An Experimental Monsoon Forecast For Water Managers In The Southwestern Us

Moisture from the North American Monsoon (NAM) can be a critical component of summertime water supply in the Southwestern US. Although key water management decisions are made in late spring and summer that could benefit from monsoon forecasts, there is a lack of robust forecasting products that provide actionable information on the time horizon needed to inform water planning and operations, and available seasonal forecasts tend to underpredict monsoonal precipitation. This is not surprising given the small-scale processes that contribute to monsoonal convection, which are not resolved at the coarse spatial scales at which most forecast models are run. One promising approach to improving seasonal monsoon forecasts is to identify large-scale atmospheric patterns that can be identified in the available forecast products and related statistically to precipitation. A collaboration between scientists at the National Center for Atmospheric Research (NCAR) and water managers at the Bureau of Reclamation (Reclamation) have used this approach to develop experimental monsoon forecasts for Arizona and New Mexico. We identified synoptic “weather types” that capture the recurring modes of NAM variability, and used hindcasting techniques to evaluate the skill of these weather types, as captured in available seasonal forecast products, for predicting monsoon precipitation in four sub-regions of the Southwest. We then developed statistical models that use the weather types to predict probabilistic precipitation characteristics. Finally, in a collaboration between scientists and water managers, we are assessing the value of the forecasts for streamflow prediction and water management. Applying the framework to a domain over Arizona and New Mexico, we find that monsoon season weather types based on synoptic-scale moisture advection are skillfully reproduced in spring forecasts by the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecast System. An online Google Colab Notebook has been developed to allow Reclamation to download real time ECMWF forecasts, assign the weather types, and to make experimental precipitation predictions using the statistical models that showed skill compared to climatology in the retrospective period. Finally, we generate conditional monsoon streamflow ensembles based on the precipitation predictions and assess the usability of the forecasts for ongoing operational processes and decision-making.