Improved modeling of canopy transpiration for temperate forests by incorporating a LAI-based dynamic parametrization scheme of stomatal slope

• The flux-derived slope (G 1 ) in the USO model was investigated in temperate forests. • G 1 displays considerable seasonal variations with a minimum value in mid-summer • Satellite-derived leaf area index (LAI) could well capture the seasonality of G 1 . • The LAI-based dynamic G 1 parametrization was used to modeling transportation (T c ). • The scheme of dynamic G 1 could effectively reduce the error in T c estimation. The ecosystem-level conductance-photosynthesis models, which represent a linearly coupled relationship between canopy stomatal conductance (G s ) and CO 2 assimilation, have been increasingly used for modeling transpiration (T c ). As a key parameter in these models, the slope parameter (G 1 ) has been observed to vary considerably over the seasons in the field, but is often parametrized with a biome-specific temporally constant G 1 , resulting in large potential uncertainty. Here we hypothesized that G 1 varies with leaf area index (LAI) phenology and soil water content (SWC) seasonality, and accurate characterization of G 1 seasonality offers an avenue to improve T c modelling. To test these hypotheses, we first investigated the seasonality of Eddy flux-derived G 1 and then explored its relationship with satellite-derived LAI and field-observed SWC seasonality at 12 temperate forest FLUXNET sites across the Northern Hemisphere. Last, we cross-compared the two schemes of model parameterization of G 1 for modeling T c : (1) a constant G 1 (FIX) and (2) a dynamic G 1 parameterized using the selected variables (DYN). Our results show G 1 displays considerable seasonal variations across all sites, with a minimum value in mid-summer. Further variance partitioning analysis demonstrates that the seasonal variations in G 1 show direct linkages with LAI phenology rather than SWC seasonality likely associated with leaf aging and ontogeny development. Last, we found relative to the FIX model, the DYN model (using LAI for G 1 parameterization) significantly reduced the model uncertainty in terms of RMSE by 24.6 ± 11.8% and 32.0 ± 8.7%, respectively for G s and T c at a daily scale. These results collectively improve our understanding of the dynamic pattern and proximate controls of G 1 seasonality, and highlight the effectiveness of using remote sensing-derived LAI phenology for improved characterization of G 1 seasonality that ultimately contributes to the improved process model simulations of the seasonal dynamics of G s and T c across temperate forest landscapes.