Soil prediction for coastal wetlands following Spartina alterniflora invasion using Sentinel-1 imagery and structural equation modeling.
Globally, biological invasions have significantly altered soil functions in coastal wetlands such as nutrient cycling and carbon restoration. Although these changes have been explored, the quantitative prediction of soil properties that respond to species invasions have not well been studied. To facilitate development of an invasion-specific soil predictive model, we used time series Sentinel-1 data to represent the invasion process regarding species establishment stages and structural equation modeling (SEM) to quantify the interacting relationships within the soil-vegetation system that was assumed to be useful in explaining the spatial variability of soils. By applying the SEM model with time series Sentinel-1 data as covariates, we predicted topsoil properties in the east-central China coast that is impacted by Spartina alterniflora Loisel. An accuracy assessment based on leave-one-out cross-validation showed the efficacy of the proposed model to predict soil properties, with a ratio of performance to deviation (RPD) of 1.47 for soil salinity, 0.99 for soil pH, and 1 for soil organic carbon (SOC). In the SEM, the plant invasion presented a direct influence on soil salinity that influenced soil pH, whereas SOC was influenced by soil pH, but its effects on soil pH and SOC were not important. These results highlight that the interacting relationships between the plant invasion and the soil system can be modeled. It also highlights the potential utility of time series Sentinel-1 data in capturing the invasion process, providing evidence that Sentinel-1 data might be useful in estimating soil properties in areas with dense natural vegetation cover.