Climate Perturbation Sensitivity
A climate perturbation sensitivity method is introduced here in which
the current climate is perturbed based on projected future
climatological changes. In this method, a climate perturbation signal of
the future atmosphere is added to high-resolution baseline hourly
observations. The general perturbation approach, and the method used
here has two main assumptions: (i) GCM outputs for current and future
climates show relative changes rather than absolute changes in climate;
and (ii) the number of precipitation events is constant in current and
future climates (Semadeni-Davies et al., 2008). The perturbation method
used only modifies the observed past and does not consider future
changes in frequency and intensity of weather patterns. The assumption
of linear scaling used for temperature in the delta method may introduce
uncertainties for non-linear variables such as precipitation,
particularly for extremes (Kay et al., 2009). It is also assumed that
the basin vegetation, and in the case of Wolf Creek, permafrost
(Williams et al., 2015), will remain unchanged.
The range of annual perturbations in precipitation and warming
considered for this study is largely based on the atmospheric changes
estimated by the SRES A2 scenario and two RCPs. The climate dataset used
for the SRES A2 scenario was obtained from the North American Regional
Climate Change Assessment Program (NARCCAP). These simulations provide
climate data for regional climate models driven with GCM boundary
conditions (Mearns et al., 2007). The range of temperature and
precipitation perturbations was chosen based on the average climate
changes that were obtained for RCPs and for the eleven NARCCAP regional
climate models for the periods 2041-2070 minus 1971-2000. The climate
dataset used for RCP scenarios was adapted from the recent Fifth
Assessment Report of the Inter-governmental Panel on Climate Change
(IPCC) (Barros et al., 2014). These four RCPs corresponding to specific
radiative forcing values of 2.6, 4.5, 6.0, and 8.5
W/m2 were used as a basis for long-term and near-term
modelling experiments in climate change studies. For the southern Yukon
Territory (Wolf Creek) a warming of up to 2°C with an increase in the
annual precipitation of less than 10%, and a warming of up to 5°C with
a 20% increase in annual precipitation are projected based on RCP2.6
and 8.5 respectively. Similar warming with smaller precipitation
increases are expected in Marmot Creek and Reynolds Mountain. Most
modelled scenarios project the future climate to be wetter, but some
SRES scenarios (Moss et al., 2010) show regional decreases in
precipitation of up to 15% for the 2080s. Rather than following any
specific RCP or SRES, the sensitivity analysis spans potential changes
in air temperature and precipitation from all RCP and SRESs; perturbing
air temperature by 0°C to 5°C in 1°C intervals and precipitation by
–20% to +20% in 10% intervals. These changes were applied to
observations from all three basins.
The degree of hydrological sensitivity to climatic changes is evident in
the resulting shape and slope of contours of change in a variable
(Figure 2). The contours were estimated by linear interpolation between
the mean responses to the 30 combinations of warming (0°C to 5°C in
intervals of 1°C) and precipitation change (-20% to +20% in intervals
of 10%). When a hydrological variable is more sensitive to air
temperature increase or precipitation change, the contour line is
perpendicular to that axis (Temperature Figure 2a; Precipitation Figure
2b), when the variable is sensitive to a linear interaction of air
temperature and precipitation changes, there will be a slope in the
contour line (Figure 2c), and if the interaction is complex, the slope
and the contours will not be straight lines (Figure 2d). Applying the
same ranges of change in air temperature and precipitation to each of
the three basins allows direct comparison of the responses of the
simulation model for each basin. Different combinations of warming and
precipitation change make it possible to estimate how much additional
precipitation is needed to offset the impacts of a specific air
temperature increase on annual runoff and peak snowpack. The additional
precipitation increases were estimated based on the interpolation of the
two contour lines above and below the present climate values.
To compare and contrast snow accumulation and ablation amongst sites,
hydrological response units (HRUs) in Wolf Creek are grouped into
alpine, shrub tundra, and forest zones based on biomes and elevation
bands; in Marmot Creek the HRU are grouped into alpine, treeline
(affected by blowing snow from alpine), forest, and forest clearings;
and in Reynolds Mountain the HRU are grouped into source, sink,
interception, and sheltered snow regimes (Figure 1). The mean annual
peak snow accumulation is defined as the average maximum snow water
equivalent (SWE) over the hydrological year and occurs in March or April
in these basins.
The probability density function (PDF) of simulated hourly SWE to
warming air temperatures and precipitation changes was used to compare
the effects of perturbed forcing meteorology on snow regime. The kernel
density estimation (a non-parametric approach) was used. This estimation
is based on a normal kernel function and a window parameter (bandwidth)
that is a function of the length of the time series (Wolf Creek, n
= 18 years × 365 days × 24 hours ; Marmot Creek, n = 9 ×
365 × 24 ; Reynolds Mountain, n = 25 × 365 × 24 ). The
Kolmogorov-Smirnov (K–S) test (Massey Jr., 195) was used to compare the
simulations to observations using a significance level of ≤ 0.05.
Sensitivities of snow and streamflow regimes per 1°C warming were
estimated by averaging responses to the 30 combinations of warming and
precipitation change in the three mountain basins. Sensitivities of five
main characteristics that describe a basin snow regime were
investigated. These characteristics are the timing of snowcover
initiation (snow season start), snow-free date (snow season end),
duration of the snow season, duration of snowmelt period, and magnitude
of the peak snowpack. The duration of the snow season is the difference
between the date of snowcover initiation and the date the basin becomes
snow free. The duration of the melt period is the difference between the
date of peak SWE and the date the basin becomes snow free.