The spatio-temporal variations of the water budget components in the Amazon region are investigated by using a combination of hydrometeorological observations and moisture fluxes derived from the NCEP/NCAR reanalyses, for the period 1970-1999. The key new finding of this study identifies the major differences in the water balance characteristics and variability between the northern and southern parts of the basin. Our results show that there is a seasonality and interannual variability of the water balance that varies across the basin. At interannual time scales, anomalies in the water balance components in the northern Amazon region show relatively stronger links with tropical Pacific interannual variability. Over the entire region, precipitation exceeds evaporation and the basin acts as a sink of moisture (P>E). However, on some occasions the basin can act as a source for moisture (P<E) under extreme conditions, such as those related to deficient rainfall in northern Amazonia during the strong El Nino of 1983. Our estimates of the Amazon region's water balance do not show a closure of the budget, with an average imbalance of almost 50%, suggesting that some of the moisture that converges in the Amazon region is not accounted for. The imbalance is larger over the southern Amazon region than over the northern region, and it also exhibits interannual variability. Large uncertainties are detected in the evaporation and moisture-convergence fields derived from the reanalyses, and in the case of evaporation it can be as large as 10-20% when compared with the few field observations across the basin. Observed precipitation fields derived from station data and from grid-box products also show some discrepancies due to sampling problems and interpolation techniques. The streamflow observed at the mouth of the river is obtained after corrections on the series observed taken at a gauging site almost 200 km inland. However, variability in the evaporation, moisture convergence, and observed rainfall and runoff matches quite well.
