MIAMI — The modern era of weather forecasting as we know it today began in the 1980s, when computer models became trustworthy enough to become the cornerstone of producing a forecast. At the WSI Commercial Aviation User Group in Andover, Massachusetts, The Weather Company’s Senior Vice President of Global Forecasting Services, Dr. Peter Neilley spoke about what he called the time-honored forecasting paradigm, and how modern technology has helped airlines to mitigate irregular operations caused by weather.
Thirty years ago, weather forecasts were updated two to four times a day, and precise within a range of “Tens of kilometers.” But since 2005, availability of data has exploded 10-fold every five years. That amounts to a 10 million percent growth of weather data availability over the past thirty years! Forecasts are now being run hourly, instead of 2-4 times per day. Forecasts are now precise to within just a few kilometers. The availability of “Big Data” has led to what WSI calls Velocity, Volume, and Variety. Velocity is the frequency at which the forecasting models are updated. Volume refers to the number of available weather stations collecting data. Variety speaks to the fact that there are nearly a dozen forecasting models routinely available to forecasters. “Big Data” enables forecasting precision, but it breaks the legacy forecast paradigm. There is now so much data available, that forecasters have to limit what they look at.
Nowadays, weather data is collected by computers which break down the data and try to replicate human cognitive decisions. It is important to remember that forecasts favor the models that have been more historically accurate recently. There is now so much data available that automated forecasts are more skillful and accurate that the legacy-based human approached forecasts of the 1980s. On-demand forecasts are now the largest application within the Amazon Cloud, generating up to 250,000 forecasts per second, based on user location. The Weather Company CEO David Kenny said Cloud computing has dropped the costs of on-demand forecasting by 95 percent, and mobile user numbers have risen 200 percent in the last two years.
Forecasts take only eleven milliseconds to generate, and only 300 milliseconds from the time they’re requested to when they are produced on the user’s smartphone. WSI reports that the all-time peak number of on-demand weather forecasts for a single day is 26 Billion. Considering the fact that there are roughly 7 billion people on Earth, we come to the realization that millions of users are requesting multiple forecasts per day.
How does this benefit airlines? By using WSI forecasting applications during flight planning, pilots can plot routes around inclement weather or between convective cells. This holds implications for airlines on both the operational and financial sides. Operationally, the ability to foresee bad weather allows pilots the opportunity to plan a flight route that may help them avoid holding patterns or diversions. Turbulence avoidance also reduces the potential for injuries to passengers and crew members. On the financial side, having an accurate weather forecast helps pilots reduce the fuel consumption by planning a more direct route and reducing the need for post-turbulence maintenance inspections and offline time for the aircraft. WSI reports that for May 2015, its PilotBrief app was used over 300,000 times.
Dr. Neilley said that for the future, forecasters will have to filter the input data. A smart front end plus human intervention will be the key. Human forecasters will manage the interpretation of the data provided by computer models, or what they call the Humans Over the Loop (HOTL) model, versus being in the loop. In aviation, the forecast paradigm will be adopted to an ensemble approach that risk-based forecasts. This probabilistic approach will incorporate the element of a chance happening, instead of the old deterministic method, which used only known data.