Abstract

Digital cameras are becoming increasingly popular in environmental sciences and proximal soil sensing as low cost, but high quality and high-resolution sensors. The high spatial resolution of the images shows a great potential for mapping predicted soil properties at small scales, from soil profiles to the microscale. To explore the potential of digital cameras for the morphometric analysis at a fine scale, we took 50 samples plus 3 microplate scale excavations from a profile wall of a Luvisol and analyzed samples for soil organic carbon (SOC) and Fe content for model calibration. Images of sieved and ground soil samples were taken under standardized laboratory conditions. After image correction, RGB colors were obtained for each sample and transformed into different color space models. Based on the obtained soil colors, different regression models were built to predict SOC and Fe content. Simple regression produced predictions with R2adj values of 0.90 for SOC (using HSV V) and 0.70 for Fe (using CIE a*) for ground soil samples. For sieved samples, R2adj values were lower with 0.69 for SOC using HSV V and 0.61 for Fe using CIE a*. Multiple linear regression models with interaction terms could improve those predictions for sieved samples to R2adj values of 0.94 for SOC and 0.89 for Fe, using the complete HSV and CIELab color tristimulus, respectively. Based on the best pedotransfer functions, we predicted SOC and Fe contents for the excavated microplates at a 1 × 1 mm resolution. With this method we were able to map the microscale variability of SOC and Fe content with great accuracy. This study showed, that accurately calibrated digital images can be a cost-effective method to map the distribution of SOC and Fe and potentially other soil physical and chemical parameters at a very fine scale.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call