Satellite-based investigation of water stress at the basin scale: an integrated analysis of downscaled GRACE estimates and remotely sensed data

Climate change is one of the main challenges of human beings in recent history which invites a real menace to different aspects of human life as well as nature. It has had perceptible ramifications for the agricultural sector in many regions of the world which jeopardizes the food safety of the residents (Gregory et al. 2005). Climate change is fundamentally manifested as changes in the established norms of climatic parameters such as precipitation (P), temperature (T), evapotranspiration (ET), etc. in a region (McCarty 2001; Dore 2005). The decreasing P and rising T and ET accompanied by increased population put tremendous pressure on water resources (Richey et al. 2015). The deterioration of the hydrological balance between inputs and outputs in a region due to natural and anthropogenic drivers engenders a gap between water supply and demand which is defined as water scarcity (Hasan et al. 2019). Water stress is regarded as a potential driver of regional insecurity that can contribute to regional unrest (Richey et al. 2015).

The spatiotemporal dynamics between human and natural water systems can be understood easily through water stress analysis. To date, different approaches have been followed by researchers to characterize and analyze water stress which yielded various water stress indicators. Water Resources Availability (Yang et al. 2003); Watershed Sustainability Index (Chaves & Alipaz 2006); Green–Blue Water Scarcity (Rockström et al. 2009); Water Scarcity Function of Water Footprint (Hoekstra et al. 2011); Water Use Availability Ration (Alcamo & Henrichs .2002); Local Relative Water Use and Reuse (Vorosmarty et al. 2005); Water Poverty Index (Lawrence et al. 2002); and Water Supply Stress Index (McNulty et al. 2010). Although these indices have made good strides in assessing water stress status, their optimal use is shackled by some restrictions mainly due to their inability to integrate all forms of ‘available water’, particularly, soil moisture and groundwater due to data dearth (Hasan et al. 2019).

To overcome the aforementioned handicap, Hasan et al. (2019) proposed a new water stress index based on the Gravity Recovery And Climate Experiment (GRACE)'s Terrestrial Water Storage (TWS) estimates and precipitation data from the Tropical Rainfall Measuring Mission (TRMM). The GRACE-observed TWS is a composite hydrological parameter consisting of different water cycle components such as surface water, soil moisture, snow water, canopy water, and groundwater (Rodell et al. 2018). Therefore, it encompasses all forms of water storage and its application for water stress assessment gives a clear insight into the total water availability in a region (Hasan et al. 2019). In their recent study, Hasan et al. (2019) tested the feasibility of the Potential Available Water Storage (PAWS) indicator to determine water stress in African countries and reported good performance. They applied this approach over a large scale (country level) because of two reasons; first, the coarse resolution (ranging from 200,000 km² at low latitudes to 90,000 km² near the poles) (Rodell et al. 2018) of the GRACE data hinders its applications over smaller scales. Moreover, they validated their results against the water availability statistics retrieved from the AQUASTAT dataset (a global water information system of the Food and Agriculture Organization (FAO)) which is available only on country scales. The novelty of this study lies in its application of the PAWS index for a smaller scale (basin level) which is more required for hydrological analysis and applications. In this premise, the GRACE-derived TWS was downscaled to 10 km of spatial resolution based on Random Forest (RF) machine learning. The downscaled TWS later was integrated with satellite precipitation estimates from the CHIRPS (Climate Hazards Group InfraRed Precipitation with Station data) mission to derive the PAWS index.

The GRACE is a twin satellite remote sensing project, working in tandem to gather the gravitational signals of the Earth (Rodell et al. 2007). The GRACE has been providing the public end-users with data on the monthly anomalies of TWS processed by different centers (Khorrami et al. 2022). Mass Concentration (Mascon) blocks are the latest data format of the mission, which are post-processed and ready to use. Moreover, the Mascon solutions are global and not tailored toward a particular application, and hence, can be utilized for all scientific topics of interest (Save et al. 2016). Therefore, the latest release of the GRACE/GRACE-FO JPL Mascon solutions with a spatial resolution of 50 km was utilized in this study. The TWS data can be downloaded from https://earth.gsfc.nasa.gov/geo/data/grace-mascons.

The FLDAS model is a large-scale hydrology model within which two Land Surface Models (LSMs) are used to simulate different hydrometeorological variables through integrated in situ, modeled, and satellite-derived observations at different resolutions (Loeser et al. 2020). The FLDAS makes simulations under the Variable Infiltration Capacity (VIC), and the Noah models (McNally et al. 2017) with a spatial resolution of 25 and 10 km, respectively. In this study, the different hydro-meteorological parameters such as P, ET, T, soil moisture storage, snow water storage, and runoff were extracted from the Noah model of the FLDAS mission via https://disc.gsfc.nasa.gov/datasets?keywords=FLDAS.

CHIRPS dataset is a relatively new quasi-global, high-resolution, daily/monthly precipitation dataset. The multisource CHIRPS project combines three different datasets (global climatology, satellite estimates, and in situ observations) to generate gridded precipitation data. Being known as one of the most potent gridded representatives of the rainfall field (Ray et al. 2022), CHIRPS is often used as an alternative source of precipitation measurements, especially where sufficient measurements are not accessible (Paca et al. 2020). The CHIRPS monthly precipitation data with a resolution of 0.05 degrees can be downloaded from http://chg.geog.ucsb.edu/data/chirps/.

