: QP, PM, WRH and AL)
38The surface of the QP fan (sample QP-fan/a, Figure 11a) consists of very poorly sorted sediments (3.14 phi) which typically characterize the surface tip of the flooding fan. The layer (QP-fan/a) is sand dominated but gravels compose ca. 25% of the bulk. Sands are mostly composed of fine and medium grain sizes of around 300 microns. Absence of coarse sand (> 500 microns) is surprising given that the gravel component is important. The sand fraction is similar in size to aeolian sands (see below) which could indicate that the latter was the main sediment source, reworked and re-deposited at the gully fan.
39The surfaces of the four studied playas represent the fresh flash-flood sediments (Figure 11). Their thicknesses varied between 3 centimeters (WRH-playa/a) and 9 centimeters (PM-playa/a). Thicker sediment tracts at the Paromeos Monastery site (sample PM-playa/a) are clearly linked to larger watersheds by comparison with the other sites. In general, the playa’s surface flood sediments are composed of silt and clay (> 95% of total bulk; AL-playa/a, QP-playa/a, PM-playa/a), indicating dominant decantation process. Mean grain size is around 4‑10 microns and with a total distribution spanning 0.2 to 30 microns. By contrast, (WRH-playa/a) has a fine sand component comprising up to 38% of the total; the mean grain size is ca. 50 microns and the sediment is poorly sorted, distributed from 0.2 to 300 microns. These variations are probably related to the dominant grain size in the gullying source area (e.g. sand stone dominant in WRH site), as well as the distance between the sediment source area and the playa. Indeed, WRH core site is located a few meters from the sediment source while the PM-playa received fine sediments transported by some long gullies descending from Qaret Paromeos. Longer sediment transport promotes a better sorting of grains between the source (gully) and the sinks (playa). Finally, the mud facies are generally dark in color, related to relatively high organic content. The latter likely derived from vegetation that developed in the flooded areas.
40We recovered sand-dominant units in all cores. Sand layer thicknesses vary between 4 to 20 cm and are mostly composed of fine to medium sands. Mean grain sizes vary from ca. 150 to 600 microns. Samples (AL-playa/b, QP-playa/b and WRH-playa/c) have symmetric peak grain size distributions (Figure 12). These well-sorted sands suggest aeolian deposition. Poor sorting in others sand facies may translate the sampling of different sand layers deposited under various wind conditions.
Fig 12 Grains size distribution for sub-surface playa sands
(a) AL-playa/b, (b) QP-playa/b, (c) WRH-playa/c
41With the exception of the site WRH, all retrieved sedimentary sequence show that subsurface playa muds have been deposited earlier to playa sand deposition at each site. By comparison with playa muds deposited during the November 4th 2015 flash flood, each subsurface mud layer should correspond to a former flash flood. Thus, playas may constitute promising sedimentary archives for palaeo-flood reconstructions in Wadi An‑Natrun.
42The occurrence of some landforms in Egyptian deserts is related to episodic surface runoff (e.g. flash floods). Occasionally, these have been used to estimate the magnitude of flash-flood events (KEHEW et al., 2010; GABR and EL‑BASTAWESY, 2015). However, their morpho-dynamics, ages and evolution are poorly understood and require further investigations. The current study has focused on the flood features formed by a well-known flash-flood event.
43The flash-flood landforms have been formed by multiple processes. The key process is gully erosion. Once a gully is formed, several processes lead to channel expansion, such as crack development, piping, plunge pool and splash erosion, fluting and mass failure (VANMAERCKE et al., 2016). The typical colluvium process (KIRKBY, 2004) is observed where gullying acts on slopes and produces sediments deposited in fans and playas.
44The interrelated factors, discussed below, control the episodic surface runoff in drylands and its geomorphic action. These are conceptualized in terms of erosivity (i.e. the erosive ability of the running surface water) and erodability (i.e. the resistance of surface topsoil to erosion). Basically, the surface runoff is initiated when the precipitation arriving at the surface exceeds surface infiltration rates (WAINWRIGHT and BRACKEN, 2011). This mechanism is dependent on rainfall behavior as well the surface characteristics (including vegetation cover, slope gradients and permeability).
