The MRB is located at the northeastern margin of the Tanzania Craton on the plateau between the eastern and western branch of the EARS, east of Lake Victoria (Fig. 2a). It gently slopes from the Mau escarpment (2940 m a.s.l.) toward Lake Victoria (1130 m a.s.l.) (Sinclair et al., 2008b) and extends from longitude 33∘47′ to 35∘47′ E and from latitude 0∘38′ to 1∘52′ S covering an area of approximately 14 000 km2. The two perennial tributaries of the Mara River, Amala and Nyangores, drain the forested headwaters and join to form the perennial Mara mainstem at ca. 1695 m a.s.l. (McClain et al., 2014). From there the mainstem flows through the Masai Mara National Reserve in Kenya and the Serengeti National Park in Tanzania into Lake Victoria at Musoma Bay (Fig. 2a).

The MRB is characterised by a unique heterogeneous geology being positioned on the northeastern margin of the Tanzania Craton at the intersection of Precambrian orogenic belts (Fig. 2b). The oldest rocks in the basin are the Archean granitic gneisses and amphibolites of the Tanzania Craton which are exposed in the centre of the basin (Grey et al., 1969). The Precambrian metasediments and meta-volcanics of the Nyanzian–Kavirondian Orogenic Belt (a.k.a. Lake Victoria goldfields) dominate the geology in the west of the basin (Henckel et al., 2016). Neogene to recent lavas and tuffs, including basalt, trachyte and phonolite, unconformably rest on the Precambrian rocks in the north and centre of the basin (Shackleton, 1946). Parts of the remaining flat-lying trachytic phonolite lava sheets extend above and below the Isuria and Utimbara escarpments. Quaternary sediment and soils dominate the southwest and northeast of the basin.

The Tanzania Craton has been tectonically stable from the Cambrian (530 Ma) until the onset of Neogene extensional tectonics, associated with the formation of the EARS (Barth, 1990). Doming followed by rifting and subsequent rift shoulder uplift, associated with the eastern branch of the EARS, combined with hydrological and erosion–deposition processes, have given rise to a diverse array of landforms across the region. On a smaller scale, the asymmetric central and southern basins of the MRB are controlled by the Plio-Pleistocene escarpments of the NNE–SSW-trending Isuria fault (a.k.a. (Soit) Oloololo and Siria; hereafter Isuria) and the roughly E–W-trending Utimbara fault (a.k.a. Utimbaru; hereafter Utimbara) (Shackleton, 1946). The Isuria escarpment extends from Gibaso in Tanzania in the southwest over a length of ∼ 90 km to Chebunyo in Kenya in the northeast (Fig. 2a). The escarpment of the southeast-dipping fault varies in height from 100 m in the northeast to ca. 400 m in the southwest. The Utimbara fault has a surface expression measuring over ∼ 100 km from the mouth of the Mara River into Lake Victoria at Musoma Bay in the west to the Tagari Hills in the east. The fault downthrows to the south with the uplifted blocks showing slight tilting to the north, and its escarpment height varies from 150 m in the east up to 400 m in the centre. The largest displacement of ca. 400 m along both faults is constrained by a discontinuous cap of a Neogene phonolite lava at the top of the escarpment, identical to the lava outcropping at the base (Saggerson, 1966). In addition to the main escarpments, the area is cut by several smaller E–W-trending faults exposed in Precambrian bedrock without showing signs of recent reactivation (Henckel et al., 2016).

The MRB is located in a region of highly complex seasonal rainfall patterns and exhibits a bimodal rainfall with two rainy seasons linked to annual oscillations of the inter-tropical convergence zone in March–May (long rains) and October–December (short rains) (McClain et al., 2014; Mwanake et al., 2019). Mean annual values of precipitation vary from 600 mm yr-1 in the lowlands to nearly 1500 mm yr-1 in the highlands (Fig. 2a) (Defersha et al., 2012). Significant differences in rainfall also occur between catchments on the escarpment and below the escarpment due to topographic effects and the regional influence of Lake Victoria (Camberlin et al., 2009). The mean daily temperature also changes with elevation from 11 ∘C on the Mau escarpment to 24 ∘C down by Lake Victoria (Fick and Hijmans, 2017).

