Multitemporal satellite imagery analysis for soil organic carbon assessment in an agricultural farm in southeastern Brazil

Soil organic carbon (SOC) plays a crucial role for soil health. However, large datasets needed to accurately assess SOC at high resolution across scales are labor-intensive, time-consuming, and expensive. Ancillary geodata, including remote sensing spectral indices (RS-SIs) and topographic indicators (TIs), have been proposed as spatial covariates. Reported relationships between SOC and RS-SIs are erratic, possibly because single-date RS-SIs do not accurately capture SOC spatial variability due to transient confounding factors in the soil (e.g., moisture). However, multitemporal RS-SI data analysis may lead to noise reduction in SOC versus RS-SI relationships. This study aimed at: i) comparing single-date versus multitemporal RS-Sis derived from Sentinel-2 imagery for assessment of topsoil (0–0.2 m) SOC in two agricultural fields located in south-eastern Brazil; ii) comparing the performance of RS-SIs and TIs; iii) using adequate RS-SIs and TIs to compare sampling schemes defined on different collection grids; and iv) studying the temporal changes of SOC (0–0.2 m and 0.2–0.4 m). Results showed that: i) single-date RS-SIs were not reliable proxies for topsoil SOC at the study sites. For most of the tested RS-SIs, multitemporal data analysis produced accurate proxies for SOC; e.g., for the Normalized Difference Vegetation Index, the 4.5th multitemporal percentile predicted SOC with an R2 of 0.64; ii) The best TI was elevation (ranging from 643 to 684 m) with an R2 of 0.70; iii) The multitemporal SI and elevation maps indicated that the different sampling schemes were equally representative of the topsoil SOC's distribution across the entire area; and iv) From 2012 through 2019, topsoil SOC increased from 19.3 to 24.1 g kg−1. The ratio between SOC in the topsoil and subsoil (0.2–0.4 m) decreased from 1.7 to 1.1. Further testing of the proposed multitemporal RS-SI analysis is necessary to confirm its dependability for SOC assessment in Brazil and elsewhere.