Floodwater harvesting to manage irrigation water and mesquite encroachment in a data‐sparse river basin: an eco‐hydrological approach

This investigation attempts to understand the eco‐hydrology of, and accordingly suggest an option to manage floodwater for agriculture in, the understudied and data‐sparse ephemeral Baraka River Basin within the hyper‐arid region of Sudan. Reference is made to the major feature of the basin, that is, the Toker Delta spate irrigation scheme. A point‐to‐pixel comparison of gridded and ground‐based data sets is performed to enhance the estimates of rainfall. Analysis of remotely sensed land use/cover data is performed. The results show a significant reduction of the grassland and barren areas explained by a significant expansion of the cropland and open shrubland (invasive mesquite trees) areas in the delta. The cotton sown area is highly dependent on the flooded area and the discharge volume in the delta. However, the area of this major crop has declined since the early 1990s in favour of cultivation of more profitable food crops. Expansion of mesquite in the delta is problematic, taking hold under increased floodwater, and can only be manged by clearance to provide crop cultivation area. There is a great potential for floodwater harvesting during the rainfall season (June to September). A total seasonal runoff volume of around 4.6 and 10.8 billion cubic metres is estimated at 90 and 50% probabilities of exceedance (reliabilities), respectively. Rather than leaving the runoff generated from rainfall events to pass to the Red Sea or be consumed by mesquite trees, a location for runoff harvesting structure in a highly suitable area is proposed. Such a structure will support any policy shifts towards planning and managing the basin water resources for use in irrigating the agricultural scheme.

favour of cultivation of more profitable food crops. Expansion of mesquite in the delta is problematic, taking hold under increased floodwater, and can only be manged by clearance to provide crop cultivation area. There is a great potential for floodwater harvesting during the rainfall season (June to September). A total seasonal runoff volume of around 4.6 and 10.8 billion cubic metres is estimated at 90 and 50% probabilities of exceedance (reliabilities), respectively. Rather than leaving the runoff generated from rainfall events to pass to the Red Sea or be consumed by mesquite trees, a location for runoff harvesting structure in a highly suitable area is proposed.
Such a structure will support any policy shifts towards planning and managing the basin water resources for use in irrigating the agricultural scheme.

K E Y W O R D S
Baraka River Basin, eco-hydrology, floodwater harvesting, land-cover classification, Mesquite, Toker Delta

