Observed trends: What to be aware of

• Different climate variables and datasets have different periods of record

Each dataset and climate variable that comprise it has its own period of record. These periods reflect the time for which sufficient quantity of records is available and, in the case of the derived variables, the time period over which calculations have been made.

 It should be noted that the Met Office reserves the right to issue periodic updates that reflect improvements in their ability to produce the gridded data sets, including the addition of subsequent observations from subsequent years.

• Daily precipitation observations are not available as a contribution to UKCP09

Although daily precipitation datasets have been generated, they are not available under UKCP09. Those wishing to have access to the daily precipitation datasets should enquire at the UK Met Office about purchasing these. For UKCP09 users, the UK Met Office will be making available various derived quantities that have been calculated for each 5 km grid, by year and season. See the Met Office website.

These derived quantities are provided to facilitate comparisons with the UKCP09 Weather Generator baseline values.

• No sub-daily climate information is available

The available datasets do not include observations at less than daily time-steps. As such, there are no observational data available through UKCP09 which can be compared with the baseline (1961–1990) hourly values generated by the UKCP09 Weather Generator.

• The gridded datasets are for a point and are representative of the climate of the 5 km grid square

The gridded datasets are estimates of the climate for the centre point of a 5 km x 5km grid cell. The method used to generate the monthly and annual gridded datasets are described in more detail in a paper published in the International Journal of Climatology (vol. 25 (2005), pages 1041–1054. The methods used to generate the daily grids are described in more detail in a report that can be downloaded from the UK Met Office website. 

• It contains some uncertainties

Uncertainties also exist in historical climatological observed data, from sampling and instrumental errors, and adjustments made in the record for discontinuities, and processing/analysis errors. These uncertainties arise from the fact that climatological instruments (mechanical and electronic) are used to sample the climate in real environments. They also arise as a result of changes in instrumentation (type, location and number/density) and from changes in the environment in which the instruments are located.

Uncertainties also arise through processing of the recorded data some of which is done to address the other sources of error (e.g. homogenising and gridding the observational record). The density of the station network used to grid the climate observation varies through time and for different climate variables. Regression and interpolation are used to generate values onto the 5 km grid from the irregular station network, taking into account factors such as latitude and longitude, altitude, terrain shape, coastal influence and urban land use. This alleviates the impact of station openings and closures on the homogeneity of the observations, but the impacts of changing station network cannot be removed entirely, especially in areas of complex topography or sparse station density.  There is also the contribution from the fact that the available network for monthly observations is different than that available for daily observations. This can be problematic particularly in areas where there is sparse station density.

These derived quantities are provided to facilitate comparisons with the UKCP09 Weather Generator baseline values.

• No sub-daily climate information is available

The available datasets do not include observations at less than daily time steps. As such, there are no observational data available through UKCP09 which can be compared with the baseline (1961–1990) hourly values generated by the UKCP09 Weather Generator.

• Trends information is only available in images and summary tables

Recent changes and trends information are provided within the UKCP09 Observed trends report in a number of different ways. These include maps illustrating the difference between the two 30-year periods (1961–1990 and 1971–2000), and maps showing changes over the period 1961–2006  In the latter case, the maps were generated by calculating at each 5 km grid square a linear trend for each season and the whole year. This gives an indication of recent trends, but the depicted value is sensitive to the period over which it was taken and in particular to any extremes at either end of the averaging period. As such the depicted changes and trends must be treated as indicative and useful for geographic patterns, but their exact values should be treated with caution. Graphs of trends are also included for the 14 UKCP09 administrative regions. The depicted regional trends were smoothed to show decadal variations (see UKCP09 Trends Report for further details). 

Tables of change over the period 1961–2006 by season and for the whole year are also provided for each of the 14 UKCP09 administrative regions. In addition, changes for temperature and precipitation are given for the period 1914-2006. Linear regression was used to calculate the trend for each of the variables and then the average rate of change from the linear regression multiplied by the length of the data period was used to provide a measure of change since the start of the period.

Users are encouraged to refer to the UKCP09 Observed trends report for further details and to view the maps, graphs and tables.

• The observed climate during this period of record does contain some climate change

Users are reminded that when using observed climate information, especially in terms of calculating return periods and other statistics related to extreme, that there are limitations under which these statistics are valid. For example, as the climate during the period of record has been changing (and is therefore non-stationary), users should be wary about calculating and interpreting return periods and other statistics that assume a stationary data set.