Fractional vegetation coverage downscaling inversion method based on Land Remote-Sensing Satellite (System, Landsat-8) and polarization decomposition of Radarsat-2

ABSTRACT

Due to multi-source data information fusion, the precision of eco-hydrology models is improving rapidly. In particular, the fractional vegetation coverage (FVC) is of great significance in the remote sensing monitoring of surface parameters. In this study, downscaling inversion was performed using Normalized Difference Vegetation Index (NDVI) and Ratio Vegetation Index (RVI) data from Land Remote-Sensing Satellite (System, Landsat-8) and RVI-Freeman data from Radarsat-2 with polarization decomposition, incorporating the scattering entropy (H) and anisotropy (α). Further, modified vegetation indices (mVIs) and corresponding calculation methods were developed to describe the FVC precisely. Two deep learning (DL) methods were used for mVI optimization. The results showed that the inclusion of H and α greatly facilitated FVC estimation and that the mVIs and DL provided higher accuracies and smaller errors than the previous methods (NDVI or RVI). HαmRVI, one of the mVIs, had the highest accuracy in FVC simulation using a vegetation index, and the particle swarm optimization neural network (PSONN) achieved the best performance. The FVC was then predicted with 8 m resolution using the mVIs and PSONN, demonstrating that the proposed method effectively compensates for the fluctuations in high-FVC valley wetlands caused by high water content, avoids overestimation in grasslands, and provides great detail while retaining the original regional variations.

Disclosure statement

The authors declare no conflict of interest.

Author Contributions

M.L. developed the initial and final versions of this manuscript and analyzed the data. T.L., L.D., Y.L., L.M., and Y.W. contributed their expertise and insights, overseeing all of the analysis and supporting the writing of the final manuscript. M.L., J.Z., Y.Z., Y.L., and Z.C. performed the experiments.

Additional information

Funding

This research was funded by the National Key R&D Program of China (No. 2018YFC0406400), the International S&T Cooperation Program of China (No. 2015DFA00530), the National Natural Science Foundation of China (Nos. 51939006 and 51620105003), the Inner Mongolia Natural Science Fund Key Project (No. 2018ZD05), the Ministry of Education Innovative Research Team (No. IRT_17R60), the Innovation Team in Priority Areas Accredited by the Ministry of Science and Technology (No. 2015RA4013), the Inner Mongolia Industrial Innovative Research Team (Grant 2012), the IMAU Innovative Research Team (Grant NDTD2010-6), and the program for Young Talents of Science and Technology in Universities of Inner Mongolia Autonomous Region (Grant NJYT-18-B11);Inner Mongolia Industrial Innovative Research Team [Grant 2012];