Estimation of change in terrestrial water storage for Abbay River Basin, Ethiopia

For water balance estimation, in situ measurement of each element of water balance gives a complete understanding of the water in the basin. But, in reality, direct measurement of all components of the water balance was not done at all. In situ data are inadequate and incomplete to indicate the water balance of the water bodies. In recent decades, different researchers have investigated changes in water resources for different regions using satellite, ancillary, and hydrological models for water resource optimization and mentoring. Satellite images were used to analyze the distribution of surface water in the river basin as static and as binary variables (Haas et al. 2009; Kuenzer et al. 2013, 2015; Feng et al. 2014; Ogilvie et al. 2015). Due to a very Advanced High Resolution Radiometer, Moderate resolution Imaging Spectro-radiometer and Visible Infrared Imaging Radiometry Suite, optimal sensors with coarse spatial resolution and daily orbit were detecting water changes at high temporal resolution. From MODIS data, water surface change was estimated across lakes in China from 2000 to 2010 using 8-day MODIS data (Sun et al. 2014). The monthly extent of inland water bodies from 1986 to 2012 in central Asia was estimated using MODIS and Advanced High Resolution Radiometer data were estimated (Klein et al. 2014). Rodell & Famiglietti (2001) used the Gravity Recovery Climate Experiment to assess terrestrial water in Illinois. Their results show soil moisture and groundwater changes caused the largest water storage changes. Tiwari et al. (2011) estimated water changes in southern India using gravity recovery and climate experiment (GRACE). They showed hydrological changes having strong impacts on changes measured by GRACE. Moiwo et al. (2011) used GRACE and an empirical model to evaluate water storage changes in the Himayas and Tibetan Plateau, and their results detected storage variation with small random errors. Nie et al. (2016) evaluated terrestrial water storage from GRACE to show the global drought index (TSDI), which was correlated with other drought indicators, and their result indicates that the total storage deficit index (TSDI) can be used to monitor drought globally. Chinnasamy & Agoramoorthy (2015) estimated changes in groundwater using the Global Land Data Assimilation System and GRACE dataset in Tamil Nadu. Their result is mainly used for agriculture in Tamil Nadu (India) and confirmed a remarkable depletion rate that is higher than the recharge rate. Syed et al. (2008) indicated that the simulations of terrestrial water storage (TWS) from Global Land Assimilation System (GLDAS) are consistent with corresponding (TWS) from GRACE. The time series assessment of terrestrial water storage and its relationship with hydro-metrological factors in the Gilgit-Baltistan region using GRACE observation and the GLDAS–NOAH model was conducted (Dostdar et al. 2021). They used GRACE to forecast the possible trends of increasing and decreasing TWS with high accuracy as compared to past studies, which did not use satellite gravity data and their result shows a decreasing trend for GRACE (Centers for Space Research (CSR), GeoForschungsZentrum (GFZ), and Jet Propulsion Laboratory (JPL)) and the GLDAS–NOAH model. Ozturk (2022) evaluated changes in water storage in Southeastern Anatolia, Turkey, using GRACE and GLDAS. Based on the estimated data, their results demonstrate decreasing water storage in all sub-areas. Shao & Lui (2023) analyzed the groundwater storage change and influencing factors in China based on GRACE data and their results showed that groundwater storage in southern China increased at a rate of 4.79 mm/a and decreased in other regions from 2003 to 2016. All these studies confirm that datasets used from GLDAS and GRACE can be important in estimating water resource variations and hazards like droughts and floods.

Few studies indicate the seasonal change of water resources in Ethiopia, mainly focusing on the characterization and forecasting of seasonal precipitation and meteorological drought analysis. Due to changing topography, terrestrial water storage estimation is a challenging task because of the complexity of water balance components. To date, there is no explicit information on the GRACE and GLDAS databases for the estimation of terrestrial water storage that exists for the Abbay River Basin, although these data products are very important for local, regional water resource management. Considering the lack of high-precision, reliable water balance components, hydrological data, insufficient knowledge of terrestrial water balance and limited technologies, our research concentrated mainly on estimating terrestrial water storage based on GRACE, TerraClimate, GLDAS dataset and, for the first time, produces a continuous yearly, monthly and seasonal terrestrial water storage change and liquid-equivalent water thickness for the Abbay River Basin using machine-learning methods. So, the general objective of this study was to estimate changes in terrestrial water storage in the Abbay River Basin using Open Access Satellite Databases and Hydrological Model. Unlike the other studies described here, our main objective is to estimate changes in terrestrial water storage in the Abbay River Basin and to identify the factors that may play a major role in the variation of water storage in the study area. For this purpose, GLDASE with other hydrological and geospatial variables were integrated to estimate TWS, liquid-equivalent water thickness from the GRACE dataset was evaluated, and spatiotemporal change of TWS was evaluated and finally, the trend variability and slopes of TWS and LWE for the Abbay River Basin using the Mann–Kendall trend test and Theil–Sen's was estimated for water resource management and planning in order to reduce over-utilization of water resources and to reduce water shortage in the study area.

