Mapping soil organic matter content in field using HJ-1 satellite image

Abstract: Soil organic matter content is one of the main factors affecting productivity of agricultural soils. Many studies have shown that the remote sensing is a good tool for estimation of soil organic matter (SOM) content. Satellite hyperspectral image or airborne hyperspectral image has been used in the last decade. However, the data derived from these images have a long revisiting period and are expensive in acquisition and processing. To solve this problem, this study aimed to estimate SOM based on HJ-1 satellite multispectral data that had revisiting period of one day and were cost free. SOM content monitoring model was built by remote sensing with the spatial resolution of 30 meters based on HJ satellite CCD's multispectral data and SVC HR-768 spectrometer hyperspectral data. There were four bands for the satellite data and their spatial resolution was 30 meters. In addition, there were 768 bands of hyperspectral data distributed from 350 to 2 500 nm. S-G filter was used to eliminate systematic errors of spectrometer during hyperspectral data-based model fitting. Spectral resolution of resampled hyperspectral data matched that of CCD data from spectral response function (SRF). Then hyperspectral data and SRF were used to analog reflectivity of CCD data at each band. The correlation between SOM and surface spectral characteristics of the samples was analyzed, from which a preliminary SOM monitoring model was established. To further improve the initial monitoring model, initial model histogram was matched to sample's histogram to correct the initial monitoring model. The final monitoring model of SOM was then established. Taking into account spatial difference between the samples and remote sensing images, ten soil samples were used to test the model. The results showed that there was a good linear relationship between estimated and measured SOM values (determination coefficient 0.93, the slope 1.2 and the standard deviation was 0.57%). Based on the model, the distribution of the farmland SOM was mapped with the spatial resolution 30 m and the temporal resolution of one day. The cost-free data of HJ-1 and the model provided an economical tool to estimate SOM in farm field.