Uncertainty of runoff sensitivity to climate change in the Amazon River basin.

We employ the approach of Roderick and Farquhar (2011) to assess the sensitivity of runoff (R) given changes in precipitation (P), potential evapotranspiration (E ), and other properties that change the partitioning of P (n) by estimating coefficients that predict the weight of each variable in the relative change of R. We use this framework using different data sources and products for P, actual evapotranspiration (E), and E within the Amazon River basin to quantify the uncertainty of the hydrologic response at the subcatchment scale. We show that when estimating results from the different combinations of datasets for the entire river basin (at Óbidos), a 10% increase in P would increase R on average 16%, while a 10% increase in E would decrease R about 6%.