The study was located in northeastern Ethiopia at 7°40′N and 12°51′N latitude and 34°25′E and 39°49′E longitude with an area of approximately 176,200 km² and an elevation difference from 483 to 4,266 m AMSL. Ethiopia has 12 river basins and this study was focused on the Abbay River Basin. The Abbay River Basin was one of the tributaries of the Nile River, the world's longest river (Cherinet et al. 2019). The first world civilization in the art of irrigation and cultivation of crops was around the Nile River (Woodward et al. 2007). The Abbay River is a transboundary river that covers the drainage basin of 11 countries, including the Democratic Republic of the Congo, Tanzania, Burundi, Rwanda, Kenya, Ethiopia, Eritrea, South Sudan and Egypt, and flows to the Mediterranean Sea (Oloo 2007). The Abbay River is an essential river for Ethiopia, and the Grand Renaissance Dam of Ethiopia was constructed on it. The river starts in the high mountainous part of Ethiopia and serves as a contributor to the Nile River. It is located in an area where water is a critical resource for domestic use and irrigation agriculture. The topographic feature at the upstream section of the river basin was mountainous, whereas the downstream section was relatively flat to gently undulating. There are varying climatic zones in the river basin due to environmental conditions. The temperature ranges from 28 to 38 °C at the upstream section and ranges from 15 to 20 °C at the downstream section of the river basin. Generally, rainfall in the study area ranges between 787 and 2,200 mm per year and the lowest rainfall recorded was less than 100 mm per year. The wet season in Ethiopia begins in April and ends in September. The dry season was from November up to February (Tarkegn & Jury 2020). The location map of the study area is presented in Figure 1.

To estimate terrestrial water storage in the Abbay River Basin, different datasets were used, such as surface soil moisture (SSM), profile soil moisture (PSM), root zone soil moisture (RSM), surface runoff (SRO), canopy water storage (CWS), temperature (T), evapotranspiration (ET) from GLDAS processed at NOAH (0.25° × 0.25°) and resample to 1° × 1° for further hydrological data analysis. Precipitation data from TerraClimate are at 0.5° resolution and resample to 1° × 1°. The data were collected from January 2010 to December 2020, with monthly temporal resolution, and are presented in Table 1.

Spatiotemporal change in water storage is not feasible by field measurements in most regions due to the high cost/limited observations. The gridded GRACE satellite data (RL-05, level-3) are processed and provided by three institutions, CSR (the University of Texas), JPL (Pasaden, CA, USA), and GFZ (Potsdam, Germany), respectively (Wahr et al. 1998; Landerer & Swenson 2012). Thus, satellites are one of the main tools that can be used for data collection in almost all regions. So, GRACE is the widely used satellite that can evaluate global water storage changes which was launched in March 2002 (Ramillien et al. 2004; Wahr et al. 2004; Schmidt et al. 2006). The measurements of GRACE can produce spatiotemporal change in the Earth's gravity field, which shows the water mass change over land (Wahr et al. 1998; Tapley et al. 2004b). GRACE data that were available include monthly anomalies from 2004-04-01 to 2017-01-07 that are computed relative to a time-mean baseline (2004–2009) (Swenson & Wahr 2006; Landerer & Swenson 2012; Swenson 2012). These products have undergone pre-processing, such as a de-striping filter, glacier isostatic adjustment, and Gaussian smoothing (Chen et al. 2019). To improve accuracy, the GRACE terrestrial water storage averaging method was used and linear interpolation was used to fill in all missing monthly data (Landerer & Swenson 2012; Long et al. 2015; Seyoum & Milewski 2017). The original signal that was lost during data processing was restored by a multiplicative scaling factor that was introduced to the monthly mass-gridded data provided by the GRACE (Seyoum et al. 2019). Scaling factors for GRACE terrestrial water storage are further explained by Landerer & Swenson (2012).

For this study, the climatological dataset was collected from the TerraClimatic dataset and this dataset was developed from three global gridded climate datasets found at high spatiotemporal resolution and available to the public through an unrestricted data repository of Idaho's Northwest Knowledge Network University (Harris et al. 2017). Temporal information is inherited from CRU Ts4.0 for most global land surfaces for temperature, precipitation, and vapor pressure. TerraClimate also produces a monthly surface water balance using a water balance model that includes reference ET, precipitation, temperature, and interpolated plant extractable soil water capacity. The dataset was validated monthly on data obtained from the Global Historical Climatology Network database (Menne et al. 2012). The dataset was available from 1958/01/01 to 2021/12/01 at 0.5° grid. The mean yearly precipitation and temperature from 2003 to 2022 were collected for this study and prepared for further data analysis.

