Inappropriate uses of the Weather Generator
Limitations on the use of each output are identified in the reports and guidance – it is important that users read through these before using the Weather Generator output.
The following are inappropriate uses of the Weather Generator:
- Looking for a specific date or period
- Time series should not be averaged
- Time series generated for adjacent grid squares from separate runs should not be averaged
- Selecting a large number of grid squares in a geographically diverse area
- Time series from two (or more) different 30-year time periods should not be combined
- Extreme statistics beyond specified limits should be used with care
- Selecting a month as a temporal averaging period to condition the UKCP09 Weather Generator is not allowed
- Using less than 100 time series of at least 30 years in length
The Weather Generator does not generate predicted weather for any specific date in the future. The time series generated are labelled with a year beginning with 3001 to help avoid confusion. This applies to both the baseline and future time series.
Averaging the Weather Generator outputs to produce a single time series negates the ability to explore different impacts, vulnerabilities and response strategies.
- FAQ: Why do I need to run the Weather Generator 100 times?
- Weather generator guidance
- See Worked example for good practice: Informing resource planning and management
- See Worked example for good practice: Modelling impacts
There is no correlation in time between the separately generated series (e.g. at any given day within the two series, it may be raining in one grid square and not in the other).
- FAQ: What can I use to look at daily climate over a large area (>1000km2)? Is there an alternative to the Weather Generator?
- FAQ:
- Weather generator guidance
- See Worked example for good practice: Investigating impacts
- See Worked example for good practice: Updating existing research
- See Worked example for good practice: Informing resource planning and management
Time series across large areas result in a set of time series that corresponds to a single point (5 km grid) that is representative of the selected region. Care must be taken in selecting areas where there are large differences in characteristics (e.g. coastal locations, areas comprised of both urban and rural areas). Users should breakdown the area of interest into a number of smaller homogeneous areas or single grid squares.
- FAQ
- Weather Generator guidance
- See Worked example for good practice: Investigating impacts
- See Worked example for good practice: Updating existing research
- See Worked example for good practice: Informing resource planning and management
The time series generated by the UKCP09 Weather Generator are stationary, so combining generated time series from two different 30-year time periods which have different stationary climates is inappropriate.
Users should avoid attempting to derive unrealistic (or Annual Exceedance Probabilities). Extreme statistics for return periods longer than 10 years should be used with caution. Hourly extreme statistics are subject to even more uncertainty, and return periods beyond 5 years should be used with caution.
- Selecting a month as a temporal averaging period to condition the UKCP09 Weather Generator is not allowed
Perturbing the Weather Generator for conditions from a future month has been deemed less scientifically valid than using a future season. There is a danger that using the change factors from a specific month to perturb the Weather Generator might skew the results towards an extreme set of conditions.
The aim is to sample the change factors frequently enough so that a stable distribution of results is obtained. In this case stability depends on the sensitivity of the impact being analysed to the different variable being generated (typically varies by region).
- Last updated: Sunday, 11 March 2012