| INTRODUCTION
Although some old and some relatively new options for African river basin management do exist, they are not without constraints. In arid environments of Africa, rainfall is characterized normally by a limited number of rain events but with high intensities that generate massive short-lasting runoff. Accordingly, spate irrigation, that is, flood diversion and spreading, is practiced involving capturing of storm floods mainly from catchments and ephemeral streams and diversion into croplands (Fadul, Masih, & De Fraiture, 2019;Ngigi, 2003). There is evidence that the criteria for sustainable intensification can be fulfilled by water harvesting since such a practice can, for example, improve water availability, increase agricultural yield, rehabilitate degraded lands with an influx of fertile sediments, and minimize adverse effects on the environment that are induced by the use of external chemical inputs (Dile, Karlberg, Temesgen, & Rockström, 2013;Rockström, Barron, & Fox, 2002;Rockström & Falkenmark, 2015). Ensuring sustainable water supply through evaluation of the potential of largescale (e.g., river basin) water harvesting and identification of suitable runoff zones/sites for feasible water harvesting structures is on one hand very useful, but on the other hand challenging, for the planners and water resources engineers and managers (Ghebreamlak et al., 2018;Jha, Chowdary, Kulkarni, & al, 2014). El-Sadek, El Kahloun, and Meire (2008) and Ligdi, El Kahloun, and Meire (2010) argued that determining the eco-hydrological status of a river basin permits integrated water resource management. Derivation of eco-hydrological properties, that is, vegetation, hydrologic, and geomorphic attributes, can support management actions and improve decision-making for ephemeral and intermittent streams (Levick et al., 2018). The principles of eco-hydrology are useful in developing sustainable approaches aiming at managing river floodplain systems and minimizing the flood risk (Kiedrzy nska, Kiedrzy nski, & Zalewski, 2015). Moreover, a better understanding of the ecological processes and water management are believed to create multifunctional agro-ecosystems (Gordon, Finlayson, & Falkenmark, 2010). The tight coupling between hydrology and ecology in the world's arid ecosystems is poorly understood; therefore, there is a need for strengthening the research in water-scarce ecosystems (Jackson, Jobbágy, & Nosetto, 2009). However, a major hurdle to progress in addressing eco-hydrological challenges in water-limited environments is "the paucity of data at multiple scales and poor quantification of spatial interactions among hydrological elements", such as topography, soils, and rainfall (Newman et al., 2006). It is generally accepted that the capability of traditional monitoring networks of capturing the spatio-temporally dynamic and complex eco-hydrological processes is usually low (Krause, Lewandowski, Dahm, & Tockner, 2015), especially in vast and physiographically intricate regions where extensive field measurements and observations are either poor or impossible in practice.
Reflecting on African river basins, it is clear from the above background that the prospects for ultimate success of development planning and management are constrained by the harsh environment, limitations of funding, lack of specialized labor, institutional difficulties, and technological problems (Mather, 1989). These constraints create problems of poor database and inadequacy of monitoring which make planning and management often being based on false assumptions (Barrow, 1998). Advances in monitoring techniques and data analysis linked within an interdisciplinary interpretive framework aid in realizing and addressing the myriad of issues facing the dryland eco-hydrological systems and effectively managing these systems (Wang et al., 2012). As indicated by Levick et al. (2018), under the situation of large geographic extent of a study area and infeasible collection of intensive gauge-based streamflow data, Geographic Information System (GIS)-derived variables and hydrologic modelling can be alternatively utilized.
Sudan is among the African countries where spate-irrigated agriculture is extensively practiced in river basins located within vast arid regions. There, seasonally and episodic heavy floods are exploited to irrigate adjacent agricultural lands (Fadul et al., 2019;Fadul, De Fraiture, & Masih, 2018;Fadul, Masih, De Fraiture, & Suryadi, 2020;Fujihara, Tanakamaru, Tada, Adam, & Elamin, 2020;Ghebreamlak et al., 2018). The system of irrigation in Toker Delta is the most complicated spate (flood) irrigation system in the world, as the amount of land irrigated depends on the available Baraka River flows (van Steenbergen, MacAnderson, & Haile, 2011). The scheme was established by the British colonial administration during the World War I for the primary purpose of irrigated cotton production.
Nevertheless, the basin is surprisingly understudied. Many changes have resulted in the deterioration of the irrigation infrastructure of Toker Delta Agricultural Scheme (TDAS)-an over-century old scheme. These changes have eventually decreased the efficiency of the scheme. According to Meybeck (2003), such changes are referred to as syndromes of riverine changes caused by natural conditions and human pressures, such as land-use change (agriculture), irrigation, reservoir construction, and water management types (flood control). Based on interviews with experts of the scheme during a field visit in January 2018 and a literature review (de Nooy, 2010;van Steenbergen, MacAnderson, & Haile, 2011), these syndromes, adapted to the situation in the Barak River Basin, are summarized in Table 1 (also with reference to Figure 1). The scheme development is subject to various constraints. Ensuring proper planning and development of the irrigation infrastructure and water distribution on the floodplain is hindered by a serious lack of meteorological and hydrological information. Another constraint is the lack of effective overall management, including inadequate flood embankment capacity and bunds along the course of the river, inadequate control of mesquite trees, insufficient design and implementation capacity of the TDAS, and exacerbation of drawdown of groundwater wells by increasing pump irrigation. The present work aims at making use of meteorological, hydrological, and land cover data (ground-based, satellite-based, and/or estimated) in an attempt to: 1. understand the current status of the ephemeral Baraka River Basin, with a special focus on the part located in the hyper-arid zone of Sudan to account for the complexity of existing interactions between hydrological and ecological processes.
2. suggest floodwater harvesting as a water management option for the basin in general and the agricultural scheme in Toker Delta in particular with the background understanding of the ecohydrology of the basin.

