Integrating Data Assimilation and Multi-modeling Within CHPS for Improved Seasonal Drought Prediction

MAPP supports developing an Integrated Drought Prediction Capability that incorporates research advances into operational intraseasonal to interannual climate and hydrologic prediction by means of multiple data sources and models, land-surface and hydrologic modeling, and data assimilation. Improved long-range hydrologic drought forecasts can help to mitigate drought impacts-they provide guidance to agencies responsible for water allocation, water conservation, and the mitigation of other adverse impacts such as wild land fires. With respect to drought characterization and prediction, the U.S. Drought Monitor (DM) and the NOAA Climate Prediction Center’s (CPC) US Drought Outlook (DO), are the primary operational tools available to water managers charged with dealing with drought contingencies. However, both the DM and DO depend on a range of tools from simple statistical indices to water balance models driven by statistical forecasts.

Remote sensing-based real-time snow extent fields can, in some regions, help characterize one of the key components used to define drought, which is critically important for identifying conditions that could flicker seasonal droughts. In operational forecasting, however, consideration has been given to methodologies that include coupling both multiple climate and multiple hydrological models, increasing the pool of streamflow forecast ensemble members and accounting for cumulative sources of uncertainty. Here, we propose a three-year collaborative research project that will quantify and reduce the major uncertainties involved in drought forecasting by implementing state of the art of land data assimilation methods and Bayesian multi-model combination in the context of seasonal hydrologic forecast system. We will focus in particular on those observed variables that have been shown to have the greatest potential for improving hydrological forecasts in the western U.S., specifically in situ observations of Snow Water Equivalent (SWE); remotely sensed observations of Snow Cover Extent (SCE) from MODIS and streamflow from USGS gauges. We will also investigate, on a more local basis, the potential for assimilation of in situ soil moisture from the USDA SCAN network given the difficulty and inaccuracy in using remotely sensed soil moisture fields for the western US.

The system we propose will use Community Hydrologic Prediction System (CHPS) as the framework to incorporate hydrologic and land surface models developed over the last three decades with methods of improving hydrologic initial conditions (i.e., land data assimilation) and its integration with Ensemble Streamflow Prediction (ESP) which are the key aspects of the proposed research. This proposal, which is a collaboration of Hamid Moradkhani at Portland State University (PSU) with Andy Wood at CBRFC and Donald Laurine at NWRFC responds to the second area solicited on the MAPP Information Sheet “Develop an Integrated Drought Prediction Capability by incorporating research advances in climate prediction, land-surface and hydrologic modeling, and data assimilation”. It will integrate past CPPA-funded research by the PI and others in the development and application of drought forecasting. In particular, project benefits from the participation of development and operation hydrologists at two RFCs to ensure the transition of research to operation in the western US with the transformable potential to other regions. Summary of work to be completed include the following tasks: 1) Establish domains for retrospective analysis. To limit the data volumes involved in proposed analyses, sub-basins or sub-regions of the domains will be identified. Criteria to be considered will include the availability of the observed verification datasets and the presence of drought; 2) Gather and pre-process in situ and historical remote sensing datasets. For the domains identified in Task 1, the necessary remote sensing products will be processed into convenient form for analysis and archived for further analysis; 3) Set up each hydrologic model for the study domains and implement the advanced model calibration to ensure the suitability and effectiveness of each model for further analysis; 4) Implement ensemble data assimilation to quantify the uncertainties associated with the land surface initial condition required for 1-12 months ensemble streamflow prediction; 5) Develop the multi-model ensemble combination of hydrologic forecasting for both soil moisture and streamflow; 6) Conduct retrospective assessment on the hydrologic/drought prediction skills using single and multi-modeling as developed in Tasks 4 and 5; 7) Incorporate the framework within the newly developed operational Community Hydrologic Prediction system (CHPS).