A 50-Day Heat Wave Forecast, and the Future of Subseasonal to Seasonal Prediction

For those of us interested in the future of long-range weather forecasting, two developments this week pair very nicely. A paper published on Monday online in the journal Nature Geoscience shows how heat waves across the midwestern and eastern U.S. may be predictable with some skill as far as 50 days out. On its heels is the Tuesday release of a report from the U.S. National Academies, Next Generation Earth System Prediction: Strategies for Subseasonal to Seasonal Forecasts. The report argues that there is great potential to improve the quality and value of forecasts in the two-week to 12-month range, if the necessary resources can be marshaled--and if researchers can develop and tailor products designed to fit the needs of users.

The Pacific Extreme Pattern: A prelude to big heat in the central and eastern U.S.
The paper in Nature Geoscience--led by Karen McKinnon, a postdoctoral researcher at the National Center for Atmospheric Research--finds that a particular arrangement of sea surface temperature (SST) can take shape across the North Pacific weeks ahead of the advent of widespread summer heat over much of the central and eastern U.S. This oceanic configuration, dubbed the Pacific Extreme Pattern (PEP), features colder-than-usual SSTs along the west coast of North America, with a warmer-than-usual area north of Hawaii and another cold anomaly toward Japan (Figure 1).

The PEP bears some of the hallmarks of the negative (cold) phase of the Pacific Decadal Oscillation, which is known to raise the odds of drought and heat across parts of the United States. However, the PDO and PEP are separate beasts, according to McKinnon. “PEP has a smaller spatial scale and varies more within a season than the PDO,” said McKinnon in an email. “While in some cases SST anomalies may project similarly onto both patterns, we do not believe that they are generally the same phenomenon.”


Figure 1. Colored areas show anomalies (departures from average) in sea surface temperature associated with the Pacific Extreme Pattern (PEP) at the 40-day lead time, when it would suggest an enhanced risk of heat 40 days later over the central and eastern U.S. Overlaid in dashed and solid black contours are the Pacific Decadal Oscillation (PDO). While a negative PDO is defined by a large warm anomaly that spans most of the North Pacific (solid contours), with a cold tongue near the western U.S. (dashed contours), the PEP has a large cold anomaly in the western part of the ocean basin (blue colors), a smaller warm anomaly in the middle part of the basin, and then a second cold anomaly in the eastern part of the basin. Image credit: Karen McKinnon, NCAR.


Figure 2. The five clusters of stations across the U.S. identified as tending to experience hot days at the same time. The cluster examined in this study (blue stations] covers much of the central and eastern U.S. Image credit: Fig. S1(a) from K.A. McKinnon et al., Long-lead predictions of eastern United States hot days from Pacific sea surface temperatures, published online on March 28, 2016, in Nature Geoscience.


By exploring where unusually hot U.S. summer temperatures tend to cluster, the study team decided to focus on a large and important region extending from the Central Plains and encompassing most of the nation east of the Mississippi River and north of Florida (see Figure 2). As the PEP evolves in a sequence of steps identified by the authors, it tends to generate high pressure off the West Coast and, further downstream, dry, hot weather over the central and eastern U.S. On a regional basis, the paper defines a hot day as one where at least 5% of the study area (at least 80 of 1613 weather stations) experiences a high of at least 6.5°C (11.7°F) above average (one standard deviation). Using this yardstick, the evolution of the PEP provides significant skill at predicting the timing of hot days for the region as a whole up to 50 days in advance. Even at the individual-station level, the PEP demonstrates significant skill at about half of all locations out to 40 days, particularly across the Mississippi Valley.


Figure 3. SST anomalies in the North Pacific Ocean 50 days in advance of June 29, 2012. The pattern inside the green box best matches the early stage of the Pacific Extreme Pattern, indicating that there would be an increase in the odds of a heat wave in the eastern half of the United States at the end of June. Temperatures on June 29 (bottom) largely bore out the forecast, with readings above 100°F covering much of the Central Plains and mid-Mississippi Valley. Image credit: Karen McKinnon, NCAR.


Hints in May of heat in July
The authors make their case further through a retroactive “hindcast” of the scorching U.S. summer of 2012, which produced record heat and grinding drought across much of the study domain. The three biggest multi-day summer heat spikes of 2012 began on June 25, July 16, and 29 July. The state of the PEP on May 15 corresponded to a more-than-threefold increase in the likelihood of hot days 40 days later (June 24). At the end of May, the PEP indicated even stronger 40-day odds for a hot period around July 9.

