Mapping inundation dynamics in a heterogeneous floodplain: Insights from integrating observations and modeling approach

Abstract

Determining the spatiotemporal dynamics of surface water in a heterogeneous floodplain is difficult, especially for its surrounding isolated lakes. Seasonal inundation patterns of these isolated lakes can be misestimated in a hydrodynamic model due to their short and erratic appearances. A surface water time series of Poyang Lake having an 8-day revisit frequency with a 30 m spatial resolution from 2000 to 2016 was conducted for the first time. This study was produced with a modified hierarchical spatiotemporal adaptive fusion model (HSTAFM) by integrating both Landsat and MODIS data. Discrepancies between model-based surface water and remotely-sensed surface water were evaluated, and possible causes were discussed. Results show that the modified HSTAFM can better detect the water features of a floodplain, thereby providing more detailed information in an seasonal isolated lake system than the MODIS MOD13Q1 product. With the fusion product, we found that Poyang Lake evidently shrank after experiencing a longer low-water period after the impoundment of the Three Gorges Dam. A large proportion of these discrepancies (averaging 36%) between model-based and remotely-sensed surface water distributed in seasonal isolated lakes, mainly occurred during high-water level periods. Uncertainties in the hydrodynamic model might attribute to smaller defined lake boundaries, bathymetric variations, human disturbance, and unconsidered groundwater recharge/discharge. These results provide a new insight into the temporally continuous and spatially dynamic assessment of simulated surface water, which is essential for the future improvement in the hydrodynamic model.