An uncertainty framework for hydrologic projections in gauged and ungauged basins under non-stationary climate conditions

Streamflow is a major freshwater resource to mankind and also provides ecosystem services such as habitat to aquatic species. Increasing threats to global water security due to rising population, land use change and global warming poses a major challenge to water resource planners. To make sound management decisions, they need reliable estimates of future water resource availability. In this dissertation, we identify and address the challenges in obtaining continuous projections of future streamflow in order to assist water management decisions. With this practical aspect in mind, we also address the knowledge gaps in hydrologic modeling under large uncertainties, under a changing climate or land use and in data sparse regions of the world.;The general approach to derive projections of streamflow is to drive a hydrologic model whose parameters are calibrated on historical streamflow with projections of changed future climate obtained from global climate models (GCMs). The first issue in this approach is that parameters of hydrologic models display a dependence on the climate of the calibration period. This makes the future projections of streamflow using parameters calibrated on historical climate unreliable. We ask -- when and by how much does this climate dependence of model parameters impact the streamflow projections? We suggest a trading space-for-time framework to account for the climate dependence of model parameters and compare the projections derived from the proposed framework to the traditional method.;The second question is whether accounting for the climate dependence of hydrologic model parameters matter in the wake of large uncertainties in the climate change projections from the GCMs. We apply the trading space-for-time framework to the Olifants Basin in South Africa, a data sparse region. Using available downscaled climate projections, we show the large uncertainties in future streamflow projections and find that the uncertainty ranges are a function of the historical climate of the watershed and its future trajectory.;The trading space-for-time method relies on the ability of the hydrologist to identify streamflow signatures that can be regionalized, so we need to know which signatures to regionalize in a given location. In the third study, we use a diagnostic approach to identify the watershed characteristics and streamflow signatures that determine successful parameter transfer in different regions of the US, and hence should be preferentially regionalized.;In the fourth and final study, we answer a more general question -- how to assist decision makers when there are large unavoidable uncertainties in future projections of streamflow? Until we reduce uncertainties in climate and hydrologic models through better understanding of the system, we will require alternative methods to aid decision making. Using an exploratory modeling design, we analyze a wide range of climate and land use settings in a watershed to assess the critical thresholds of climate and land use change, after crossing which the watershed is likely to transition to vulnerable regimes.;This approach provides the decision maker an estimate of the vulnerability of a particular aspect of the hydrologic regime (magnitude, duration, frequency, or rate of change) to environmental change. After that, the available information on future climate change can be integrated to assess the plausibility of the watershed actually witnessing such changes. We also compare the results of this approach with the traditional approach to derive hydrologic projections and show they agree well with each other but the new approach provides more information to the decision maker.