Introduction
Recent reports from the Intergovernment Panel on Climate Change (IPCC), based
on the coupled atmosphere–ocean general circulation models (GCMs), on the
condition of increasing concentration of greenhouse gases
indicate that climate change by the end of this
century (e.g., increase in the mean temperature and change in the
precipitation amount) is expected to occur irregularly in space and time but
to mostly affect some specific and critical regions ,
including the vicinity of the Mediterranean – well known as one of the
world's climatic hotspots
. Within
this region, the Alpine and adjacent areas are expected to undergo a
relatively larger temperature increase , which has been
generally confirmed since the IPCC Fourth Assessment Report
AR4;.
In a generic mesoscale basin, such potential changes will influence
hydrologic budget, thus altering the amount of available water and acting as
climate feedback. Previous studies conducted over the alpine areas
demonstrated amplification of the climate
change signal by topography through local hydroclimatic and land surface
feedbacks: the snow cycle plays a key role as variations in the cycle of snowpack accumulation and melting affect the generation of snowmelt-driven
runoff. In addition, the temperature and seasonal precipitation pattern
changes can affect the permanent or seasonal snowmelt, thus affecting
streamflow timings, groundwater recharge, and runoff, and consequently the water
availability. Even where the precipitation will increase, the concurrent
warming will favor a further increase in evapotranspiration. The decrease in
water supplies, conjunctly with the likely increase in the demand, could
significantly influence agriculture (the largest consumer of water) as well
as municipal, industrial, and other uses . Nevertheless, to evaluate
the local net effect of changing climate on water resources, the hydrologic
budget must be detailed .
Seasonal variations of temperature and precipitation also drive changes in
runoff and streamflow: for instance, the spring peak streamflow may occur
earlier than the present in places where snowpack significantly determines the water
availability . Such changes may seriously influence the water
and flood management, often with significant economic consequences, though
the resulting effects may differ for regions even at similar latitudes, as
evidenced by for the high latitudes of North America and
Eurasia.
Usually GCMs are calculated in relatively coarse grid spacings, thus
inadequately representing the regional topography and climate
. Therefore, downscaling of the GCM variables to regional
scale is essential for a better depiction of regional climate: the dynamic
downscaling uses the regional climate models (RCMs) with a higher resolution
(typically 10–50 km) and the same principles of dynamical and physical
processes as GCMs e.g.,. It is
demonstrated that RCMs significantly improve the model precipitation
formulation e.g.,. In this
context, a project called the Prediction of Regional scenarios and
Uncertainties for Defining EuropeaN Climate change risks and Effects
(PRUDENCE; http://prudence.dmi.dk/, last access: 15 September 2017) was undertaken, aiming at providing
high-resolution climate change scenarios for Europe at the end of the 21st
century via dynamical downscaling of global climate simulations
. found that, over Europe, GCMs and
RCMs behave similarly for the seasonal mean temperature with higher spread in
GCMs; however, during summer, the spread of the RCMs – particularly in
terms of precipitation – is larger than that of the GCMs, which indicates
that the European summer climate is strongly controlled by parameterized
physics and/or high-resolution processes. They also concluded that the
PRUDENCE results were confident because the models had a similar response to
the given radiative forcing. showed that the signal from the
PRUDENCE ensemble is significant in terms of the minimum expected 2 m
temperature and precipitation responses. demonstrated that
RCMs in PRUDENCE generally reproduce the large-scale circulation of the
driving GCM. found a broad agreement, in the 21st century
climate projections over Italy, between the results obtained from the
ensembles of PRUDENCE and the Coupled Model Intercomparison Project (CMIP;
https://cmip.llnl.gov/, last access: 15 February 2018) Phase 3 (CMIP3); however, the CMIP3 GCMs showed a
much larger range of bias for temperature and precipitation than the PRUDENCE
RCMs. These studies indicate that results from the PRUDENCE and CMIP3/CMIP
Phase 5 (CMIP5) experiments are roughly equivalent for the Mediterranean
region and the Alpine sector.
The GCMs represent the large-scale atmospheric and oceanic processes. Even if
they include sophisticated atmospheric physics and feedbacks with land
surface and ocean conditions, they only show conditions averaged over large
areas. Hydrologic processes normally operate at relatively smaller scales, i.e.,
mesoscale and storm scale in meteorology and basin scale in hydrology, and local
conditions can be most extreme than those suggested by the areal mean values
see, e.g., the analysis on the groundwater use and recharge
in. Several recent studies attempted to evaluate the hydrologic
effects of climate changes in individual small-scale catchments using a
variety of water balance models and climate change scenarios
e.g.,. Despite of some differences in results, due to the
different forcing data or scenarios used , they have gathered
some suitable information at basin or regional scale.
