Premium
Assessment of Carbon Stocks in the Topsoil Using Random Forest and Remote Sensing Images
Author(s) -
Kim Jongsung,
Grunwald Sabine
Publication year - 2016
Publication title -
journal of environmental quality
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.888
H-Index - 171
eISSN - 1537-2537
pISSN - 0047-2425
DOI - 10.2134/jeq2016.03.0076
Subject(s) - environmental science , soil water , moderate resolution imaging spectroradiometer , wetland , topsoil , thematic mapper , remote sensing , hydrology (agriculture) , spatial distribution , soil science , satellite imagery , satellite , ecology , geography , geology , geotechnical engineering , engineering , biology , aerospace engineering
Wetland soils are able to exhibit both consumption and production of greenhouse gases, and they play an important role in the regulation of the global carbon (C) cycle. Still, it is challenging to accurately evaluate the actual amount of C stored in wetlands. The incorporation of remote sensing data into digital soil models has great potential to assess C stocks in wetland soils. Our objectives were (i) to develop C stock prediction models utilizing remote sensing images and environmental ancillary data, (ii) to identify the prime environmental predictor variables that explain the spatial distribution of soil C, and (iii) to assess the amount of C stored in the top 20‐cm soils of a prominent nutrient‐enriched wetland. We collected a total of 108 soil cores at two soil depths (0–10 cm and 10–20 cm) in the Water Conservation Area 2A, FL. We developed random forest models to predict soil C stocks using field observation data, environmental ancillary data, and spectral data derived from remote sensing images, including Satellite Pour l'Observation de la Terre (spatial resolution: 10 m), Landsat Enhanced Thematic Mapper Plus (30 m), and Moderate Resolution Imaging Spectroradiometer (250 m). The random forest models showed high performance to predict C stocks, and variable importance revealed that hydrology was the major environmental factor explaining the spatial distribution of soil C stocks in Water Conservation Area 2A. Our results showed that this area stores about 4.2 Tg (4.2 Mt) of C in the top 20‐cm soils. Core Ideas Remote sensing‐supported models produced excellent predictions in a carbon‐rich system. Finer spatial resolution images did not produce more accurate soil carbon predictions. Vegetation indices (SPOT, Landsat ETM+, and MODIS) served as major predictors in soil carbon models.