Predicting the weather – what’s involved?

by Motoko Kakubayashi

The philosopher Aristotle once mused that weather is the second greatest problem to man, second only to the study of human nature itself.  As long as humans have ventured outside, we’ve tried to anticipate weather changes.  From ancient hunters and farmers to modern commuters and the country’s GDP, our survival and economy has largely depended on the rhythmic arrival of sun, wind, and rain.

The first and foremost order of business when predicting the future weather is to understand what is happening to it now.  The practice of predicting the weather began with the Babylonians in 650 BC, but their methods were informal and relied on observing patterns in sunsets, clouds, or rainbows.  Today, these methods have been abandoned because of their statistical unreliability and because we know a lot more about the science of meteorology.

Formal weather forecasting began around the mid-1800’s when electric telegraphs were invented.  Telegraphs made it possible for people to report and collect atmospheric data such as pressure, rain and wind from locations over a large area, and then use scientific knowledge to predict how the weather would evolve.

Today MetService, a State Owned Enterprise, is responsible for weather forecasting and weather warnings in New Zealand on behalf of the Ministry of Transport.

Another company, the National Institute for Water and Atmospheric Research (NIWA), which is a Crown Research Institute, provides a scientific basis for the sustainable management and development of the country’s atmospheric, marine, and freshwater systems and associated resources.  NIWA’s contribution to weather prediction is to carry out research that will improve the accuracy of forecasts.

Keeping an eye on the weather

Until a few decades ago, atmospheric data was measured and reported mostly by trained observers including MetService staff, lighthouse keepers, and aircraft pilots.  Today, most surface weather observations are carried out by automatic weather stations 24 hours a day.  There has also been a huge increase in additional information from weather satellites and weather radar networks providing data over much larger areas at all times of the day .

Satellites have become a very important tool in weather observation.  Their global view from above makes them useful for watching what weather is moving where.  Geostationary satellites orbit with the earth, remaining at the same position above the equator and provide at least hourly images of the earth’s disc beneath them.  Polar orbiting satellites orbit north and south over the poles as the earth turns beneath them, so their view is always changing, and global in coverage.

The main geostationary satellite used for New Zealand is the Japanese MTSAT-1R satellite.  This spacecraft carries sensors (called radiometers) that can sense radiation reflected from and emitted by the Earth and the atmosphere at different wavelengths. The visible light sensor uses visible wavelengths to capture detailed snapshots of the skies from space.  This allows the texture and shape of clouds to become clearly visible and identifiable.  However, as with any normal camera, the satellite relies on the sun’s light to reflect off the earth to capture an image, making night time images impossible.

So meteorologists also make use of imagery from infrared (IR) wavelengths, which can be used both day and night.  At wavelengths where the atmosphere does not itself radiate, IR radiation is stronger from warmer surfaces, such as the Earth’s surface and clouds at lower levels, while it is weaker from higher and colder clouds, so IR images provide valuable information about clouds and their height both day and night.  At other IR wavelengths the atmosphere itself radiates, and so it is possible to make images of quantities such as water vapour in the air, which can then be used to deduce the amount of water vapour in an area.

Polar orbiting satellites provide the most useful information for improving weather forecasts, since they carry radiometers that operate in both in the infrared and microwave and provide detailed information about the temperature and water vapour structure of the atmosphere – both horizontally and vertically.  In the infrared the sensors measure the radiation emitted by carbon dioxide and water vapour molecules.  In the microwave spectrum they measure emissions primarily from oxygen and water vapour.

These instruments are carried on satellites with names like NOAA15, NOAA16, NOAA17, NOAA18, Aqua, MetOp-1A.  The instruments on these satellites scan the atmosphere below the spacecraft as they orbit around the Earth approximately every 100 minutes, and provide thousands of spectral observations for every 20 by 20 km region of the Earth’s surface and the atmosphere above it.  These satellites are the great secret of weather prediction, and have led to weather forecasts over the largely ocean covered southern hemisphere that are as good as or better than those available for the largely land covered northern hemisphere.  NIWA has satellite receivers in Wellington and Lauder that receive direct broadcast data from these satellites as they pass over New Zealand.

