Scottish Road Network Climate Change Study: UKCP09 update Autumn 2011
4 Methodology
The UKCP09 projections for the more general variables such as mean annual\seasonal temperature or precipitation have been made available as pre-prepared images that can be downloaded from the UKCP09 website. However many of the climatic variables of importance to this study have not been the subject of these pre-prepared graphics and it has been necessary to undertake bespoke analysis of climate change time-series generated by the UKCP09 Weather Generator. The Weather Generator tool creates synthetic time-series of weather variables at 5 km resolution, which are consistent with the climate projections. It cannot provide regionally varying time-series. This section describes how the output from the UKCP09 Weather Generator was used to derive the necessary understanding.
The variables derived using Weather Generator data, together with their definitions, are listed in Table 4.1.
| Variable | Definition |
|---|---|
| Frost days | Average annual number of days with min temperature below 0oC |
| Freeze-thaw | Average annual number of days when temperature fluctuates above 2°C and below -1°C |
| Winter duration | Road managers currently work to a winter period defined as 1 October to 15 May |
| Hot days | Average annual number of days with max temperature above i) 25oC and ii) 30oC |
| Growing season | Growing season start: This is the start date for the growing season (calculated as Julian days), where the growing season is assumed to start on the 5th consecutive day with a mean daily temperature of 5°C or greater. Growing season end: This is the end date for the growing season (calculated in Julian days), where the growing season is assumed to end on the 5th consecutive day with a mean temperature of 5°C or less. Growing season length: the number of days between the start and end of the growing season |
| 10-year rainfall | The extreme daily rainfall depth that has a return period of 1 in 10 years. (Equivalent to an annual occurrence probability of 0.1) |
| 2-year rainfall | The extreme daily rainfall depth that has a return period of 1 in 2 years. (Equivalent to an annual occurrence probability of 0.5) |
| Soil moisture deficit | Average annual pattern of the development and replenishment of soil moisture deficit for a grass vegetation cover |
| Groundwater recharge | Average annual pattern of recharge periods. Determined as the period when the soil moisture deficit has been replenished above field capacity of the soil. |
4.1 Approach
Three locations (Glasgow, Aviemore and Dundee) were selected as being broadly representative of the climatic and geographic range across Scotland. For each site the Weather Generator was run for the future time horizon of interest to produce a series of simulated data sets. This large amount of simulated data is provided both for the future time period requested and also for the simulated baseline 1961-1990 period. Table 4.2 summaries the data sets used.
| Location | Emission scenario | ||
|---|---|---|---|
| Low | Medium | High | |
| Glasgow | 2020s & 2080s | 2020s & 2080s | 2020s & 2080s |
| Aviemore | 2020s & 2080s | 2020s & 2080s | 2020s & 2080s |
| Dundee | 2020s & 2080s | 2020s & 2080s | 2020s & 2080s |
Each of the 18 data sets has the equivalent of 3000 years of daily data. Given the large quantity of data, bespoke computer programmes were developed to manage the data and also to undertake the analysis for each of the climatic variables given in Table 4.1. A summary of the function of each model is presented in the list below:
- Days of frost
Daily minimum temperature 30-year time series were extracted from the Weather Generator output. Occurrence of temperatures < 0°C were established for the baseline and future scenarios. Probability analysis was undertaken to provide the probability density functions and to supply the 10, 50 and 90 percentile values of the number of frost days and the change between baseline and scenario runs. - Days of freeze-thaw
Daily maximum and minimum temperature 30-year time series were extracted from the Weather Generator output. The model counted the days of occurrence of maximum temperature above 2°C and minimum temperature below -1°C (also combining temperature estimates of the previous and following day) for the baseline and for the future scenario. Probability analysis was undertaken to provide the probability density functions and to supply the 10, 50 and 90 percentile values of the number of freeze thaw events and the change between baseline and scenario runs.
Daily minimum and maximum temperature doesn't actually provide enough information to properly determine the number of freeze-thaw cycles. The measure described is therefore a surrogate which will tend to over estimate the number of events. Nevertheless it should provide a reasonable indication of the relative change in the number of freeze-thaw events. - Winter duration
The baseline winter period was set as 1 October to 15 May. The variable analysed was sub-zero temperature days. For each location, 100 baseline and 100 scenario runs of 30-year daily simulated data was used. The average occurrence of sub-zero temperatures per calendar day was derived for the baseline and the 2080s medium scenario. The start and end thresholds for the winter period for each location were set to mimic those given by the 1 October beginning date and the 15 May end date. From this the predicted shift in the winter period was obtained. - Hot days
Daily maximum temperature 30-year time series were extracted from the Weather Generator output. Occurrence of temperatures > 30°C (or 25°C) were established for the baseline and future scenarios. Probability analysis was undertaken to provide the probability density functions and to supply the 10, 50 and 90 percentile values of the number of hot days and the change between baseline and scenario runs. - Growing Season
Daily maximum and minimum temperature 30-year time series were extracted from the Weather Generator output. The mean temperature was calculated as an average of the two values . The model determined the start and finish day as prescribed in Table 4.1 for each of the 100 runs for the baseline and future scenario. Probability analysis was undertaken to provide the probability density functions and to supply the 10, 50 and 90 percentile values related to the growing season[2]. - Extreme Rainfall analysis model
Daily rainfall 30-year time series were extracted from the Weather Generator output. The model carries out the extreme value frequency analysis on each of the 30-year time series from which the 10-year rainfall (or the 2-year rainfall) is derived. Probability analysis was undertaken to provide the probability density functions and to supply the 10, 50 and 90 percentile values of the 10-year (or 2-year) extreme daily rainfall depth and the change between baseline and scenario runs. - Soil moisture deficit and groundwater recharge
Daily potential evaporation and rainfall were extracted from the Weather Generator output. These form the input to the HYLUC soil water balance model (Price, 2000) which was run for a lowland grass cover. Each of the paired time-series simulations were manually put through the model and the resultant average soil moisture deficit profile for each run calculated. The separate profiles were used to determine the 10, 50 and 90 percentile values. Similarly the period of recharge was determined as the period when the soil moisture deficits were replenished.
The UKCP09 team recommend analysing actual observed data to check the reliability of the baseline output from the Weather Generator[3]. Greater confidence in the performance of the Weather Generator is gained if it can simulate the current climate accurately. Where this does not occur the findings and conclusions need to be treated with caution. To this end the observed climate data for Paisley was obtained as a surrogate for Glasgow and the variables of interest calculated from this record for comparison. Similarly both the 10-year and 2-year extreme rainfall estimates for each of the 3 target sites was obtained from the Flood Estimation Handbook software and compared to those values derived from the Weather Generator. The results of these comparisons are discussed in the relevant sections.
The projected changes to the various climate variables, resulting from the analysis of the weather generator data, have been tabulated in the following sections in two ways:
i) Provision of the baseline and future 10%, 50% and 90%-tile values. This allows the reader to understand the magnitude of the particular climatic variable being considered in the appropriate units of measurement. The actual values of the weather generator data are presented in the table.
ii) Provision of the projected change in the climatic variable compared to the baseline. This allows the reader to understand the magnitude of the change of the variable being considered. This is derived from multiple runs of the weather generator for the variable of interest. The predicted changes can be used to create a probabilistic distribution of the change in that variable.
Although there will be reasonable consistency between these two sets of tables it is possible for slight "apparent" differences to occur. This is a function of the slightly different basis of the calculations and should not be interpreted as inconsistencies. Both ways of presenting the data are valid and the principles under discussions are not affected by these slight differences.