MOD16 is an algorithm used by MODIS (Moderate Resolution Imaging Spectroradiometer) mission to generate high-resolution ET data by integrating MODIS observations and MERRA meteorological data (Souza et al. 2019). In this study, ET data from the MOD16A2 products with a spatial resolution of 500 m and the temporal resolution of 8 days were used. MOD16A2 ET estimates are freely available at https://modis.gsfc.nasa.gov/data/dataprod/mod16.php.

The coarse spatial resolution of the GRACE data remains a handicap for local studies with a spatial extent of less than the GRACE resolution (Wang et al. 2019; Khorrami & Gunduz 2021a). Recent developments in downscaling techniques make GRACE data suitable to assess the local scale variations in water storage (see e.g., Chen et al. 2019; Ali et al. 2021; Arshad et al. 2022; Yin et al. 2022; Khorrami et al. 2023b). In this study, the RF machine learning technique, which is proven to be a very successful downscaling technique for the GRACE data (Chen et al. 2019), was applied to enhance the spatial footprint of the GRACE TWS from to over the entire country. The RF benefits from the average values obtained from decision trees generated based on a set of homogenous subsets of random predictors (Rahaman et al. 2019). The spatial downscaling is implemented as follows: (i) All the input variables were aggregated to the spatial resolution of the GRACE-JPL Mascon (50 km), and then the statistical associations between the TWS, geospatial variable (elevation), and FLDAS-derived variables (soil moisture, snow water, rainfall, surface runoff, ET) at 10 km resolution were evaluated by developing the RF model to predict TWS. Later, (ii) the residual values were calculated by deducing the model-derived TWS from the GRACE-derived TWS. Afterward, (iii) the developed model was applied to the hydrological and geospatial variables at a spatial resolution of 10 km to attain the estimated 10 km TWS. Finally, (iv) the residual correction was performed at 10 km by adding residuals to the estimated TWS at 10 km to obtain the downscaled TWS.

The PAWS is a satellite-derived hydrological index proposed by Hasan et al. (2019) to investigate water stress status at large scales. The PAWS is calculated as follows:

The water stress level of the basin based on the PAWS threshold was determined according to the classification given in Table 1.

To validate the accuracy of the results, the variations of PAWS were validated against the WSD and the net water flux (NWF) in the basin. NWF is calculated as precipitation minus ET and describes the effective flux of water between the atmosphere and the earth's surface (Khorrami et al. 2023a) and therefore provides important information regarding the interaction of the atmosphere with the land surface. This parameter is also important from a hydroclimatic perspective because it is a measure of water stress (Gao & Giorgi 2008). Decreasing NWF signals water scarcity thus diminishing available water in a basin. Therefore, higher NWF corresponds to a condition where water is more available for other compartments of the hydrological cycle. In this study, the annual ET was subtracted from the annual P to estimate the NWF over the study area.

The WMB is one of the basins most susceptible to drought in Türkiye (Özüpekçe 2021). The recent water stress status of the basin was investigated based on the PAWS index. According to the results, The WMB has suffered from diminishing water storage from 2003 to 2020 with a storage loss of 0.09 km³ per annum which is equal to 1.62 km³ of total water storage loss during the last 18 years. Since the GRACE signals are more sensitive to agricultural and hydrological droughts (Okay Ahi & Jin 2019), the results can be interpreted taking the drought analysis results into account. The critical storage loss of the basin in 2007, 2008 2018, and 2020 can be ascribed to the harsh country-wise drought incidents (Türkeş et al. 2009; Kurnaz 2014; Özüpekçe 2021; Khorrami & Gündüz 2022). The hydrologic status of the WMB has been in particular very critical in 2008 during which the basin lost −1.79 km³ of its water storage. This is justified by taking the drought analysis results of the region into consideration. Özüpekçe (2021) reported extremely dry conditions for the basin in 2008 and stated that the wetlands of the basin had been in peril of complete disappearance due to the drought side effects as well as anthropogenic impacts.

The basin-wise PAWS results reveal that except in 2003, the WMB has had a critical status with a per capita available water of less than 1,200 m³. Since the water stress threshold is 1,700 m³, it can be inferred that the basin has been under great pressure from 2004 onward. Although the GRACE estimates reflect the impacts of both natural and anthropogenic forces on the variations of water storage (Mandea et al. 2020), in light of the high associations between the PAWS and NWF over the basin, it can be inferred that the variations of water storage and thus water stress in this basin are fundamentally ascribed to climate change consequences.

The prevalent impacts of climate change and population increase have culminated in the detriment of the hydrological balance in many regions of the world. As a result, water resources were put under exorbitant pressure challenging the availability of freshwater, especially in the regions with semiarid to arid climates. Since water availability plays a crucial role in sustaining the world food supply and socio-economic development, investigation of water availability and analysis of water stress are very important. Remote sensing observations provide a good alternative for the traditionally used field data with their global coverage, affordability, and acceptable accuracy. The PAWS is a new index developed based on the GRACE estimates and satellite precipitation data to investigate water stress. In this study, downscaled GRACE data and fine-resolution precipitation from the CHIRPS mission were integrated to evaluate water stress at a local scale. The results suggested a very good performance of the index for a local scale analysis. The PAWS agrees well with the NWF and WSD in the basin suggesting the reliability of the index for basin-wise assessments of water stress.

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

All relevant data are available from an online repository at https://figshare.com/s/c0899ea7796c74810e83.

The authors declare there is no conflict.