Another important weather observing tool is weather radar.  Unlike satellites, which look down on weather conditions from high above, radars look outwards and upwards at the weather from ground stations.  Currently, MetService operates five weather radars across the country: Mount Tamahunga (near Warkworth), New Plymouth Airport, Outlook Hill (near Wellington), Rakaia (southwest of Christchurch) and Invercargill Airport, and has plans for future expansion.

These weather radars send out microwave beams, and if precipitation (rain, snow or hail) is within the radar’s range then a radar echo returns to the source.  The data is processed every few minutes, so locating precipitation and tracking its motion and development is carried out efficiently.

The most common problem with radar is that large barriers such as high mountains block the microwaves.  Another problem is that because of the curvature of the earth, a radar beam sent out horizontally will be well above the surface by the time it is 100 or 200 km away, so it misses precipitation that is below the beam.

Model weather

Computer weather models start with an initial snapshot of what the weather is like over the whole globe or a limited area.  The process of estimating this snap shot is, called data assimilation (DA) and it combines information from a previous model forecast with the latest observations from satellites and weather stations.  Once that step is complete the forecast model predicts the weather anywhere from the next few hours (short-range forecasts) to the next few weeks (long-range forecasts).

Because the weather is controlled by the laws of physics and fluid dynamics, its evolution can be expressed as a series of mathematical equations.  These non-linear partial differential equations describe how the atmosphere will change from its initial conditions (current weather), and are extremely complex.  To make the highest resolution solutions about how the weather changes globally or in a limited region requires a supercomputer to work through the calculations faster than the actual weather evolves.  In fact, the world’s most sophisticated supercomputers are needed to solve these kinds of complex environmental forecasting problems.

Currently, MetService uses weather information from three global weather models: the UK’s Met Office Unified Model (UM), US Global Forecast System (GFS), and European Centre for Medium-Range Weather Forecasts (ECMWF).  These numerical weather prediction models provide data about global weather, some of which is freely available online for the public and sharing among the world’s meteorological centres.

Scientists at MetService use the initial conditions produced by the global models to initiate their limited-area model forecasts.  The limited-area models are based on the Fifth-Generation Pennsylvania State University/National Center for Atmospheric Research mesoscale model (MM5) and the Weather Research and Forecasting (WRF) model.  These models look at the finer details of weather within a small area and take into account local geography more than global models are currently able.  The result is an accurate picture of the future state of the atmosphere over New Zealand.

Local DA is currently being tested experimentally by MetService and is due to become operational in the next year or so.

Scientists at NIWA have developed the New Zealand Limited Area Model (NZLAM) – which is based on the UM.  NZLAM is a data assimilating model that makes full use of all local observations, both satellite and ground based, leading to the best possible estimate of the initial conditions.  Therefore, improved forecast accuracy NZLAM uses about half of NIWA’s supercomputer every 6 hours, making accurate forecasts out to 48 hours ahead.  Weather forecasts from this model are also being used to forecast many weather impacts – from river flood to storm surge and damaging waves.

Improving a model

The results from weather models however, are seldom 100% correct.  Problems arise for several reasons, although the most obvious is because the Earth’s atmosphere is chaotic.  As popularly described in the butterfly effect of chaos theory, minor changes in initial conditions (well within the limits of how well we can observe the weather over the whole globe) for a weather model can grow into large differences in model forecasts over a few days or weeks.  But as our understanding of the atmosphere increases, and our computers get better, scientists are continually able to create new or improved weather models, which incorporate a better understanding of uncertainty rather than just giving one “best” forecast.