45The rainfall intensity favors desert flash floods. In the southern Mediterranean, intensive rainfall in a few minutes or hours can yield high energies with the potential for a significant amount of erosion (WAINWRIGHT and BRACKEN, 2011). The current case study has demonstrated that rainfall was concentrated for seventeen hours, with peak rain over the study area up to 30 mm/hour (SALEEM, 2017). Rainfall thresholds initiating runoff and gully erosion are variable (POESEN et al., 2003). Less is known about threshold values in hot deserts. However for soil loss in Belgium, 15 mm has been defined as the rainfall threshold for gully development (KNAPEN and POESEN, 2010).
46The patchy vegetation cover in arid lands provides extensive bare surfaces that enhance runoff production (WAINWRIGHT and BRACKEN, 2011). Also, sparse vegetation encourages the formation of soil crusts and thus runoff production and gullying (VALENTIN et al., 2005). In the current study, we have observed the vegetation cover of the gully head cuts. At both the sites of QP and WRH, the gully areas are almost free of vegetation. The AL site shows sparse annual vegetation (mainly, Senecio glaucus and Cotula cinerea) which flourished in response to a recent rainfall event.
47Topography controls the incision of rills and gullies. At the four studied sites, the observed gullies are found on steep slopes that increase runoff velocity and thus water erosivity. Also, topography determines the inundation positions. This is evident from the studied gully ends, and the topographical change from steep to gentle slopes. At a larger scale, the topographic setting of Wadi An‑Natrun depression is conducive to the inundation of the depression floor.
48The geological characteristics impacted the initiation and development of the studied landforms. For example, at the Al site, large steep slope hollows have been formed by surface water erosion and mass failure processes which acted on the internal structure of the sandstones (i.e. cross bedding clearly appears on the gully-hollow sides). This is proven by the phenomenon of collapsed fracture-bounded gravels. At the QP site, the variations in lithology (i.e. alternations of sand and silt-clay deposits on the steep slope) yield gullies of different patterns and dimensions. Also, the occurrence of stones embedded on the surface acts to reduce infiltration (WAINWRIGHT and BRACKEN, 2011; POESEN 1992), and favor runoff. This could explain the significant erosion dominating the AL site (steep slope section). At two sites (QP and AL), we observed that ephemeral lakes and playas are found due to the presence of low permeability muddy subsurface units (e.g.‑ AL‑playa/c, QP‑playa/c and QP‑playa/d). These units probably slowed down the infiltration processes and favored trapping of the flash-flood surface water.
49Land-use patterns play an important role in mediating gully erosion. For example, tillage and grazing reduce topsoil coherence, raising erodability and promoting runoff production. At the AL site (most upper-slope section) the rill incisions are spatially correlated with plowed fields and wheel tracks. At the QP site, we observed a strong relationship between quarrying activities and gully erosion.
50The 4th November 2015 episodic flash flood has produced characteristic landforms of both erosional and depositional origins. These landforms are of variable size and patterns and they have been mainly reshaped by gullying and mass failure.
51At Wadi An-Natrun, lake size doubled, dried lakes were rapidly inundated and new ephemeral ponds appeared on the depression floor. All ponds were dry by the next summer and lake margins were transformed into playas. The depression received floodwater from gullies cut into adjacent hillsides. For example, at Qaret Paromeos (QP) we report a gully 35 m long, 0 to 3.3 m large and 0 to 2.1 m deep, cut during the flood event through hydraulic erosion and mass sediment failures. Sediments were deposited where the gully slope breaks to a flat plain, in the form of a depositional fan. In this case study, sediments were sorted along the profile from heterogeneous materials made of gravels and sands at the gully mouth down to decantation muds deposited within the ephemeral pond at the edge of the gully fan. The pond then evolves into a playa with mud cracks and salt-tolerant vegetation (ca. six months after the flood event). Other studies have shown how the geomorphological impacts of the same flash-flood event vary in terms of gully size and sediment transport, depending on local soil erodibility. Our study participates in better understanding the geomorphological impacts of flash flooding in desert areas. The sedimentary record of flood events is key to understanding flood histories.