Wetlands are responsible for 20%–31% of global methane (CH₄) emissions and account for a large source of uncertainty in the global CH₄ budget. Data‐driven upscaling of CH₄ fluxes from eddy covariance measurements can provide new and independent bottom‐up estimates of wetland CH₄ emissions. Here, we develop a six‐predictor random forest upscaling model (UpCH4), trained on 119 site‐years of eddy covariance CH₄ flux data from 43 freshwater wetland sites in the FLUXNET‐CH4 Community Product. Network patterns in site‐level annual means and mean seasonal cycles of CH₄ fluxes were reproduced accurately in tundra, boreal, and temperate regions (Nash‐Sutcliffe Efficiency ∼0.52–0.63 and 0.53). UpCH4 estimated annual global wetland CH₄ emissions of 146 ± 43 TgCH₄ y⁻¹ for 2001–2018 which agrees closely with current bottom‐up land surface models (102–181 TgCH₄ y⁻¹) and overlaps with top‐down atmospheric inversion models (155–200 TgCH₄ y⁻¹). However, UpCH4 diverged from both types of models in the spatial pattern and seasonal dynamics of tropical wetland emissions. We conclude that upscaling of eddy covariance CH₄ fluxes has the potential to produce realistic extra‐tropical wetland CH₄ emissions estimates which will improve with more flux data. To reduce uncertainty in upscaled estimates, researchers could prioritize new wetland flux sites along humid‐to‐arid tropical climate gradients, from major rainforest basins (Congo, Amazon, and SE Asia), into monsoon (Bangladesh and India) and savannah regions (African Sahel) and be paired with improved knowledge of wetland extent seasonal dynamics in these regions. The monthly wetland methane products gridded at 0.25° from UpCH4 are available via ORNL DAAC (https://doi.org/10.3334/ORNLDAAC/2253).

Diffuse radiation can increase canopy light use efficiency (LUE). This creates the need to differentiate the effects of direct and diffuse radiation when simulating terrestrial gross primary production (GPP). Here, we present a novel GPP model, the diffuse‐fraction‐based two‐leaf model (DTEC), which includes the leaf response to direct and diffuse radiation, and treats maximum LUE for shaded leaves (ɛ_msh defined as a power function of the diffuse fraction (D_f)) and sunlit leaves (ɛ_msu defined as a constant) separately. An Amazonian rainforest site (KM67) was used to calibrate the model by simulating the linear relationship between monthly canopy LUE and D_f. This showed a positive response of forest GPP to atmospheric diffuse radiation, and suggested that diffuse radiation was more limiting than global radiation and water availability for Amazon rainforest GPP on a monthly scale. Further evaluation at 20 independent AmeriFlux sites showed that the DTEC model, when driven by monthly meteorological data and MODIS leaf area index (LAI) products, explained 70% of the variability observed in monthly flux tower GPP. This exceeded the 51% accounted for by the MODIS 17A2 big‐leaf GPP product. The DTEC model's explicit accounting for the impacts of diffuse radiation and soil water stress along with its parameterization for C4 and C3 plants was responsible for this difference. The evaluation of DTEC at Amazon rainforest sites demonstrated its potential to capture the unique seasonality of higher GPP during the diffuse radiation‐dominated wet season. Our results highlight the importance of diffuse radiation in seasonal GPP simulation.