GRACE: Gravity Recovery and Climate Experiment long-term trend investigation over the Nile River Basin: Spatial variability drivers

Abstract

GRACE (Gravity Recovery and Climate Experiment) long-term terrestrial water storage anomaly (TWSA) is attributed to the complex interaction of climatic, physical and anthropogenic drivers. This paper, therefore, explores how different hydroclimatic and anthropogenic processes interact and combine over “space” to produce the mass variations that GRACE-TWSA detects. Using the Nile River Basin (NRB) as a case study, it explicitly analyzes nine hydroclimatic and anthropogenic processes, as well as their relationship to the TWSA in different climatic zones. The analytic method employed the long-term trends derived for both the dependent (TWSA) and independent (explanatory) variables via applying two geographically multiple regression (GMR) approaches: (i) an ordinary least square regression (OLS) model in which the contributions of all variables to TWSA variability are deemed equal at all locations; and (ii) a geographically weighted regression (GWR) which assigns a weight to each variable at different locations based on clustering occurrences. The models’ efficacy was investigated using standard goodness of fit diagnostics. The OLS explains that the basin at large TWSA spatial variability significantly attributed to five variables, i.e., precipitation, runoff, surface water storage, soil moisture storage, and population density, (p < 0.0001). The OLS model, however, produced an R2 value of 0.14 with skewed standardized residuals. In contrast, the GWR model retained varying explanatory variables by different climate zone. For instance, the results showed that all nine variables contribute significantly, with varying ranking, to the trend in TWSA in the tropical zone. The evapotranspiration (ET) and population density are the only significant variables in the semiarid zone; population density contributes significantly to TWSA variability in all zones. The GWR model yielded R2 values with a median of 0.71 and normally distributed standard residuals. To evaluate the robustness of the GWR approach, the basin-wide TWSA pattern was simulated using the GWR model outputs. Herein, the GWR highlights the importance of the spatial locations to attribute the spatial variability in GRACE TSWA long-term trends. This spatial information, therefore, is critical for developing robust statistical models for reconstructing time series of proxy GRACE anomalies that predate the launch of the GRACE and for gap-filling between GRACE and GRACE Follow-On (GRACE-FO) mission.