Quantifying the dynamics of water bodies, wetlands and biomass in the Poyang Lake region: A multi-polarization SAR remote sensing approach.

摘要

Field measurements were combined with synthetic aperture radar (SAR) images to evaluate the use of C-band multi-polarized radar remote sensing for estimating plant parameters (plant height, fresh biomass, dry biomass and vegetation water content) of wetland vegetation, and mapping the dynamics of water bodies, wetlands (natural wetlands and rice paddies) and flooding extents in the Poyang Lake region. The capacity of L-band SAR in land cover mapping was also investigated by integrating with optical imagery.;Hydrological patterns in Poyang Lake are the dominant factor controlling the spatial and temporal variations of wetland species in Poyang Lake. Water levels in this region are primarily governed by five rivers (Ganjiang river, Xiushui river, Raohe river, Fuhe river, and Xinjiang river). Its northern region is also influenced by the backflow from Yangtze River. The above-ground total biomass increased steadily from March following the hydrological cycle. Wetland species colonizing at different altitudes were gradually flooded from late spring to summer. Carex spp. died during flooding periods and started another growth cycle in autumn after flooding receded. Canopy volume dominates the radar backscattering mechanism in Carex spp. wetlands during their growth period, but the temporal variation of radar backscatter from these wetlands is mainly influenced by flooding. Tall wetland species (Miscanthus sacchariflorus, Phragmites communis Trin., and others) still emerged above water surfaces during flooding peaks and started to senesce in autumn. Surface backscattering mechanism is dominant during the early growing stage and the senescent period of tall vegetation. Plant canopy variation controlled the temporal dynamics of radar backscatters from Phragmites communis Min. Radar backscattering mechanisms from Miscanthus sacchariflorus wetlands were more complicated during the flooding periods. The variations of ground water depth and plant structure of Miscanthus sacchariflorus during its growth period result in over 10 dB spatial and temporal variation in ASAR backscatter in HH- and HV-polarization.;The relationship of canopy height with ASAR backscattering coefficient is the most significant among the influencing factors (plant height, fresh biomass, dry biomass, vegetation water content) on radar backscattering mechanism (R2=0.9 for HH-polarization and R2=0.59 for HV-polarization) from Phragmites cummunis Trin. HH- and HV-backscatters are more sensitive to the variation of dry biomass (R2=0.76 for HH and R2=0.56 for HV) than to that of fresh biomass (R 2=0.07 for HV and R2=0.42 for HH). Plant water content plays a negative role and attenuates the backscattering signals in both polarizations. For Phragmites communis Trin. with tall stalks (over 2m) and long, blade-like leaves, HH-polarization is more sensitive to vegetation parameters than HV-polarization for C-band SAR signals. Similar to Phragmites communis Trin., ASAR backscattering coefficient in both polarizations is more sensitive to plant height and dry biomass of non-flooded Miscanthus sacchariflorus, and their regression coefficients (R2) are over 0.5 for HH-polarization and over 0.4 for HV-polarization. Plant water content has no evident effect on the variation of ASAR backscatter. HV-polarization is more sensitive to the variation of above-water canopy parameters than HH-polarization for flooded Miscanthus saccharifiorus. HH- and HV-polarized radar backscatters from Carex spp. wetlands increased significantly with the variation of plant height, fresh biomass and dry biomass, but they reach saturated when vegetation grows up to 30cm. Compared with those tall grass with stalks and long blade-like leaves, the correlation of fresh biomass with HV-polarization is more pronounced (R 2=0.78) than that with HH-polarization (R2=0.41) for Carex spp. Vegetation structure play a more important role in radar backscattering mechanism than plant water content for these three wetland species.;Temporal profiles of C-band multi-polarized backscatter coefficients for individual land cover types over the period of December 2004 to November 2005 were studied and described in the context of the ecology and seasonal dynamics of biophysical parameters of individual land cover types. A knowledge-based hierarchical land cover mapping method was developed to quantify the dynamics of paddy rice, natural wetlands and floods using the time series of HH- and HV-backscatters. The specific phenological and ecological characteristics of wetlands including paddy rice are the most important data in mapping their spatial and temporal patterns. The classification accuracy is over 90% for water bodies, rice paddies and Carex spp. wetlands, but it is not high for tall wetlands (68%). A decision tree approach was adopted to evaluate the capacity of L-band SAR in land cover mapping by combining with optical imagery. Classification errors were mainly induced by the mixed spectrum between and covers, and lack of independent training data and validation data also caused uncertainty in the results.