Earth System Models (ESMs) predict increased frequencies of extremely wet and dry periods in the Amazon over the next century, resulting in very uncertain Amazonian carbon budgets. Because the water cycle strongly influences the carbon cycle and other climate processes and feedbacks, we contend that accurately estimating CO₂ and CH₄ emissions from upland and floodplains in ESMs requires progress on three fronts: improved hydrologic descriptions, aquatic biogeochemistry, and improved spatial scaling of hydrologic and biogeochemical processes. This project will address these three issues by developing in the Community Land Model (CLM) a multi-scale hydrological and biogeochemical modeling framework based on a subgrid-parameterization scheme and scale-aware downscaling techniques. This approach will bridge the gap between coarse-resolution and fine-scale hydrological and biogeochemical predictions. The goal of the multi-scale framework is to predict CLM gridcell-scale states and fluxes consistent with fine-resolution simulations at orders of magnitude lower computational cost. The hydrologic descriptions will be built upon a highly efficient, physically-based model (Process-based Adaptive Watershed Simulator; PAWS) that has been applied and tested in several temperate watersheds. This framework will greatly reduce uncertainties of hydrologic-biogeochemical simulations due to spatial scaling, enhance our simulation capabilities, and enable uncertainty quantification and multi-objective optimization at climate scales.  We will also integrate a novel aquatic ecosystem model based on previous studies of floodplain carbon dynamics. With these new process representations in CLM we will evaluate the impact on Amazonian carbon budgets of increased frequency of wet and dry periods. We will also develop our scaling approach in a generalizable manner so that our results will be applicable to global simulations over decadal to centennial time scales.