For years, scientists have warned of heightened risks of extreme storms due to climate change, andthose storms have already begun devastating communities.
Weather forecasting will play an increasingly important role in helping prepare communities for dangerous weather, but how are extreme snow and rain events forecasted? Research by professors Ania Panorska and Tom Kozubowski in the Department of Mathematics and Statistics, along with colleagues from Scripps Institution of Oceanography (Scripps), provides a new predictive model for extreme rain and snow events as well as the return periods (the average amount of time between events) and return levels (amount of precipitation) of extreme events. The study waspublished last month in Scientific Reports.
Previously, precipitation was predicted using hourly or daily models, which, for heavy precipitation events that last longer than a day, may not be particularly useful or accurate.
“Standard modeling techniques for extreme precipitation or storms use a fixed time frame (e.g. 5 days, 30 days, etc.) for analysis,” Panorska said. “However, duration of actual storms does not conform to these arbitrary timeframes.”
The researchers developed a model that uses a more holistic approach to defining storms as they actually are. The approach used in this research is called trivariate event distribution, or TED. TED produces a more accurate assessment of rain or snow accumulation probabilities and takes into account duration, maximum intensity and the overall magnitude of weather events over the course of the actual duration of an event, instead of the typically fixed timeframe such as 5 or 30 days. The model also incorporates the possibility of observations with wide variability of daily downpour.
“TED allows us to model whole physical events, rather than temporally isolated data points or aggregations over prescribed, arbitrary durations,” Panorska said. “It allows us to assess the probabilities, return levels, and return periods of the most relevant constituents of an event—duration, maximum intensity, and overall total magnitude—independently, simultaneously, or conditional on each other.”
Using data from the National Oceanic and Atmospheric Administration’s Global Historical Climatology Network-Daily database, the researchers tested their model on data from the western U.S. and found that it was a good fit for 87% of events.
“Although our present application of TED is to precipitation, it can be applied to many other events including heat waves, floods, as well as growth or decline events of financial markets,” Panorska said.
Kozubowski and Panorska will continue developing the model with their doctoral students Robert Chastain, Erick Luerken, and Tanner Miller to include truncated observations, covariates, and develop its spatial version. The professors also have master’s students and undergraduates contributing to the model.
“The project provides an excellent opportunity for the statistics and data science students to work on real world problems and tackle questions of current impact,” Panorska said. “We are continuing thecollaboration with researchers at Scrippsto make sure that the models we develop are consistent with observations, science, and needs of the public. It is exciting multidisciplinary teamwork!”
The research was supported by the California Department of Water Resources Atmospheric River Program and included researchers from Wroclaw University in Poland.