# Analysing the information

We have presented our analysis in three ways.

- Firstly, we calculated linear trends for each variable for Scotland as a whole and for three regions (see figure 1). We did this for annual and seasonal average periods. Each region encircles an area of similar characteristics in terms of climate. For the purposes of this report we have been termed them North, West and East Scotland, using the capital letter to indicate one of the three regions rather than an area of Scotland. We used a method called linear regression to work out the trend in each variable. We then used the average rate of change from the linear regression multiplied by the length of the data period to provide a clear measure of change since the start of the period. We used a statistical test, known as the Mann-Kendall test, to show whether significant changes had taken place. If these trends are statistically significant we show them in bold (significant at the 5% level).
- Secondly, we produce graphs of the time-series for each variable for each of the three regions of Scotland. These show the year to year variability for each variable. We also include a smoothed version to show the running average as an indication of the longer-term trends and variations (see figure 3 for an example).
- Finally, where appropriate, we produced a map of trends we have already seen showing the spatial variation which was not easy to see from the national average figures often presented elsewhere. (See figure 4 as an example).

We analysed the information using linear regression, and tested the significance of the trends using the non-parametric Mann-Kendall tau test (Sneyers, 1990). The Mann-Kendall test is a rank-based non-parametric test. For each value in the series, we work out the number of values before it which are higher than it to gain evidence of a trend in the series. You need to be careful when interpreting results from linear trends as the assumption that the trend is linear is not always valid. However, the trends are often close to linear, and the combination of linear trends with the Mann-Kendall significance test has been widely used when analysing of climate trends (for example, Domroes and El-Tantawi, 2005; Shen et al, 2005).

In this study the spatial analysis of growing season length, start and end dates is slightly different to that of the other calculated variables, in that it is based directly on the baseline recorded climate of the UK dataset provided by UKCIP with the UKCIP02 scenarios. This information provides monthly average temperatures rather than the daily values needed to work out growing season, so we used a method called sine curve interpolation to estimate daily values based on the method of Brooks (1943). We then used this daily information within the calculations for growing season length, start and end dates. Also, the standard dataset UKCIP02 is for 1961 to 2000, but for this application we have updated it to 2004.

**For a more detailed technical analysis please see the accompanying SNIFFER publication “Patterns of Climate Change across Scotland: Technical Report.”**