Clearly, the PEP is not a perfect predictor, but it may serve as an useful new avenue toward probabilistic heat and drought forecasts over a key part of the U.S., with more specific timing than now offered by today’s leading techniques. “The Pacific Extreme Pattern appears to provide a cohesive framework for improving seasonal prediction of summer precipitation deficits and high temperature anomalies in the eastern U.S.” the paper asserts. “The identification of predictive skill at a seven-week lead time is an important advance over current seasonal forecast models that tend to under-predict the probability of extremes.” It’s possible that the PEP is associated with one or more factors that also influence eastern U.S. heat and dryness. The authors add that “it would be useful to better determine whether the ocean forces, feed backs on, or simply acts as a passive recorder of atmospheric anomalies in the months preceding hot weather.”

Dr. Todd Crawford, chief meteorologist at The Weather Company, said: “These results confirm the importance of North Pacific SST patterns in modulating summer temperature patterns over the US on seasonal time scales, and for the first time suggest predictability of extreme heat events on sub-seasonal time scales.  There is now the potential for the addition of another useful statistical forecasting technique to accompany dynamical model output in the sub-seasonal forecaster's toolbox.” Crawford, whose TWC Energy group carries out seasonal and sub-seasonal prediction for a range of customers, told me he’s being “peppered with client questions/comments” about the paper.

Going operational (experimentally)
We’ll soon find out how well this new technique works in real time. Based on daily PEP calculations, McKinnon and colleagues are planning to make predictions for this summer, beginning in early May. These forecasts will be available through a link to be posted at McKinnon’s website. “Our hope is that these can be used directly by, e.g., city leaders to ensure that enough cooling rooms are available for those without air conditioning if there are increased odds of a heat wave, and utility companies who may want to make sure they have sufficient power to bring online in case of spikes in electricity demand,” McKinnon told me. “We also hope to interface with operational and seasonal forecasters to see if the information from PEP could complement that provided by the current dynamical models.”


Figure 4. Leniel Fields of K&K Maintenance wipes his face in the heat as he trims and maintains the grounds at the Franklin School Apartments near downtown St. Louis on July 23, 2012. The city hit 106°F that day, with temperatures remaining above 80°F at night for three consecutive days. Image credit: AP Photo/St. Louis Post-Dispatch, Erik M. Lunsford.

A comprehensive strategy for improving forecasts up to a year out
The PEP study is one step in a direction encouraged by the National Academies report released on Tuesday. This report serves as an update to a similarly themed 2010 study, Assessment of Intraseasonal to Interannual Climate Prediction and Predictability, this time dropping the multi-year component and focusing on the interval from two weeks to 12 months in advance, a period dubbed S2S (seasonal to subseasonal). Another new tack is broadening the kinds of phenomena that might be predicted, encompassing extreme weather events such as the heat waves analyzed in the new paper above. The study examines recent progress in using such phenomena as the Madden-Julian Oscillation and the North Atlantic Oscillation as forecasting tools, along with such efforts as the North American Multi-Model Ensemble. “However, an associated U.S. national research agenda aimed at strengthening the contributions of S2S forecasts to public and private activities has not yet emerged,” the report noted.

The study’s vision--that “S2S forecasts will be as widely used a decade from now as weather forecasts are today”--includes 16 recommendations to get us there, as well as four research strategies:

• Engage users in the process of developing S2S forecast products
• Increase S2S forecast skill
• Improve prediction of extreme and disruptive events and consequences of unanticipated forcing events
• Include more components of the Earth system in S2S forecast models

“It is easy to envision the potential value of high-quality predictions two weeks to 12 months ahead for any number of industries--for example, energy, water resource management, and agriculture,” noted committee chair Raymond Ban (Ban and Associates) in the report’s preface. “Even if such information never matches the level of confidence associated with tomorrow’s weather forecast, it could still be used by individuals, businesses, and governments to plan and make a large array of important decisions.”


Figure 5. Severe-weather risk areas for Wednesday (left) and Thursday (right), March 30 and 31, 2016, as designated by NOAA’s Storm Prediction Center on Tuesday morning, March 29.

Tornadoes may spin up across Central/Southern Plains on Wednesday
We’ll be back with a new post by Thursday at the latest. We’re also keeping an eye on a fairly classic set-up for early-spring severe weather in the nation’s midsection, especially from around Kansas City to the Dallas-Fort Worth area. At midday Tuesday, NOAA’s Storm Prediction Center was calling for a slight risk of severe weather on Wednesday over much of the southern and central Great Plains, shifting on Thursday into the central Gulf Coast area. The chance of a major tornado outbreak did not appear large, given the moderate amounts of instability expected and a tendency toward southwest winds at most levels, which would tend to reduce vertical wind shear. Early-morning storms on Wednesday may also cut back on daytime heating. Still, it’s late March, and all of the ingredients should be present for the full gamut of severe weather, including tornadoes in some areas. SPC noted in its Tuesday update that the risk for Wednesday could be upgraded in subsequent outlooks.

Bob Henson