Effect of soil crust on the prediction of soil organic matter based on soil colour

Abstract

Soil organic matter (SOM) plays a pivotal role in enhancing soil quality and structure. Given the darkening effect of SOM on soil colour, rapid SOM prediction can be achieved by quantifying soil surface colour. However, owing to heterogeneity in soil surfaces and differences in colour acquisition between laboratory and field environments for various sensors, understanding the transferability of laboratory findings to field applications is essential. Therefore, this study aimed to uniquely evaluate the effects of soil crusts formed in field environments and soil moisture on SOM prediction using different sensors. 125 soil samples were collected from the 0–20 cm topsoil layer on the Bashang Plateau, North China. Dispersed and crusted soil samples were prepared in the laboratory to simulate controlled and natural conditions, and colour data were obtained using digital cameras and Nix colour sensors. The study results showed that crusted soil samples exhibited a better correlation with SOM than did dispersed samples, while slight soil moisture enhanced this correlation. Among the red, green and blue bands, the red band exhibited the highest correlation with SOM. Mathematical transformations, particularly the excess red index (ExR), further improved this relationship, achieving a correlation coefficient of 0.87. Comparing the digital camera and Nix sensor prediction results revealed that integrating soil surface variation information facilitated significant model accuracy improvement. Among the modeling results, the digital camera provided the best prediction for the crusted soil samples, with a coefficient of determination (R²) of 0.80 and a root-mean-square error (RMSE) of 0.50 %, while R²_val = 0.80 and RMSE_val = 0.61 % were obtained for the validation results. Due to the unevenness of the soil surface and the importance of the sampling area size for stable predictions, the non-contact digital camera is more suitable for the acquiring soil surface colour information and predicting SOM content than the Nix sensor.