How accurately can we predict the weather?
The process of predicting weather patterns is a very complicated science. It requires the need to analyze and decode massive data sets gathered from thousands of sensors and weather satellites every day.
Identifying patterns in collected data to predict the future is a very strenuous task. For best results, it also needs to be done in real-time.
But like any kind of forecast, weather forecasting is something of an educated guess. Since we cannot control the weather, the best meteorologists can do, is to use past and present data and patterns to attempt to predict the future.
This is especially true to provide information for disaster events.
The accuracy of weather predictions has increased over time, but it is still not 100% accurate. According to some estimates, a seven-day weather forecast is about 80% reliable.
Shorter timescales are more so, with a five-day weather forecast about 90% correct. Anything longer than seven days, especially ten-day forecasts or longer tend to be only about 50% accurate.
As the atmosphere is constantly changing, estimates over long periods have proved to be very difficult to model and predict.
Meteorologists achieve this by using computer programs called weather models to make these forecasts.
What do meteorologists use to predict the weather?
Meteorologists use a variety of sensors, satellites and computer models to predict future weather patterns. Most people tend to be familiar with basic instruments like thermometers, barometers, and anemometers for recording temperature, air pressure, and wind speed respectively.
But they do also employ more sophisticated pieces of equipment like weather balloons. These are special balloons that have a weather pack on them that measures temperature, air pressure, wind speed, and wind direction in all the layers of the troposphere.
Radar systems are also used by meteorologists to measure precipitation around the world.
Some of their most powerful tools are environmental satellites like NOAA, the National Oceanic and Atmospheric Administration, which operates three types of environmental satellites that monitor Earth’s weather.
One of these is polar-orbiting satellites. Satellites as part of NOAA’s Joint Polar Satellite System (JPSS) orbit approximately 500 miles (805 km) above Earth.
These satellites constantly orbit the Earth from pole to pole up to 14 times a day. The combination of the Earth's spinning about its axis and the satellites quick orbits enable every part of the planet to be monitored twice a day.
This enables the satellites to provide enormous data sets about the Earth's entire atmosphere including clouds and oceans at very high resolution. Using this kind of data, meteorologists are able, in theory, to predict long-term weather patterns.
These satellites have a variety of instruments on-board that record information on the planet's albedo (or reflected radiation).
This data is very useful for making assessments on air quality over time. This information is incorporated into weather models, which in turn leads to more accurate weather forecasts.
Other instruments can also be used to map sea surface temperature—an important factor in long-term weather forecasting.
This data can then be used to help predict the weather including large-scale seasonal changes like El Nino and La Nina. They also collect data vital to helping forecast severe weather patterns like hurricanes, tornadoes and blizzards days in advance.
Data is also used to help assess environmental hazards like droughts, forest fires, and harmful coastal waters.
The next type of satellite used by meteorologists is called deep space satellites. For example, NOAA’s Deep Space Climate Observatory (DSCOVR) orbits one million miles (1,609,344 km) from Earth.
These kinds of satellites provide space weather alerts and forecasts whilst also monitoring solar energy absorbed by the Earth every day. DSCOVR is also able to record information about the Earth's ozone and aerosol levels in the atmosphere.
How is AI being employed to help predict the weather?
The enormous data sets required and inherent unpredictability of the Earth's atmosphere makes predicting future events very tricky indeed. Current computer models are required to make judgments of several large-scale phenomena.
These include things like how the Sun heats the Earth's atmosphere, how are pressure differences affect wind patterns and how water-changing phases (ice to water to vapor) affect energy flow through the atmosphere.
They also need to consider the Earth's rotation in space which helps churn the atmosphere throughout the day. Any tiny change in one variable can profoundly change future events.
This fact inspired MIT meteorologist Edward Lorenz to coin his now famous phrase "The Butterfly Effect" back in the 1960s. This refers to how a butterfly flapping its wings in Asia could drastically alter the weather in New York City.
Today, Lorenz is known as the father of chaos theory. Because of this Lorenz believed the maximum limit to accurate weather prediction is likely somewhere in the order of two weeks.
But this is where AI could be employed to improve the accuracy and reliability of weather forecasting. AI can be used to use computer-generated mathematical programs and computational problem-solving methods on vast data sets to identify patterns and make a relevant hypothesis, generalizing the data.
Given the inherent complexity involved in weather prediction, scientists are now using AI for weather forecasting to obtain refined and accurate results, fast! By using deep learning mathematical models, AI could learn from past weather records to predict the future.
One example is the Numerical Weather Prediction (NWP). This model studies and analyses vast data sets from satellites and other sensors to provide short term weather forecasts and long term climate predictions.
Other companies are also currently investing heavily in AI weather prediction. IBM, for example, recently purchased The Weather Company and combined its data with their in-house AI development Watson.
This led to the development of IBM's Deep Thunder which provides customers with hyper-local weather forecasts within a 0.2 to 1.2-miles resolution.
Monsanto has also been investing in AI for weather forecasting. Monsanto's Climate Corporation is used to provide agricultural weather predictions.