These studies also reveal that the land surface has been recognized as a
critical component for the climate. Key points are the partitioning of solar
radiation into sensible and latent heat fluxes, and that of precipitation
into evaporation, soil storage, groundwater recharge, and runoff. Despite the
increased consideration of such processes, the land surface parameters are
not systematically measured at either large scale or mesoscale, making it
hard to perform hydrologic analyses. To overcome such a problem, we have used
a methodology called the CLImatology of Parameters at the Surface (CLIPS),
proposed by some other studies e.g.,.
According to CLIPS, the output of a land surface model (LSM) is used as a
surrogate of surface observations, to estimate the surface layer parameters.
The goal of this study is to investigate the effects of climate change, based
on high and low emission scenarios, on the hydrologic components in the
Alpine and adjacent areas, including the Po Valley in Italy, near the end of
this century. Section 2 describes the details of RCM and LSM employed in this
study, and Sect. 3 describes the experiment design. Results concerning the
hydrologic budget are reported in Sect. 4, and conclusions are provided in
Sect. 5.
Description of the models
In this study, calculation of the future hydrologic budget
components has been performed through the University of Torino model of land
Processes Interaction with Atmosphere UTOPIA;:
meteorological inputs to drive UTOPIA under the current and future climate
conditions are obtained from the Regional Climate Model version 3 (RegCM3).
Since the details of the RegCM3 run have already been published
, here a short description of RegCM3 will be
given. Regarding UTOPIA, just some portions relevant for this study are
described.
Despite the availability of the products for Europe within the World Climate
Research Program COordinated Regional Downscaling EXperiment (EURO-CORDEX;
http://www.euro-cordex.net/, last access: 5 February 2018), which includes a newer version of RegCM (i.e.,
RegCM4), we decided to employ RegCM3 for the following reasons: (1) RegCM3 had
been employed in several important projects, including PRUDENCE, ENSEMBLES
(http://ensembles-eu.metoffice.com/, last access: 5 February 2018), and the Central and Eastern europe
Climate change Impact and vulnerabiLIty Assessment (CECILIA;
http://www.cecilia-eu.org/, last access: 5 February 2018), whose outputs had been used in numerous studies
focusing on Europe
e.g.,; (2) RegCM3 had
also been widely used, even most recently, for the studies of climate
projections, model evaluations, and sensitivities over the target areas in our
study – the Alpine and adjacent areas
e.g.,;
(3) since plenty of model outputs were available from several relevant projects
(e.g., PRUDENCE, ENSEMBLES, CECILIA, etc.) and we had limited resources for
exploring all available data sources, we decided to select a well-known model
which had been extensively used for such kind of studies.
RegCM3
The earliest version of RegCM was originally proposed by
and to use limited-area models as a tool for regional
climate studies, with the aim of downscaling the GCM results. In this way, the
GCM runs could provide the initial conditions and time-dependent boundary conditions to
RCMs.
The dynamical core of RegCM3 is based on the hydrostatic version of the National
Center for Atmospheric Research/Pennsylvania State University Mesoscale Model version 5
(MM5: ). The RegCM is a hydrostatic and compressible primitive
equation model with a σ vertical coordinate. More details on RegCM3 are referred
to in the MM5 documentation
and some papers describing previous versions of RegCM
e.g.,. The RegCM3 is documented in
.
The RegCM3 includes several physical process packages. Precipitation involves
both grid- and subgrid-scale processes e.g.,, which are
crucial as a source of errors in climate simulations
e.g.,.
The implemented subgrid precipitation schemes are described in , ,
, , and .
The physics of the surface processes is described according to the Biosphere–Atmosphere Transfer
Scheme (BATS) manual .
Subgrid differences in topography and land use are taken into account using a
mosaic-type approach . Two kinds of water bodies are
considered – open (e.g., oceans) and closed (e.g., lakes).
The open-water bodies are described by the water temperature, introduced as a
boundary condition for the model. The closed ones are treated as the open
bodies, or using a specific one-dimensional lake model interacting in
two ways with the atmosphere . Aerosols and chemical
compounds are considered, accounting for their diffusion and removal processes,
as well as the radiative effects; details about the RegCM3 chemistry are found in
, , and .
The RegCM3 has been employed and tested in various contexts, on various space
scales, for a broad range of scientific problems, including climate change
, air quality
, water resources , extreme events
, agriculture , land cover change
, and biosphere-atmosphere interactions .
