Weather Generator
A Weather Generator is a statistical method of creating projections of future daily (or sub-daily) climate that are consistent with climate change projections for longer temporal averaging periods (e.g. or seasonal).
In detail
Weather Generators (WGs) traditionally use existing weather data and random number sampling to produce long time series of statistically plausible daily and hourly weather data. The UKCP09 Weather Generator creates a statistical representation of a series of daily future climate which have similar statistical properties as the observed data on which they are trained. For UKCP09, daily (and hourly) time series for future periods is generated based on perturbing the Weather Generator according to the probabilistic projections. The UKCP09 Weather Generator is based around a stochastic model that simulates future rainfall sequences. Other weather variables are derived from the rainfall states. Statistical measures within the Weather Generator are then perturbed according to the UKCP09 probabilistic projections to generate time series for future time periods.
At a simple level the UKCP09 Weather Generator is essentially trained by using the 5 km, daily observed baseline of 1961–1995. This means that the Weather Generator model baseline is fitted to the 1961–1995 historical observations.
Change factors, which provide projections for the change between the baseline climate and the future climate, are randomly taken from the sampled data. These change factors are then used to perturb the statistically derived time series to generate statistical expressions of what daily time series of future climate may look like.
- An overview of the UKCP09 Weather Generator is given in Chapter 4 of the UKCIP09 Briefing report. A small selection of prepared results is given in Section 5.11.
- The UKCP09 Weather Generator is described in more detail in the Reports & guidance section, with further information in the UKCP09 Projections of future daily climate for the UK from the Weather Generator report.
It is not statistically robust to only use one set of change factors from the sampled data to drive the Weather Generator. Users should instead use a minimum of 100 time series as suggested within the User Interface. It is also recommended that the minimum length of time the Weather Generator should be run for is 30 years.
- Further limitations of Weather Generators are described in Chapter 5 of the UKCP09 Weather Generator report.
- Some worked examples were developed prior to the launch of UKCP09 that used dummy outputs to examine how the UKCP09 Weather Generator might be used. The worked examples should not be interpreted as a best practice method for using UKCP09, but as one way of using the outputs.
- A selection of inappropriate uses of the UKCP09 Weather Generator is also provided.
- Last updated: Sunday, 11 March 2012
