Search site

About Weather Generator 2.0

What has changed?

Weather Generator, Version 2.0 includes four main updates. These include:

A new statistical distribution has been implemented to better replicate historical extreme periods across the range of return periods at which it is applicable.

Improvements have been made to heat wave duration by adding an extra dry spell transition.

Modifications have been made to shorten the simulated day length to be consistent with measurements of effective sunshine.

A new baseline has been introduced for sunshine and vapour pressure to produce more realistic changes in future projections. 

Why and how has it changed?

  • Rainfall extremes

Experience with the original UKCP09 Weather Generator has shown that it underestimated hourly extreme rainfall amounts by up to 50% in the 30 year return periods. This was due to the simplified nature of the rainfall model which had been used for its robustness. To improve the performance, a number of different test distributions were trialled before a gamma distribution was selected. The performance at daily and hourly resolutions return period estimates were validated against the Flood Estimation Handbook Depth Duration Frequency model and were found to be more consistent (generally within the 90% confidence limits). It should be noted that some under estimation remains at the highest return periods (see examples in Figure 1 and 2 below).

   
Figure 1. The Weather Generator at hourly resolution. Comparison of Weather Generator rainfall estimates with Flood Estimation Handbook (FEH) Depth Duration Frequency model estimates for a site in north east England.  
   
eather Generator 2.0 Figure 1
   
Figure 2. The Weather Generator at daily resolution. Comparison of Weather Generator rainfall estimates with Flood Estimation Handbook (FEH) Depth Duration Frequency model estimates for a site in north east England.  
   
eather Generator 2.0 Figure 1   
   
  • Temperature extremes

The original UKCP09 Weather Generator underestimated daily maximum and hourly extreme temperatures. This has been improved in two ways: the use of an extra dry spell transition and the use of a different temperature distribution.

Extra dry spell transition

Originally, the UKCP09 Weather Generator considered four different two-day sequences:

  • DD    Dry today/Dry yesterday
  • WW    Wet today/Wet yesterday
  • DW    Dry today/Wet yesterday
  • WD     Wet today/Dry yesterday

Experience showed that these sequences were not able to simulate the build-up of temperatures during longer dry spells. For example, if a two-day dry spell fell between two wet days (WDDW), it would show the same results as if it were part of a longer dry spell (DDDD). As such, the original UKCP09 Weather Generator tended to underestimate daily maximum temperature values during a period of successive dry days. In these cases the maximum temperature may build up over the successive dry days and there may not be overnight cooling in between these days. A fifth sequence of days was therefore added to improve the performance of the Weather Generator for maximum temperatures and heat wave duration.

  • DDD    Dry today/Dry yesterday/Dry day before

Improved temperature distribution

The original Weather Generator assumes a normal distribution (i.e. a bell-shaped curve) for temperature between 1961-1990, however, temperature extremes can cause a distribution to skew in some conditions (e.g., heat waves and cold spells). A calculation (power transform) has been applied within the Weather Generator and validation testing using observed data has shown that this alleviates the problem evident in the original Weather Generator.

Back to top

  • Improvement in sunshine hours

The UKCP09 Weather Generator estimated daily sunshine, producing values from the moment the sun rose above the horizon. This over-estimated effective sunshine as observational equipment does not begin recording until the sun is higher in the sky. To deal with this shortcoming, modifications were made to include a more realistic (slightly shorter) day length. The updated version now more accurately reflects how sunshine has been measured in the observations.

  • Changes to the climatology's baseline in sunshine and vapour pressure

In 2010, users noted that outputs from the UKCP09 Weather Generator showed unrealistic changes in future sunshine hours. It was also discovered that there was a related issue with vapour pressure. These issues are due to the choice of observed climatology used in generating the change factors and do not stem from errors in either the UKCP09 probabilistic data or the UKCP09 Weather Generator. A new observed baseline has therefore been set up to produce more realistic changes.

Most of the variables required in the change factors (such as temperature and precipitation) are available in the projections. Sunshine and vapour pressure are two variables that are not available and have therefore been derived using variables that are available (Sunshine is derived from cloud cover, and vapour pressure is derived from relative humidity and temperature).

For sunshine, the solution implemented was to create a new baseline climatology from observational cloud cover so that the relationship (SUNSHINE FRACTION = 1 - CLOUD FRACTION) remains constant for both the baseline and future climate. The baseline climatology for vapour pressure was derived from observational relative humidity and temperature rather than directly from observational vapour pressure data.

These updates to the baseline climatologies have been implemented for use by the updated UKCP09 Weather Generator. It should be noted that the updating of these baseline climatologies will affect the future simulated absolute values for these variables resulting from the UKCP09 Weather Generator.

The technical note on sunshine and water vapour can be downloaded from the Technical Notes table on the Reports and Guidance section.

Back to top