Effects of Varying Spatial Resolutions of Remote Sensing Images On Digital Prediction Models for Soil Biogeochemical Properties in the Everglades, Florida.

There is an increased use of remote sensing (RS) images in digital soil mapping due to the capability to directly or indirectly capture the soil forming factors such as biotic properties and parent materials. Comprehensive studies have been conducted on the use of RS images to assess soil properties, and integrated approaches using the images with various statistical methods have shown great potential for developing soil prediction models. However, the effect of different spatial resolutions of the RS images modeling soil properties is relatively unknown, especially in wetlands. The objectives of this study were to (i) develop prediction models for soil properties (total phosphorus, nitrogen, and carbon) utilizing RS images and environmental ancillary data and (ii) elucidate the effect of different spatial resolutions of RS images on inferential modeling of soil properties in a subtropical wetland: Water Conservation Area-2A, the Florida Everglades, U.S. Soil cores were collected (n=108) from the top 10 cm. The spectral data and derived indices from RS images, which have different spatial resolutions, included: MODIS (250 m), Landsat ETM+ (30 m), and SPOT (10 m). Block kriging (BK) and random forest (RF) ensemble trees were used to derive soil models at varying spatial scales mirroring the RS image resolutions (250, 30, and 10 m). An entropy analysis was conducted to quantify the amount of spatial information of each of the prediction models. RF models using RS derived spectral input variables showed improved prediction results (> 89%) when compared to BK. The RF models using finer spatial resolution images showed better predictive capability than the coarser resolution models. Results suggest that the spectral data derived from RS images can improve the predictive quality of soil properties in the Everglades. However, the effect of varying spatial resolutions of RS images was relatively small.