Development and application of large-scale hydrologic and aquatic carbon models to understand riverine CO2 evasion in Amazonia

Many researchers are investigating the topic of CO₂ efflux to the atmosphere from waters of the Amazon basin at several scales. We are developing a physically based modeling system to simulate this flux throughout the whole basin as a function of time-transient climate, vegetation, and hydrology. This modeling system includes an ecosystem land surface model (IBIS; Foley et al. 1996, Kucharik et al. 2000), a hydrological routing model (HYDRA; Coe 2000, Coe et al. 2002), and a new aquatic carbon processing module that we are incorporating into HYDRA (Howard et al. in prep). HYDRA has been recently modified to better represent river discharge and flood extent and height throughout the Amazon Basin. These modifications include: 1) using empirically derived equations representing stream width and height at flood initiation (Costa et al. 2002) to provide more accurate estimates of the initiation and cessation of flood conditions; and 2) using spatially explicit river sinuosity data (Costa et al. 2002) and a stream velocity function based on the Manning equation to provide more realistic representation of stream flow timing and magnitude. HYDRA has been calibrated and validated with observations of river discharge, water height, and flooded area at numerous locations in the mainstem and headwaters of the basin. Results of this validation show better agreement with observations than the previous version of HYDRA but also indicate the need for improved land surface topography and precipitation datasets. The aquatic carbon processing module prototype is currently implemented as an aspatial STELLA/textregistered model, decoupled from HYDRA, that simulates individual grid cells (at ∼ 9 km resolution). We drive the model with IBIS-derived hydrological inputs from the land, and with empirically derived estimates of C inputs (from CAMREX and LBA sources). To allow for seasonal fluctuations in the aquatic-terrestrial transition zone, for each timestep we simulate the volume of water contained in each of four chemically-distinct zones in the grid cell: pelagic (open water), littoral (near-shore), floodable lowland, and terra firme (upland). With this information the model simulates the dynamics among six different pools of aquatic C: autotrophs; coarse (CPOC) and fine (FPOC) particulate organic carbon; dissolved organic carbon (DOC); dissolved inorganic carbon (DIC); and sediment. The amount of CO₂ efflux from the water surface is calculated for the grid cell in each timestep. We identify 9 environments that are hydrochemically distinct at this coarse scale: the Amazonas mainstem downstream of the Rio Negro; small, medium, and large whitewater rivers; small, medium, and large blackwater rivers; whitewater floodplain/lake environments; and blackwater floodplain/lake environments. We use the aspatial prototype of our aquatic carbon model to determine the CO₂ efflux from each of these different environments. We use imagery classified by Hess et al. (2003) to determine the area of lowland Amazonia falling into each of these categories. Finally, we use these model results and data to extrapolate the aquatic CO₂ efflux across the Amazon basin, and then put our exploratory results in the context of previous and on-going studies in this area.