Multi-decadal time series of remotely sensed vegetation improves prediction of soil carbon in a subtropical grassland.

Soil carbon sequestration in agroecosystems could play a key role in climate change mitigation but will require accurate predictions of soil organic carbon (SOC) stocks over spatial scales relevant to land management. Spatial variation in underlying drivers of SOC, such as plant productivity and soil mineralogy, complicates these predictions. Recent advances in the availability of remotely sensed data make it practical to generate multidecadal time series of vegetation indices with high spatial resolution and coverage. However, the utility of such data largely is unknown, only having been tested with shorter (e.g., 1-2 yr) data summaries. Across a 2,000 ha subtropical grassland, we found that a long time series (28 yr) of a vegetation index (Enhanced Vegetation Index; EVI) derived from the Landsat 5 satellite significantly enhanced prediction of spatially varying SOC pools, while a short summary (2 yr) was an ineffective predictor. EVI was the best predictor for surface SOC (0-5 cm depth) and total measured SOC stocks (0-15 cm). The optimum models for SOC in the upper soil layer combined EVI records with elevation and calcium concentration, while deeper SOC was more strongly associated with calcium availability. We demonstrate how data from the open access Landsat archive can predict SOC stocks, a key ecosystem metric, and illustrate the rich variety of analytical approaches that can be applied to long time series of remotely sensed greenness. Overall, our results showed that SOC pools were closely coupled to EVI in this ecosystem, demonstrating that maintenance of higher average green leaf area is correlated with higher SOC. The strong associations of vegetation greenness and calcium concentration with SOC suggest that the ability to sequester additional SOC likely will rely on strategic management of pasture vegetation and soil fertility.