Fine-scale variation in soil properties in the MRB gives rise to “catenas”, soil sequences that vary from ridge top to hill slope to foot slope to valley. On ridge tops, soils are shallow and sandy, often exposing bedrock. Soils become progressively deeper and heavier further downslope, reaching maximum depths and levels of organic matter on valley floors (Jager, 1982). The granitic rocks and quartzites in the area are associated with grey to brown and sometimes reddish sandy soils (Cambisols). Residual soils on meta-volcanics and metasediments are finer-grained, redder and often lateritic (Mulgrew, 1966). Ferruginous laterite and associated loams are generally found above and adjacent to the banded ironstone but can also develop on the Neogene volcanics. Black clay soils (mbuga) mainly develop on basaltic lavas, amphibolite and chlorite. Grey clastic alluvial deposits are mainly found in the Mara valley and surrounding riverbeds (Thomas and Kennedy, 1977). The soils of volcanic, fluvial and colluvial origin are generally higher in Ca, Mg and plant-available P than those covering the rest of the system (Olff and Hopcraft, 2008).

The basin vegetation has undergone severe changes during the last century: the region was previously covered by montane forest in its headwater regions and a mixture of grasslands and shrublands in its middle and lower section. This pristine vegetation is today limited to the nature conservation areas of the Serengeti–Mara ecosystem (e.g. the Masai Mara National Reserve in Kenya and the Serengeti National Park in Tanzania) (Sinclair et al., 2008b). The main land cover in the basin can today be classified as agriculture (cropland and range lands), herbaceous vegetation and shrubland (savannah), residual montane forest, and wetland (Kalensky, 1998). Cultivated land accounted for approximately 35 % of the basin in 2000, composed mainly of small-scale farms and tea plantations. Due to a continuing agricultural expansion and urbanisation, today large-scale farms and irrigated crops are increasing in the area (McClain et al., 2014). Along the Mara River and its tributaries, gallery forest with closed canopy prevails (Sinclair et al., 2008b).

We used the digital elevation model Copernicus DEM GLO-30 with 30 m horizontal resolution as input for the regional morphometric analysis of the drainage basin. For this application, this DEM is superior to the higher-resolution TanDEM-X DEM because it is free of water-generated noise. Using QGIS (QGIS.org, 2021) with the SAGA plug-in (Conrad et al., 2015), the DEM was pre-processed to fill the sinks, and a flow accumulation raster was developed. The drainage network was calculated with a pixel threshold of 500 cells. To ensure the generated stream network is representative of the natural stream, drainages have been visually confirmed using high-resolution ESRI World Imagery (ESRI et al., 2017).

Sentinel-2 multispectral imagery (Drusch et al., 2012) was used to extract land cover information. The Earth observation mission developed by the European Space Agency (ESA) acquires 13 bands in the VNIR (visible and near-infrared) to SWIR (short wavelength infrared) range, with 4 bands at 10 m, 6 bands at 20 m and 3 atmospheric correction bands at 60 m spatial resolution (Drusch et al., 2012).

For the spectral indices, 69 Sentinel-2A scenes (granules T36MXD, T36MYD) with a cloud coverage lower than 5 % acquired between 14 October 2016 and 11 January 2022 were selected. As Level-2A processed products have only existed since December 2018, the previous data needed atmospheric, terrain and cirrus correction of the top of the atmosphere using the Sen2Cor processor (Louis et al., 2016). A common method to process remote-sensing data is producing band ratios. By taking the ratio of two bands or several bands the method highlights features of interest in the original grey-scale image (Rowan et al., 1976; Inzana et al., 2003). We used a combination of three spectral indices – the normalised difference vegetation index (NDVI), the clay mineral ratio (CMR) and the moisture stress index (MSI) – as a proxy for vegetation stability and soil quality, which allows conclusions about ecosystem functions. The NDVI was calculated from the 10 m resolution Sentinel-2 imagery to assess the spatial and temporal changes in primary production in the study area. The NDVI is a measure of the amount of vegetation at the surface and is related to the health of the vegetation as healthy (photosynthetically active) vegetation reflects a high amount of energy as compared to the unhealthy and sparse vegetation (Grebby et al., 2014).