UTOPIA
The UTOPIA is a diagnostic one-dimensional model, formerly named the Land
Surface Process Model LSPM;. It can be
used on a stand-alone basis or be coupled with an atmospheric circulation
model or an RCM, serving as the lower boundary condition. All specific
details about its use and features are fully described in .
The land surface processes in UTOPIA are described in terms of physical
fluxes and hydrologic states of the land. The former includes radiation
fluxes, momentum fluxes, sensible and latent energy fluxes, and heat transfer
in multi-layer soil, while the latter includes snow accumulation and melt,
rainfall, interception, infiltration, runoff, and soil hydrology. All the
fluxes are computed using an electric analogue formulation, in which the
fluxes are directly proportional to the gradients of the related scalars and
inversely proportional to the adequate resistance.
The UTOPIA domain is vertically subdivided into three main zones – the soil,
the vegetation, and the atmospheric layer within and above the vegetation
canopy layer. Variables are mainly diagnosed in the soil and in the
vegetation layers. The canopy itself is represented as a single uniform layer
(i.e., big leaf approximation), whose properties are described by vegetation
cover and height, leaf area index, albedo, minimum stomatal resistance, leaf
dimension, emissivity, and root depth. The soil state is described by its
temperature and moisture content. These variables are calculated by the
integration of the heat Fourier equation and conservation of water mass
equation using a multi-layer scheme. The main parameters include thermal and
hydraulic conductivities, soil porosity, permanent wilting point, dry heat
capacity, surface albedo, and emissivity. The UTOPIA can have as many soil
layers as a user specifies; however, a sufficient number of layers is
required for numerical stability. Note that numerical stability is strictly
related to the integration time step – the model becomes unstable and blows
up eventually with an inadequately large time step. This is particularly true
in the presence of strong moisture gradients, which could lead to errors in
the representation of soil moisture profiles.
Finally, the presence of snow is parameterized with a single layer assumption.
Snow can cover vegetation and bare soil separately and possesses its proper
energy and hydrologic budgets, thus interacting with the other components.
The UTOPIA is a diagnostic model; thus, some observations in the atmospheric
layer are required as boundary conditions, including air temperature,
humidity, pressure, wind speed, cloud cover, longwave and shortwave
incoming radiation, and precipitation rate. Usually these observations are
measured values, along with the reconstruction of some missing data
using adequate interpolation techniques.
The UTOPIA, as well as its predecessor LSPM since 2008, has been tested with field
campaigns and measured data either by itself or as coupled with an atmospheric
circulation model. Examples of its use can be found in several studies.
compared LSPM and BATS in the Po Valley, Italy. studied
its dependence on initialization. used LSPM to study
surface energy and hydrologic budget on the synoptic scale.
used the LSPM to analyze extreme flood events in Piedmont, Italy. In ,
LSPM has been used to study the 2003 heat wave in Piedmont. Studies with LSPM on non-European
climates have also been accomplished, related to very dry sites ,
to the onset of the Asian monsoon ,
and to the soil temperature response in Korea to a changing climate
.
The UTOPIA was also coupled with the Weather Research and Forecast (WRF) model, version 3, and
applied to a flash flood caused by a landfall typhoon, as well as to the exceptionally wet
period 2008–2009 in the northwestern Italy .
Recent applications include studies on the parameterization of soil freezing ,
and the cold spells over the Alpine area and the Po Valley .
It has also been applied to studies on vineyards environment, including canopy resistance ,
energy and hydrologic budgets ,
sensitivity to vegetation parameters ,
and an analysis of turbulence .
Experimental design
The goal of this study is to evaluate the components of the surface hydrologic budget
on a mesoscale area from a climatic point of view, and to compare the effects of the
climate change on these values. Two 30-year periods have been considered: the first
one (1961–1990) is the baseline period or reference climate (RC), whereas the other
is the last 30 years of the 21st century (2071–2100), named the future climate
(FC) here. The period 1961–1990
has been employed in numerous previous studies on climate change projections and impacts,
even very recently
e.g.,. It had also been used in various climate projection projects using GCMs
and/or RCMs, such as CMIP3/CMIP5, PRUDENCE, ENSEMBLES, and CECILIA.
The climate projections are obtained through the IPCC Special Report on Emissions
Scenarios (SRES) A2 and B2 emission scenarios .
Note that the A2 scenario assumes large increases in population and economical development while the B2 scenario assumes more
sustainability and consequent smaller increases in those; thus, the
concentration of carbon dioxide are projected to be higher for A2 than for B2. The
future climates based on the A2 and B2 scenarios are hereafter referred to as
FCA2 and FCB2, respectively.