| Study area
The Baraka River Basin (Figure 2), also named Khor Baraka in Sudan, originates from the Eritrean Highlands and covers a catchment area of about 66,400 km 2 -most of which is located on the Eritrean side of the basin. It runs for 300 km in the north-west of Eritrea and through Sudan for 200 km, and flows finally to Toker Delta at Shidin before it reaches the Red Sea (de Nooy, 2010). It constitutes the most important ephemeral river in Eastern Sudan. The river channels are dry for most of the year. Baraka's hydrological pattern can be distinguished by high intermittent flood events occurring between June and October (Anderson, 2008;de Nooy, 2010). Even though there are some winter rains in the lower catchments, the Sudanese part of the basin receives most of its water from the runoff generated from the rains falling in the Eritrean part of the basin during the summer.  Table S1.
Toker Delta is located in the hyper-arid zone of Sudan (Elagib, 2002), covers an area of 170,520 ha and has its base lying at the Red Sea ( Figure 2). The delta receives large river deposits of fertile silt during the flood season of June to September. It is divided into three areas: western, eastern, and central, where the central part represents the main irrigation scheme, that is, the TDAS. Current profitable crops for farmers are sorghum, millet, and vegetables. The total irrigable area is 80,000 ha.
2.2 | Data collection and processing 2.2.1 | Rainfall estimation and evaluation Due to the limited availability of and access to ground-based meteorological and hydrological observations over the study area, this study utilized mostly open-source data sets for rainfall as will be described below and land cover data. Data on irrigated and cotton sown areas were provided by the TDAS for the period 1900-2016 during the field visit in January 2018. Ground-based meteorological data across the basin are very scarce. In the Sudanese territory of the basin, many of the rain gauges located inside or close to the basin belong to Sudan Meteorological Authority (SMA), but have been out of operation for at least two decades (See Table 2). Rainfall data from six ground rain gauges (see Figure 2) within or near the Sudanese part of the study area were acquired from SMA for different periods (Table 2). In addition, as access to data from the Eritrean site was not feasible, the decision was made to use rainfall estimates. For this purpose, Global Precipitation Climatology Center Full Data Reanalysis Version 8.0 (GPCC V8) of Deutscher Wetterdienst was used due to the temporal coverage needed for the analysis. This updated product of GPCC is a gridded precipitation data set based on data from more than 116,000 rain gauges, and is freely available for the whole world on a monthly basis for the period spanning from 1891 to 2016 . It has a spatial resolution of 0.25 and is freely available in the Network Common Data Form (NetCDF) file. Basheer and Elagib (2018) evaluated the previous version of this product (GPCC V.7) in South Sudan. Using several metrics, they ranked GPCC V.7 as the best performing product on monthly, and annual scales based on the linear fit measures, namely lowest intercept (nearest to zero), nearest slope to 1.0 and highest R 2 , and smallest errors (mean bias error (MBE), mean absolute bias error (MABE), and root mean square error (RMSE)). In addition, data on the irrigated and cotton-sown areas over 1900-2011 and 1900-2016, respectively, for Tokar Delta were obtained from the TDAS administration in Toker city.
To evaluate the rainfall product, the point-to-pixel method was used to compare the data sets from both the selected available rain gauges and the corresponding pixel of the GPCC product on a monthly basis. This method is commonly used for validating rainfall estimates within East Africa (Basheer & Elagib, 2018Gebrechorkos, Hülsmann, & Christian, 2018). Several performance indicators were used to evaluate the agreement of the GPCC product with the rain gauge data, such as the RMSE, the MABE, the MBE, the range of bias error, the slope, and the y-intercept of the linear regression of the two data sets.
T A B L E 1 Key syndromes and associated symptoms of Barak River changes in relation to marked influences of humans

Syndrome Symptoms
Tree invasion Presence and spread of invasive mesquite trees throughout the delta (Figure 1(a) and (b)) and the river channels and banks, thus blocking or diverting the water.

Sedimentation
Local dune and sand ridge formation forced by erosive winds, which gradually block the flood flow channels. Poor land management by the farmers leading to the creation of uneven topography of small sand and silt hills.

Fragmentation
Inhibition of the movement of sheet flow and irrigation of land when crop residues (sorghum stalks), bushes or other obstacles are not removed. These conditions result from the creation of uneven topography, as described above, and eventually result in uneven water distribution.

