Construction of land surface dynamic feedbacks for digital soil mapping with fusion of multisource remote sensing data

Authors:
Publisher website
Google Scholar
Abstract
Summary: The use of environmental covariates to predict soil spatial variation is a widely adopted approach to digital soil mapping. However, commonly used covariates such as topography, landform and vegetation are often ineffective for estimating soil variation in areas of low relief. Recent studies have shown the effectiveness of a new covariate called land surface dynamic feedback (LSDF) for digital soil mapping in such areas. The construction of LSDF relies on remote sensing (RS) data with high temporal resolution to record the drying process after a rain event. The trade‐off of obtaining high temporal resolution with RS data is that they are often of low spatial resolution. To overcome this limitation, our study uses the ESTARFM (enhanced spatial and temporal adaptive reflectance fusion model) algorithm to fuse MODIS and Landsat 8 data to obtain images that benefit from the high temporal resolution of MODIS and high spatial resolution of Landsat. The LSDF was then derived from the fused images to predict soil texture in a case study in north Xuancheng, Anhui Province. Compared with particle‐size fractions estimated with LSDF derived from the original MODIS data, the results were more accurate and produced more spatial detail when mapped. We conclude that the ESTARFM algorithm can improve the spatial resolution of high temporal resolution RS data and offers an effective approach to derive more accurate measures of LSDF for digital soil mapping in areas of low relief. Highlights: We propose a new method to combine different remote sensors to predict soil properties in a low relief area. We constructed a land surface dynamic feedback model with high spatial resolution for digital soil mapping. The fusion process improved prediction the most in areas with complex and heterogeneous land cover. Accuracy improved with the mapping of finer spatial detail compared to the original MODIS data.
Funding Information
  • Nanjing Normal University (YBPY16_006)
  • National Natural Science Foundation of China (41431177)

This publication has 29 references indexed in Scilit: