Interpreting the model results requires consideration of uncertainties, includin

Interpreting the model results requires consideration of
uncertainties, including the input assumptions of the
analysis (ie uncertain trends in water withdrawals and climate, which determine future river discharge and stress).
We therefore analyzed and report results from two contrasting cases for 2050, based on two distinct sets of
socioeconomic driving forces and climate projections
(see Alcamo et al. 2007). Estimates are also uncertain
because of the uncertainty of the underlying model
(WaterGAP). Based on the “goodness” of the model calibration (see WebFigure 1), the uncertainty of estimated
river discharge is “medium” in approximately 43% of the
world’s large river basins, “lower” in 32%, and “higher” in
25% (WebFigure 1). Using stochastic simulation, Kaspar
(2004) quantified the parameter and input uncertainties
of WaterGAP and determined that the latter had a larger
impact on estimates of future river discharge than parameter uncertainty. Using qualitative ranking based on
goodness-of-fit and other criteria, Alcamo et al. (2003b)
estimated the geographic variation of different types of
model uncertainty. Collectively, these analyses suggest
that we have covered a substantial part of the expected
uncertainty by examining two contrasting cases of socioeconomic driving forces and climate projections.
Furthermore, while the reliability of the results is lower
for some regions, the output is sufficient for the coarsescale analysis required to explore trends at a global scale.