PIAHSProceedings of the International Association of Hydrological SciencesPIAHSProc. IAHS2199-899XCopernicus GmbHGöttingen, Germany10.5194/piahs-369-157-2015Extreme values of snow-related variables in Mediterranean regions:
trends and long-term forecasting in Sierra Nevada (Spain)Pérez-PalazónM. J.mj.perez@uco.esPimentelR.https://orcid.org/0000-0001-6990-4874HerreroJ.https://orcid.org/0000-0002-5741-6301AguilarC.PeralesJ. M.PoloM. J.https://orcid.org/0000-0002-6296-2198Fluvial Dynamics and Hydrology-Andalusian Institute of Earth System Research, University of Cordoba, Cordoba, SpainFluvial Dynamics and Hydrology-Andalusian Institute of Earth System Research, University of Granada, Granada, SpainM. J. Pérez-Palazón (mj.perez@uco.es)11June201536936915716221April201521April2015This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://piahs.copernicus.org/articles/369/157/2015/piahs-369-157-2015.htmlThe full text article is available as a PDF file from https://piahs.copernicus.org/articles/369/157/2015/piahs-369-157-2015.pdf
Mountain areas in Mediterranean regions constitute key monitoring points for
climate variability and its impacts, but long time datasets are not always
available due to the difficult access to high areas, relevant for capturing
temperature and precipitation regimes, and the predominance of cloudy remote
sensing images during the snow season. Sierra Nevada National Park (South
Spain), with altitudes higher than 3500 m a.s.l., is part of the Global
Change in Mountain Regions network. Snow occurrence just 40 km from the
seaside determines a wide range of biodiversity, a snowmelt fluvial regime,
and the associated ecosystem services. This work presents the local trend
analysis of weather variables at this area together with additional
snow-related variables. For this, long term point and distributed
observations from weather stations and remote sensing sources were studied
and used as input and calibration datasets of a physically based snow model
to derive long term series of mean and maximum daily fraction of snow covered
area, annual number of days with snow, annual number of days with
precipitation, mean and maximum mean daily snow water equivalent, and
snowmelt and evaporation volumes. The joint analysis of weather and snow
variables showed a decrease trend in the persistence and extent of the snow
cover area. The precipitation regime, rather than the temperature trend,
seems to be the most relevant driver on the snow regime forcing in
Mediterranean areas. This poses a constraint for rigorous scenario analysis
in these regions, since the precipitation pattern is poorly approximated by
climatic models in these regions.
Introduction
Climate variability impact on the hydrological regime is more
evident over mountain regions due their particular extreme conditions
(Beniston, 2003; IPCC, 2007). Hence, a continuous monitoring of snow state
variables can help to assess climate variability and early detection of
shifts at significant time scales. For example, the decrease of snow cover
area is an indicator of a warming climate phase conditions. Moreover, in
regions where the typical alpine-mountain climate is modified by additional
meteorological drivers(i.e. the recurrence of drought periods and torrential
rainfall events), these impacts are enhanced, which makes them crucial areas
to monitor these climate variations. This is the case of Mediterranean
mountainous areas, where alpine and semiarid conditions coexist (Giorgi,
2006).
However, snow monitoring constitutes arduous work. On one hand, the limited
accessibility to these sites during the snow season, and the hard working
conditions for the instrumentation, make in situ continuous monitoring
systems difficult to maintain (De Walle and Rango, 2008). On the other hand,
the potential use of satellite remote sensing information (i.e. Landsat TM,
ETM+ and OLI with temporal resolution of 16 days and spatial resolution of
30m×30m, MODIS, daily images with
250m×250m spatial resolution, and NOAA,
1km×1km daily images) to capture snow evolution at
medium and large scales is limited to recent decades, poses a constraint for
snow applications in heterogeneous areas due to their fixed and not always
high enough spatial resolution, and is not feasible during a significant
fraction of the snow season because of the recurrent presence of cloud cover.
Therefore, a lack of detailed snow information is common over mountainous
areas, especially at high altitudes.