The NDVI is calculated from the amount of red and near-infrared light reflected from the Earth's surface and gives values between 1 and -1. Higher values of the NDVI indicate rich and healthier vegetation, while the lower values indicate poor and sparse vegetation (Pettorelli et al., 2005). For Sentinel multispectral data, the NDVI is calculated by (B08 - B04)/(B08 + B04). The average NDVI was calculated for the selected time span. The NDVI was used in this study as it is a reliable and well-studied ecological indicator. Other vegetation indices (generalised difference vegetation index, GDVI; soil-adjusted vegetation index, SAVI) were calculated but did not show significant differences compared to the NDVI. In order to examine how the NDVI varies along and perpendicular to strike, 3 longitudinal and 17 transverse profiles of both faults were extracted.

Additionally, a moisture stress index (MSI) was calculated from the 20 m resolution Sentinel imagery to determine vegetation water stress and draw conclusions about plant-available water. It was calculated as the ratio between the near-infrared (NIR) and shortwave infrared (SWIR1) values (B11/B8). The values of this index range from 0 to over 3, whereas green vegetation commonly shows values between 0.4 to 2 (Hunt and Rock, 1989).

The clay mineral ratio (CMR) is a geological index, which helps identify geological features containing clay and alunite as hydrous minerals absorb radiation in the shortwave infrared region (2.17–2.63 µm). It is calculated by the ratio between the 20 m resolution SWIR1 (B11) and SWIR2 (B12) bands (Alasta, 2011; Nath et al., 2019). As active fault zones are potential regions for primary clay mineral formation within the fault zone as well as for the accumulation of clay minerals through mineral weathering and downslope transport (Barton et al., 1995; Solum et al., 2005), this index can help identify fault activity. However, this band ratio can also indicate carbonate mineralisation and recently burnt areas, which demonstrate a high reflectance in the SWIR2 (Alasta, 2011).

The morphology of the IUFZ is characteristic of a region of active normal faulting (Fig. 3), with an uplifted plateau on the footwall separated by a steep escarpment from a wide plain on the downthrown hanging wall. The plain is crossed by several smaller fault scarps mostly subparallel to the main escarpment. The E–W-trending escarpment of the Utimbara fault can be clearly traced on the satellite imagery over a length of ca. 65 km (Fig. 3a). A strongly eroded escarpment in the west towards Lake Victoria builds the western tip of the Utimbara fault adding to its total length of ca. 86 km. Where the Utimbara fault forms a step-over to the Isuria fault, a complex array of minor normal faults connects the two escarpments. The NE–SW-trending escarpment of the Isuria fault is clearly visible in the TanDEM-X dataset over a length of ca. 86 km. If the eroded northern offshoot is included, the fault has a total length of ca. 93 km.

The striking Neogene escarpments of the Isuria and Utimbara faults are associated with surface displacements between 100 and 400 m measured from the TanDEM-X digital elevation model. The amount of vertical displacement varies over the length of the Utimbara escarpment. For our analysis, we focused on measurements along clear geomorphic expressions of the escarpment. From west to east, the height of the escarpment builds up from ∼ 100 m, increasing rapidly to 300 to 400 m in the centre. After a decrease down to ca. 100 m, the height increases again up to ca. 230 m. The eastern offshoot of the fault slowly decreases in height from west to east. In general, the displacement curve shows a typical half-moon shape with the highest displacement in the fault centre.

