Global monitoring of soil multifunctionality in drylands using satellite imagery and field data

Models derived from satellite image data are needed to monitor the status of terrestrial ecosystems across large spatial scales. However, a remote sensing-based approach to quantify soil multifunctionality at the global scale is missing despite significant research efforts on this topic. A major constraint for doing so is the availability of suitable global-scale field data to calibrate remote sensing indicators (RSI) and, to a lesser extent, the sensitivity of spectral data of available satellite sensors to soil background and atmospheric conditions. Here, we aimed to develop a soil multifunctionality model to monitor global drylands coupling ground data on 14 soil functions of 222 dryland areas from six continents to 18 RSI derived from a time series (2006-2013) Landsat dataset. Among the RSI evaluated, the chlorophyll absorption ratio index was the best predictor of soil multifunctionality in single-variable-based models (r = 0.66, P < 0.01, NMRSE = 0.17). However, a multi-variable RSI model combining the chlorophyll absorption ratio index, the global environment monitoring index and the canopy-air temperature difference improved the accuracy of quantifying soil multifunctionality (r = 0.73, P < 0.01, NMRSE = 0.15). Furthermore, the correlation between RSI and soil variables shows a wide range of accuracy with upper and lower values obtained for AMI (r = 0.889, NMRSE = 0.05) and BGL (r = 0.685, NMRSE = 0.18) respectively. Our results provide new insights on assessing soil multifunctionality using RSI that may help to monitor temporal changes in the functioning of global drylands effectively.

Automated summary

Models derived from satellite image data are needed to monitor the status of terrestrial ecosystems at a large scale. However, there is a lack of remote sensing-based approach for quantifying soil multifunctionality globally. This study aimed to develop a soil multifunctionality model using field data and remote sensing indicators (RSI) from a Landsat dataset. The results showed that a multi-variable RSI model improved the accuracy of quantifying soil multifunctionality. The correlation between RSI and soil variables varied across different RSI.