The use of physically based snowmelt-accumulation model, locally calibrated
and validated by means of remotely sensed data, allows estimating the
snowpack evolution and its significant state variables during these gap
periods, over the non-monitored areas, and beyond the snow monitoring period
(Pimentel et al., 2015). For that, a correct representation of the
meteorological forcings, the main inputs in snow modelling, is crucial to
obtain accurate snow simulations, the weather data series available in the
study area, and their quality, being a relevant limitation for a satisfactory
model performance. Weather variables long-term series can, thus, be used to
generate long-term snow-related variables series, which allows the
understanding of the past snow dynamics, and its uncertainty, the estimation
of future trends, and the forecasting of snow behaviour under given climatic
scenarios. The objective of this work is to study the meteorological patterns
in Sierra Nevada Mountain, an alpine climate region in Southern Spain, in
relation to selected snow-related variables to obtain long-term trends in
this area as a basis to assess future climate scenarios. Special emphasis has
been done in the analysis of the occurrence of extreme events and its impact
on snow dynamics.
Study site and available data
Sierra Nevada Mountains, in Southern Spain, are a linear mountainous range
parallel to the shoreline of the Mediterranean Sea (Fig. 1), where the
highest summits of the Iberian Peninsula can be found (Mulhacen peak,
3479 m). It is the second highest range in Europe, only surpassed by the
Alps. Consequently, snow usually appears throughout the year in the areas
above 2000 m. Its proximity to the sea, only 40 km south, generates a very
specific climate as the result of the interaction between sea and mountain
conditions. The typical alpine climate is then modified by the Mediterranean
surrounding features and, thus, snow is significantly affected. Annual
precipitation fluctuates widely and can range from 300 to 1000 mm, with a
high spatial variability throughout the area due to topographic effects. The
average temperature ranges from -5 to 5 ∘C during the snow season,
although minimum values of -20 ∘C can be found at certain times in
winter. Due to these particular conditions, it belongs to the Global Change
in Mountain Regions network (GLOCHAMORE) and it is recognizes as both Natural
and National Parks.
Location of Sierra Nevada Mountain Range in Spain; limits of the
protected areas: National Park (dark grey coloured area) and Natural Park
(light grey coloured area); location of the different weather station (black
dots); and limits of the five regions in which the study area has been
divided: R1 – Adra, R2 – Andarax, R3 – Fardes, R4 – Genil and R5 –
Guadalfeo.
Different weather station networks are available in this area, lacking, in
general, data over 2000 m a.s.l.; in this work, validated daily datasets of
precipitation, temperature, wind velocity, and relative humidity from
selected stations of the Spanish Meteorological Agency (AEMET) were employed
as input of a distributed physically based snow model (Fig. 1), calibrated
and validated in this region in previous works (Herrero et al., 2010;
Pimentel et al., 2014; Pérez-Palazón et al., 2014). The reference
period 1960–2000 was analysed for long-term trends and basis of future
scenarios assessment. The study is limited to area in which five different
regions were identified to assess the local impact of weather trends on snow
(Fig. 1).
Methods
The average annual precipitation and mean, maximum and minimum average daily
temperature for the reference period, 1960–2000, were obtained over the
study area in Sierra Nevada, together with selected annual snow
related-variables that were simulated during this period using the physical
and distributed hydrological model, WiMMed (Herrero et al., 2011;
http://www.ugr.es/~ herrero/wimmed/), developed and
calibrated in this region (Herrero et al., 2009, 2011; Millares et al., 2009;
Aguilar et al., 2010; Aguilar and Polo, 2011; Pimentel et al., 2014). The
model includes specific spatial interpolation algorithms for the weather
variables to include the abrupt topography of this area (Herrero et
al., 2010), and an energy and water balance snow model. Long-term
trends of both the meteorological and snow-related variables were analysed
emphasizing the importance of extreme precipitation events.
Snow modelling
The snowmelt-accumulation model for Mediterranean sites developed by Herrero
et al. (2009) is a physical model based on a point mass and energy balance,
which is extended to a distributed way by means of depletion curves (Herrero
et al., 2011; Pimentel et al., 2014).
