The concepts behind the weather generator methodology are well-established and have been used for many years in such areas as flood risk estimation, climate impacts assessments and water resources management.
The UKCP09 Weather Generator has benefited from previous work some of which involved members of the UKCP09 Weather Generator project team. This previous weather generator work included BETWIXT (Built Environment: Weather scenarios for investigation of impacts and extremes) and EARWIG (Environment Agency Rainfall and Weather Impacts Generator).
The UKCP09 Weather Generator methodology reflects a further development in that it is based on advances in our understanding and that it produces outputs consistent with the underlying UKCP09 . Extensive testing and verification has been undertaken to validate the resulting methodology the results of which are reported on within the Weather Generator report.
Weather generators are designed to produce plausible time series of daily climates consistent with baseline observed conditions (i.e. similar statistics and other characteristics to that of the underlying observed climate). They are partly informed by one of the basic premises that the weather on one day informs but does not necessarily determine the weather on the subsequent day. For example, it is more likely, but not necessary, that it will rain tomorrow if it has rained today.
There is also a random element contained within the methodology to deal with those aspects of daily climate that are not explained by persistence. This random element allows different (but statistically equivalent) series to be generated and allows different climate eventualities to be evaluated.
The UKCP09 Weather Generator provides the capability for users to generate statistically plausible daily time series for defined 5 km grid square (or for a location representative of an area up to forty 5 km grid squares) that are consistent with the UKCP09 baseline climate (1961–1990) and the UKCP09 probabilistic projections (monthly, seasonal and annual projections at a spatial resolution of 25 km).
Users can also generate possible hourly time series. These hourly values are estimated from the daily time series using simple disaggregation rules (see the Weather Generator report ) taken from the baseline observed climate and which remain the same for future time periods. Users are reminded that hourly time series should only be requested if an hourly resolution is specifically needed for their intended application. They are also reminded that hourly time series files will be quite large.
Traditionally weather generators have been used to generate long time series (up to 10,000 years) of daily climates determined by a given 30-year baseline observed climate. For UKCP09 Weather Generator, users will be able to generate daily time series for both the present and the future with durations of between 30 and 100 years (integer number of years only). In the case of hourly time series, the only option available is for time series of 30 years duration. This restriction is believed the best compromise considering that it provides users with time series that are of the same duration as the 30-year time period of observations and is a standard period normally used for deriving climatological statistics, while at the same time providing a manageable volume of data.
Users should be aware that, as the 25 km probabilistic climate projections are stationary, (do not contain a changing climate during any given 30-year period), the generated time series are also stationary for each chosen future. Experience and testing has suggested that 100 time series of at least 30 years length for each 30-year future is a minimum requirement.
The daily time series produced using the UKCP09 Weather Generator can also be used to estimate the implications of climate change on the basis of a number of derived indices (see Table 1 in the Weather Generator report). Widely used temperature and rainfall indices have been identified and users will be able to derive some additional ones from the generated time series using the associated Threshold Detector. The time series produced using a weather generator are more suited to this purpose than those available from climate models (e.g. Regional Climate model output) as the weather Generator time series for the baseline period are based on observed data and therefore do not require bias-correcting. If output from the 11-member RCM time series were used for this purpose, bias-correcting (considering the difference between the modelled and actual baseline climate) would be necessary.