In the last decade, numerous studies on climate projections and impacts
have been conducted using the SRES scenarios, which were the base
scenarios in the CMIP3 experiments. After the emergence of new scenarios – Representative
Concentration Pathways RCPs;,
which were employed in the IPCC Fifth Assessment Report AR5; and
the CMIP5 experiments, there have been many studies either to check similarities and/or differences
between the two scenario sets for a given projection period
e.g.,
or to address the value of using both scenario sets for future climate projections
e.g.,.
It turns out that both SRES and RCP scenarios are generally in
good agreements, for pairs of closest counterparts, in projecting
climate in the 21st century. For example,
mentioned that SRES A2 was comparable to RCP 8.5.
found that the RCP 4.5 and SRES B1/A1T scenarios were broadly consistent with
the fossil fuel production forecasts.
pointed out that the RCP scenarios spanned a larger range of temperature
estimates than the SRES scenarios, and indicated similar temperature
projections for pairs between the two scenario sets: RCP 8.5 similar
to A1FI, RCP6 to B2, and RCP 4.5 to B1, respectively.
showed that the cumulative CO2 emission and corresponding warming in the
short term (2030) are approximately the same across all emission scenarios,
whereas those in the longer term (2100) are similar between close counterparts
of the selected SRES and RCP scenarios: A1FI to RCP 8.5, A1B to RCP 6, and
B1 to RCP 4.5, respectively.
reported a common drying trend, over the Mediterranean region, between
the CMIP3 simulations based on SRES A1B and the CMIP5 simulations based
on RCPs 4.5 and 8.5. It is also indicated by
that SRES A2 has similarities to RCP 8.5 in terms of radiative forcing,
future trajectories, and changes in global mean temperature. In ,
differences in warming rates existed between the two scenario sets due to
different transient forcings; however, with a 30-year average for each
scenario as in our study, the results and conclusions of using the SRES A2/B2
scenarios would not be significantly different from those of using the
closest RCP counterparts.
To obtain a broad range of projections,
projected global warming through all available emission scenarios,
showing that RCP 8.5 and SRES A1FI and A2 lead to the highest temperature
projections. Most recently,
and
assessed impacts of climate change on temperature and rainfall,
respectively, by the mid-21st century in Ireland using both the SRES
and RCP scenarios, and provided a wide range of possible climate
projections.
found that future summers had the largest projected warming
under RCP 8.5 while future winters had the greatest warming under A1B and A2.
created a medium-to-low emission ensemble using the RCP4.5 and B1 scenario
simulations and a high emission ensemble using the RCP8.5, A1B, and A2
simulations, which enabled 25 high and 21 medium-to-low emission
ensemble comparisons: they found significant projected decreases in mean
annual, spring, and summer precipitation amounts – largest for summer, with
different reduction ranges for different scenario ensembles.
Furthermore, the SRES scenarios by themselves have often been adopted in most recent
studies, even long after the release of the RCP scenarios, because the old scenarios
were in accord with their objectives
e.g.,.
We employed the SRES marker scenarios because of their long-term consistency in assessing
the impact of climate change on global and regional factors of socioeconomy and environment
during the last decade – including air quality ,
water quality and resources ,
energy ,
agriculture and forestry ,
fisheries ,
health and disease ,
climate and weather extremes ,
wildfires ,
ecosystems and biodiversity ,
and so forth. Although an ensemble approach with all possible scenarios would
increase the spread of hydrologic budget simulations, due to the limited
resources, we decided to select two representative marker scenarios: A2 as
the higher-end and B2 as the lower-end emission scenario, respectively.
Simulations of RegCM3 for the two periods (i.e., 1961–1990 and 2071–2100)
are fully referenced in and ,
and have been chosen for this study because they are still among the
highest-resolution datasets currently available. As shown in ,
the RCM outputs with high resolution can allow the
hydrologic cycle to be efficiently reconstructed at a large-basin scale, even in an orographically complex
area such as the Alps.
The domain for this study involves most of the Alpine region and the Po River
basin, as shown in Fig. 1. It is bordered by the meridians 5 and
15∘ E and the parallels 43 and 48∘ N. We have chosen this
domain for two main reasons: (1) the Alps represent a critical environment
that already answered most effectively to the recent climate warming
e.g.,; and (2) the Alps are the source of the longest
and greatest European rivers (e.g., Rhyne, Rhone, Danube, Inn, Arc, Po,
etc.). Under these considerations, it is essential to evaluate potential
changes in the soil variables and the hydrologic budgets, induced by
climate change.
