Frequently Asked Questions

Q: What anomaly period have you used?

A: Anomalies have been calculated relative to the 1961-1990 average. The method used to derive the climatology is described in Rayner et al. 2006.

Q: Where do the raw observations come from?

A: The observations on which the dataset is based come from version 2.5 of the International Comprehensive Ocean Atmospere Data Set (ICOADS). The ICOADS data and associated products are freely available for users around the world. From 2007 to the present, the observations are taken from the Global Telecommunication System (GTS).

A: HadSST3 is an update of HadSST2. It is based on a later version of ICOADS that contains many more SST observations and therefore the coverage has improved during most periods, but particularly in the 1880s and 1940s. The methods used to adjust for changes in instrumentation have been extended to cover the whole of the record from 1850 to present. In HadSST2 bias adjustments were only applied from 1850 to 1941. We have also improved our estimates of the uncertainties in the data and we are presenting the bias uncertainty as a set of 100 interchangeable realisations that capture the spatial and temporal correlations in the bias-uncertainty fields.

HadSST3 is very different from HadISST. HadISST is a globally-complete, interpolated SST and sea ice concentration product designed primarily for use as a forcing data set for reanalyses and atmosphere-only GCM runs. A new version of HadISST2 is planned that incorporates the new ICOADS data, new satellite data and the bias adjustments used in HadSST3.

A: The bias adjustment algorithm has a number of inputs - for example the estimated sizes of the engine intake biases and their evolution over time - which are somewhat uncertain. The 100 realisations of the data set are produced by varying the inputs to the bias adjustment algorithm within a plausible range for that input given what we know about the data. For each realisation a random set of these inputs is chosen. The 100 realisations together give some idea of the uncertainty in the fields arising from uncertainties in the bias adjustment procedure. The ensemble of 100 data sets does not include the effects of uncertainties due to measurement and sampling uncertainties. These are contained in a separate file.

A: The ensemble of 100 data sets represents uncertainty in the bias adjustments, estimated by varying uncertain parameters that are used to estimate the bias adjustments. Uncertainties arising from other measurement errors (both correlated and uncorrelated), sampling errors at grid-box scale and sampling errors at a larger scale are not included in the ensemble.

A: When we calculate the trend in global average SST, we calculate the global annual average and the trend for each of the 100 realisations. The distribution of the 100 values gives an idea of the uncertainty in that derived quantity arising from uncertainty in the bias adjustments. Likewise, it should be possible to run your analysis on each of the realisations and the spread in results gives an idea of the sensitivity of the analysis to observational uncertainty. A separate file contains the measurement and sampling uncertainties that are not strongly correlated over long time periods.

If you are interested in exploring the observational uncertainty, we recommend that you use a variety of SST data sets in addition to HadSST3. In the early record and in data sparse regions, the differences in the treatment of missing data, random measurement errors and quality control become significant components of the overall uncertainty.

A: We also provide a file that contains fields of the median SST anomaly from the 100 members in each grid box. If you use only the median, you will not be able to explore the sensitivity of your analysis to the estimated observational bias uncertainties.

A: The unadjusted SST anomalies are supplied from the data download page. Bear in mind that the unadjusted SST anomalies contain significant uncompensated biases and should not be used for climate trend analyses.

A: Please read the HadSST3 papers first (part 1 and part 2). Broadly speaking, there are two kinds of errors in observational measurements: random errors and systematic errors. Random errors are relatively benign. Although they might confound a single observation, they tend to cancel out when large numbers of observations are averaged together. In any year, the global annual average is based on tens of thousands of observations. When averaged together, random errors tend to vary as the inverse of the square root of the number of observations that contribute to the average so the random error on the annual average is around 100 times smaller than the random error on a single observation (assuming each observation takes roughly the same weight in the average).

Systematic errors are much more problematic because their effect becomes more pronounced as greater numbers of observations are aggregated. In HadSST3, systematic errors are dealt with in two ways. The first is to make adjustments based on how ships made SST measurements. These adjustments are themselves uncertain because we have imperfect knowledge concerning the size of the biases and the exact methods used to make the measurements. This uncertainty has been estimated by allowing uncertain parameters in the algorithm to be varied within their plausible ranges thus generating a range of bias adjustments. The second is to assume each ship, or buoy to have an independent bias that is peculiar to that ship.

The validity of the bias adjustments and their uncertainties is difficult to assess. Since the late 1940s, there is consistency between estimates of global SST change made using only observations collected using buckets and measurements made using Engine Room Intake observations. From the 1990s, there are also plentiful observations from drifting and moored buoys as well as SSTs retrieved from satellite instruments. However, during the Second World war and in earlier periods, the majority of observations were made using a single method - either Engine Intake Measurements during the War, or buckets before the War.

In the pre-war period, some confidence might be derived from consistency with marine air temperature measurements, but such conclusions are necessarily tentative because marine air temperature measurements are used to some extent in all current schemes for adjusting SST prior to the Second World War. For example, in HadSST2 and HadSST3, the fraction of canvas and wooden buckets used prior to 1920 is determined by a comparison between NMAT and SST data. Removing this dependence, and considering the effects of assuming that either all buckets were canvas, or all buckets were wooden suggests that global average SST in 1850 might have been 0.2°C higher than the current estimates or 0.1°C lower.

The bias adjustments account for the large scale differences in averages between different measurement methods, but do not account for differences in biases between ships using the same general method. The effect of the differences of biases between ships using the same general measurement method is estimated by allowing each individual ship to take its own bias. Averaging together observations from a single ship will not reduce this uncertainty, but averaging together observations from different ships (which will each have different biaes) will reduce this uncertainty. It varies approximately as the inverse of the square root of the number of ships contributing to the average (assuming each ship makes similar numbers of observations). For annual averages based on observations from hundreds or thousands of ships, this effect is generally small.

A. Diagrams are Crown Copyright. Source should be acknowledged as Met Office Hadley Centre.

Commercial and media enquiries

You can access the Met Office Customer Centre, any time of the day or night by phone, fax or e-mail. Trained staff will help you find the information or products that are right for you.
Contact the Met Office Customer Centre