How Are Weather Forecasts Made?

Weather forecasts are so familiar to all of us that one might never pause to think how they are made. If your business is somehow effected by weather, it might be a good idea to learn some basics behind the weather forecasts. That way you will be in a better position, for instance, when evaluating the quality of potential weather providers.

Most people might think that weather forecasts are always done by a meteorologist. In fact, forecasting the weather has a lot more to do with computers than humans – even though humans are certainly involved. Weather forecasters have been dealing with Big Data long before the term was invented. For instance, Foreca processes hundreds of gigabytes of raw weather data every day, and the science today is more in choosing the right data sources and managing the data processing and evaluation methods, rather than a human trying to figure out the route of a low pressure system.

1. Measuring the Weather

Measuring the weather

To predict weather, we must first measure the current state of the entire Earth System. Measurements are made from space, in the air, on the ground and in the sea. Global coverage is required, since weather systems do not respect national borders and quickly travel from one continent to another.

Ground Observations

On the ground, national weather institutes maintain weather station networks. The stations measure the state of the ground and the lower atmosphere. The stations are standardized and the data is freely exchanged between all the nations of the world. The station locations are carefully chosen to represent the average climate in the region.

Observations from the Space

The importance of satellites has grown over the years. The public is probably familiar with the impressive hurricane animations seen in media, but satellites also have a more important role: taking measurements for initializing computer models. Modern weather satellites carry astonishing equipment that makes it possible to remotely measure a wide range of parameters such as sea surface temperature, snow coverage, amount of water vapour and even wind speed and direction. The most important feature of satellite data is uniform global coverage; satellite data can fill-in the holes in the sparse ground-based networks.

Observations from the Air

Measurements from the air are important because most of the atmosphere is beyond the reach of ground-based equipment. Larger commercial aircraft and weather balloons carry weather measurement instruments and send the information back to the ground.

Observations from the Sea

Sea surface temperatures have a substantial effect on weather. Measurements of water temperature, waves and other sea parameters have traditionally been made using ships and buoys. Weather buoys are expensive to maintain and are therefore relatively sparse. Since 70% of the Earth is covered by water, satellite measurements are crucially important, especially over the Seas.

2. Analysing the Current State of the Earth System

All the measurements are fed into a super computer, which uses them to form a picture of the complete Earth System, as it is “now” – the current state of the atmosphere, the sea and the ground.

It is possible to know the state of the atmosphere with certainty only at points where there is measurement equipment such as a weather station. Everywhere else, the computer relies on sophisticated physical analysis to get as probable a value as possible for every single point of air, sea and ground.

For example, the European Center for Medium-range Weather Forecasts (ECMWF) creates an analysis of the measurements in which the Earth’s atmosphere is split into air cubes of approximately 8 km in diameter, i.e. a grid of 8 km resolution.

3. Calculating the Global Forecast

Numerical weather prediction is the calculation of what will happen next based on what the weather is like now. The required physics and mathematics have been known since the early 1900s, but it was only after the invention of modern computers that numerical weather prediction became possible.

When calculating the forecast, the computer steps through time in tiny increments, e.g. 30 seconds. Starting from the initial situation based on measurements, the computer first makes a forecast 30 seconds into the future. Then, it uses that as a starting position and forecasts another 30 seconds, resulting in a one-minute forecast and so on. In each of these steps, the computer calculates how much air needs to be moved from one air cube to the next, how much moisture moves with the air, how much the sun heats the ground etc.

Errors gradually accumulate in the forecasting calculations, so the longer the forecast is run, the larger the errors become. Typically, the first three to five days can be predicted fairly reliably.

Did You Know?
Global weather forecast models are run only by a handful of organizations on some of the largest super computers in the world. The two most widely used global models are being run by the ECMWF in Europe and by the NOAA in the USA. The improvement of computers and forecast models over the years has resulted in a steady increase in model resolution and forecast quality. Forecast quality is not completely determined by the resolution, but the grid size determines the smallest phenomena that can be forecasted. At the moment (2017), the most accurate global forecast model is the ECMWF, which is run at the resolution of 8 km.

4. Improving the Global Forecasts

Global forecast models are calculated every six or twelve hours, and since the calculations take several hours, they are already a little out-of-date at the time of publication. There are also small-scale weather phenomena, which the coarse global model grid cannot represent. There are a few approaches for addressing these shortcomings of the global forecast.

Local High Resolution Weather Models

Global weather models are still limited to approximately 10 km resolution, even when using the fastest supercomputers, but it is possible to have very different weather in different parts of 10-by-10 km area. This is especially true next to large bodies of water or in mountainous areas.

Calculating a higher-resolution weather model inside the global model makes it possible to achieve a resolution of just a couple of kilometres or even better. Such models are called local models and most national weather services run a local model for their own country.

Foreca obtains many local models from all over the world and also calculates its own local models for important areas.

Statistical Methods

For locations where measurements are available, it is possible to keep track of errors in the forecasts. In addition to quality control, this information can be used to improve future forecasts.

For example, if the weather model forecasts for a location are always two degrees too cold, the forecast can be improved by adding two degrees to the weather model prediction. In reality, the errors are seldom this simple. Therefore, the statistical analyses must use a lot of data over a long time period in order to be able to find patterns in the errors.

Foreca uses sophisticated statistical analysis to reduce the forecast errors as much as possible by using all the available data from multiple weather models, both global and local, and all available measurement parameters. The system has been developed over the course of many years.

NowCasting

It is possible to fine-tune the model forecasts with measurements that have been taken after the model was initialized. This is called nowcasting, and it is an effective technique when forecasting the next couple of hours into the future.

If the location of a predicted rain cloud is off even by a couple of kilometres at the edge of the rain area, a forecast user on the ground will experience totally different weather – a failed forecast.

The Foreca Nowcasting System fuses radar, satellite and weather station measurements. Some institutions are even able to provide measurements every 5 minutes, so Foreca runs a new high-resolution nowcast several times per hour, improving the accuracy of the next hours’ forecast.

5. Finalising the Forecasts

The forecast needs to be presented in an easy-to-understand way for different uses. Foreca creates forecasts globally for consumers and for businesses. Business customers include car manufacturers, road winter maintenance operators, TV stations, newspapers, mobile application developers, energy companies and many others.

Point Forecasts

A point forecast shows what the weather is going to be like at a single coordinate point. It is often easier to understand than other types of forecasts, but does not offer as much information about the general weather situation as maps do. Typically, point forecasts are visualized as a graph or a row of weather symbols or formatted into a short text.

Forecast Maps

Forecast maps show what is happening in a larger area around the location of interest. This helps understand the general weather situation, which is useful, for example, for travel planning and for understanding the uncertainties in the forecast. Foreca offers weather maps as layers, which can be placed over a background map and animated. The weather layers are compatible with all industry standard map providers such as Google, Here or Bing. Foreca can also provide custom background maps which are designed to visually work together with the weather layers.

Weather Data APIs

Various businesses use weather forecasts to optimize their operations or to serve their customers better, who need weather data to integrate into internal systems. For these cases, Foreca offers weather data in XML, JSON or custom formats. Data transfer is possible both via on-line APIs and as push or pull batch transfers, using most of the common protocols.

In addition to weather data products, Foreca provides turn-key solutions for a variety of industries.

Foreca Weather Services