Jökull - 01.01.2013, Blaðsíða 76

P. Crochet
tude (AMF) and timing (DAMF) were defined as the
largest daily flow discharge occurring in the hydrolog-
ical year and the corresponding time within the year.
The spring flow peak magnitude (SMF) and timing
(DSMF) were defined as the largest daily flow dis-
charge occurring in the period from 1 March to 16
July and the corresponding time within the year, as in
Jónsdóttir et al. (2008). The center of volume date is
defined as the date on which half or more of the total
volume of water for a given period of time passed a
river gauging station (Hodgkins et al. 2003). It is ex-
pected to be a more robust indicator of the timing of
the bulk of high-flows than the flow peak. This con-
cept was applied to annual streamflow measurements
(CTQ) and extended to annual snowmelt (CTS) and
total water input (CTW) by summing the respective
daily volumes over the hydrological year. Other in-
dices like the flow-weighted timing or center of mass
of streamflow can be used to detect streamflow tim-
ing changes (e.g. Stewart et al., 2004). All dates are
expressed as Julian days since Sept. 1st of the hydro-
logical year.
The flood occurrence rate (FOR) represents the
average number of flood events per year and was
analysed using a peak over threshold (POT) sampling
technique. The magnitude and date of occurrence of
all independent streamflow peaks exceeding a certain
threshold were extracted from the daily discharge se-
ries. The selected flood events are assumed to be in-
dependent and distributed as a Poisson process (Lang
et al., 1999). The POT modelling approach comple-
ments the AMF analysis. The two main difficulties
concern the choice of the threshold and the definition
of a criteria for selecting independent peaks, espe-
cially when they are clustered. Various methods have
been suggested to extract independent peaks. Bayliss
and Jones (1993) used the criteria that the time period
between two consecutive peaks must be at least three
times the time of the rising limb and the minimum
discharge between two peaks must be less than 2/3 of
the first peak discharge. Similar criteria were used by
Silva et al. (2012). In a POT analysis considering 21
catchments around the world, Svensson et al. (2004)
used a peak separation time between peaks of at least
five days for catchments smaller than 45000 km2, so
as to allow for the flow to recede appreciably between
peaks. Lang et al. (1999) presented a review of ob-
served practices for POT analysis and suggested com-
prehensive guidelines for threshold selection, based
on several tests. This approach was adopted in this
study using the tests implemented in the R-package (R
Development Core Team, 2010) POT (Ribatet, 2006).
The criteria used to extract independent peaks are sim-
ilar to those in Bayliss and Jones (1993) and Silva et
al. (2012), except that as the streamflow data used in
this study had a daily temporal resolution, an exact
evaluation of the time to peak was not possible, but
assumed to be one day, at most. The following crite-
ria were used:
- The time period between two consecutive peaks
must be more than three days.
- The minimum discharge between two peaks
must be less than 2/3 of the first peak discharge.
Two different thresholds were selected and respec-
tive POT series extracted, leading to an average of
three independent peaks per year (POT-3) and an av-
erage of two independent peaks per year (POT-2) at
each gauging station.
Method of analysis
A common period of analysis was defined for all
catchments, for which temperature, precipitation and
discharge data were available, i.e. 35 years, from
Sept. 1st 1971 to August 31st 2006. The hydrological
series (Table 2) were divided into four subsets, con-
stituting the 25% coldest, 25% warmest, 25% wettest
and 25% driest years, with 9 years per subset. Note
that the years in each subset were not necessarily the
same for the various catchments, and that they are not
mutually exclusive so that for some catchments, some
years belonged to more than one subset. In the fol-
lowing text, the different subsets will be referred to
as the “cold”, “warm”, “wet” and “dry” subsets, re-
spectively. The decade 2001–2010 was the warmest
of the past 60 years in Iceland (Crochet and Jóhann-
esson, 2011) and all catchments have between three
to five of their warmest years within the 2002–2006
period. The sensitivity of the river basins to tem-
perature variations was examined by comparing the
hydrological characteristics between warm and cold
76 JÖKULL No. 63, 2013