The model assumes a horizontally uniform snow cover distributed in one
vertical layer. In the control volume defined by the snow column per unit
area, which has the atmosphere as an upper boundary and the ground as a
lower one, the water mass and energy balance can be expressed by:
dSWEdt=R-E+W-M.dSWE⋅udt=dUdt=K+L+H+G+R⋅uR-E⋅uE+W⋅uW-M⋅uM,
where SWE (snow water equivalent) is the water mass in the snow column, and
u is the internal energy per unit of mass (U for total internal energy).
In the mass balance, R defines the precipitation rate; E is water vapour
diffusion rate (evaporation/condensation); W represents the mass transport
rate due to wind; and M is the melting water rate. On the other hand,
regarding energy fluxes, K is the solar or short wave radiation; L the
thermal or long-wave radiation; H the exchange of sensible heat with the
atmosphere rate; G the heat exchange with the soil rate; and
uR, uE, uW and uM are the
advective heat rate terms associated with each one of the mass fluxes
involved in Eq. (1).
This approach permits an easy extension to a distributed model by means of
making calculations simultaneously in each cell. However, a direct extension
cannot be done when the cell area is not completely covered by snow. In such
cases, the parameterization of subgrid processes is made by including
depletion curves (Luce and Tarboton, 2004; Herrero et al., 2010; Pimentel et
al., 2014).
The point model was calibrated and validated by means of local in situ
measurements (Herrero et al., 2009) and the distibuted model thanks to remote
sensing information (Herrero et al., 2010; Pimentel et al., 2014).
Long term data analysis
Annual and decadal regimes for the reference 40-year period (1960–2000) were
analysed for the average annual precipitation (P) and mean, maximum and
minimum average daily temperature (Tmean, Tmax,
Tmin) in the area, obtained from daily measurements at the selected
weather stations, and the following snow related variables obtained from the
validated snow model: (1) mean and maximum daily fraction of snow covered
area (SCmean and SCmax), obtained as mean and maximum
value of daily percentage of snow covered area over the whole area
(m2m-2); (2) annual number of days with snow (nday
with snow), calculated as the number of day in which the snow exceeds
10th percentile of SCmean distribution (days); (3) annual number of
days with precipitation (nday with precipitation), counted as
days with precipitation greater than 1 mmday-1 (days); and
(4) mean and maximum mean daily snow water equivalent (SWEmean
and SWEmax), calculated as mean and maximum values of daily average
SWE (mm). Their individual temporal trend and main statistic descriptors were
performed. The significance of these temporal trends was tested by means of
Mann–Kendall test (Gibbons and Chakraborti, 2010). The test assumes as null
hypothesis: non-monotonic trend (H0), which was tested versus the
alternative hypothesis: presence of monotonic trend (H1). A level of
significance (α) equal to 0.05 is used to define the rejection region.
The snow regime dependence on weather variables was also assessed.
Extreme data analysis
The extreme data analysis was performed on a precipitation event definition
basis. A precipitation event is defined as a period in which rainfall over
1 mmday-1 is registered at some weather station within the study
area (Polo et al., 2010). The event is characterized by two main variables:
D, duration (days) and Pevent, cumulative precipitation over the
study area (mm). Different snow variables selected from the long term
analysis results were also obtained for each event in the resulting 40-year
event series at the region. The extreme data analysis was done on the subset
events over the 95th percentile of Pevent.
Results
This section shows the different results of the long-term and extreme data
analysis performed over selected annual and decadal variables, average over
the study area (4584 km2). The distribution over the five identified
regions in the area (Fig. 1) is also presented to further asses some of the
results.
Long term data analysisAverage annual precipitation and temperature during 1960–2000
On a regional basis, the 40-year mean annual values of P and T are 510
and 93 mmyear-1, for the total precipitation and snowfall,
respectively, and 26, 12.5 and 0.4 ∘C for the average daily maximum,
mean and minimum temperature, respectively. Figure 2a and b show the
evolution during the reference period (1960–2000) of the annual
precipitation (total precipitation and snowfall) and the maximum, mean and
minimum daily temperature averaged over the study area by the model from the
available daily datasets measured at the weather stations in Fig. 1. The
decadal trend for these variables has also been included (Fig. 2c, d).
