Linking GRACE-Derived Water Storage Accelerations to Changes in Hydro-Mteorological Fluxes over West Africa

Abstract:

Several researchers have postulated that, under a changing climate, an intensification of the water cycle is under way. This is usually related to increases in hydrological fluxes such as precipitation (P), evapotranspiration (E), and discharge (R). It is under debate, however, whether such observed or reconstructed flux changes are real for West Africa and on what scale. Large-scale increase or decrease of the flux deficit P-E-R, i.e. flux changes that do not compensate, would lead to acceleration or deceleration of water storage potentially visible in GRACE-derived time series. In agreement with earlier studies, we do find accelerations in global maps of gridded GRACE water storage anomalies (Eicker et al., submitted). For sub-Sahel West Africa such accelerations amount up to 7 mm/a²; i.e. statistically significant even seen the short GRACE record. However, W Africa water storage variability is strongly controlled by sea surface temperature and precipitation (Forootan et al., 2014), and interannual/decadal climate variability may mask long-term changes e.g. related to land use change. Yet, here we show that even after isolating and removing a global climate mode in the GRACE data that appears ENSO-related, using a new method, the observed signal over W Africa remains. We can explain this effect only partly with changing surface water levels (Volta reservoir). We then repeat our analysis with flux fields from global atmospheric reanalysis that include land surface models in online (ERA-Interim, CFSR) and off-line (MERRA-Land) mode, and TRMM precipitation data. We find that these fields show surprising skills in reconstructing water storage variability at the monthly timescale when compared to GRACE. In contrast, in particular ERA-Interim falls short in displaying trends that would correspond to GRACE accelerations. We hypothesize this may be due to time-varying biases in the reanalysis-generated fluxes as noticed in other studies. Essentially we conclude that GRACE provides, even though its data record is short, new information that should be used to constrain future regional reanalyses towards a better representation of the water cycle over West Africa.