Estimating temporal changes in soil pH in the black soil region of Northeast China using remote sensing

Abstract

Soil pH is a dominant soil variable because it influences many physical, chemical, and biological soil properties and plant growth processes. This study aimed at identifying the potential for field hyperspectral reflectance, multispectral satellite imagery and environmental variables to predict soil pH changes for croplands in regional areas located in the black soil region of Dehui in Northeast China. First, hyperspectral reflectance was acquired from a small area of Dehui; then, Landsat images with a 30 m resolution were acquired on 10 Oct 2005 and 22 Oct 2016 for the whole study region. This showed the scope for estimating soil pH with remote sensing derived variables both at the field scale and the regional scale for croplands in black soil regions. We selected 18 spectral variables from hyperspectral reflectance, 17 spectral variables from multispectral imagery and 6 environmental variables, respectively, and found relationships between the controlling variables and the measured soil pH values. Subsequently, 7 spectral indicators from hyperspectral reflectance, 7 spectral indicators from multispectral reflectance and 3 environmental indicators had higher correlations with soil pH among those variables. Finally, we used the partial least squares regression (PLSR) model to build soil pH prediction models for the 7 spectral indicators from hyperspectral reflectance and for the 7 spectral indicators from Landsat image with 3 environmental indicators, then we combined the two models to obtain a soil pH prediction model at the regional scale. The results showed that several thermal reflectance, vegetation indices and environmental variables, including the ratio vegetation index (RVI), the normalized difference vegetation index (NDVI), the normalized difference water index (NDWI), land use, mean annual precipitation and mean annual temperature were strong predictors of soil pH distribution at the regional scale. The combined soil pH prediction model had an R2 of 0.50, a root mean square error (RMSE) of 0.30 and a residual prediction deviation (RPD) of 1.55, indicating reliable precision and average prediction abilities. We estimated that soil pH values showed obvious heterogeneity over the study region, and soil pH mainly changed with crop type. Our results also suggested that the average decrease in soil pH for most of the study area of was approximately 0.50 units during 2005–2016. This study provides new approaches and knowledge for the better investigation of soil properties at a regional scale by using RS technologies.