Multi-model perspectives and inter-comparison of soil moisture and evapotranspiration in East Africa—an application of Famine Early Warning Systems Network (FEWS NET) Land Data Assimilation System (FLDAS)

Convergence of evidence from different agro-hydrologic sources is particularly important for drought monitoring in data sparse regions. In Africa, a combination of remote sensing and land surface modeling experiments are used to evaluate past, present and future drought conditions. The Famine Early Warning Systems Network (FEWS NET) Land Data Assimilation System (FLDAS) routinely simulates daily soil moisture, evapotranspiration (ET) and other variables over Africa using multiple models and inputs. We found that Noah 3.3, Variable Infiltration Capacity (VIC) 4.1.2, and Catchment Land Surface Model based FLDAS simulations of monthly soil moisture percentile maps captured concurrent drought and water surplus episodes effectively over East Africa. However, the results are sensitive to selection of land surface model and hydrometeorological forcings. We seek to identify sources of uncertainty (input, model, parameter) to eventually improve the accuracy of FLDAS outputs. In absence of in situ data, previous work used European Space Agency Climate Change Initiative Soil Moisture (CCI-SM) data measured from merged active-passive microwave remote sensing to evaluate FLDAS soil moisture, and found that during the high rainfall months of April-May and November-December Noah-based soil moisture correlate well with CCI-SM over the Greater Horn of Africa region. We have found good correlations (r>0.6) for FLDAS Noah 3.3 ET anomalies and Operational Simplified Surface Energy Balance (SSEBop) ET over East Africa. Recently, SSEBop ET estimates (version 4) were improved by implementing a land surface temperature correction factor. We re-evaluate the correlations between FLDAS ET and version 4 SSEBop ET. To further investigate the reasons for differences between models we evaluate FLDAS soil moisture with Advanced Scatterometer and SMAP soil moisture and FLDAS outputs with MODIS and AVHRR normalized difference vegetation index. By exploring longer historic time series and near-real time products we will be aiding convergence of evidence for better understanding of historic drought, improved monitoring and forecasting, and better understanding of uncertainties of water availability estimation over Africa