MIT Open Access Articles
Advances in weather prediction
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation: Alley, R.B., K.A. Emanuel and F. Zhang. “Advances in weather prediction.” Science, 365,
6425 (January 2019): 342-344 © 2019 The Author(s)
As Published: 10.1126/SCIENCE.AAV7274
Publisher: American Association for the Advancement of Science (AAAS)
Persistent URL: [Link]
Version: Author's final manuscript: final author's manuscript post peer review, without
publisher's formatting or copy editing
Terms of use: Creative Commons Attribution-Noncommercial-Share Alike
Submitted Manuscript: Confidential
Profitable Prophecy: Advances in Weather Prediction
R.B. Alley1*, K.A. Emanuel2 and F. Zhang3
1
Department of Geosciences, and Earth and Environmental Systems Institute, The Pennsylvania
State University, University Park, PA 16802, USA
2
5 Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of
Technology, Cambridge, MA 02139, USA.
3
Department of Meteorology and Atmospheric Science, and Penn State Center on Advanced
Data Assimilation and Predictability Techniques, The Pennsylvania State University, University
Park, PA 16802, USA.
10 *Correspondence to: 517 Deike Building, University Park, PA 16802, USA, rba6@[Link] .
Rapid advances in weather-forecast accuracy are paying large benefits for society, with potential
for even greater payoffs from targeted investments.
15 Abstract. Weather forecasting provides societal benefits that greatly exceed costs. Advances in
forecasting have been rapid, arising from improved observations and models, and better
integration of these through data assimilation and related techniques. Further improvements are
not yet constrained by limits on predictability, and can be enabled by targeted investments in this
successful public-private partnership. Better forecasting in turn can contribute to a wide range of
20 valuable environmental forecasting, from forest-fire smoke to bird migrations, improving well-
being in numerous and sometimes surprising ways.
1
Submitted Manuscript: Confidential
Eighty years ago, an intense hurricane struck New England without warning, killing more than
600 people. Since then, even though coastal populations have swelled, death tolls have dropped
dramatically.
Many people and organizations contributed to this improvement, from first-responders to
5 architects and drafters of zoning codes. But, as the American Meteorological Society prepares to
celebrate its 100th anniversary, the improvement in forecasting stands out. Modern 72-hour
predictions of hurricane tracks are more accurate than 24-hour forecasts just 40 years ago (Figure
1a), giving sufficient time for evacuations and other preparations that save lives and property.
The authors routinely hear jokes about prediction errors from people who wouldn’t dream of
10 planning an outdoor activity without first checking the weather forecast. Those forecasts from
leading numerical weather prediction centers such as NOAA’s National Centers for
Environmental Prediction (NCEP) and the European Centre for Medium-Range Weather
Forecasts (ECMWF) have been improving rapidly—a modern 5-day forecast is as accurate as a
1-day forecast in 1980, and useful forecasts now reach 9-10 days into the future (Figure 1b) (1).
15 Predictions have improved for a remarkably broad range of hazardous weather conditions,
including hurricanes, blizzards, flash floods, hail, and tornadoes, with skill emerging in
predictions of seasonal conditions.
A 2009 study found that US adults obtain weather forecasts an astounding 300 billion times per
year, from a great range of sources (2). That study also found the value of weather forecasts to
20 US households is $31.5 billion, from public expenditures of just $3.4 billion and private
expenditures of $1.7 billion, yielding benefits more than 6 times costs (2). Studies in other
nations similarly have found that the investment in weather forecasting pays large dividends,
ranging from 3 to 10 times costs (3). The value of weather predictions is further confirmed by
the rapid growth of businesses purchasing customized commercial forecasts from the private
25 weather enterprise, which is increasing in capitalization, employment and innovation (4).
Why forecasts are improving
Three key developments enabled these advances in forecast skill: Better and more extensive
observations, better and much faster numerical prediction models, and vastly improved methods
30 of assimilating observations into models. Remote sensing of the atmosphere and surface by
satellites provides valuable information around the globe many times per day. Much faster
computers and improved understanding of atmospheric physics and dynamics allow greatly
improved numerical prediction models, which integrate the governing equations using estimated
initial and boundary conditions.
35 At the nexus of data and models are the improved techniques for putting them together. Because
data are unavoidably spatially incomplete and uncertain, the state of the atmosphere at any time
cannot be known exactly, producing forecast uncertainties that grow into the future. This
“sensitivity to initial conditions” can never be overcome completely. But, by running a model
over time and continually adjusting it to maintain consistency with incoming data, the resulting
40 physically consistent predictions can greatly improve on simpler techniques. Such data
assimilation, often done using four-dimensional variational minimization, ensemble Kalman
filters, or hybridized techniques, has revolutionized forecasting.