The Mann–Kendall test is a nonparametric statistical test used for the analysis of trends in climatologic and hydrologic time series (Sharma et al. 2019). The Mann–Kendall test is a rank based method used to analyse the trends in time series. This test has been found to be reliable even for nonnormal time series. Further, the Mann–Kendall test statistics are not significantly affected by the presence of outliers. In this test, the null hypothesis H0 assumes that the realizations are independent and that no trend exists in the data series which is tested against the alternative hypothesis H1, which assumes that the monotonic trend exists in the time series. Assuming X_i and X_j are two subsets of data series where i = 1,2, 3,…,n − 1 and j = i + 1, i + 2, i + 3,…,n.

If is greater than Z_{100_α}, where α represents the chosen significance level at (5% significance level or 95% confidence level with Z_95%) then the null hypothesis is invalid, implying that the trend is significant. Positive values of Z statistics indicate an increasing trend, while negative values of Z statistics represent the negative trend (Timbadiya et al. 2013).

If N is an even number and Sen's slope estimator calculated as if N appears odd, it was considered as if N is even. Finally Q_med was calculated by two sided test 100% confidence interval, and then a slope was calculated by the nonparametric test. The positive value of Q_i represents an increasing trend and the negative value shows a decreasing trend in the time series.

Estimation of ET for a large river basin like Abbay requires complex modeling and data analysis. There are various methods and models to estimate ET, including the Penman–Monteith method, the Hargreaves method, and the Priestley–Taylor method. These methods generally require data on weather variables such as temperature, humidity, wind speed, solar radiation, and atmospheric pressure, as well as data on vegetation cover, soil moisture and other factors. For this study, to estimate ET, remote sensing data from satellites were used. This remote sensing can provide information on vegetation cover, land surface temperature, and other variables that are important for estimating ET. The moderate resolution imaging Spectro-radiometer and Landsat satellites are commonly used for this purpose. The analyzed result for the spatial (Figure 6(b)) distribution of ET experienced increasing trends in the eastern, northern, and western parts of the river basin, whereas a few areas in the southern part of the river basin show decreasing trends and temporal (Figure 7(b)) distribution of ET, a general decreasing trend for the study period.

CWS is a relatively new concept that involves capturing rain water in the canopy of trees and plants, which is then slowly released into the soil and groundwater over time. This technique has been shown to be effective in increasing water availability, especially in areas with low rainfall and high ET rates. In the Abbay River Basin, CWS could be used to improve water availability for agriculture, which is a major source of water for livelihood for many people in the region. By planting trees and other vegetation that have high water storage capacities, farmers can capture rain water and reduce runoff, allowing more water to infiltrate into the ground and recharge the groundwater aquifer. In addition to improving water availability, CWS can also help to reduce soil erosion, enhance biodiversity, and provide other ecosystem services. However, the success of this technique depends on a range of factors, including the type of vegetation used, the soil conditions, and the hydrological characteristics of the area. Overall, canopy water storage has significant potential as a water management strategy in the Abbay River Basin and other regions with similar environmental conditions. The analyzed result shows CWS for the Abbay River Basin (Figure 7(c)) was high in the eastern, western and northern parts of the river basin, whereas the central and southern parts of the river basin experienced low CWS. Therefore, the CWS result was affected by land use cover and a similar report was confirmed in Osun Basin Nigeria, West Africa (Pypker et al. 2005). However, further research and testing are needed to determine the most effective approaches and to quantify the benefit of this technique.

The surface runoff in the Abbay River Basin is influenced by various factors such as climate, topography, land use and soil types. The basin receives significant rainfall between June and September, which results in high surface runoff. In addition, the topography of the basin is characterized by steep slopes, which further contribute to high surface runoff. The land use is dominated by agriculture, which also affects the surface runoff by altering the natural vegetation cover. According to the study conducted by the Ministry of Water Research, surface runoff in the Abbay River Basin is 49.6 billion cubic meters. However, it is important to note that the surface runoff varies greatly depending on the season and year, and is influenced by various factors. Figure 6(d) shows that the western, eastern and northern parts of the river basin experience low runoff, whereas very few central parts of the river basin experience high runoff and the other part indicates medium runoff. The temporal distribution of surface runoff was evaluated and the result was presented (Figure 7) and the average minimum yearly surface runoff of 145.71 mm/year occurred in 2016, whereas the average maximum runoff of 576.56 mm/year occurred in 2012. The study shows that the Abbay River Basin experienced varying runoff during the study time. Based on the study conducted by Yenehun et al. (2017), the spatiotemporal variation of surface runoff occurred because of land use land cover, soil type, and the intensity of precipitation, and this study was confirmed by those previous studies (Chaemiso et al. 2021).