From the southern tip towards the centre, the height of the Isuria escarpment increases constantly from ca. 50 m up to a maximum of ca. 390 m; towards the north the height decreases steadily and the escarpment ceases. The northern offshoots of the Isuria fault show the smallest amount of vertical displacement. The total displacement could likely be higher than the measured values because, especially along the Utimbara fault, displacement markers are rare due to erosion of the Neogene volcanics and the thickness of the volcanic rocks is poorly known.

The Isuria and Utimbara faults both show a continuous linear front (Fig. 3a), with minor interruptions because of the presence of normal faults oblique to the ridge. We split the escarpments into nine segments, here named with a first letter representing the main tectonic structure they belong to. Six different segments can be identified along the Utimbara escarpment: a strongly eroded discontinuous western front (U1), a less eroded front striking E–W (U2), a straight and continuous front striking E–W (U3), a slightly curved front striking E–W (U4), a continuous ESE–WNW-striking front (U5), and, after a step-over, a slightly more eroded eastern front (U6). Three different segments can be identified along the Isuria front: a continuous southern front with a NE–SW main strike (I1), a continuous central front with a NNE–SSW strike (I2) and a less continuous northern front that shows N–S-trending sections (I3). Smf index values were calculated for these nine mountain fronts. The highest value for the Utimbara front is 1.89, and the lowest value is 1.25. Values for front sinuosity for the Isuria fault range from 1.06 to 1.2.

Instrumental seismicity and geodetic data are sparse but generally suggest extensional to transtensional fault kinematics (Saria et al., 2014; Glerum et al., 2020; Stamps et al., 2021).

Long-term fault displacement rates are bracketed by the exposed trachytic phonolite lavas that are vertically displaced along the IUFZ by 390 (±10) m. As the original geologic reports of the area predate the common use of radiometric age dating, Shackleton (1946) estimated these lavas to be of Pliocene (3.5 to 4.5 Ma) age, which still needs to be verified by radiometric dating. Field observations by Shackleton (1946) suggest an interval of tectonic quiescence after the eruption of the Neogene lavas, since quartz gravels occur on the lava surface, close to the edge of the Utimbara escarpment, southwest of Tarime. It therefore seems that the eruption of the Isuria lavas occurred some time before the onset of faulting activity, but the exact timing is yet to be determined. Assuming a maximum age of the onset of faulting of 3.5 to 3 Ma, the long-term displacement rate for both faults is approximately 0.1 mm yr-1, which falls within the typical range of fault slip rates in continental rift systems (e.g. Zielke and Strecker (2009) for EARS, Gold et al. (2017) for Lower Rhine Graben or Friedrich et al. (2003) for Basin and Range). The morphology of the Isuria escarpment, with Smf index values lower than 1.4 (Fig. 3a), indicates that today, too, tectonic activity exceeds erosion, especially in its central part (Anand and Pradhan, 2019; Pérez-Peña et al., 2010; Soria-Jáuregui et al., 2019). The Utimbara escarpment is similarly characterised by low Smf index values from 1.25 to 1.38 apart from two segments in the western portion showing values higher than 1.4. These segments coincide with the Mara Wetland, where alluvial deposits with high sedimentation rates occur.

Despite the subtropical climate and inferred low fault displacement rates of 0.1 mm yr-1, geomorphic indicators of the fault activity are preserved, likely because of the low erodibility of the Neogene volcanics. Tectono-geomorphological mapping along the IUFZ revealed several signs of recent surface rupturing activity. The most compelling geomorphological evidence for active faulting is the clearly traceable fault scarp ranging in height from approx. 2 to 35 m, which can be found in Pleisto-Holocene footwall sediments for several kilometres parallel to both escarpments of the IUFZ (Fig. 3b, c). The fault scarp segments are parallel to the escarpments and are frequently crossed by drainage channels, which show both seemingly sinistral and dextral horizontal offsets of 10 to 20 m when crossing the scarp (Fig. 3e).