2
Submitted Manuscript: Confidential
Sensitivity to initial conditions places strong limits on long-term forecast skill—the details of
weather cannot be predicted accurately, even in principle, much beyond roughly two weeks. But
weather forecasts are not yet strongly constrained by this limit, leaving optimism for continuing
forecast improvement (5). Sensitivity to initial conditions varies greatly in space and time, and an
5 important but largely unsung advance in weather prediction is our growing ability to quantify the
uncertainty by using large ensembles of numerical forecasts starting from slightly different but
equally plausible initial states, together with perturbations in model physics.
Several features of the weather system are notably more persistent than day-to-day weather,
allowing accurate predictions further into the future with even greater scope for improvement.
10 For example, the Madden-Julian Oscillation (MJO) moves eastward around the Earth’s tropics
over 30-90 days, affecting rain, wind, clouds and air pressure, the onset and demise of summer
monsoons (6), and more, with important agricultural and other implications. El Nino/La Nina
(ENSO) variability, with a typical 4-6-year recurring cycle and varying intensity, significantly
affects economies around the world by impacting most weather-dependent human activities, so
15 increased predictive skill could be used in myriad ways to improve well-being (7).
In parallel with improving forecasts, communication of the growing wealth of weather data has
expanded greatly, enabling a timely flow of ever more detailed and accurate information to a rich
diversity of users. Only a few decades ago, one had to wait for the morning newspaper or the
evening news to get the latest forecast, and warnings were delivered mostly by flags, sirens, and
20 police bullhorns. Today, detailed weather information is available at the touch of a finger on a
smartphone, which is rapidly becoming a vehicle for instant, geographically targeted weather
alerts.
Opportunities for Further Improvement
25 Weather-forecast improvement is the essential first step that will enable valuable improved
predictions of a great range of related environmental phenomena. For example:
àHurricanes Harvey and Florence, among others, highlighted the immense damages from
precipitation-driven flooding. With over 40 million people in the US living in the 100-year
floodplain, far more than previously believed (8), the value of improving flood forecasts is high,
30 requiring improvements in forecast models to better predict flooding from hurricanes (9) and
other sources.
àStorm surge, so important in events including Superstorm Sandy, is increasing with sea-level
rise, but depends sensitively on tides, wind, and atmospheric pressure interacting with the
detailed coastal configuration, requiring dedicated forecasting driven by accurate weather
35 predictions.
àSummertime sea-ice loss is opening the Arctic to shipping, recreation, resource extraction and
other activities, but seasonal sea-ice regrowth heightens dangers; recent work shows bright
prospects for sea-ice forecasts extending more than a month into the future (10).
àWildfire activity has been high recently, with important air-quality impacts. NOAA is
40 developing the High Resolution Rapid Refresh-Smoke (HRRR-Smoke) system to provide timely
warnings to people vulnerable to health impacts, and to aviation and other affected groups.
3
Submitted Manuscript: Confidential
àWeather drives many changes in animal behavior, which can be anticipated by weather
forecasters to good effect. For example, forecasts of bird migrations can now “…reduce
collisions with buildings, airplanes, and wind turbines; inform a variety of monitoring efforts;
and engage the public” (11).
5 àAs renewables come to play an increasing role in power systems, forecasting the availability
of sun, wind, and river flow will take on increased importance, as will forecasts of energy
demand, a large part of which is driven by weather.
Modeling centers are increasingly integrating efforts across timescales, from short-term weather
to climate. Forecasters are progressing rapidly at the subseasonal to seasonal time scales, on
10 MJO (12) and other phenomena. In the US, additional efforts by NOAA to improve forecasts at
the subseasonal to seasonal scale were legally mandated in 2017.
Climate models have shown skill in projecting many changes, including global mean surface
temperature, and the rise of absolute humidity with the associated increase in especially intense
precipitation when conditions are favorable. The models, however, remain challenged to project
15 regional shifts. While one cannot guarantee that improved weather forecasts will lead to
improved regional projections, progress seems likely (13).
Ways forward
To take advantage of the growth opportunities listed above, and more closely approach the limits
20 of predictability for weather and associated hazards at various temporal and spatial scales, the
entire community can strategize research and investments. The following areas are likely to be
especially fruitful:
àMaintain and improve data collection, targeting regions and times of special interest, and
assimilating targeted observations into models. For example, satellites and UAV observations
25 can focused on hurricanes, providing better predictions of their tracks and intensities.
