Taking the planet’s temperature: How are global temperatures calculated?

Thermometer

Grantham intern Peter Davies (Department of Physics) delves into global temperature records.

Statements such as “2014 earth’s warmest year on record” or “No global warming for 18 years 1 month” are conclusions from different atmospheric temperature data sets.  Before assessing which is true it is important to understand how temperatures are measured, how data sets are created and used to calculate global temperatures, and the strengths and weaknesses of each approach.

Using a variety of ways of measuring is crucial to avoid introducing the same errors in data sets created in similar ways. The combination of surface temperature data sets and satellite data sets provides this variety.

Land and sea surface temperatures

Thermometers for measuring land temperatures are not in contact with the ground, but with the air, typically 1.5 metres above the ground within a weather station. Thus, strictly, land temperatures should be referred to using the term “near surface temperatures” though we will use the words “surface temperatures” in this blog. Readings are now automated but previously were taken manually.

Measuring sea surface temperatures requires contact with water at, or taken from, between 1 mm to 20 m depth. In the past these observations were gathered by suspending a thermometer over the side of a ship; measuring the water temperature in a ship’s intake port; or collecting samples in wooden or canvas buckets. For each method, varying the measuring procedure leads to variation in the results.

Modern measurements are from ships, and static or drifting buoys. Measurement processes are supposed to be standardised, but there are still small differences between nations.

Land surface temperatures can vary significantly in a small area, but we are generally interested in anomalies – the change in temperature at a single place between one date and another. Unlike individual readings, temperature anomalies on land or sea are consistent across regions many hundreds to thousands of miles wide. Hence sparse coverage of surface temperature measurements in a region is not necessarily a major problem.

The World Meteorological Organisation recommends defining the temperature of a location for a 24 hour period as the average of the maximum and minimum temperatures recorded during that period. Although not the best calculation now available, it is the easiest to apply consistently for the calculation of anomalies.

Temperature adjustments

Land and sea surface temperature data is quality-checked and adjusted to remove the biases from each different measurement process.

On land these include changes in the time of day of observations and moves or changes to locations. Observations from modern, well-sited, automated equipment are treated as accurate and historical data is adjusted to use the baseline set by these modern observations.

For sea surface temperatures from ships, another check performed is that consecutive readings recreate a sensible ship’s course, allowing time or location errors to be spotted.

Global temperature graph with and without adjustments
Land data from GHCN, sea data from HADSST3

So, how do these adjustments affect the bigger picture? Land surface temperature adjustments increase the global land temperature trend slightly. Sea surface adjustments decrease the sea trend considerably. Overall the surface temperature adjustments cause a significant reduction in trends over a century or more, while making little difference to the conclusion that global warming is real.

From local to global: global temperature data sets

Many global surface temperature data sets pull together land and sea temperature data. Each data set processes raw temperature observations differently. First, temperature data is generally laid out on a grid. To calculate global temperature, researchers average the readings for each grid point, weighted by the area associated with that grid point. Latitudes further away from the equator have smaller areas.

Missing regional observations

Some land and sea regions have inadequate numbers of weather stations or buoys to provide safe temperatures for particular grid points. The grid points with missing temperatures can be handled by not providing a value, estimating a missing value based on surrounding values, or using patterns from the satellite observations to aid in estimating values from the surface temperature data sets.

Some key surface temperature data sets

GISTEMP 

The NASA Goddard Institute for Space Studies surface temperature data set uses the US NOAA (National Oceanic and Atmospheric Administration) GHCN (Global Historical Climatology Network) of land-based meteorological stations, ERSST (Extended Reconstructed Sea Surface Temperature) for ocean areas, and SCAR (Scientific Committee on Antarctic Research) for Antarctic stations.

Missing observations from a grid region are estimated from surrounding readings using kriging.

HadCRUT4

This uses CRUTEM4 (Climate Research Unit of the University of East Anglia and the UK Met Office Hadley Centre) land temperatures and the HadSST3 (UK Met Office Hadley Centre) sea surface temperatures.

Where observations are missing from a grid region the value is left empty.

Cowtan & Way

Kevin Cowtan and Robert Way produce two data sets which supply temperatures missing from the HadCRUT4 data set. The “kriging” version inserts estimates based on surrounding HadCRUT4 grid points with data. The “hybrid” version uses snapshots of satellite data containing differences between neighbouring grid points, which are combined with the existing HadCRUT4 data to estimate missing values.

Satellite Temperature Data Sets

The satellite “total lower tropospheric” (TLT) temperature data sets provide a view of temperatures in the lower atmosphere, providing a broad comparison with surface temperature data sets.

