How can we reasonably compare the water surface area between models & satellites

Abstract:

Land surface water is one of the key components in the global water cycle. It exists in various forms (e.g., rivers, lakes, wetlands, etc.) and is impacted by different land surface conditions and anthropogenic activities. Compared to sensing the water from satellites, modeling of those water bodies is superior in temporal and spatial continuity. However, the model ability to represent various water bodies has not be evaluated over the globe at an adequate spatial resolution. This study estimated the daily global land water surface area by a global hydrodynamic model (CaMa-Flood) from 2001 to 2014. The characteristics of the maximum water surface (occurrence > 10%) are compared with the water occurrence dataset derived from Landsat (90m). Their differences are explained from physical processes, land cover types and discussed with various uncertainties.

Results show that model estimations exhibited similar spatial patterns of the global land water surface area with satellite-derived results. Though, limited by the original model spatial resolution (10km), the small water depressions away from main river channels and small parallel rivers in a same unit catchment are not represented in CaMa-Flood. This results in the underestimation of water surface in CaMa-Flood than Landsat in the high-latitudes (e.g., Canadian Shields) and the coastal areas. Water surface in the irrigated area (e.g., deltas regions and irrigated districts) is generally overestimated due to the ignorance of some natural processes (e.g., re-infiltration, evaporation) and human water regulation (e.g., canals, levees, water consumption) in the CaMa-Flood. Though, the ignorance of irrigation in paddy fields leads to underestimation of CaMa-Flood since these seasonal water bodies can be captured by Landsat. The water bodies under thick vegetation (e.g., Amazon, Indonesia) and frequent cloud covers (e.g., lower Mekong) can be better represented in the model since they are difficult to be detected by optical sensors because of the obstructions. Uncertainties in the runoff forcing, model parameters and baseline topography are the remained reasons for their difference. This study from a global study tells how a global hydrodynamic model can represent different water forms and how a reasonable comparison can be made between models and satellites.