The predictability of hydrologic drought depends in large part on land surface memory; in particular, the capability to simulate the seasonal evolution of snow and soil moisture. The skill of seasonal drought predictions in the western USA is therefore intimately linked to the validity of hydrologic and land-surface models that are used to produce the predictions. The goal of this proposal is to improve predictions of hydrologic drought in the western USA; specifically to both improve model representations of snow processes and quantify uncertainty in snow simulations. We focus on snow because a large amount of predictability in seasonal streamflow in the western USA is derived from knowledge of the accumulated snowpack.
Our proposed research has four elements:
1) Develop and test a framework for improving model selection/specification in hydrologic modeling. This involves summarizing the different decisions regarding process selection and representation, and integrating multiple representations of all important snow processes into a common modeling framework.
2) Apply this framework to assess appropriateness of current modeling approaches, and provide recommendations on methods to improve existing snow models. Testing model hypotheses requires data that is not routinely collected in standard observing networks, and, as such, we propose to collate data from existing experimental studies. Using these data we will identify appropriate quantitative metrics to evaluate model decisions and identify a range (ensemble) of scientifically defensible model representations.
3) Quantify uncertainty in model simulations by allowing for ambiguity in the representation of different processes. Explicitly formulating the multiple working hypotheses at the level of model sub-components (rather than entire models) can mean that inter-model differences in our system provide a superior estimate of structural uncertainty than in multi-model systems that include a small number of individual models of varying complexity.
4) Incorporate these modeling approaches in the CHPS system to evaluate the value of physically realistic snow models for seasonal drought predictions. Use of flexible modeling approaches and multi-model approaches provide a thorough test of CHPS as an integrative modeling system.
This proposed project directly addresses the desired elements of the FY2011 call, specifically by (i) providing a capability to link multiple model components; and (ii) evaluating uncertainties based on model formulation.