Soil erosion
Formation of small gullies and erosion channels.

Natural conditions
Unpredictable nature of Baraka River as regards its course and flood direction, magnitude, timing, and width. See the Tomasay bund in Figure 1(c) which functions as flood protection-works of Toker town and water withholder from the noncultivable western part of the delta.

RMSE =
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi 1 n where X i and X o are the estimated and the observed values of rainfall, respectively, and n is the data size.
A regression line of the GPCC data against ground rain gauge data, with a slope closer to 1 and y-intercept closer to zero indicates better performance (Allen, 1996). The other indicators (Equations (1)-(3)) are well known and are used in several studies (Allen, 1996

| Land cover
For the purpose of floodwater harvesting (FWH) analysis, a readily available long-term average land-cover released in 2010 for global coverage with a spatial resolution of 300 m from the European space agency (ESA) was used in this study (ESA, 2010). This product has been used in many studies around the world (e.g., Eberenz et al., 2016;Herold, See, Tsendbazar, & Fritz, 2016). However, to obtain long-term annual land-cover data, which were needed to is advantageous in that the data set was derived using supervised classifications of MODIS Terra and Aqua reflectance data (Friedl & Sulla-Menashe, 2015). The MCD12Q1-IGBP classification scheme has 75% overall land cover classification accuracy (Friedl et al., 2010) and has been utilized in many studies worldwide (e.g., Weber, Sadoff,  (Kanji, 1997) was used to examine the significance of trends of the time series of land-cover classification areas. Linear regression was also employed to establish eco-hydrological relationships between land-cover areas and irrigated areas or discharge volumes.

| Runoff estimation
In areas with no runoff data or inadequate historical discharge records to validate rainfall-runoff models, the best way to overcome this problem is by using a method requiring few parameters. Therefore, the Soil Conservation Service-Curve Number (SCS-CN) method (NJDEP, 2004) (NJDEP, 2004). This method is also recommended for arid and semi-arid regions and has been applied in Sudan to calculate the runoff generated from a rainstorm for rainwater harvesting applications (Mahmoud, Elagib, Gaese, & Heinrich, 2014). It can be summarized as follows: where Q s is the surface runoff (discharge) in mm, P is the rainfall in mm, I a is the initial abstraction in mm, and S is the potential maximum retention in mm after rainfall has begun. The initial abstraction concerns the early losses before runoff has begun, and S is related to the soil type and is a function of CN. Both I a and S can be calculated using the following two empirical equations (USDA, 2010): In this study, the CN was determined for each sub-watershed according to its specific soil type, land use, and vegetation cover characteristics (USDA, 1972) using GIS. It was then applied in Equation (6) to calculate S. The estimated maximum potential runoff in each watershed was calculated for different rainfall reliabilities using Equation (4).
As outlined in Section 2.1, the amount of land irrigated in Toker Delta depends on the amount of discharge that flows into the delta from Baraka River. With this background, and due to the unavailability of measured discharge data for the basin, the estimates of runoff obtained herein were validated indirectly by understanding the ecohydrology of the basin and delta, that is, the relationship between both the ground-based cotton sown area and the satellite-based shrubland area and the estimated discharge.

