Han et al. (2025) FluxHourly: global long-term hourly 9 km terrestrial water-energy-carbon fluxes

Identification

• Journal: Earth system science data
• Year: 2025
• Date: 2025-12-11
• Authors: Qianqian Han, Yijian Zeng, Yunfei Wang, Fakhereh Alidoost, Francesco Nattino, Yang Liu, Zhongbo Su
• DOI: 10.5194/essd-17-7101-2025

Research Groups

• Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, 7522 NH Enschede, the Netherlands
• Netherlands eScience Center, 1098 XH Amsterdam, the Netherlands

Short Summary

This study presents FluxHourly, a global long-term (2000–2020) hourly 9 km dataset of terrestrial water-energy-carbon fluxes, generated by integrating model simulations, in-situ measurements, and machine learning to enable analysis of ecosystem responses to climate extremes at unprecedented spatiotemporal scales.

Objective

• To generate a global, long-term (2000–2020), hourly, 9 km resolution dataset of terrestrial water-energy-carbon fluxes (FluxHourly) to address the scarcity of high spatiotemporal resolution flux data for understanding ecosystem functioning and climate interactions.

Study Configuration

• Spatial Scale: Global, 9 km (0.1°).
• Temporal Scale: Long-term (2000–2020), hourly.

Methodology and Data

• Models used:
  • STEMMUS-SCOPE (Simultaneous Transfer of Energy, Mass, and Momentum in Unsaturated Soil; Soil Canopy Observation, Photochemistry and Energy fluxes radiative transfer)
  • Random Forest (RF) regression algorithm (specifically RF_OI, Random Forest with Optimal Interpolation)
  • Optimal Interpolation (data assimilation method)
• Data sources:
  • In-situ measurements: PLUMBER2 dataset (170 globally distributed flux tower sites from FLUXNET2015, La Thuile, OzFlux networks), eddy covariance observations.
  • Meteorological data: ERA5-Land (9 km, hourly).
  • Remote sensing and ancillary data: Canopy height, Leaf Area Index (LAI), Maximum carboxylation rate (Vcmax), Carbon dioxide concentration (CO2), Surface soil moisture (SSM), International Geosphere–Biosphere Programme (IGBP) land cover type.
  • Comparison datasets: FLUXCOM, FLUXFORMER, GLEAM (v4.2a), TROPOMISIF.

Main Results

• The RF_OI emulator achieved high accuracy, with Pearson Correlation Coefficient (r-score) values higher than 0.88 for most variables (Rn 0.99, LE 0.88, H 0.92, G 0.92, SIF685 0.99, SIF740 0.99), and 0.80 for Gross Primary Productivity (GPP).
• Root Mean Square Error (RMSE) values were below 36.03 W m⁻² for net radiation (Rn), 0.04 W m⁻² µm⁻¹ sr⁻¹ for solar-induced fluorescence (SIF), and between 4.30–4.86 µmol m⁻² s⁻¹ for GPP.
• Testing on independent stations showed r-score values consistently higher than 0.8.
• Feature importance analysis indicated that incoming shortwave radiation (Rin), surface soil moisture (SSM), and leaf area index (LAI) are the top predictor variables, with Rin being the most important for all seven target variables.
• FluxHourly demonstrated a high degree of agreement with in-situ Rn (95.3% overlapping probability density function) and comparable agreement with other products for latent heat flux (LE) (86.7%), sensible heat flux (H) (86.9%), and GPP (86.6%).
• The dataset successfully captures diurnal variations and shows strong consistency with in-situ measurements, offering superior temporal resolution compared to existing products.
• Uncertainty estimates are provided for each timestep and grid, with the RF_std2 model showing better performance in quantifying uncertainty.
• Scale mismatch in predictor variables, particularly incoming shortwave radiation from coarse-resolution gridded data, can lead to an underestimation of fluxes.

Contributions

• Generation of FluxHourly, the first global long-term (2000–2020) hourly 9 km dataset of terrestrial water-energy-carbon fluxes, addressing a critical gap in high spatiotemporal resolution flux data.
• Development of the RF_OI emulator, which integrates the physically-based STEMMUS-SCOPE model, optimal interpolation with in-situ measurements, and machine learning with remote sensing and meteorological data.
• Provides simultaneous predictions for seven key water-energy-carbon flux variables (net radiation, latent heat flux, sensible heat flux, soil heat flux, gross primary productivity, and solar-induced fluorescence at 685 nm and 740 nm) with high accuracy.
• Offers uncertainty estimates for the predicted fluxes, enhancing data reliability and utility for scientific analysis.
• Enables detailed analysis of ecosystem responses to climate extremes at unprecedented spatiotemporal scales, supporting advancements in understanding ecosystem functioning and climate interactions.

Funding

• China Scholarship Council (grant no. 202004910427)
• SURF Cooperative (grant no. EINF-6614 and EINF-12364)
• The Netherlands Organisation for Scientific Research (NWO) KIC, WUNDER project (grant no. KICH1.LWV02.20.004)
• Netherlands eScience Center, EcoExtreML project (grant ID. 525 27020G07)
• Water JPI project “iAqueduct” (Project number: ENWWW.2018.5)
• ESA ELBARA-II/III Loan Agreement EOP-SM/2895/TC-tc
• ESA MOST Dragon V and VI Program

Citation

@article{Han2025FluxHourly,
  author = {Han, Qianqian and Zeng, Yijian and Wang, Yunfei and Alidoost, Fakhereh and Nattino, Francesco and Liu, Yang and Su, Zhongbo},
  title = {FluxHourly: global long-term hourly 9 km terrestrial water-energy-carbon fluxes},
  journal = {Earth system science data},
  year = {2025},
  doi = {10.5194/essd-17-7101-2025},
  url = {https://doi.org/10.5194/essd-17-7101-2025}
}