In the parched southwestern United States, few forecasts are as important as the future height of Lake Mead, which tells federal authorities how much water to release to the 20 million people living downstream of the giant reservoir. This year, the U.S. Bureau of Reclamation is testing out a new tool it hopes will make those projections a little better: A model that can predict - months in advance - the summer rainfall associated with the North American Monsoon.
The ability to forecast monsoon rains that far in advance has long eluded meteorologists. But if the new approach proves successful, the bureau believes it could lead to better summer projections of Lake Mead’s January water level - a key metric the agency uses to plan water releases during the coming year. With water levels in Lake Mead running a record low and the bureau implementing its first-ever water shortage declaration in 2022, even small improvements in these reservoir projections can make a difference.
The monsoon forecasting tool, detailed in a recent study published in the journal Advancing Earth, Space and Science, is one of several emerging efforts by researchers to predict key weather patterns, or weather-related hazards, months ahead of time. Other research teams are also rolling out experimental, long-range forecasts that can predict how many acres will burn during an upcoming wildfire season and where ocean heat waves will occur up to a year in advance, providing more explicit and comprehensive information than existing tools. Collectively, these new tools are pushing the boundaries of seasonal forecasting, a realm of weather forecasting that has traditionally focused on predicting patterns in temperature and rainfall.
“When we’re thinking about really long lead times, and the ability of models to forecast if there’s moisture available and the atmospheric patterns are in the right configuration, advancing beyond this traditional approach of looking myopically at temperature and precipitation is the way to go,” said Chris Castro, an expert on the North American Monsoon at the University of Arizona who was not involved with the new monsoon paper.
Runoff from snowpack in the Rocky Mountains supplies most of the water to reservoirs that feed the Southwest. But as drought persists, water managers have taken an increasing interest in the rain that falls over Arizona and New Mexico in the summer due to the monsoon, a seasonal climate pattern. But monsoon rainfall, which falls in bursts as thunderstorms quickly form and dissipate over mountains, is notoriously difficult to forecast with even just a few days’ lead time.
“Our modeling systems are very bad at simulating this kind of thunderstorm rainfall, especially in regions where you have mountains like the desert Southwest,” said Andreas Prein of the National Center for Atmospheric Research, lead author of the recent study.
To try to predict the rainfall further in advance, Prein and his colleagues turned to a weather model from the European Centre for Medium-Range Weather Forecasts. This model, the researchers found, can reliably simulate large-scale surges of atmospheric moisture from the Gulf of Mexico or the tropical Pacific over the southwestern United States several months ahead of time. Those moisture surges correlate with how much rain actually falls, in a given month, over a given drainage area. That allows researchers to predict how rainy it will be even though they can’t say exactly where or when the thunderstorms will form.
As the Bureau of Reclamation puts that new tool to the test this summer, a separate research team with the National Oceanic and Atmospheric Administration is developing seasonal forecasts for a climate hazard far offshore: Ocean heat waves.
Ocean heat waves, which like their counterparts on land are increasing in severity as the planet warms, can impact everything from the productivity of commercial fisheries to the migration patterns of endangered species. The ability to predict them far in advance would help resource managers prepare for these impacts - which is exactly what the NOAA scientists now say is possible with their new seasonal forecast tool.
“The high level finding is that in a lot of cases, these events are predictable,” said Michael Jacox, an oceanographer NOAA’s Southwest Fisheries Science Center in Monterey, Calif., who led the research published recently in the journal Nature.
Jacox and his colleagues compared 30 years of forecasts produced by six North American climate models with actual sea surface temperature measurements collected at the same time, and found specific ocean regions where heat waves are predictable months ahead of time. These include the eastern tropical Pacific, where heat waves are closely tied to the El Niño climate pattern and are predictable nearly a year in advance, as well as the West Coast of North America, where forecasts are possible up to six months out. In general, ocean regions influenced by El Niño and La Niña have heat wave patterns that are predictable several months out, while the western edges of oceans that are more influenced by rapidly changing currents don’t have such patterns, according to the research.
For predictable ocean regions, the authors believe a heat wave forecasting system could be readily developed because the underlying climate models are already being run every month.
Ocean managers are keen to see that system built.
Steven Lonhart, a research ecologist with the Monterey Bay National Marine Sanctuary, said long-range heat wave forecasts could help natural resource agencies monitor the health of species critical to their environment, like kelp, and prepare to assist heat-vulnerable animals like sea otters. Fisheries managers, meanwhile, could use heat wave forecasts to alter annual catch quotas or the timing of fishing season to account for how warm water will impact population numbers.
Wildfires, another hazard intensifying across the country, are also becoming more predictable at seasonal time scales. A recent study published in Environmental Research Letters found climate conditions in the winter and spring help drive the severity of the summer fire season out West, accounting for more than half of the year-to-year variability and trend in summer burned area.
“What we found is winter and spring climate sets the stage for summer fire activity,” said lead study author and NCAR scientist Ronnie Abolafia-Rosenzweig.
With that insight in mind, Abolafia-Rosenzweig and his colleagues developed an experimental tool for predicting, at the end of April, how much land in the West will burn between June and September. Researchers feed data on precipitation, temperature, atmospheric moisture and other climate metrics from the preceding winter and spring into the tool, which is based on machine-learning models. Based on these conditions, the models predict how much land will burn in a given summer, a forecast researchers can validate by comparing it with satellite-measured burned area.
This summer, their models are predicting 3.8 million acres of land will burn, a forecast the researchers will validate as the summer progresses. While still experimental, such predictions go beyond current operational fire season outlooks, which focus on whether the fire activity out West will be above or below average.
John Abatzoglou, a fire expert at the University of California, Merced, said the authors’ findings “add another tool to the toolbox” to help fire managers take preventive measures in advance of fire season. But Abatzoglou, who was not involved in the paper, said summer weather is of “paramount importance” to fire season severity - conditions not currently integrated into the tool.
Study co-author Cenlin He acknowledged that predicting and integrating summer weather, as well as summer activities that cause fires, into seasonal forecasts is a “very challenging problem and different groups in the scientific community are working on resolving this problem.”
All of these seasonal forecasting advances could benefit communities affected by weather and climate extremes - though it’s not the same as being able to say exactly when or where a specific weather event or hazard will occur with months of notice.
“The state of the science is not there yet,” said Jon Gottschalck, who heads the Operational Prediction Branch at NOAA’s Climate Prediction Center.
Still, Gottschalck said that the types of approaches the new studies are taking - mining data with machine learning; predicting the large-scale patterns that correlate with specific types of weather - could be important for advancing seasonal forecasting at large, with numerous potential applications.
For instance, the authors of the monsoon study are exploring whether their approach can be used to forecast winter snowfall over the Rockies several months out. Winter snowpack is “the leading factor” in annual runoff and water supply predictions, according to bureau hydrologist and study co-author Shana Tighi.
“Any tool that improves the forecasts for snow accumulation will be extremely useful for water management planning,” Tighi said.