EUSFlux: Mapping Forest Carbon Uptake in the Eastern US from Upscaled Flux Tower Data

This repository provides a wall-to-wall, spatially explicit dataset of Net Ecosystem Productivity (NEP) across eastern United States forests at 500 m spatial resolution and 16-day temporal resolution, spanning 2003–2023. NEP is estimated using a Random Forest model trained on eddy covariance observations from AmeriFlux and NEON flux tower networks and applied to gridded remote sensing, solar-induced fluorescence (SIF), and meteorological predictor variables. Predictor variables include MODIS-derived surface reflectance bands (MOD09A1 Bands 1–7), NDVI (MOD13A1), LAI and FPAR (MOD15A2H), NIRv (computed as (NDVI − 0.08) × NIR), GOSIF solar-induced fluorescence, and Daymet v4 meteorological variables (Tmin, shortwave radiation, precipitation, daylength, and log-transformed VPD). All predictors are temporally composited to 16-day intervals and harmonized to a 500 m spatial grid. Feature selection was performed using permutation importance across 100 iterations of 5-fold spatial cross-validation; SHAP analysis provides model interpretability. The final model uses 11 predictor variables: Category (forest type), month, NDVI, SIF, surface reflectance Bands 2, 6, and 7, LAI, FPAR, daylength, and shortwave radiation, and Tmin. The Zenodo repository includes: • The preprocessed flux tower training dataset (NEE_flux_unh_sif_2.csv), containing 16-day site-level NEE observations and co-located predictor variables from AmeriFlux/NEON sites in eastern US forests; • An independent evaluation dataset of monthly FluxCom NEE estimates at seven held-out flux tower test sites (test_sites_monthly_fluxComNEE_2003_2021.csv; column NEE_gC_m2_d1); • An aboveground carbon benchmark raster from Harris et al. (2021) / Global Forest Watch, masked to the eastern US study domain (GFW_masked.tif); • The study area boundary shapefile (StudyArea.shp); • Seven Python notebooks implementing the full analysis workflow, from feature selection through temporal anomaly detection; • Annual NEP GeoTIFFs (YYYY_yearly_sum.tif, 2003–2023) and a 20-year mean map (NEP_20yr_Mean_Map.tif) at 500 m resolution; NEP is reported in units of gC m⁻² year⁻¹ in annual rasters, with positive values indicating net carbon uptake and negative values indicating net carbon release. The sign convention follows NEP = −NEE; flux tower data and intermediate prediction rasters are in NEE convention and are sign-flipped in the analysis scripts. The Zenodo repository has been updated to include comprehensive file-level and variable-level metadata describing the contents, formats, column names, units, and provenance of all shared data files and code. A complete README is provided within the deposit. Predictor variables are derived from the following external datasets, which are not included in this deposit and must be obtained separately: MODIS MOD09A1, MOD13A1, and MOD15A2H (NASA LP DAAC / Google Earth Engine); GOSIF SIF (University of New Hampshire); Daymet v4 (ORNL DAAC / Google Earth Engine); AmeriFlux/NEON flux tower observations; FLUXCOM RS+METEO (Max Planck Institute for Biogeochemistry); NLCD 2010 (USGS); and a forest stand age raster for 2010.