Characterization of Soil Variability By Digital Mapping Based on Fourier Transform Mid-Infrared Photoacoustic Spectroscopy.

Geographic mapping is a straightforward mean for characterization of soil variance, exploitation and management of soil resources. However, the soil features and variances presented in current soil maps are not enough for practical application, due to the limitations on acquisition of soil attribute information. Traditional laboratory chemical method is time- consuming and laborious. There is a pressing need to determine soil fertility rapidly for catering precision agriculture development requirements. To solve the problem, a soil map with transform infrared photoacoustic spectroscopy (FTIR-PAS) technology, which is widely used due to rapid, nondestructive and well-rounded characterization of soil properties, were employed to characterize soil variation. In this study, a total of 933 soil samples were collected from agricultural fields located in the Lishui County in the Jiangsu Province of China, and FTIR-PAS was utilized to record the soil spectra and the spectral principal components (PCs) were extracted by principal component analysis (PCA). PCs extracted contained most of soil variance information and were then used for soil mapping using the methods of Mahalanobis distance (MD) and red–green–blue (RGB) theory, respectively. The entire soil maps are completed by ArcMap on ordinary Kriging interpolation.  The MD soil map, which the color changed from red to yellow and blue from south to north gradually, met the actual circs better and more external. Undetermined soil can be classified roughly according to the color of the MD soil map which facilitates further quantitative analysis. However, there was a mismatch between the soil type displayed by traditional soil map and the soil type revealed by MD soil map, which was due to the inherent defects of MD principle. The RGB map was more like a 3D map in a sense, in which red-green-blue was equivalent to the x-y-z of rectangular coordinate system. The first three PCs were assigned to the red, green and blue respectively. About 90% of soil spectral information will be passed onto the map in situ The zonal characteristics and a rough classification was in the same tendency of the two graphs as a whole Nevertheless, the prominent different can also be found. The RGB map provided more detailed and comprehensive soil variation conformation and matched conventional soil map better. Then Self-adaptive spectral predict models base on the RGB principle was built to determine the soil properties such as pH, SOM, AP, TN, TP and TK. One soil sample was randomly selected and 50 samples with the similar colors were chosen to build a model. The results showed that RPD values of all the models reached 3.3 and the mean of R was 0.80. The study demonstrated that FITR-PAS spectra and RGB imaging based on soil mapping could character and predict soil properties with different soil variation, which would be useful in soil management and utilization of soil resources.