Evapotranspiration in the Congo Basin Determined by a Remotely Sensed Water Balance

Despite its importance to regional climate and ecohydrology, the seasonal and interannual trends of evapotranspiration (ET) in tropical Africa remain poorly understood. ET models of the region are constrained by very few direct observations, as the sole previous direct measurements have only examined non-native vegetation over short timescales. Remote sensing approaches offer the ability to collect great amounts of data at large spatial scales, but many remote-sensing based estimates of ET are still largely influenced by hydrologic modelling assumptions that have not been validated in the Congo Basin. This study combines remotely sensed hydrological data, in-situ river discharge data, and a basin-wide water balance approach to estimate the magnitude and trends of ET in the Congo Basin. We estimate a basin-wide annual mean ET of 115 cm/year plus or minus 5 cm/yr for the period of 2002-2013. This value compares favorably to historical estimates of Congo Basin ET.

In addition, a variety of precipitation and evapotranspiration model products are compared to determine their suitability for the understudied tropical African system. Global land surface models and machine learning products (i.e. FLUXNET-MTE) both significantly underestimated ET over the same time span. While global ET products capture the shape of the seasonal ET cycle relatively well, they also greatly underestimate the magnitude of seasonal and interannual variability. Precipitation data products also vary greatly in magnitude and seasonal cycle over the Congo, contributing a significant source of uncertainty in water balance estimates of ET. Solar-induced fluorescence (SIF) data suggest a marked decline in regional photosynthetic activity after 2007, while photosynthetically-active radiation (PAR) appears to increase steadily over the same period, indicating changes in seasonal water availability may be increasingly inhibiting plant productivity over the past decade.