Soil moisture refers to the amount of water held in the soil particles and the spaces between them. It is an important aspect of the water cycle and plays a crucial role in plant growth, nutrient availability, and ecosystem health. Soil moisture can vary greatly depending on factors such as precipitation, temperature, vegetation cover, soil type and land use. Monitoring soil moisture level is important for agriculture, forestry, water resource management, and weather forecasting, as well as for understanding the impacts of climate change on the ecosystem and human activities. For this study to estimate terrestrial water storage, surface soil moisture, RZSM, and profile soil moisture were evaluated and the spatial distribution is presented in Figure 6(e)–6, whereas the temporal variations are presented in Figure 7(d)–7(f), respectively. Figure 6(e) shows high surface soil moisture in the central and southern parts of the river basin, whereas a few parts of the northwestern river basin experience low surface soil moisture. The spatial distribution of the surface soil moisture ranges from 3.12 to 8.0 mm. As presented in Figure 6(f), the RZSM ranges from 154.55 to 405.4 mm, with high concentrations in the central and southern parts of the river basin. A few areas experienced low RZSM (red color). Figure 6 presents the profile of soil moisture for the Abbay River Basin. The result shows the high spatial distribution in the southwestern parts of the river basin, the northeastern and northwestern parts experience low profile soil moisture, whereas the central parts of the river basin experience medium profile soil moisture. The classification of the soil moisture was based on the global land data assimilation system (Li et al. 2019). The temporal variation of soil moisture (Figure 7(d)–7(f)) shows similar variation, except for profile soil moisture, which shows an increasing trend for the study time.

The Blue Nile Basin is the largest watershed in Ethiopia and covers an area of more than 190,000 km². The region is generally characterized by a tropical climate, with temperatures varying based on the time of the year and the altitude of the area. In terms of temperature trends, it is important to note that global warming is having an impact on temperatures worldwide, and Ethiopia is no exception. According to data from the World Bank, the annual average temperature in Ethiopia has increased by about 1.3 °C since the early 1960s. This increase could have an impact on the Abbay River Basin, as rising temperatures could lead to changes in precipitation patterns and river flow, all of which could have significant impacts on the local environment and the people who rely on it. Figure 7 shows the spatial variation of temperature in the Abbay River Basin. Higher temperatures occurred in the northwestern part of the river basin, whereas the rest experienced low to medium temperatures and the temporal variation of the temperature (Figure 7) shows varying trends, and generally a decreasing trend for the study time.

Presently, many regions of the world are concerned about their available water supply, irrigation, and industrial activity. A growing population, human activities and inadequate management make access to clean water increasingly difficult. The shortage of efficient and effective observation in countries, especially in developing countries, results in improper management. Due to this, the major tasks of the satellite and hydrological model in the analysis of water storage in areas with limited observations are emphasized.

In this paper, the change in terrestrial water storage is estimated using the GLDAS dataset and TerraClimate dataset for the Abbay River Basin. Additionally, liquid-equivalent water thickness using the GRACE dataset was evaluated. The analyzed spatiotemporal variation of the terrestrial water storage was changing based on time and space; this is due to change in precipitation, temperature and other human activities experienced in the river basin. Based on the result, the change in terrestrial water storage shows a varying trend from the study time. The spatial variation shows a decreasing trend in high elevation regions in the upstream of the river basin, but increases in the downstream low-land area.

The terrestrial water storage was evaluated for the Abbay River Basin using the GLDAS dataset and TerraClimate. Liquid-equivalent water thickness for the river basin was estimated using the GRACE dataset. Spatiotemporal variation in the river basin was analyzed and water decline in the upstream of the river basin, and increasing of the terrestrial water storage in the downstream section of the river basin. Seasonality of the terrestrial water storage was evaluated and high water storage was observed in the study area for the autumn season, whereas low water storage was observed in the spring season. Wet and dry season water storage was evaluated for the study area for the study duration and the result demonstrates a negative value (water crisis) for the dry season, but a positive value (water surplus) for the wet season. The obtained result was essential for drought monitoring during the dry season, flood monitoring during the wet season and water resource optimization and management.

Climate change and anthropogenic activities largely affect water resources. The result obtained here shows that water storage is easily influenced by climatic variables such as rainfall and temperature, both directly and indirectly. In addition to climate change, there are notable human activities in the study area, such as irrigation, hydroelectric and others that affect water resources.

This study provides evidence that simulated GLDAS and TerraClimate provide valid data even in regions where observation is limited. The observed result shows that long-term monitoring of terrestrial water storage is necessary to prevent possible water shortages in the future during the dry season, for flood management and to conserve natural resources. For optimal utilization of the available water resource continuous monitoring was required, which is a requirement of sustainable water management and optimization for saving freshwater to deliver it to future generations.

This research did not receive any specific grant from funding agencies in the public, commercial, or nonprofit sectors.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.