| This research analyzes hydrological variability in association with climate variability and climate change; and using this information it develops a statistical forecasting framework which is then integrated with a decision support system to manage water resources efficiently. This research has three major components: (i) Climate diagnostics of hydroclimate variables; (ii) Development of a statistical forecasting framework; and (iii) Evaluation of water resources decision strategies. As part of the first component, interannual variability of various hydroclimate variables is studied and large-scale climate features that drive the variability are diagnosed. The second component develops a framework for ensemble seasonal streamflow forecasts using the large-scale climate information obtained from the diagnostics. This uses nonparametric methods (assumption free and data driven) in association with an objective criterion, Generalized Cross Validation (GCV, selects a suite of best models) to issue ensemble forecasts of seasonal streamflow at several lead times (i.e., one month to five months at monthly intervals), and at several locations simultaneously. In the third component, forecasted streamflow ensembles drive a water resources Decision Support System (DSS) and the skills of various decision variables are evaluated. The DSS used is RiverWare, a flexible tool in which operational policies are easily incorporated.; The utility of this integrated climate diagnostics-to-DSS framework is demonstrated by successfully implementing it on the Gunnison River Basin, in Colorado, USA. This basin is in a semi-arid climate and has several competing water resources including agriculture, water supply, energy and environmental considerations. The framework can also be used for future planning and management of water resources in the basin. This offers an attractive tool for integrated water resources management and planning, especially in the western US, which is severely water stressed. Last, but not least, a methodology is proposed to generate realistic streamflow scenarios using paleo reconstructions (tree ring based reconstructions of past streamflows) in conjunction with the limited observational record. This approach provides the ability to generate a richer variety of drought and flood scenarios than those observed in recent history. It, thus, provides useful insight into streamflow variability, which is crucial for water resources management and planning in the basin. |