Globally decreasing annual rates were obtained for both the total
precipitation and snowfall (Fig. 2a), with values of 4.136 and
1.250 mmyear-1, respectively. Their decadal cumulative values
analysis (Fig. 2c) shows the same trend, with snowfall dropping from
108 mmyear-1 in the first decade to the final decadal
75 mmyear-1. A greater dispersion of both annual variables can
be observed in this last decade.
In the case of temperature, an increasing trend of the annual mean of the
daily mean and maximum values was found (Fig. 2c), with global rates of 0.035
and 0.021 ∘Cyear-1, respectively. On the contrary, a
decreasing global rate of 0.009 ∘Cyear-1 can be observed
in the minimum values. The same trends were found in the decadal analysis
(Fig. 2d), without any clear pattern in the dispersion of the annual values
for each decade.
Evolution of the annual cumulative precipitation (a) and
the annual mean, maximum and minimum daily temperature (b), averaged
over the study area. Boxplot of the decadal mean annual cumulative
precipitation (c) and the decadal mean maximum, mean and minimum
daily temperature (d); the central mark represents the median, the
upper and lower edges of the box are the 25th and 75th percentiles,
respectively, the whiskers extends to the most extreme data point not
considered outlier, crosses are the outliers, and dots represent the decadal
mean values.
Evolution of the annual mean and maximum daily fraction of snow
covered area over the whole study area (a). Boxplot of the decadal
maximum (b) and mean (c) daily fraction of snow covered
area; the central mark represents the median, the upper and lower edges of
the box are the 25th and 75th percentiles, respectively, the whiskers extends
to the most extreme data point not considered outlier, crosses are the
outliers, and dots represent the decadal mean values.
Evolution of the annual number of days with snow and with
precipitation over the whole study area (a). Boxplot of the days
with precipitation (b) and the days with snow (c); the
central mark represents the median, the upper and lower edges of the box are
the 25th and 75th percentiles, respectively, the whiskers extends to the most
extreme data point not considered outlier, crosses are the outliers, and dots
represent the decadal mean values.
Evolution of the annual mean of the maximum and mean daily SWE over
the study area (a). Boxplot of the decadal maximum (b) and
mean (c) daily SWE; the central mark represents the median, the
upper and lower edges of the box are the 25th and 75th percentiles,
respectively, the whiskers extends to the most extreme data point not
considered outlier, crosses are the outliers, and dots represent the decadal
mean values.
Snow-related variables
The evolution of SCmean and SCmax at the study region is
represented in Fig. 3a, which 40-year means values are 0.055 and
0.38 m2m-2, corresponding with areas of 254.02 and
1736.50 km2, respectively. A decreasing trend is found for the mean
value (0.0007 m2m-2year-1) against the growing trend found
in the case of the maximum value trend
(0.0007 m2m-2year-1). Figure 3b and c show the decadal
analysis for both variables; where similar trends are also observed, with
values over 0.35 and between 0.05 and 0.06 for the decadal SCmax
and SCmin, respectively. Nevertheless, a high dispersion is found
between decades.
Figure 4a includes the evolution of the annual number of days with
precipitation and with snow over the whole study area. Mean 40-year values of
173 and 40 days are found respectively. Both variables experiment a global
decreasing trend, which represents a loss of 42 days with precipitation and
11 days with snow over the area during the analysed period. Figure 4b and c
also show decreasing trend with variable dispersion between decades, which
for example is very high in the last decades for days with precipitation.
The evolution of the annual mean of the maximum and mean daily SWE values is
represented in Fig. 5a, in which a high variability can be observed, with
mean 40-year values of 38.73 and 10.91 mm, respectively. The global
decreasing trend obtained for both variables is higher for SWEmax,
which doubles the rate obtained for the SWEmean (0.625 against
0.235 mmyear-1). The boxplots (Fig. 5b, c) with the decadal
analysis also shows decreasing trends, with a high dispersion in both the
first and last decades; this last decade shows the highest value for the
SWEmax in the reference period, and the widest interval of values
for both SWEmax and SWEmean variables.
Extreme data analysis
Following the extreme event definition in Sect. 3, 82 extreme precipitation
events were identified during the reference period, with a mean duration and
cumulative precipitation of 12 daysevent-1 and
120 mmevent-1. Figure 6a shows the evolution of the annual
number of extreme events throughout the study period, with a mean value of
2.1 eventsyear-1 in a range of 0–5 eventsyear-1.
