Predicting Soil Organic C and N in the Russian Chernozem from Wireless Color Sensor Measurements.

Color sensor technologies offer opportunities for affordable and rapid assessment of soil organic carbon (SOC) and total nitrogen (TN) in the field, but applicability of these technologies may vary by soil type and land use. The objective of this study was to use an inexpensive color sensor to develop SOC and TN prediction models for the Russian Chernozem (Haplic Chernozem) under two management regimes: a native grassland (not cultivated for at least 300 years), and an adjacent continuous fallow field in the Kursk region of Russia. Twenty-one dried soil samples were analyzed using a Nix Pro^TM color sensor that is controlled through a mobile application and Bluetooth to collect CIEL*a*b* (darkness to lightness, green to red, and blue to yellow) color data. The data were used to develop the natural log of SOC (lnSOC) and TN (lnTN) prediction models using depth, L*, a*, and b* for each sample were used as predictor variables in regression analyses. Resulting residual plots, root mean square errors (RMSE), and coefficients of determination (R²) were used to asses model fit for the SOC and total N prediction models. The final models for all soil samples, which included depth and color parameters, for lnSOC (R²=0.968, RMSE=0.006, p-value<0.001) and lnTN (R²=0.959, RMSE=0.009, p-value<0.001) for all samples contained depth, L*, and a* as predictor variables. The final models for native soil samples, which included depth and color parameters, for lnSOC (R²=0.980, RMSE=0.004, p-value<0.001) and lnTN (R²=0.971, RMSE=0.007, p-value<0.001) for all samples contained depth and a* as predictor variables. The final models for all soil samples, which included only color parameters, for lnSOC (R²=0.817, RMSE=0.033, p-value<0.001) and lnTN (R²=0.792, RMSE=0.042, p-value<0.001) for all samples contained L*, and a* as predictor variables. The final models for native soil samples, which included only color parameters, for lnSOC (R²=0.891, RMSE=0.022, p-value<0.001) and lnTN (R²=0.877, RMSE=0.029, p-value<0.001) for all samples contained L* and a* as predictor variables. The results suggest that soil color may be used for rapid assessment of TN and SOC regardless of soil management practices.