Online Weather Generator report 2
Many users are engaged in assessments of changes in the severity and frequency of extremes and this is one of the most challenging aspects of climate research, beset by fundamental limitations and inadequacy in theory and practice. First of all, extremes are, by their nature, infrequent events, so that we only have small sample numbers from observational records of the most important extremes. This limits the power of statistical approaches to the estimation of extremes, and inevitably requires the user to estimate uncertainties or confidence limits around such estimates. Second, frequency analysis assumes stationarity in the observed data, and this is clearly not true for temperature (and possibly rainfall) extremes in the recent record. Third, the processes causing extremes (such as floods and droughts) are complex and their representation is at the limit of the current capability of climate models (see Annex 3 and Annex 5 from Murphy et al. 2009). Nonetheless, it is important to provide estimates of future extremes. The approach taken in UKCP09 is to use its WG to allow estimation of extremes and other properties of climate where credible, by generating long, stationary series to provide large statistical samples.
A range of extremes and derived properties may be generated using the WG. Table 1 provides some possible examples that have been used recently. Users can define their own specialist indices or measures of extremes and extract the values from WG series. They should follow the method of comparison of future values of their chosen index with the same index calculated from the control version of the WG generating baseline (1961–1990) climate, and wherever possible also validate this against their own observed data. To achieve this, when running the WG, users will be supplied with at least 100 30-year generated sequences of the baseline climate and a similar number for their chosen 30-year future period. Some widely used indices have been identified and can be extracted using the Threshold Detector (see the Annex). There are various types of weather indices or extremes that can be derived from the WG which are not available directly from the probabilistic projections, including maxima and minima (e.g. the median of annual maximum rainfall (Rmed)), threshold exceedances (e.g. occurrences of temperature above 28°C), spell lengths (e.g. mean dry spell length) and cumulative measures (e.g. GDD — cumulative growing season degree days).
A crucial measure of rainfall extremes is the annual maximum daily rainfall. An example of what can be extracted from the UKCP09 WG is presented in Figure 3, showing the model performance in reproducing the median annual maximum one-day rainfall (Rmed).