We identified a series of small, fault-bound alluvial fans from the DEM exclusively on the Tanzanian side of the IUFZ (Fig. 3b, d). Nine inconspicuous alluvial fans could be identified along the Utimbara escarpment, directly underneath the escarpment. Along the Isuria escarpment 16 alluvial fans were found along the escarpment. The fans range in area from less than 0.1 to almost 0.5 km2 and in thickness from 1 to 9 m as estimated from the DEM. They have a conical shape and are located close to the escarpment. The deposits along the Utimbara fault are generally thicker and larger than the deposits along the Isuria fault. At all fans sediment thickness appears to be highest in proximity to the fault scarp as is expected for fault-controlled alluvial fan deposits (Jackson and White, 1989). Where the rate of uplift is greater than the rate of drainage channel erosion, providing a continuous supply of fresh debris, alluvial fans are deposited adjacent to the mountain front and continued tectonic uplift results in the accumulation of thick alluvial fan deposits (Bull, 1977). The detected series of sub-circular and not interlocked alluvial fans south of the IUFZ (Fig. 3b, e) provide geomorphic evidence for active extensional faulting predating alluvial fan deposition along the IUFZ. Given their limited extent and subtle geomorphic expression, tectonic activity likely proceeded in the Late Pleistocene or Early Holocene. Deposition ages of the deposits would enable a precise calculation of recent displacement rates.

Another indicator of ongoing tectonic activity is a number of wetlands and swamps throughout the basin, mainly concentrated in the river's floodplain (e.g. the Mara Wetland). They form because of subsidence related to movement on the IUFZ and subsequent channel migration. Since the mid-Pleistocene, tectonic activity significantly disrupted and reorganised the drainage network in the area leading to a periodic rejuvenation (Peters et al., 2008; Sinclair et al., 2008a). Approximately 2 km south of the North Mara Gold Mine, where the Mara River enters the Mara Wetland, the river channel visibly migrated towards the Utimbara escarpment in the north (Fig. 4). It shows an asymmetric mosaic of meander loops and oxbow lakes indicating down-tilt combing (Holbrook and Schumm, 1999). The further migration of the channel in the tilt direction, though, is inhibited by the massive granitic gneiss around the North Mara Gold Mine.

Thermal springs often occur along tectonically active faults and are closely associated with the intensity and scale of modern activities of fault zones (Li et al., 2021; Delvaux et al., 2017). The association of the Isuria fault with the Majimoto hot spring (Fig. 3a) on the hanging wall therefore provides further evidence of current activity along the fault.

The NDVI in the study area varies from -0.22 to 0.84 with maximum values in the rainy season (March and April) (Fig. 5). Spatially, the NDVI also varies with low values in (barren) agricultural fields and grasslands and high values in forested areas and wetlands (Fig. 6a). The largest differences between rainy and dry season can be observed within the boundaries of the Serengeti National Park with dry-season values varying between 0.2 and 0.4 and rainy-season values ranging from 0.6 to 0.8 (Fig. 5).

The average clay mineral ratio (CMR) in the study area varies from below 1 in water-covered areas to maximum values of over 2.5 along the escarpments, in wetlands and forested areas (Figs. 6b, S1 in the Supplement). Recently burnt areas show anomalously low values in dry-season scenes (displayed in red). The CMR also reveals increased values along mapped fault scarps, as well as in areas dominated by volcanic rocks.

The average moisture stress index (MSI) values in the study area vary from 0.13 in the wetlands up to maximum values of 5.69 in agricultural fields (Figs. 6c, S2). Most pixels cluster around 0.5 to 1.5. Higher values indicate greater water stress and less water content (displayed in red). The average moisture stress index revealed the highest moisture stress to be in the centre of the northern extension of the Serengeti National Park around the middle reaches of the Mara River. Similarly high values are found on agricultural fields on the Utimbara footwall.

Despite the complex precipitation patterns influencing vegetation distribution, the geological–tectonic signal can be extracted from spectral indices by time-series analysis. We evaluated a spectral signal time series to separate the seasonal meteorological from the long-term geological signal (Figs. 5, S1, S2). All indices highlight several locations with permanently favourable values for plant growth along the escarpments (Fig. 6a).

