Topsoil Organic Carbon Stocks and Uncertainty in Florida Grazing Lands Derived from Quantile Regression Forest (30 m Resolution)

This dataset provides spatially explicit predictions of topsoil organic carbon (SOC) stocks and associated prediction uncertainty for grazing lands across Florida, USA. The data were generated using a parsimonious, uncertainty-aware digital soil mapping framework based on Quantile Regression Forest (QRF) modeling and cross-temporal integration of legacy and contemporary soil observations.

The dataset consists of five raster layers in GeoTIFF format, including a binary grazing land mask, SOC predictions, and associated uncertainty bounds, all at 30 m spatial resolution:

• Grazing land mask (binary): Spatial extent of Florida's grazing lands used for modeling (1 = grazing land, 0 = non-grazing land)
• Mean SOC stock map (t ha⁻¹): Expected value of predicted topsoil SOC stocks
• 5th percentile map (Q5): Lower bound of the prediction interval
• 95th percentile map (Q95): Upper bound of the prediction interval
• 90% prediction interval (PI) width: Difference between Q95 and Q5, representing spatial uncertainty magnitude

These layers collectively enable both deterministic and probabilistic interpretation of SOC spatial variability, supporting applications in carbon accounting, grazing land management, and uncertainty-aware decision-making. The grazing land mask defines the spatial domain of analysis and can be used to subset SOC predictions.

Data characteristics:

• Spatial resolution: 30 m
• Spatial extent: Florida, USA (grazing lands)
• Coordinate reference system: EPSG:4326
• File format: GeoTIFF (.tif)
• Units: t ha⁻¹ (SOC stock layers)

Notes:

• This dataset supports a manuscript currently under review. 
• The dataset represents modeled estimates rather than direct measurements and should be interpreted accordingly.
• Prediction uncertainty is derived from the QRF model and reflects the spread of the conditional distribution; it does not capture all sources of uncertainty (e.g., sampling bias, measurement errors).
• Users are encouraged to consider both mean predictions and uncertainty bounds in analyses and decision-making.

Citation: if you use this dataset, please cite:

1. Zhao, C., Song, J., Dubeux, J., Grunwald, S., Bretas, I. L., Liao, H.-Y., Tziolas, N., Harley, J. B., Zare, A., Babaeian, E., Garcia, L., Queiroz, L., & Mendes, C. T. E. (2026). Topsoil Organic Carbon Stocks and Uncertainty in Florida Grazing Lands Derived from Quantile Regression Forest (30 m Resolution) (v1.0) [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.19192952
2. Zhao, C., Song, J., Dubeux, J., Grunwald, S., Bretas, I. L., Liao, H.-Y., Tziolas, N., Harley, J. B., Zare, A., Babaeian, E., Garcia, L., Queiroz, L., & Mendes, C. T. E. (2026). Spatiotemporal controls on soil organic carbon stocks in subtropical grazing lands: An uncertainty-aware digital soil mapping approach. Available at SSRN: https://ssrn.com/abstract=6459841 or http://dx.doi.org/10.2139/ssrn.6459841

Associated resource:

1. Web GIS application: An interactive web-based GIS application for exploring and visualizing the spatial distribution of topsoil organic carbon stocks and associated uncertainty is available at: https://es-geoai.rc.ufl.edu/agroes-grazing-soc/
2. GitHub repository: 
   • Source code: https://github.com/Ecosystem-Services-GeoAI/florida-grazing-soc-qrf

Funding sources:

• This research is funded by the following grants: 2022-2023 Florida state legislative budget AI-HARVEST; Florida Milk Checkoff; Florida Cattle Enhancement Board (P0326003); USDA-NIFA Research Capacity Hatch Funds (FLA-AGR-006393); Florida Agricultural Experiment Station UF/IFAS Archer Early Career Seed Grant (P00133052); Startup funds from the Florida Agricultural Experiment Station UF/IFAS, University of Florida; The FSCPD was funded by USDA-CSREES-NRI grant award 2007-35107-18368 (PI: Grunwald).