Reconstruction of Net Land Surface Energy Flux Constrained by Energy Conservation and Land Surface Model Simulations

Net land surface energy flux (F_LS) is a key variable controlling land–atmosphere coupling and climate feedback mechanisms, yet observational coverage is limited and model-simulated F_LS values are quite spread. To address this deficiency, a novel energy conservation method is applied to CMIP6 LS3MIP (Land Surface, Snow and Soil Moisture Model Intercomparison Project) simulations to establish the relationship between F_LS and surface temperature change at each grid point, and reconstruct gridded monthly F_LS from 1985 to 2024. Systematic validation against eddy covariance station observations demonstrates that our reconstructed F_LS outperforms existing products. The reconstructed F_LS maintains consistency with reanalyses and model simulations in terms of globally averaged anomaly variability during the common period of 1985–2012, with an average correlation coefficient of 0.77. The reconstructed F_LS data show statistically significant correlations with observations at all stations used, a performance superior to that of reanalysis data. For turbulent fluxes derived from the reconstructed F_LS, approximately 71% of sites exhibit higher anomaly correlation coefficients with observations than those from reanalysis data. The high F_LS uncertainties are mainly distributed in Northern Hemisphere high latitudes and high-altitude zones due to large errors in snowmelt. The difference between LS3MIP and our estimation is situated mainly over 50°–70°N, where the land area mean standard deviation from the multiannual mean F_LS is 0.80 W m⁻² for LS3MIP and 0.55 W m⁻² for our estimation. The F_LS constrained by the simple energy model provides a useful reference for diagnosing land–atmosphere interactions and evaluating climate models.