Skill and Usability: Subseasonal to Seasonal Predictability of 14-day Extreme Precipitation Periods in the CONUS

Abstract

Long periods of extreme precipitation can cause significant impacts to many economic sectors, including transportation, water resource management, agriculture production, and infrastructure. Planning for these impacts would ideally start between the typical weather (out to 10-14 days) and seasonal (1-3 months) forecast periods. This gap in forecasting periods, referred to as the subseasonal to seasonal timescale (S2S), is known for relatively low accuracy in precipitation prediction when compared to other extremes, such as heatwaves and drought. Thus, very few forecast products exist for stakeholder use for decision-making regarding S2S precipitation. This dissertation (1) quantifies the current prediction capabilities of atmospheric conditions associated with extreme precipitation within S2S models, (2) demonstrates how extra-tropical cyclones play a role in driving extreme periods, and (3) engages with stakeholder communities to learn what is needed for usable S2S products. Studies have shown that S2S precipitation forecasts have very low skill beyond 2 weeks. Yet studies quantifying the skill associated with atmospheric variables coincident with and preceding extreme precipitation periods have not been conducted. A database of 14-day extreme precipitation periods across the US was used to assess the current predictive skill of synoptic variables during the extreme periods for the first time. The highest skill is seen for geopotential height (representative of atmospheric pressure variations). Since the anomaly correlation coefficient drops below climatology by Week 2 for all variables, model bias during the extreme periods was examined. During the extreme periods, there is a dry bias that increases with lead time for specific humidity. The models also have an inability to realize the magnitude of 500hPa geopotential height dipole, common to all extreme periods. While, on average, skill is low past Week 2, there are cases where the precipitation period is driven by large-scale atmospheric conditions and has skill out to Week 4. Case studies of these 14-day extreme precipitation periods revealed that they are often created by multiple rounds of storms, which are associated with extratropical cyclone (ETC) activity. Studies have shown that ETC activity accounts for 50% of Northern Hemisphere precipitation; thus, understanding ETC activity during the 14-day extreme periods would help reveal future avenues of S2S prediction of long-duration precipitation extremes. Using a database of ETC tracks within ERA5 observational data, frequency, speed, intensity and event precipitation contribution of ETC activity during our extreme precipitation periods was analyzed and compared to non-event periods ETC activity. Characteristics of event ETC activity was highly dependent on the region. The Northern Plains events have slow moving, intense ETC that, on average, contributed more than 50% of the events precipitation. However, the West Coast, Great Lakes and Southeast have intense ETCs, that are more frequent during event periods than non-event periods. While the Northeast has many events with intense ETC activity, with higher precipitation contributions, the speed and frequency of ETCs was not significantly different than non-event periods. The Southern Plains did not have any differences between event and non-event ETC activity, with high variance of ETC precipitation contribution. Therefore, there is a strong connection to ETC activity within precipitation events in many regions that would need to be explored within S2S models in the future. Since S2S forecasts have had low skill, forecast products have been underutilized. This creates an opportunity to work alongside stakeholders to develop forecast products that are useful and usable. The Prediction of Rainfall Extremes at Subseasonal to Seasonal Periods (PRES2iP) project team conducted three workshops over six years to engage with stakeholders to learn what is needed for decision-making for subseasonal precipitation. The PRES2iP team used the knowledge gained from the first two workshops to design a suite of extreme precipitation subseasonal forecast products for the third and final workshop in April 2023. Participants identified that impact-based text products alongside quantitative precipitation forecasts helped to improve their interpretation of products at longer lead times and greater ranges of uncertainty. Participants also noted that regional time-series of precipitation were useful when a measure of uncertainty was also included in the product. Uncertainty tolerance relating to the subseasonal forecast location and size of the 14-day extreme precipitation period was also found using an object-oriented verification scheme. Overall, there was no single product that worked best, but instead, a combination of spatial and temporal products with associated narratives, including impacts is needed. This work helps to close the gap associated with S2S prediction of impactful extreme precipitation, while also engaging potential users throughout the research process to co-produce usable and trustworthy S2S precipitation forecast products.