Satellite sensors

The curve labelled TLT shows the fraction of the TLT temperature obtained from various heights
The curve labelled TLT shows the fraction of the TLT temperature obtained from various heights

The observations come from a series of mainly weather (and some climate) satellites in polar orbits, which measure temperatures using microwaves. The original microwave sensor used was the MSU (Microwave Sounding Unit). In 1998 the AMSU (Advanced Microwave Sounding Unit) replaced it on all new satellites. Although the AMSU sensors are an improvement over the MSU they do not monitor exactly the same things. Readings from the two have to be merged carefully for any time span which includes both MSUs and AMSUs

Microwaves from the earth’s surface itself make only a small contribution to the overall satellite temperature readings. Hence the satellite TLT data sets represent temperatures higher in the atmosphere than the surface temperature data sets do.

Merging records from different satellites

Processing the satellite MSU and AMSU readings to create temperature data sets over time is complex. Trends over decades require temperature records from up to 16 satellites to be combined into a single record with accurate matching of temperatures measured by MSUs and AMSUs. Yet there is often little overlap in time between early satellites launched consecutively, which may never have passed over the same point at the same time of day to allow a direct cross-check. The calibration of individual satellites may also drift over time.

A satellite orbit crosses the equator at a particular time of day, but, with no thrusters on board, the orbits gradually drift slightly due to friction with gas molecules in space. Hence satellites pass a given point at increasingly different times of day as time passes, affecting the estimate of daily temperatures. The AQUA satellite is an exception as it has thrusters.

 Top graph contains satellite temperature changes not corrected for orbital decay drift for a mid (not lower) tropospheric data set. Middle graph is after standard corrections for orbital drift. Bottom chart shows the benefit of a more advanced correction technique. From “Removing Diurnal Cycle Contamination in Satellite-Derived Tropospheric Temperatures: Understanding Tropical Tropospheric Trend Discrepancies” Po-Chedley, Thorsen & Fu, Journal of Climate Volume 28, Issue 6, March 2015
Top graph contains satellite temperature changes not corrected for orbital decay drift for a mid (not lower) tropospheric data set. Middle graph is after standard corrections for orbital drift. Bottom chart shows the benefit of a more advanced correction technique. From “Removing Diurnal Cycle Contamination in Satellite-Derived Tropospheric Temperatures: Understanding Tropical Tropospheric Trend Discrepancies” Po-Chedley, Thorsen & Fu, Journal of Climate Volume 28, Issue 6, March 2015

The satellite temperature data set providers

UAH (University of Alabama, Huntsville)

Staff at the University of Alabama, Huntsville were the first to analyse weather satellite observations to determine atmospheric temperature changes at different altitudes.

RSS (Remote Sensing Systems)

The company Remote Sensing Systems set up a second analysis of weather satellite observations using alternative methods to those chosen by the UAH team.

Strengths and weaknesses of the satellite temperature data sets

Satellite readings provide nearly global coverage of the earth’s surface and have a consistent error across most regions.

The analysis processing of satellite observations requires a distinction between unfrozen land, surface water (lakes and oceans, ice melt) and ice and snow. Melt water above sea ice increases the error most at the poles where warming has been the fastest.

Recent research has cast doubt on whether satellite temperatures are measured accurately in the presence of clouds and rain.

The earth’s surface is the most significant altitude for determining the effects of climate change, but the surface temperatures have only a small influence on the satellite data set lower tropospheric temperatures.

The method of merging together overlapping satellite records differs between the satellite data set providers. Given the complexity of the analysis, different satellite temperature data sets and research analyses reach different conclusions from the same available satellite observations. The two main production temperature data sets now agree reasonably well with each other but not necessarily with estimates of temperatures and trends derived by other researchers from the same data, nor with the surface temperature data sets. Carl Mears of RSS believes that the surface temperature data sets are likely to be more accurate than the satellite data sets.  The UAH team beg to differ.

How surface and satellite temperature data sets compare

Processing of surface temperature data sets is straightforward and adjustments have only a small effect on calculated global temperature trends. Most show similar trends. However, coverage is sparse for some areas of the globe.

By contrast, the satellite data sets provide consistent global coverage, except for issues at the poles, but require complex processing to merge observations from different satellites. They include only a small contribution from surface temperatures. Different processing decisions by individual researchers and data set producers tend to result in significantly different satellite-derived trends so the resulting calculations have more inherent uncertainty.

Conclusion

All these data sets are required to present a balanced picture of what is happening to surface and lower atmospheric temperatures. When evaluating statements about surface temperatures a heavier weight should be given to the more accurate surface temperature data sets, but the satellite data sets should not be dismissed, though they provide a picture of the lower part of the atmosphere including only a small component from the surface.

The statement “2014 earth’s warmest year on record” is a conclusion drawn from multiple surface temperature data sets.  It is likely to be true as it is based on the more accurate surface temperature data sets and the natural assumption would be that it applied to surface temperatures (see Jo Haigh’s blog)

The statement “No global warming for 18 years 1 month” is based on two satellite temperature data sets only.  It should be qualified by “in the lower atmosphere counting only a small contribution from the surface”.  The qualified statement may be true, but there is lower confidence in this because of the complexity of processing required to derive the satellite temperature data.

Find out more about the influence of ocean heat uptake on global temperatures in our briefing paper

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s