Improving physiological simulations in seasonally dry tropical forests with limited measurements

Semiarid regions and seasonally dry tropical forests play a critical role in global carbon exchange. In the Southern Hemisphere, these areas can contribute over 80% of positive global carbon storage in wet years due to their sensitivity to precipitation. It is therefore vital that climate models accurately represent land surface processes in these regions, including the contribution of vegetation seasonality to the water budget and carbon cycle. However, the simulation of phenological processes introduces new uncertainties associated with vegetation parameters, especially in biomes with a lack of field experiments, such as the Caatinga biome, a seasonally dry tropical forest. Furthermore, the global land cover maps used in land surface models and their associated parameters do not accurately reflect the diversity of vegetation in these regions. In this study, we improved the Noah-MP leaf area index simulation through parameter calibration and sensitivity analysis for the Caatinga biome in Brazilian semiarid region. To assess parameter uncertainty, we applied the generalized likelihood uncertainty estimation (GLUE) method with MODIS LAI as reference data. The best-performing models from GLUE improve the LAI simulation for BSA natural vegetation. The results indicated that the most sensitive parameters in LAI simulation are the field capacity, the specific leaf area index, and the leaf turnover rate. These parameters regulate water stress, conversion from leaf mass to LAI, and leaf carbon allocation to leaves. Additionally, it was observed that some vegetation parameters exhibit seasonal behavior, suggesting that allowing parameters to vary within the year could enhance the simulations.