Prediction and map-making of soil organic matter of soil profile based on imaging spectroscopy: A case in Hubei China

In this paper, the content and distribution of soil organic matter (SOM) along a soil profile was predicted using two hyperspectral imaging cameras, a visible and near infrared (VNIR) camera (400-1000 nm) and a near infrared (NIR) camera (900-1700 nm). Three to five soil samples were taken from the different horizons of each soil profile. Each sample was divided into four subsamples of different soil particle sizes and one not-messed subsample. All samples were placed under the cameras to obtain hyperspectral data to establish different SOM partial least square regression (PLSR) prediction models. The performance of the spectral prediction models of five subsamples was acceptable. And the coefficient of determination of validation dataset (R²_val) and the ratio of prediction to deviation (RPD) of the not-messed soil samples were 0.833 and 2.005, respectively. Then the model was used to predict the SOM content and make the distribution map of SOM along profile.