Informing an integrated hydrological model with SWOT in the tropical Usangu wetland

Tropical wetland hydrology is often poorly monitored by in situ gauges, yet it is critical for understanding the global carbon budget and water resource management. Earth observation (EO) data has long been used to calibrate and update hydrological and hydraulic models, providing valuable insights in poorly instrumented catchments. The recently launched Surface Water and Ocean Topography (SWOT) mission provides simultaneous observations of surface water extent and height at near global coverage. SWOT has been extensively used for informing river hydraulic and hydrodynamic models, but much less for integrated wetland hydrological models that represent surface water – groundwater interaction. Here we focus on modeling the hydrology of the Usangu wetlands in Tanzania. Usangu is representative of a wide range of tropical wetlands featuring a variety of land cover types (grasslands, forests, marshes, crops, permanent and seasonal flooding) and presents strong changes in hydrology both seasonally and interannually due to human impact. 

We developed a MIKE SHE integrated hydrological model for the Usangu wetlands and surrounding alluvial fans. The model is coupled with a 1D river routing model and forced by a lumped-conceptual rainfall runoff model at all major river inlets to the wetland. Model forcing data includes daily CHIRPS v2.0 precipitation data and FAO reference evapotranspiration data. From Sentinel-2 multispectral imagery we extract river widths used to inform cross section shape. Vegetation maps are built from a combination of MODIS leaf area index (LAI), maps from aerial surveys, and global land cover maps. For calibrating the base model, we use three river discharge stations located along three separate rivers feeding the Usangu wetlands. SWOT pixel cloud data is processed to make dynamic flood extent maps over the wetland area. To improve flood extent estimation under dense vegetation, additional radar satellites (PALSAR-2, Sentinel-1) are used in combination with SWOT. SWOT pixel cloud data is also used to estimate river heights and establish an updated rating curve at the main outlet of Usangu, along the Great Ruaha River.

We present the first results characterizing the Usangu wetland hydrology as seen from multiple earth observation satellites (SWOT, Sentinel-2, other radar satellites) and compare with predictions from the integrated MIKE SHE model. SWOT-derived flood extent maps are compared with the modeled flood extent over the wetland domain using overlap-based metrics such as CSI and F-score. River heights from SWOT are compared with modeled river water levels.