Applications of passive microwave data to monitor inundated areas and model stream flow

Abstract

The observation of surface water bodies in all weather conditions and better knowledge about inundation patterns are important for water resource management and flood early warning. Microwave radiometers at 37 GHz were applied to observe and study the inundation pattern in large subtropical floodplains in China, i.e. the Poyang Lake and Dongting Lake floodplains, due to the trade-off between the capability to penetrate hydrometeors and vegetation, revisiting time, and spatial coverage and resolution. Taking the shallow sensing depth at 37 GHz into account, open water, inundated area and water saturated soil surface all determine the surface emittance measured by the radiometer. Thus, Water Saturated Surface (WSS) is defined as the combination of these three land surface elements. In subtropical regions, seasonal changes in vegetation cover and various surface roughness conditions are the major challenges for the observation of surface water bodies with microwave radiometers. Atmospheric attenuation, observation gaps and errors in the microwave observations reduce the quality of daily radiometric observations. To deal with the attenuation due to vegetation and surface roughness, a two-step model was developed: the first step is to retrieve the polarization difference emissivity from Polarization Difference Brightness Temperature (PDBT) at 37 GHz with the simplified radiative transfer model and the vegetation optical thickness at 37 GHz parameterized from Normalized Difference Vegetation Index (NDVI) ; the second step is to retrieve the fractional area of WSS from the emissivity difference with a linear model, which can be parameterized according to the Qp surface roughness model. To remove the noise and extract the surface signal (including surface emittance and vegetation attenuation) from the daily PDBT time series, the Time Series Analysis Procedure (TSAP) was developed to identify the spectral features of noisy components in the frequency domain and remove them with a proper filter. The overall method combined the TSAP and the two-step model to derive daily observation of WSS area. The retrieved WSS area in the Poyang Lake floodplain was in a good agreement with the lake area observed from MODerate-resolution Imaging Spectroradiometer (MODIS) and Advanced Synthetic Aperture Radar (ASAR). The observations and analysis of the inundation patterns in the Poyang Lake and Dongting Lake floodplains with this method illustrated the close relationship between inundated area, precipitation and stream flow. Furthermore, a lumped hydrological model, named the discrete rainfall-runoff model, was developed to fully use the retrieved WSS area and to study the role of inundated area in stream flow production. This model simulates stream flow as the integration of contributions of antecedent precipitation in a certain period. Three implementations of the model were developed with the help of ground water table depth and the retrieved WSS area. The case study in the Xiangjiang River basin (upstream catchment of the Dongting Lake floodplain), China, illustrated that: 1) the longest duration of antecedent precipitation is a key parameter to determine model performance; 2) long duration would increase the model uncertainty and lead to overfitting; 3) the application of the WSS area can reduce the duration required to achieve a reasonable accuracy. The model parameters indicated the interaction between stream flow and various water storages, and the calibration results of three implementations implied the recharge period of ground water.