Quantitative prediction of soil organic matter content using hyper spectral remote sensing and geo-statistics

Soil organic matter (SOM) content was predicted at two different scales. At a large scale, the geo-statistical method was applied to interpolate the spatial distribution of SOM throughout Hengshan County of China. Additionally, at a small scale in Hyperion image, by analyzing the correlation between spectrally reflective data and SOM concentrate, the ratio of the reflectivity reciprocal-logarithm’s first derivative of 623.6 nm against the reflectivity reciprocal-logarithm’s first derivative of 564.4 nm was selected as the sensitive regression variable, and the best multivariate retrieval model was developed. Then the retrieval model was utilized to the hyper-spectral data for SOM quantitative mapping, and the adjusted R square coefficient of 0.8684 revealed a precise result. For objective comparison, 30 soil samples were used for spatial interpolation in geo-statistical way at the same scale in Hyperion imagery. After comparing and analyzing the two methods, it indicates that the predicted result of geo-statistics is not so good as that by hyper-pectral retrieval way due to the influences of sample quantities, sample distributions, sample intervals together with the inner-inclusion hypothesis.