Climate Science


by IPCC Expert Reviewer Dr Vincent Gray

17TH MARCH 2014


Climate science began with philosophical writings in India around 3000 B.C. and writings on oracle bones from the Shang Dynasty in China 2000 BC. In 350 BC Aristotle described the hydrologic cycle in his book, Meteorology and the Greek scientist Theophrastus compiled a book on weather forecasting, called the Book of Signs

Measurement instruments such as rain and wind gauges, barometer, thermometer, hygrometer were added over the years. Networks of weather observations were set up in Italy as early as 1654. Joseph Henry at the Smithsonian Institute in the USA set up a United States network in 1849.

In Britain, Francis Beaufort developed his Wind Force Scale in 1806. He set up weather notation coding and organised reliable tide tables around British shores.

Admiral FitzRoy Barometer
Antiques Atlas
His protégé Robert FitzRoy was appointed in 1854 as chief of a new department within the Board of Trade to collect weather data at sea from ship captains using loaned tested instruments.

The first daily weather forecast in the “Times” appeared in 1860. The following year a system was introduced of hoisting storm warning cones at principal ports when a gale was expected. Fitzroy invented a barometer which still occasionally turns up on the Antiques Roadshow.

Meteorological Offices and weather stations were established all over the world. Observations were exchanged using the electric telegraph.

The first professional meteorologist to obtain a university post was Sir George Simpson, who was appointed lecturer in the University of Manchester in 1905. He accompanied Captain R F Scott expedition to the South Pole in 1910-13 and he was Director of the British Meteorological Office from 1920 to 1938.

The quantity of climate observations became so large that it was only with the advent of the computer in the 1950s that they could be organised into semi empirical models to provide a reliable forecasting service.

The climate depends on the behaviour of fluids, of the atmosphere and of the oceans. There is currently no comprehensive theoretical physics of fluids and the best that can be done involves the use of non linear equations with second order differential quantities.

Edward Lorenz at

showed that if this method is applied to the climate a very slight error in the boundary conditions (for example, the movement of a butterfly’s wing) would be escalated by the equations, making a long term forecast impossible. Lorenz concluded that for the climate the prediction of the sufficiently distant future is impossible by any method.

Meteorologists use a variety of computer services from all over the world and compile numerical models at several different levels of complexity for different purposes. Some of these are free on the internet, others may be sold commercially.

Modern weather forecasting involves a combination of computer models, observations, and knowledge of trends and patterns. Using these methods, reasonably accurate forecasts can be made, up to about five days in advance. Beyond that, detailed forecasts are less useful, since atmospheric conditions such as temperature and wind direction are very complex.

The accuracy of weather forecasts has been recently studied by
Ripley E A & Archibold O W 2002 Accuracy of Canadian short and mebium-range weather forecasts Weather 57 (12)448-45 at

They give a very useful summary of previous studies. They say
The present limit of deterministic weather predictability is a few weeks at most (Hoskins and Sardeshmukh 1987; Ripley 1988). The major limiting factors are incomplete knowledge of the atmosphere’ s initial state (Gilchrist1986) and imperfect understanding of atmosheric processes (Somerville 1987). The first factor is most important at short lead times while modelling errors become the dominant limitation in longer forecasts (Anthes and Baumhefner 1984).
Their own study covers the entire Canadian system for the year 2000. They say
In this analysis, we assess the accuracy of short- and medium-range forecasts of minimum and maximum temperatures and precipitation for lead times of 1 to 5 days, issued daily by the Meteorological Service of Canada (MSC, formerly the Atmospheric Environment Service) for selected cities during 2000.
Their detailed results are tabulated They found that temperature forecasts had a bias of about ±1ºC and were rarely better than ±2ºC.

The results are plotted in the following graphs:

Fig. 3  Distribution of errors in minimum and maximum temperature
forecasts at 7 Canadian sites in 2000

The British Met Office at comes to a similar conclusion. They say:
Forecasts for both maximum and minimum temperature are compared to the actual values observed at an agreed list of 119 sites across the UK. The sites used for verification are those where we have quality-controlled data and where we produce forecasts for. 
The early morning forecast on our website is used to produce a percentage number of the times when the forecast is accurate to within +/- 2°C. This is based over a rolling 36-month period to smooth out extremes and give a representative averageTemperature forecast - performance
Temperature forecast - performance 
This information will be updated every month. 
Maximum temperature - first day of forecast

  • 93.8% of maximum temperature forecasts are accurate to within +/- 2°C on the current day (36-month average).
  • Target for 2013/14 is 85.0%.
       Minimum temperature - first night of forecast

  • 84.3 % of minimum temperature forecasts are accurate to within +/- 2°C on the first night of the forecast period (36-month average).
  • Target for 2013/14 is 80%. .
Climate is essentially local. Numerical climate models have to be supplemented with specific local characteristics to provide a reliable local forecast. The Global Climate is an array of the climate properties at each individual locality, as influenced by more general features. It is portrayed to us every day by the TV weather forecasts.

An example may be viewed at

Since Global climate properties have to be obtained at present from an assembly of individual local determinations there is no possibility of a plausible global climate model which could be capable of taking these into account.

The best forecasting accuracy for maximum and minimum temperature for each locality is rarely better than a couple of degrees Celsius for a few weeks ahead at most. Global climate properties could not possibly be more accurate or further into the future than this.

Any scientific organisation or individual scientist who claims the ability to predict the temperature change of the local or global climate beyond a few weeks, even to decimals of a degree, has to be dismissed as deluded or even fraudulent.