The relation between the annual cumulative precipitation associated to
extreme events and the annual cumulative precipitation for each year is
included in Fig. 6b; again, a globally decreasing rate was found
(0.006 mmmm-1year-1). Finally, the torrential nature of
these events can be analysed from Fig. 6c quantifying the rainfall intensity
for the extreme event per year; a global increasing trend is found of
0.061 mmday-1.
Discussion
The Mann–Kendall test performed to assess the significance of the trend
shows that only the trends found for, P, snowfall, Tmean,
SCmean, nday with precipitation, SWEmax and
SWEmean are statistically significant (Table 1).
The results show a clear decrease of the annual snowfall on a regional basis
(-1.25 mmyear-1), which is also reflected in a reduction of
the annual mean SWE (0.23 mmyear-1).Therefore, the shift in the
precipitation regime seems to be the most determining driver of the snow
variables annual trends. The annual snowfall correlates linearly with P(r2=0.50), but this correlation is very poor with Tmean(r2=0.18),
although some increasing trend could be observed with decreasing annual mean
daily maximum temperature values. This behaviour differs from the trends
observed in the Alps, where precipitation and snow trends have opposing
trends and seem to be more related with temperature regime (Laternser and
Schneebeli, 2003). The annual fraction of snowfall over the total
precipitation in the study area does not directly correlate with any of the
weather variables analysed in this work.
Annual evolution of the number of extreme event in the 40-year study
period (a); relation between annual cumulative precipitation
associated with extreme events and the annual cumulative total
precipitation (b); and intensity for the extreme
events (c).
Results of Mann–Kendall test with a level of significance α=0.05 realized for each one of the variables analyzed to evaluate the
significance of the trends found.
The annual mean of SWEmax experiments a higher decrease than the
daily SWEmean value. Nevertheless, this is not accompanied by a
great reduction in the annual mean of the maximum snow covered area. Thus, an
average regional reduction in the thickness of the snowpack must be
associated to these peak moments, which supports the precipitation regime
being the main responsible of the snow dynamics over this reference period.
The torrential events are more common in the last decades, as the intensity
of the extreme events showed. This is also appreciable in the greater
dispersion associated to the extreme events variables found when compared to
most of the variables during this last decade. The non-linearity of the snow
behaviour makes it necessary a sound modelling of these regions if future
scenarios assessment is to be performed.
During this reference period, a global decrease in the annual precipitation
rate and an increase in the annual mean temperature values has been
quantified. The dataset of temperature and precipitation downscaling for the
different IPCC scenarios over the study area for this reference period
(1960–2000) overestimate both the precipitation and temperature annual rates
of decrease and increase to -2.1 mmyear-1 and 0.03 ∘C,
and thus fail to reproduce the measured values. The direct use of these
projected data sets is never advisable but for snow regions must be made with
extreme caution. The combination of long term weather variables observations
and snow modelling constitutes a sound alternative for historical time series
trends over mountainous areas and is an adequate basis for correcting and
simulating future scenario projections.
Conclusions
The joint analysis of weather and snow variables showed a
decrease trend in the extent and persistence of the snow covered area over
Sierra Nevada range. The precipitation rather than the temperature regime
seems to be the most relevant driver on the snow regime forcing in
Mediterranean areas over the reference period (1960–2000). This poses a
constraint for rigorous scenario analysis in these regions, since the
precipitation pattern is poorly approximated by climatic models in these
regions and further assessment by means of a sound snow modelling is
required.
Acknowledgements
This work has been supported by the Spanish Ministry of Agriculture, Food and
Environment (Biodiversity Foundation, Project “Study of the effect of global
change on snow and high mountain hydrology in Sierra Nevada National Park”)
and the Spanish Ministry of Science and Innovation (Research Project CGL
2011-25632, “Snow dynamics in Mediterranean regions and its modelling at
different scales. Implications for water management”). Moreover the present
work was partially developed within the framework of the Panta Rhei Research
Initiative of the International Association of Hydrological Sciences (IAHS)
(Working Group Water and energy fluxes in a changing environment). Finally,
we thank the support provided by the National and Natural Park of Sierra
Nevada.