| Development of floodwater harvesting potential map
In this study, the Multi-Criteria Decision Rules (MCDR) in GIS, namely the Weight Linear Combination (WLC) and the Boolean techniques, were used for integrating and analyzing thematic layers. The latter was introduced by Robinove (1986) and was used in this study to obtain a constraints map using the variables "true or false". This method was used to eliminate some sites that are considered unsuitable for FWH. The FWH structures can only be located within the river channels to collect floodwater. Therefore, in this study, a buffer zone of 1 km on each side of the selected rivers was suggested as an Effective Distance (ED) for floodwater availability. All suitable areas for FWH site selection in the study area were determined to lie T A B L E 2 Performance indicators of GPCC V8 seasonal rainfall data at the ground rain gauges (upper entry) and corresponding rank † (lower entry) among the stations within this ED for each river. Areas outside the ED, including areas where the 1 km buffer zone extends beyond national borders, were considered as constraints for FWH. Thus, a value of 1 was assigned to the area located inside the ED and a 0 if it was otherwise. The WLC method involved classifying each factor map and assigning a score to each class depending on its suitability to allocate a FWH structure.
Once all the required data were obtained, organized, and formulated, ArcGIS 10.4.1 was used to analyze the data and produce suitability thematic layers for FWH site selection in the study area using different functions and decision rules. The main criteria adopted in this study for selecting suitable sites for FWH were rainfall, soil type, slope, land cover, and drainage density. Each one of these criteria was assessed individually. Thematic layers were created for each suitability component, which was ranked in classes and weighted based on their importance to the selection of the FWH sites (Table S2). The weighting system used was based on the available literature (Gavade & Patil, 2011;Kahinda, Lillie, Taigbenu, Taute, & Boroto, 2008) and common practices in Sudan (Khidir; personal communication, 2017). Then, the weights were normalized using the pairwise comparison matrix (Table S3).
All the generated thematic layers were overlaid based on their normalized weights from the pairwise comparison (Table S3) using the WLC method, and with the aid of Equation (7), a final FWH suitability map was created with four classes of potential water harvesting suitability within the ED in the Sudanese part of the study area. These classes included low suitability, moderate suitability, high suitability, and very high suitability for water harvesting.
where GS i is the grid cell suitability value, W i is the criteria normalized weight, and Y i is the criteria grid cell.

| Rainfall frequency analysis and reliable discharges
Using the SCS-CN method and the runoff parameters (Table S4), subwatershed discharge at different probabilities of exceedance (reliabilities) were estimated. Because Baraka River flows only for a short period during the rainy season (June to September), the analysis of runoff was based on seasonal rainfall during this part of the year using the GPCC data over the period extending from 1976 to 2016. To this end, the rainfall probability curves were plotted by attempting different transformation functions to eliminate the skewness and retain a (near) normal distribution of the data (Mansell, 2003;Shaw, 1994).
The runoff in mm generated from each sub-watershed was estimated for reliability values of 50, 67, 80, and 90% and transformed into discharge volumes in MCM. FAO (1991) recommended the use of design rainfall of 67% reliability for agricultural purposes to account for rainfall variability and, accordingly, design a runoff harvesting system that can meet the irrigation demand.

| Rainfall estimates and variation across the basin
Based on point-to-pixel scatter plots of the June-September monthly rainfall for the nearest six rain gauges to the basin in the Sudanese territory, all the comparative performance indicators, together with the corresponding rank, are shown in Table 2. The variations in rainfall between the GPCC and the rain gauges are mostly explainable by the results for the rain gauges located north, or in the northeast, of the basin. This performance is demonstrated by coefficients of determination (R 2 ) of ≥.91. At the rain gauges located far from the Red Sea, that is, the two located southwest of the basin in the arid zone of Sudan (Elagib, 2002), the GPCC explains 76% and 85% of the variations in the rain gauge data. All the error metrics, that is, MBE, RMSE, and MABE, together with the y-intercept (closest to zero) indicate the best performance of GPCC at the rain gauges located in Toker Delta. An opposite performance is exhibited by the two rain gauges located in the arid area (Kassala and New Halfa). The overall score (smallest) demonstrates the best performance of GPCC at the rain gauges situated inside the Delta. It can be inferred from the results presented in Table 2 that the GPCC data could be used reasonably to estimate the seasonal (June to September) rainfall in the inner part of the basin, especially in view of the current state of unavailability of rainfall data.
The intra-annual rainfall behavior over the different watersheds of the Baraka River Basin is shown in Figure 3. It is noticeable in general that rainfall reduces from the watersheds located in the Eritrean Highlands towards those situated in the Sudanese territory, reaching its smallest amount in Watershed 1 at the border of Toker Delta.
Most of the rainfall contributing to the runoff of Baraka River occurs within the season extending from June to September with the peak reached in August. The total rainfall received during this period has a median of 68.1 mm (Watershed 3) to 259.5 mm (Watershed 5). This rainfall contributes about 56% (Watershed 2) to 79% (Watershed 5) of the annual rainfall. Rainfall over the watershed bordering Toker Delta occurs during November to January (median is only~42 mm); thus, this amount does not contribute to the runoff of Baraka River.
Only~11 mm of rain fall over this watershed from June to September of the year thus representing only 16% of the total annual rainfall.
Rainfall is also highly variable during the year. The seasonal (June to September) rainfall has a coefficient of variation of 37% for Watershed 5 to 66% for Watershed 3.