A very common way to improve the spot forecasts from weather models is to calibrate their output using real observations.  For example, if one model was predicting the next day’s temperature and produced an outcome that on average was 1.5°C lower than the actual temperature observed the next day, a more accurate forecast could be produced by adding 1.5°C to subsequent model predictions.

Since 2005, researchers at MetService have been working on a 15-day model-derived forecast system along these lines, but using much more sophisticated statistical methods.  Supported by the New Zealand government’s Technology for Business Growth, the project was completed in June 2008 and will replace the existing system by early 2009.

The system makes optimal statistical combinations of many different computer weather models, providing the best single forecast (calibrated to actual weather observations) as well as a probability distribution for the likely outcome, taking into account the uncertainties inherent in all the models.

The Reducing the Impacts of Weather Related Hazards (RIWRH) research programme being carried out by scientists at NIWA aims to safeguard lives and reduce economic losses to the country from weather hazards such as flooding or landslides.  Largely funded by the Foundation for Research Science and Technology, it is developing the next generation of advanced environmental forecasting systems for New Zealand.

NIWA’s research is focusing on the fundamental factors that contribute to weather forecast errors.  This has meant putting effort into improving the methods used to estimate the initial conditions (i.e. DA) used by the forecast model, and in particular on making best use of the polar orbiting satellite data described above.  NIWA scientists are also conducting research on ultra high resolution weather prediction models.  The spacing between the grid points of these models can be as small as 333m in the horizontal (NZLAM currently uses a 12 km grid spacing), and 76 levels in the vertical. These models require vast supercomputing resources to run, but can be used to better simulate significant weather hazards such as tornados, very heavy rain, and damaging down-slope winds.

EcoConnect is another outcome of RIWRH, and is NIWA’s initiative to provide true multi-hazard forecasting for New Zealand.  Their unique weather model NZLAM provides information for other weather hazard models such as wave forecasts, river flow, landslides, storm-surge, and snow avalanches.  The final outcomes from all of these models are available on EcoConnect.

Similarly to MetService, NIWA has also been conducting research on 15-day forecasts, producing probabilistic forecasts of weather variables for around 140 sites across the country for several years.

As a member of the World Meteorological Organization (WMO), a specialized agency of the UN, New Zealand is also part of an international research program, the Observing System Research and Predictability Experiment (THORPEX).  The aim of this program is to accelerate improvements in one-day to two-week high-impact weather forecasts.

The future

The way forward looks bright for weather prediction.  Weather models are predicted to continually improve while satellites, radar and weather stations improve in number, quality and range.

New Zealand scientists are already experimenting with four-dimensional DA (3 spatial dimensions plus time dimension).  NIWA scientists are carrying out the research needed that would allow models to make even better use of observations than the present three-dimensional DA approach used in NZLAM.  However four-dimensional DA also requires vast amounts of super computing resources, so its introduction must await a computer upgrade.

Similarly NIWA researchers expect to begin experimenting with an ensemble of high resolution NZLAM models which will lead to improved methods for estimating the errors in forecasts.

The human role in meteorology will also change.  Although it seems increasingly inevitable that all weather data obtained will become accessible online for the general public, the meteorologist’s role will remain intact as an interpreter, advisor, and generally someone to better inform you about what you should expect when planning to head off on that outdoor adventure.

Useful Links:


NIWA – From weather prediction to forecasting hazards.

Natural Hazards. and


World Meteorological Organization.


MTSAT-1R satellite.

European Centre for Medium-Range Weather Forecasts.

US National Weather Service.

UK Met Office.

Weather measurement instruments.

This article was reviewed by Chris Webster, Manager Meteorological Advice & Training at the Meteorological Service of New Zealand Limited (MetService), Dr Neil Gordon, General Manager Science R&D at MetService and Dr Michael Uddstrom, Principal Scientist: Meteorology, Environmetal Forecasting and Remote Sensing at the National Institute of Water and Atmospheric Research (NIWA).