Soil Phosphorus and Nitrogen Predictions Across Spatial Escalating Scales in an Aquatic Ecosystem Using Remote Sensing Images

The incorporation of remote sensing (RS) data into digital soil models has shown success to improve soil predictions. However, the effects of multiresolution imagery on modeling of biogeochemical soil properties in aquatic ecosystems are still poorly understood. The objectives of this study were the following: 1) to develop prediction models for soil total phosphorus (TP) and total nitrogen (TN) utilizing RS images and environmental ancillary data at three different resolutions; 2) to identify controlling factors of the spatial distribution of soil TP and TN; and 3) to elucidate the effects of different spatial resolutions of RS images on inferential modeling. Soil cores were collected (n = 108) from the top 10 cm in a subtropical wetland: Water Conservation Area-2A, the Florida Everglades, USA. The spectral data and derived indices from RS images, which have different spatial resolutions, included the following: MODIS (500 m resampled to 250 m), Landsat ETM+ (30 m), and SPOT (10 m). Block kriging and random forest (RF) were employed to predict soil TP and TN using RS-image-derived spectral input variables, environmental ancillary data, and soil observations. The RF models showed R² between 0.90 and 0.93 and root mean square error between 100.4 and 115.9 mg · kg^-1 for TP and between 1.45 and 1.52 g · kg^-1 for TN. Soil TP was mainly predicted from RS-derived spectral indices that infer on biotic/vegetation characteristics, whereas soil TN was predicted using a combination of biotic/vegetation, topographic, and hydrologic variables. Results suggest that the spectral data informed soil models have excellent predictive capabilities in this aquatic ecosystem. Interestingly, there was no noticeable distinction among different spatial resolutions of RS images to develop prediction models for soil TP and TN in terms of error assessment. However, the variability and complexity of soil TP and TN variations were much better exp- essed with the finer resolution RFSPOT model than the coarser resolution RFMODIS model as demonstrated using entropy.