David Parker, Climate Monitoring Expert
We need long-term worldwide observations of the atmosphere, oceans and land surface to understand the world's climate and how it has changed.
There are many challenges which have to be tackled when we are observing the climate, including:
Incomplete geographical coverage of measurements.
Gaps in historical climate records.
The need to use some indirect measurements of climate change.
Biases and errors in data.
Varying standards for taking observations.
Collecting information to assist interpretation of climate records.
Calculating changes in climate.
Estimating uncertainties in climate observations.
A major obstacle in assessing past climate change has been the fact that a lot of observations aren't complete. Before the 1950s, climate observations were mainly limited to weather stations and ships, and only included measurements made at or near the land or ocean surface. In the 19th century, many parts of the world were not monitored at all. There are few complete daily instrumental records stretching as far back as the eighteenth century, though one of these - Central England Temperature (CET) - has monthly data back to 1659.
In recent years things have improved:
Since the 1950s weather balloon soundings have been widespread over land.
There has been nearly worldwide coverage from satellites since the 1970s - greatly improving our ability to monitor climate.
Since 2000, there has been a massive increase in sub-surface ocean monitoring through the Argo project.
Scientists have also greatly improved the availability of instrumental data for the last two centuries. Millions of observations made from ships in the late 19th and early 20th centuries, which until recently were only available from paper logbooks, are being put in computer databases. Some of these have been included in the latest assessments of climate, like the Intergovernmental Panel on Climate Change (IPCC) 2007 report. However, many more logbooks have yet to be included.
As there are no records of climate from direct measurements before the 1600s, scientists have used other types of information to investigate further back.
Tree-rings and ice-cores can be used to infer changes in temperature and precipitation.
Depth profiles of temperature in oil-drilling boreholes can be used to estimate the changes in air temperature over recent centuries.
Corals can be used to estimate oceanic temperature and sea-level changes.
None of these indirect, or 'proxy' methods are as precise as direct instrumental measurements. Also, there are very few proxy datasets, so it is very difficult to obtain reliable estimates of past global temperatures. However, long-term temperature trends derived from borehole and other independent proxy data are in reasonable agreement, confirming the climate in the past two thousand years was not as warm as it has been in recent decades.
Reliable analysis of climatic change is made difficult by historically-varying biases in the data. Basic quality checks and correction of obvious errors in current observations are relatively simple using modern computers. However, it is much more difficult to account for the systematic differences between modern and historical observations.
Sea surface temperatures in the early 20th century were mainly measured using canvas buckets. When the buckets were drawn out of the water and the wind blew on them, the water cooled, leading to biased temperature recordings, so it is necessary to correct the data.
More recent sea surface temperature measurements are less biased because they have been made using thermometers attached to ships well below the water-line, or using purpose-built buoys.
Since the early 1980s sea surface temperatures have also been measured from satellites, which detect the infra-red or microwave radiation coming up from the oceans. Satellite measurements also need careful treatment because dust and moisture in the atmosphere can cause systematic differences with measurements made from ships and buoys.
Satellites also only measure the temperature of the surface skin of the ocean, whereas ships and buoys measure the temperature lower down. All of this is taken into account when studying the data. Corrections and adjustments are made where necessary to ensure biases are ironed out to create a reliable data set.
Satellites do not measure temperature and other climate indicators directly. They measure visible, infrared and microwave radiation of different colours and these are used to calculate temperature, humidity, rainfall, soil moisture, and vegetation characteristics, either via statistical algorithms or via physical equations expressing how radiation is transferred through the air up to the satellite. The same equations can be applied in weather-forecasting and climate-prediction models of the atmosphere allowing us to compare the radiation measured by the satellite instrument with that simulated by the model. This can help us to uncover and correct any biases or errors in the measurements and the models.
Errors can arise in combining climate data from different sources unless careful account is taken of how the original measurements were taken and how these methods may have varied. International standards for observing practices have been developed and documented over the past 150 years, and are now co-ordinated and reviewed through the World Meteorological Organization. For climate, the Global Climate Observing System climate monitoring principles have been agreed and adopted by the Conference of Parties to the UN Framework Convention on Climate Change.
These standards, for both satellite and ground-based data, include the requirement to overlap records when observing systems are changed so scientists can reliably estimate and remove any relative biases. This requirement can be expensive to implement for satellites because it means launching a new one before the old one has expired rather than when the old one fails. However, it makes the investment in satellites much better value for money. Information about how observations are made and processed (known as 'metadata') should also be kept in full, and treated with as much care as the data themselves. If a scientist knows how an observation was made, it is much easier to make a fair comparison with other observations.
An accurate record of a land-based observing station's environment, and how it may have changed, is an important part of collecting observations. This information can be vital in assessing, for example, whether a temperature record has been affected by urban development (known as the 'urban heat island' effect), or whether rainfall patterns have been affected by the growth of trees or the construction of new buildings. The station records used to develop the global temperature trends presented in the IPCC's Fourth Assessment Report (IPCC 4AR) have been screened for biases and adjusted or rejected if necessary. The IPCC's report shows that while urban warming is a significant local phenomenon, it has contributed very little to the global warming trend.
Climate observations are often expressed as differences (known as anomalies) relative to a particular period. In the IPCC's reports, anomalies have been expressed relative to the years from 1961 to 1990, the reference period approved by the World Meteorological Organization. When working out global or national average temperatures, anomalies are used instead of taking the average of absolute (or observed) temperatures. This is because anomalies vary less from one place to another and can, therefore, be averaged more reliably when there are gaps between observing stations.
If we averaged actual temperatures worldwide and the observation from a normally warm (e.g. tropical) station was missing, then the average would be biased cold by the absence of its data. When we average anomalies, any bias is much smaller because anomalies do not vary as much as actual temperatures, and, by definition, are not systematically higher or lower in one place than in another.
Scientists and mathematicians have recently developed methods to estimate the uncertainties in weather data and global averages. The uncertainties are estimated from the random and systematic errors, as well as the geographical gaps in the observations. The random and systematic errors in turn are estimated from the statistical properties of large numbers of observations.
The estimates of observed global and regional warming in the IPCC 4AR include uncertainty estimates which show the observed warming far outstrips the uncertainties.