ReferencesAguilar, C. and Polo, M. J.: Generating reference evapotranspiration surfaces
from the Hargreaves equation at watershed scale, Hydrol. Earth Syst. Sci.,
15, 2495–2508, 10.5194/hess-15-2495-2011, 2011.Aguilar, C., Herrero, J., and Polo, M. J.: Topographic effects on solar
radiation distribution in mountainous watersheds and their influence on
reference evapotranspiration estimates at watershed scale, Hydrol. Earth
Syst. Sci., 14, 2479–2494, 10.5194/hess-14-2479-2010, 2010.
Beniston, M.: Climatic change in mountain regions: a review of possible
impacts, Climate Change, 59, 5–31, 2003.
DeWalle, D. and Rango, A.: Principles of snow hydrology, Cambridge University
Press, Cambridge, UK, 428 pp., 2008.
Gibbons, J. D. and Chakraborti, S.: Nonparametric statistical inference,
Taylor & Francis/CRC Press, Boca Raton, USA, 2010.Giorgi, F.: Climate change hot-spots, Geophys. Res. Lett., 33, L08707,
10.1029/2006GL025734, 2006.
Herrero, J., Polo, M. J., Moñino, A., and, Losada, M. A.: An energy
balance snowmelt model in a Mediterranean site, J. Hydrol., 371, 98–107,
2009.
Herrero, J., Aguilar, C., Polo, M. J., and Losada, M. A.: Mapping of
meteorological variables for runoff generation forecast in distributed
hydrological modeling, Proceeding Hydraulic Measurements and Experimental
Methods, New York, 606–611, 2010.
Herrero, J., Polo, M. J., and Losada, M. A.: Snow evolution in Sierra Nevada
(Spain) from an energy balance model validated with Landsat TM data,
Proceedings SPIE 8531, Remote Sensing for Agriculture, Ecosystems, and
Hydrology XIV, Prague, Czech Republic, 2011.
IPCC: Climate Change 2007: Synthesis Report, Contribution of Working Groups
I, II and III to the Fourth Assessment Report of the Intergovernmental Panel
on Climate Change, Core Writing Team, Pachauri, R. K, and Reisinger, A.
(Eds.), IPCC, Geneva, Switzerland, 104 pp., 2007.
Laternser, M. and Schneebeli, M.: Long-term snow climate trends snow climate
trends of the Swiss Alps (1931–99), Int. J. Climatol., 23, 733–750, 2003.
Luce, C. H. and Tarboton, D. G.: The application of depletion curves for
parameterization of subgrid variability of snow, Hydrol. Process., 18,
1409–1422, 2004.Millares, A., Polo, M. J., and Losada, M. A.: The hydrological response of
baseflow in fractured mountain areas, Hydrol. Earth Syst. Sci., 13,
1261–1271, 10.5194/hess-13-1261-2009, 2009.
Pérez-Palazón, M. J., Pimentel, R., Herrero, J., and Polo, M. J.:
Analysis of snow spatial and temporary variability through the study of
terrestrial photography in the Trevelez river valley, Proceedings SPIE 8531,
Remote Sensing for Agriculture, Ecosystems, and Hydrology XIV, Edimburgh,
2014.
Pimentel, R., Herrero, J., and Polo, M. J.: Snow in a semiarid mountainous
area combining snow modelling and Landsat spectral mixture analysis,
Proceeding RSHS14 and ICGRHWE14, Guangzhou, China, 2014.Pimentel, R., Herrero, J., Zeng, Y., Su, Z., and Polo, M. J.: Study snow
dynamics at subgrid scale in semiarid environment combining terrestrial
photography data assimilation techniques, J. Hydrometeorol., 16, 563–578,
10.1175/JHM-D-14-0046.1, 2015.
Polo, M. J., Herrero, J., Aguilar, C., Millares, A., Moñino, A., Nieto,
S., and Losada, M. A.: WiMMed a distributed physically based watershed model
I: Description and validation, in: Environmental Hydraulics – Theoretical,
Experimental and Computational Solutions, CRC Press, Taylor and Francis,
360 pp., 2010.