COMBINING PLSR AND KRIGING METHOD TO RETRIEVE SOIL SALINITY WITH ALI DATA IN YELLOW RIVER DELTA

摘要

Soil salinization is one of the most common land degradation progresses in Yellow River Delta (YRD) of China. In this study, deterministic and geostatistical methods were jointly applied to estimate soil salt content (SSC) in this area on the basis of field spectra and Advanced Land Imaging (ALI) data. SSC and field spectra of 50 soil samples were collected to investigate the relationship between soil salinity and ALI-convolved field spectra using partial least square regression (PLSR) method. Significant correlation was observed with determination coefficient (R~2) of 0.837. Subsequently, SSC map was predicted using ALI reflectance data and the PLSR model. The predictions show the similarity to the field observations. To detect soil salinity in the vegetated area, 314 samples were systematically collected on PLSR-derived SSC map and were then interpolated using universal kriging (UK) method. Cross validation results show that 95.9% of the 314 samples are included in the confidence intervals under confidence level of 95% through a Bland-Altman plot. Therefore, it is convinced that the combination of methods provides an inexpensive and labor-saving approach to the estimation of soil salinity in the entire study area with desirable accuracy.