| Runoff estimates and variation for different sub-watersheds
The estimated values of runoff parameters used in calculating the discharge on each sub-watershed are shown in Table S4. The results reveal appreciable amount of seasonal (June to September) discharge, varying between 4.6 and 10.8 billion cubic metres (BCM) with 90% and 50% reliabilities, respectively, as estimated at the outlet of the basin ( Table 3)

| Eco-hydrological process in Toker Delta
The temporal and spatial variations of the land-cover areas detected across Toker Delta during the period 2001-2018 are shown in To understand the ecohydrological processes in Toker Delta, Figure 5 shows the interactions between these land covers and the hydrology of the delta. High inter-annual variability in the water volume, irrigated area, and cotton area can be seen in Figure 5 A scatter plot of cotton area versus direct-flood irrigated area reveals a highly significant correlation using the full records ( Figure 5 (c)). An even higher correlation was found with Kendall tau value of 0.810 (p < 0.0001) for the data up to 1992. In addition, the cotton area declined thereafter in a non-matching pattern with that of the irrigated area ( Figure 5(b)). Thus, the correlation became insignificant for the last two decades of overlapping data (1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012). The shrubland decreases with increasing sown cotton area ( Figure 5(d)) since shrubs are removed to make the area available for crop cultivation.
The planted cotton area increases as water discharge volume increases ( Figure 5(e)), a relationship confirmed by the dependence of the cotton area on the irrigated area as explained above.
Shrublands seem to decrease non-linearly with increasing water discharge volume until a certain discharge value of~13,450 MCM is reached ( Figure 5

| Baraka River Basin water management in Sudan
Based on the thematic layers for the development of the site suitability map for seasonal floodwater harvesting ( Figure S1), the final FWH site suitability map in the Sudanese part of the basin was developed as shown in Figure 6. The site suitability for FWH is very high, high, and moderate in 13%, 41%, and 35% of the area within the ED, respectively, while the areas with low potential represent only around 11%. Therefore, the study areas can be said to have a great potential  [This technique will allow the water to pen-etrate…] the cultivable areas rather than being lost to the Red Sea".
Based on the above results and recommendation, we propose a site for FWH as shown in Figure 6 to be located at the entrance of Toker Delta, 14 km upstream of the Iron Bridge near Shidin rock. This site is highly suitable for FWH and is close to the outlet where the discharge volume reaches the largest magnitude. It also considers the possibility of utilizing the proposed storage structure for regulating the river flows into the irrigation scheme. In addition, it can also act as water storage for use in irrigation during the dry spells rather than being left to pass to the Red Sea or lost for use by mesquite trees. This way of exploiting the available water resource will help increase the agricultural productivity of the scheme. Building dams for water harvesting in India, for instance, proved effective in playing a crucial role in farming by overcoming water scarcity during the summer irrigation season (Balooni, Kalro, & Kamalamma, 2008). The proposed structure herein will also contribute to the flood control of the river, which is characterized by high discharge during a short runoff period.

| DISCUSSION
The efforts intended for managing and developing the African river basins are inherent with difficulties associated with observing, measuring and recording unpredictable hydrometeorological characteristics, especially in remote and harsh regions (Mather, 1989).

