Land surface water is a key component of the global water cycle. Compared to remote sensing by satellites, both temporal extension and spatial continuity is superior in modeling of water surface area. However, overall evaluation of models representing different kinds of surface waters at the global scale is lacking. We estimated land surface water area (LSWA) using the Catchment-based Macro-scale Floodplain model (CaMa-Flood), a global hydrodynamic model, and compared the estimates to Landsat with 3″ spatial resolution at the global scale. Results show that the two methodologies show agreement in the general spatial patterns of LSWA (e.g., major rivers and lakes, open-to-sky floodplains), but globally consistent mismatches were found under several land surface conditions. CaMa-Flood underestimates LSWA in high northern latitudes (e.g., the Canadian Shield) and coastal areas, as the presence of isolated lakes in local depressions or small coastal rivers is not considered by the model’s physical assumptions. In contrast, model-estimated LSWA is larger than Landsat estimates in forest-covered areas (e.g., Amazon basin) due to the opacity of vegetation for optical satellite sensing, and in cropland areas due to the lack of dynamic water processes (e.g., re-infiltration, evaporation, water consumption) and constraints of water infrastructure (e.g., canals, levees). These globally consistent differences can be reasonably explained by the model’s physical assumptions or optical satellite sensing characteristics, and applying filters (e.g., floodplain topography mask, forest and cropland mask) to the two datasets allows the remaining local-scale discrepancies to be attributed to locally varying factors (e.g., channel parameters, atmospheric forcing).
