Geography Reference
In-Depth Information
extremely cold and warm conditions, whereas for precipita-
tion only the 98th percentile was utilized. In the literature the
usage of 1st/99th percentile (instead of 2nd and 98th) is more
common. However, using somewhat lower thresholds can
yield more robust results. The index time series to deal with
are relatively short, e.g. one index value per season yields
150 values for the 150-year period. In this way three values
exceed the threshold (98th) instead of one (99th). Consider-
ing the characteristics of extremes connected to issues re-
lated to data quality this can be a useful measure.
Not all the indices are real extremes indices. Some of them
represent mean conditions to make it possible to illustrate
similarities and differences between changes in extremes and
in mean conditions. In addition to the mean indices, the fol-
lowing indices are not based on percentiles. The Heat Wave
Duration Index (HWDI), daily precipitation intensity indices
(SDII98 and SDII), highest 5-day total rainfall (R5d), high-
est daily rainfall (R1d) and the highest number of consecu-
tive dry days (CDD).
All indices were computed separately for each of the
3-month seasons (MAM = March-May, JJA = June-August,
SON = September-November, and DJF = December-Feb-
ruary). Further details on the indices used can be found in
Moberg et al. (2006) and Beniston et al. (2007).
Trend Analysis
The ordinary least squares method was used to estimate linear
trends of the temperature and precipitation indices. The trend
was computed for each particular period in question: 200
years, 150 years, and 100 years. For easy comparison among
different periods, all trends were given in unit/100years . The
significance of a particular trend estimated was determined
by a t-test on the estimated slope of the regression as done in
Moberg et al. (2006). The lag-1 autocorrelation in each series
has been taken into account to adjust the degrees of freedom
accordingly. Trends that are determined to be significant at
the 5 and 1 % level were flagged in diagrams and tables that
present the results.
Fig. 2.3 Overview map showing the meteorological precipitation ob-
servations used for each period and the regional division used for the
period 1901-2000
the frequency distribution of a variable can be captured. For
example, the 2 % highest values exceed the 98th percentile.
Some indices use thresholds calculated for a reference pe-
riod to be exceeded or fallen below. For those 1961-1990
was used as reference. For the temperature variables both
the 2nd and 98th percentile were chosen in order to capture
Search WWH ::

Custom Search