Maas, E. D. and R. A. Lal. 2023. A case study of the RothC soil carbon model with potential evapotranspiration and remote sensing model inputs. Remote Sensing Applications: Society and Environment 29:100876.

Citable PDF link: https://lter.kbs.msu.edu/pub/4102

Soil carbon modeling is an important tool to inform scientists, land managers, and policy makers about the impacts of land management decisions on soil health and climate change mitigation. Models can require a wide range of data inputs to make predictions, not all of which are readily available but can often be substituted with estimates. Understanding the impact that inputs have on the model output is critical to understand the uncertainties of their predictions. This study evaluated the output of the RothC soil carbon model using three potential evapotranspiration (PET) models (Tegos et al., 2017; Droogers and Allan, 2002; and Thornthwaite and Mather, 1955; Te, DA, and TM, respectively) to estimate open pan evaporation, and three vegetation indices (NDVI, SAVI, and MSAVI) from satellite imagery to estimate corn (Zea mays L.), soybean (Glycine max (L.) Merr.), and wheat (Triticum aestivum L.) yield to convert to C input for the RothC model to predict soil organic carbon over 11 years at a research farm in Michigan. The objectives of this study were to evaluate the correlation with observed data for each estimated input as well as the variability in RothC output they produced. The TM, Te, and DA PET models resulted in 13%, 38%, and 40% wet bias, respectively, resulting in increases of 1.6, 5.5 and 7.4 Mg C ha−1, respectively, over the baseline model output. All three vegetation indices estimated corn yields poorly (R2 = 0.25, 0.26, and 0.26 for NDVI, SAVI, and MSAVI, respectively), but soybean yields exceptionally well (all R2 = 0.91). However, despite NDVI producing significantly higher values (p < 0.001), there were no significant differences between each index’s estimated yield used to calculate C input for RothC, and the resulting model runs are almost indistinguishable.

DOI: 10.1016/j.rsase.2022.100876

Associated Treatment Areas:

Modeling

Download citation to endnote bibtex

Sign in to download PDF back to index
Sign In