Identifying landscape hot and cold spots of soil greenhouse gas fluxes by combining field measurements and remote sensing data

Upscaling chamber measurements of soil greenhouse gas (GHG) fluxes from point scale to landscape scale remain challenging due to the high variability in the fluxes in space and time. This study measured GHG fluxes and soil parameters at selected point locations (n = 268), thereby implementing a stratified sampling approach on a mixed-landuse landscape (∼ 5.8 km²). Based on these field-based measurements and remotely sensed data on landscape and vegetation properties, we used random forest (RF) models to predict GHG fluxes at a landscape scale (1 m resolution) in summer and autumn. The RF models, combining field-measured soil parameters and remotely sensed data, outperformed those with field-measured predictors or remotely sensed data alone. Available satellite data products from Sentinel-2 on vegetation cover and water content played a more significant role than those attributes derived from a digital elevation model, possibly due to their ability to capture both spatial and seasonal changes in the ecosystem parameters within the landscape. Similar seasonal patterns of higher soil/ecosystem respiration (SR/ER–CO₂) and nitrous oxide (N₂O) fluxes in summer and higher methane (CH₄) uptake in autumn were observed in both the measured and predicted landscape fluxes. Based on the upscaled fluxes, we also assessed the contribution of hot spots to the total landscape fluxes. The identified emission hot spots occupied a small landscape area (7 % to 16 %) but accounted for up to 42 % of the landscape GHG fluxes. Our study showed that combining remotely sensed data with chamber measurements and soil properties is a promising approach for identifying spatial patterns and hot spots of GHG fluxes across heterogeneous landscapes. Such information may be used to inform targeted mitigation strategies at the landscape scale.