Long-term simulations of discharge and floods in the Amazon Basin Michael T. Coe, 1 Marcos Heil Costa, 2 Aure ´lie Botta, 1 and Charon Birkett 3 Received 13 April 2001; revised 21 August 2001; accepted 18 December 2001; published 23 August 2002. [1] A terrestrial ecosystem model (integrated biosphere simulator (IBIS)) and a hydrological routing algorithm (HYDRA) are used in conjunction with long time series climate data to simulate the river discharge and flooded area of the Amazon/ Tocantins River basin over the last 60 years. Evaluating the results of this modeling exercise over the entire basin yields three major results: (1) Observations at 121 stations throughout the basin show that discharge is well simulated for most tributaries originating in Brazil. However, the discharge is consistently underestimated, by greater than 20%, for tributaries draining regions outside of Brazil and the main stem of the Amazon. The discharge underestimation is most likely a result of underestimated precipitation in the data set used as model input. (2) A new flooding algorithm within HYDRA captures the magnitude and timing of the river height and flooded area in relatively good agreement with observations, particularly downstream of the confluence of the Negro and Solimo ˜es Rivers. (3) Climatic variability strongly impacts the hydrology of the basin. Specifically, we find that short (3–4 years) and long (28 years) modes of precipitation variability drive spatial and temporal variability in river discharge and flooded area throughout the Amazon/Tocantins River basins. INDEX TERMS: 1833 Hydrology: Hydroclimatology; 1860 Hydrology: Runoff and streamflow; 3322 Meteorology and Atmospheric Dynamics: Land/atmosphere interactions; 9360 Information Related to Geographic Region: South America 1. Introduction [2] The Amazon/Tocantins River system of South Amer- ica is the largest river system on the planet. It covers about 6.7 million km 2 and transports about 20% of the world’s river discharge. Although the Amazon Basin is relatively undisturbed today, rates of land conversion are increasing rapidly throughout the basin [Nepstad et al., 1999; Skole and Tucker, 1993; Skole et al., 1994]. Additionally, increas- ing atmospheric CO 2 concentrations threaten to alter the water budget through changes in temperature and the physiological responses of plants. Therefore, it is important to gain a clear understanding of how the Amazon River system behaves on seasonal to interannual timescales in order to gauge how future changes may impact the water budget of the basin. [3] Water balance and water transport models provide a means of investigating the water balance of the Amazon Basin because they are able to derive spatially and temporally consistent estimates of the energy and water budget from simple climatological data (such as precip- itation and temperature). Previous modeling studies of the Amazon Basin include a study by Vorosmarty et al. [1989], which used water balance and water transport models at 1/2 degree spatial resolution to demonstrate the feasibility of large-scale simulation of the mean dis- charge and flooding in the Amazon Basin. That study compared the simulated discharge to observations at six locations within the basin. Costa and Foley [1997] used a coupled land surface and water transport model (also at 1/ 2 degree spatial resolution) to simulate the discharge of the basin and compare it to 56 discharge locations throughout the basin. [4] Recently, a number of new long time series data sets for model input and validation have become available. In addition, more powerful computers have made higher resolution, time-transient simulations possible. Therefore, the objective of this study is to simulate the hydrology of the Amazon River basin at 5-minute horizontal resolution (about 9 km) and to evaluate the simulations with diverse data throughout the Amazon River basin. This study is an extension of previous simulations in the resolution and complexity of the models used, the time-transient nature of the simulations, and the spatial extent and diverse range of data used for evaluation. [5] To simulate the river discharge and seasonal flooding throughout the Amazon River system over the last 60 years we use the integrated biosphere simulator (IBIS) [Kucharik et al. , 2000] and the hydrological routing algorithm (HYDRA) [Coe, 2000] with long-term mean monthly climate data provided by the Climate Research Unit of the University of East Anglia, Norwich [New et al., 2000]. We JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. D20, 10.1029/2001JD000740, 2002 1 Center for Sustainability and the Global Environment, Gaylord Nelson Institute for Environmental Studies, University of Wisconsin, Madison, Wisconsin, USA. 2 Department of Agricultural Engineering, Federal University of Vic ¸osa, Vic ¸osa, Minas Gerais, Brazil. 3 ESSIC, University of Maryland at College Park, Greenbelt, Maryland, USA. Copyright 2002 by the American Geophysical Union. 0148-0227/02/2001JD000740$09.00 LBA 11 - 1