à Improve understanding of key physical processes that are currently not well represented in
numerical prediction models (e.g., air-sea and cloud-aerosol interactions). Explore the potential
use of big-data science including machine learning and neural networks to identify model
uncertainties, perform bias corrections, and automate the forecast process.
30 à Further develop advanced data assimilation techniques, to more accurately and efficiently
assimilate existing and future observations. For example, cloud radiances from satellites are not
commonly utilized at appropriate scales. Data assimilation can be applied more broadly to data-
model integration and model development.
àUse advances in computer technology to produce higher resolution forecasts with larger
35 ensembles, leading to improved probabilistic forecasts, including those of hazardous weather.
This can be realized if governments maintain a steady schedule of investment in high-speed
computing, recognizing the strong evidence that such investments will be repaid many times
over in savings to the economy.
àAccelerate research and development of operational weather forecasting by improved
40 integration with academic research and education, providing smoother career tracks and
improved incentives for bright young researchers to bring their talents to the enterprise.
4
Submitted Manuscript: Confidential
àBring advanced communication methods to bear on the important task of delivering weather
forecasts. Our ability to quantify forecast uncertainty in probabilistic terms has arguably outrun
our ability to communicate such forecasts of uncertainty, but increasing engagement of
communication experts is leading to much improved ways of communicating probabilistic
5 weather information to those who need it.
With strategic investments, the future of weather forecasting and related environmental services
is bright, as public-private partnerships improve our well-being.
5
Submitted Manuscript: Confidential
References
1. P. Bauer, A. Thorpe, G. Brunet, Nature 525, 47 (2015).
2. J.K. Lazo, R.E. Morss, J.L. Demuth. Bull. American Meteorological Society 90, 785 (2009).
3. A. Perrels, Th. Frei, F. Espejo, L. Jamin, A. Thomalla. Advances in Science and Research 10,
5 65 (2013).
4. National Weather Service Enterprise Analysis Report, [Link]
Final_NWS%20Enterprise%20Analysis%20Report_June%[Link] (2017).
5. K. Emanuel, F. Zhang, J. Atmospheric Sciences 73, 3739 (2016).
6. Taraphdar, S., F. Zhang, L.R. Leung, Z. Chen, O.M. Pauluis, Geophys. Res. Lett. (early online
10 release) [Link] (2018).
7. P. Cashin, K. Mohaddes, M. Raissi, J. International Economics 106C, 37 (2017).
8. O.E.J. Wing et al., Environ. Res. Lett. 13, 1 (2018).
9. F. Zhang, Y. Weng, Bull. American Meteorological Society 96, 25 (2015).
10. L. Zampieri, H.F. Goessling, T. Jung, Geophys. Res. Lett. (early online release)
15 10.1029/2018GL079394 (2018).
11. B.M. Van Doren, K.G. Horton, Science 361, 1115 (2018).
12. H. Kim, F. Vitart, D.E. Waliser, J. Climate, (early online release,
[Link] (2018).
13. G. Krinner, M.G. Flanner, Proc. Natl. Acad. Sci. U.S.A. 115, 9462 (2018).
20
6
Submitted Manuscript: Confidential
Acknowledgments: We thank colleagues at the American Meteorological Society and
elsewhere. We acknowledge partial support from the National Science Foundation under grant
OPP-1738934. All authors contributed equally to the content and writing of this perspective. We
know of no competing interests. No original data were used in this Perspective.
5
7
Submitted Manuscript: Confidential
Figure 1a. NOAA National Hurricane Center forecast errors for Atlantic basin tropical storms
and hurricanes, showing the rapid improvement in accuracy.
[Link]
5 [Link]
8
Submitted Manuscript: Confidential
Day 7 NHem Day 3 NHem
500hPa geopotential height Day 7 SHem Day 3 SHem
Anomaly correlation Day 10 NHem Day 5 NHem
12-month running mean Day 10 SHem Day 5 SHem
(centered on the middle of the window)
100
90
80
70
%
60
50
40
30
1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017
Figure 1b. Improvement in forecast accuracy over time. The correlation between forecasts and
subsequent weather is shown for forecasts 3, 5, 7 and 10 days into the future, for northern and
southern hemispheres. A value of 60% is generally considered useful, and 80% is highly
5 accurate. Updated from (1).
[Link]
ined&time=2018101100