Evaluation of using digital photography as a cost-effective tool for the rapid assessment of soil organic carbon at a regional scale

Abstract

A rapid and cost-effective soil organic carbon (SOC) monitoring is vital for soil management. While SOC laboratory analysis is expensive and time-consuming, spectroscopic sensor technology allows for much faster and inexpensive acquisition of data. Still, the initial cost of spectroscopic sensors is high for farmers. With soil color as a proxy, digital cameras could be used as SOC sensors. We compared the performance of SOC prediction based on soil color on a regional scale. Soil color measurements were made on samples covering the German state of North Rhine-Westphalia. SOC ranged between 0.03 and 4.74%. Images of the samples were taken under standardized conditions with four different sensors. Various soil color space models and indices were derived for SOC prediction. Modeling was performed using multiple linear regression (MLR) and random forest (RF). Best SOC prediction results were achieved using color of wet soils. Best MLR and RF models gave similar validation R2 values for the prediction of SOC at 0.66 and 0.63 with RMSE values of 0.57 and 0.61%, respectively. The MLR models using one color space showed the same performance as the RF models using the full vis spectrum. This suggested that the vis spectrum does not contain more information than the three variables of a common color space. A smartphone had the same accuracy as a more professional camera, making it a potential soil sensor. With increasing scale, soil mineralogy influences soil color weakening the performance.

Keywords

• Soil color
• Soil organic carbon
• Digital photography
• Proximal soil sensing