Amanda Maxwell

Jun 10th 2020

From Clouds to Cloud: Weather Forecasting for Long-Range Prediction


Even in the 21st century with all its technical wizardry, the weather is still difficult to predict correctly more than a week out. However, this is changing; as meteorologists move from observation into the world of big-data modeling, they increase the accuracy of this tricky prediction science.

Benefits of Accurate Prediction

Humans have always wanted to push the range of long-term and accurate weather forecasting. Knowing what the weather is going be has important consequences beyond wardrobe and leisure activity choices — public safety, agriculture and business rely on accurate forecasts. Benefits include more efficient renewable energy usage from enhanced solar and wind forecasts. The National Renewable Energy Laboratory found that a 25 percent increase in accuracy for solar forecasting could save $46.5 million from effective and efficient energy-grid management. Pushing the range of tornado warnings, even by a few more minutes, would make a difference for residents trapped in their path, allowing them more time to get to the shelters and survive. Even building design benefits from better long-range forecasting.

Where Do the Data Come From?

Weather forecasting is part scientific analysis, part pattern recognition and part chance due to the chaotic, unpredictable nature of the Earth’s atmosphere and the “butterfly effect” of one remote system on another. Historically, forecasting relied on ethno-meteorology that combines observation and local knowledge; farmers familiar with subtle cues such as cloud formation would know to harvest before a storm. Although experience and observation still matter today, the instruments used to gather data such as radar and satellites are a lot more sophisticated.

  • Radar is getting more sophisticated in the way it operates. Dual-polarization beams send out radio pulses both vertically and horizontally. Where previously only one beam scanned for data, phased-array systems in combination with dual-polarization reduce operating time with multiple beams in action. It now takes less time to scan a slice of atmosphere; as little as 30 seconds down from 6-7 minutes. Furthermore, since phased array systems also work in tandem with Doppler radar, they gather more information. For example, the systems can detect weather speed and direction, and discriminate raindrops from hailstones. This kind of information gives meteorologists a lot of predictive power when assessing tornado risk or imminent flooding.
  • Remote Sensing also gathers large volumes of weather-systems data. Drones and other unmanned reconnaissance vehicles fly their sensors right into storms and weather fronts to relay information safely back to operators on the ground. One advantage for remote sensing vehicles, for example, is that they can change course and respond to developing areas of interest.
  • Satellite Surveillance is the next layer in data gathering for weather forecasting. Both geostationary and orbital satellites operated by various nations and private enterprises scan the Earth on a daily basis, sending back data on weather systems, water indices, ocean currents and more. The NASA Earth-sensing satellite mission, Aqua has been gathering water-cycle statistics for just over 15 years. Four of its six sensors are still working beyond their intended six-year lifespan, collecting data on ocean currents, sea ice, water vapor, phytoplankton and more.
  • Crowdsourcing weather data isn’t a fantasy anymore. With the rise of smartphones, GPS and onboard systems in cars, and other devices all connecting back to the Internet of Things, each of us could be a weather station relaying news back to a central hub. Of course, there’s an app for that too — in fact several. Even NOAA (National Oceans and Atmospheric Administration) is in on the action with mPING, which gathers data from citizen scientists. Since people can see at ground level what the radar data streams cannot, the smartphone info feeds back into observational radar data to fine tune weather forecasting ability.

Complex Data + Supercomputing = Weather Forecasting Models

Bigger volumes of data switch weather prediction from probabilistic to certainty through powerful computer modeling. Although human-driven pattern recognition — where meteorologists choose the best interpretation — still plays a role, computer modeling is taking over and extending the range.

  • Models and algorithms, digital databases and RSS feeds: To crunch all these numbers, it’s no surprise that meteorologists need supercomputing to stay ahead in weather science analysis. One result is numerical weather prediction (NWP), which uses current conditions to forecast or predict future weather events. However, since weather follows a “butterfly effect,” it’s not enough just to follow local events; the NOAA Global Forecast System gathers world weather data. Not only can computational, or “in silico,” modeling handle much greater volumes of data faster than a human can crunch the numbers, they also combine the different streams from multiple sensors and GPS locators to build more powerful models.
  • Analysis and visualization: Researchers are looking for new ways to analyze the weather data to give more meaningful results. For example, the USGS considers fractal mathematics to be better for hurricane prediction.
  • Forecasting the weather: What good is a weather forecast if nobody knows about it? Speedy communications maintain the flow for both analysis and publishing the results. Federal weather monitoring agencies pull digital database streams from sensors automatically. They then feed the analyses out to public forecasting systems by RSS feeds, webpages and other rapid communication systems as fast as it is produced.

Predicting this chaotic system more accurately and further into the future is possible thanks to more data and better data handling. Future exploitation, as IBM suggests, means linking all of these sources of weather data for real-time analysis and output. In other words, the future of weather is indeed in the cloud.