Problems of poor databases and inadequacy of monitoring beset the
African river basins and hinder the understanding of the structure, functions and exploitation opportunities of the basins (Molle, Wester, & Hirsch, 2010). However, the present study has presented an approach to overcome this troublesome scarcity of data. The approch integrates observed and estimated rainfall data, satellitebased land cover data, a flexible rainfall-runoff method and a field visit to collect useful data and information to understand the bidirectional land use-water interactions. Therefore, an option for managing the vast, dynamic, and data-scarce river basin under study has been suggested. As emphasized by Krause et al. (2015), real-time monitoring and research enhance our knowledge about the complexity of spatially and temporally dynamic eco-hydrological systems. It is argued that the umbrella of eco-hydrology provides a means to embrace diverse topics at the interface between hydrology, ecosystem science and the human activities found in different settings (Bonell, 2002;Nuttle, 2002). In catchment eco-hydrology research, the spatial distribution of vegetation is often presumed as a signature of the status of water availability in a catchment (Thompson, Harman, Troch, Brooks, & Sivapalan, 2011). The eco-hydrology principle "enhances the resilience of the river basin against climate and anthropogenic change", flood safety and ecosystem services for society (Kiedrzy nska et al., 2015).
Therefore, the eco-hydrological approach proposed in this study can be framed in Figure 7 for replication in similar river basins. For instance, unavailability of data is a typical constraint to the understanding of the actual conditions of similar spate irrigation schemes in Sudan, such as the Gash Delta in the east of the country (Fujihara et al., 2020).
Management of river basin water and treating water as an economic good are considered as two of three policy prescriptions for sustainable water management (Wester & Warner, 2002). The present study area relies heavily on highly variable rainfall amounts and a short rainy season that is fundamental to the generation of runoff, which is in turn a key source of water for the land uses and vegetation growth in the basin, particularly in Toker Delta. Although mesquites are highly resistant to harsh environments and are a source of fuel and animal food, their eco-hydrological impacts are formidable. Shrubland encroachment at the expense of other land covers can cause hydrological alterations and imbalance with remarkable consequences on the ecosystem structure and function (Wang et al., 2012). A study carried out by Yasuda et al. (2014) in an arid environment in Khartoum North (Al Kadaru) showed that, due to roots that extend deep into the aquifers, the type of invasive mesquite in Sudan alters subsurface water dynamics. As indicated earlier, the growth of mesquite in the irrigable part of Toker Delta is one of the major challenges. Hence, we expect that a similar negative effect is taking place in the study area, although no data are available to confirm this effect. The field visit Delta by the local authorties during the last two decades. Clearance of mesquite was employed manually; however, the large-scale eradication plan was halted due to the high cost of the cleaning process.
As concerning agricultural management in African, production is always beset by low soil fertility (Tittonell & Giller, 2013). A recent study found a positive correlation between the crop yield and frequency and quantity of fertilization in an underperforming irrigated scheme within the Nile River basin in Sudan, thus emphasizing the importance of soil properties in determining crop productivity (Khalifa, Elagib, Ahmed, Ribbe, & Schneider, 2020). Drawing on the TDAS, it is recommended not to pursue any flood irrigation plans in the inappropriate western delta due to its characteristic sand and silt dunes (de Nooy, 2010).
Reforms for future river basin management are thought to be grounded on a couple of approaches. These approaches are identified by Molle et al. (2010) as the water development approach for the human benefit through water infrastructure and the approach that promotes restoring and maintaining ecosystems. However, Molle et al. (2010) argue that river basin management in developing countries, where poverty is widespread, will need a strong development dimension. This viewpoint is clearly reflected in the present study through the need for a water harvesting structure, the purpose of which will be to control the flood and at the same time to make better use of the scarce, seasonal water received in the basin.

| CONCLUSIONS
This study has attempted to account for complex factors that will be required to improve our understanding of, and ability to predict, the eco-hydrological processes in dryland agro-ecosystems, such as the ephemeral Baraka River Basin in the hyper-arid zone of Sudan.
The present research has succeeded in integrating ground-based, remotely-sensed and hydrological data and the assessment of floodwater harvesting potential to achieve this goal. Due to the unavailability of up-to-date gauge rainfall records, a global rainfall product was validated with past rain gauge data within and around the basin to give an overview of the rainfall pattern. This rainfall product was then employed to estimate the discharge volume over the basin sub-watersheds using the SCS-CN method, also due to unavailable rainfall-runoff data. The use of estimates of rainfall and discharge volume, together with the classification of remotely-sensed land-cover areas and ground-based data on irrigated and major crop sown areas, has been successful in establishing and improving the understanding of the underlying eco-hydrological processes in Toker