These include tectonic structures like the fault escarpments, the Mara Wetland south of the Utimbara fault and the marsh around the Majimoto hot spring south of the Isuria fault. Several smaller tectonic wetlands can be found along the Isuria fault exposing higher NDVI values than the surrounding area. On a smaller scale, tectonically controlled alluvial fans on the hanging wall appear on each of the indices (Fig. 6d, e, f) showing favourable conditions for plants to grow. Also, some tectonically quiescent, forested regions show favourable conditions for plant growth, such as the Ntagare Hills, the Mara gallery forest and the Trans Mara Conservation Area.

Our analysis of spectral vegetation indices in the area suggests that active faulting of the IUFZ improves conditions for vegetation along the escarpment. As most of the area today is influenced by human changes, separating the tectonic and geological signal from anthropogenic influences is a difficult task. Some of these patterns today are largely overprinted by anthropogenic activities. However, through the systematic analysis of the spectral signals from a geological perspective, natural factors like lithological signals and increased erosion or deposition rates could be identified and delimited from anthropogenic signals. Strong differences between protected areas and agricultural regions are obvious in all three evaluated spectral indices, especially in the rainy season. This indicates a negative anthropogenic influence on vegetation growth, which tectonic and geological processes cannot buffer. Several studies, for example, examined the anthropogenic impact on soil erosion in the MRB and mainly attributed the increase to land use changes. The conversion of natural forest to farmlands and pasturelands along the course of the Mara River provokes soil erosion and leads to a reduction in soil nutrients (Mati et al., 2008; Dwasi, 2002). Nonetheless, tectonic activity could also play a role in increased soil erosion.

The MRB is controlled by a delicate balance of tectonic subsidence related to the IUFZ and sediment input by the Mara River stabilising the climatically sensitive ecosystem. According to climate change studies (Kendon et al., 2019; Shongwe et al., 2011), the basin will experience an increase in annual river volume and rainfall amounts with wetter rainy seasons and drier dry seasons. This variability will result in higher peak flows in the wet period further increasing soil erosion and lower flows in the drier months leading to severe droughts (Osoro et al., 2018). Climate change in the area will likely lead to accelerated habitat desiccation and deterioration of vegetation quality for wildlife and livestock with the potential to disrupt the large-scale ungulate migration.

Today, there are several economic sectors that are strongly linked to a healthy MRB. The MRB supports tourism, agriculture and mining, which are three of the most profitable economic activities in Tanzania and Kenya, collectively contributing between 10 %–15 % to both countries' gross domestic products (Nelson et al., 2012). The Mara Wetland south of the Utimbara fault is the most valuable ecosystem service provider with a calculated total economic value of approximately USD 5 million a year with crop agriculture, water for commercial use, livestock and fisheries being the major contributors. The seasonally flooded areas around the permanent wetland provide good grass for extensive livestock grazing and fertile soils for agriculture (WWF-ESARPO, 2010). At least 73 % of households around the Mara Wetland in Tanzania harvest fish for both subsistence and commercial purposes. Furthermore, the wetland provides important water cleansing services in the form of the uptake of pollutants by species of phragmites and papyrus which could otherwise accumulate up the food chain. The plants can trap heavy metals in their roots and likely prevent heavy-metal pollution reaching Lake Victoria (Matagi et al., 1998; Mati et al., 2008).

This balance is not only threatened by climate change, but also by anthropogenic impacts, such as an increase in sediment input, as analysed by Dutton et al. (2019). As the soil erosion on the footwall of the IUFZ is already increased due to tectonic uplift, an increase in the erosion rate on the footwall and deposition rate on the hanging wall could have serious consequences. Anthropogenic land use and cover changes progressively reduce the potential for wildlife to spread to future climatically suitable, tectonically stabilised habitats like the Mara Wetland (Ogutu et al., 2008).