UC Berkeley

The Amazon basin is vulnerable to extreme drought events that can severely impact the local economy and ecosystems, with consequences for the global climate and ecology. However, hydrologic droughts have been mostly studied in the context of semi-arid environments and temperate regions and little is known on droughts on tropical, wet environments such as the Amazon. Also, the limited hydrologic data have been an obstacle in advancing the understanding of Amazon hydrologic systems and their response to droughts. This work employs new data analysis and stochastic simulation methods to reveal important features of droughts in the Amazon basin.

First, spatial-temporal patterns of hydrologic droughts are identified from existing streamflow data, using principal component analysis and Monte Carlo simulations to account for the uncertainty due to missing data. Second, hydrologic droughts are investigated at the watershed scale. A conceptual model is developed to quantify groundwater and subsurface water storages and fluxes during droughts, using existing hydrologic data at four representative Amazon watersheds. Then, the functioning of the typical deep soil columns and deep-rooted vegetation systems during droughts is further examined at the point scale. A one-dimensional, soil-vegetation hydrologic model is coupled with a stochastic rainfall and evapotranspiration model, calibrated with hourly data from conventional gauges and eddy covariance towers, to compute statistical characteristics of water balance components: evaporation of intercepted water, transpiration, deep percolation, and change in soil water storage. Finally, the dependency of evapotranspiration on cloudiness and soil moisture is analyzed using statistical hypothesis testing and stochastic simulations of rainfall and evapotranspiration, using observations of heat fluxes at the land surface, obtained from eddy covariance towers.

Results show distinct modes of spatial-temporal variability of droughts, with marked differences among sub-basins, especially between northern and southern rivers. A significant trend towards more intense droughts is found in the southern sub-basins. Correlations of drought indices with climatic anomalies originating from the Pacific and Atlantic Oceans are significant, but do not explain all the temporal variability of the major patterns of droughts. That suggests that hydrologic processes at smaller scales might play important roles in shaping drought characteristics. At the watershed scale, the propagation of rainfall deficits through the hydrologic systems highlights the importance of Amazon subsurface hydrology in promoting delays of timing, amplification of duration and attenuation of magnitude of deficits in subsurface, groundwater, and river systems. When annual rainfall is below average, the greater release of watershed water storage contributes to streamflow, helping to attenuate deficits. Moreover, propagation of streamflow deficits from upstream to downstream river portions affects drought characteristics in some rivers. At the point scale, the simulations of the soil-vegetation system reveal that, when rainfall is scarce, the drainage out of the bottom of the deepest soil layer is severely reduced and soil water storage decreases, while evapotranspiration is less impacted. The relationship between deep percolation and average soil moisture is found to vary in the wet and dry seasons. Finally, evapotranspiration is found to be more dependent on cloudiness, rather than soil moisture. The association of rainfall deficits with increased evapotranspiration can potentially exacerbate droughts, although the magnitude of that effect is not large at the interannual time scale. Those results highlight the potential of statistical and stochastic simulation methods in the study of the hydrology of data-scarce regions. Furthermore, they identify and quantify important hydrologic processes that shape the magnitude and duration of droughts in wet environments, which helps to better assess the vulnerability of such sensitive systems.