Computational domain with grid points (a) and geographic
map (b) with boundary of the study area (black solid lines). Grid
points represent, in terms of the grid elevation (h), the plain area (h
≤ 500 m a.s.l.; blue), the normal mountains (500 < h ≤
2000 m a.s.l.; grey), and the high-mountain area (h > 2000 m a.s.l.;
red). The map on the right panel is modified from Google Maps.
The RegCM3 outputs are provided on a Lambert grid, with a 20 km spatial
resolution, containing 720 land grid points on the analyzed domain (Fig. 1).
The domain is divided into three sets of grid points in terms of elevation: (1) one
representing the plain or low-hill areas lower than or equal to 500 m above sea
level (a.s.l.), occupying 34 % (blue); (2) another depicting normal mountains
between 500 and 2000 m a.s.l., occupying 57 % (grey); and (3) the other
belonging to the high-mountain areas higher than 2000 m a.s.l., occupying
9 % (red). In this study, among all the possible outputs available from UTOPIA,
we give particular attention to the state of soil moisture and the components of
hydrologic budget – precipitation, evapotranspiration, drainage, and runoff.
Note that some of those values were already included in the RegCM3 output database.
However, the land surface model of RegCM3 employs an old force-restore method included
in the BATS scheme which was demonstrated to be insufficient to properly evaluate
hydrologic budget .
Therefore, we made an offline run with UTOPIA in order to allow a more
realistic evaluation of the soil and budget components, and to have a
self-consistent set of variables in equilibrium among themselves.
The UTOPIA has been driven using the following output of RegCM3 over each
grid point of the domain – precipitation; short- and longwave radiation;
and temperature, humidity, pressure and wind at surface (i.e., the lowest
level of RegCM3). This procedure has been used for all three climates (RC,
FCA2, and FCB2).
The UTOPIA has been configured to represent 10 soil layers, following , who suggested the use of multiple soil layers to represent well the vertical heterogeneity
in soil properties. The thickness of soil layers starts from 5 cm in the top layer, then
doubles for every layer going to higher depths. The last layer must be interpreted as a
boundary relaxation zone. The soil characteristics have been taken from the ECOCLIMAP
database .
No soil-freezing scheme is used, and initial values of soil moisture and temperature
have been set following .
In terms of vegetation, short grasses are assumed to cover the whole domain.
Actually the domain includes the Alps, the Apennines, off-alpine and hilly
areas, and plains; thus, there is a wide range of vegetation. Regarding plains
and hilly areas, vegetation includes pastures, grasslands, and some forested
areas: mountain areas are mostly covered by trees, and the highest parts are
without vegetation or covered by permanent ice (few grid points). We decided
to set the vegetation type equal for all grid points (i.e., short grasses)
for the following reasons: (1) for the “reference climate”, to avoid any
problem in interpretation of results due to the differences in vegetation,
and (2) for the “future climate”, to alleviate the uncertainty in vegetation
type at the end of 21st century. With regard to meteorological variables,
this is not a bad assumption because most observation stations are normally
installed over short grasses. Moreover, considering plant height, root depth,
and vegetation characteristics, short grasses can be roughly regarded as most
common cereals (wheat, maize, etc.), and would not be quite different from
such kind of agricultural products. Finally, we have also performed
simulations using the “true” vegetation (as deduced by detailed databases),
and the results with the pastures and agricultural areas have generally been
confirmed, though the numerical values of the variables were slightly
different (not shown).
Although UTOPIA could be driven by the real observations in RC, it is driven
by the RegCM3 output in order to keep consistency among the RC and FC
simulations and to exclude any possible source of errors caused by
differences in input data, irregularity of grid, and/or interpolation of
missing observations. In this way, we can compare the FC representations with
an analogous RC representation. Thus, here the RegCM3 outputs for each grid
point have been used as if they were observed data.
All RegCM3 outputs were available with a time resolution of 3 h, and used as
input data to UTOPIA. In order to ensure numerical stability of the UTOPIA
simulations, these input data, except precipitation, have been interpolated
at a rate of one datum per hour: we applied a cubic spline
to the non-intermittent variables like temperature, humidity, and radiation
(flux). The intermittent variable like precipitation was simply redistributed
to keep its sum, assuming a constant rate: the input data of precipitation
were the precipitation cumulated over the timesteps of the RCM output, and
could not be interpolated with splines. Although we could have converted
precipitation to precipitation rates, interpolated them using splines, and
then reconverted to cumulated precipitation over the smaller timestep of
UTOPIA, the result of such a complicated procedure was almost equivalent
to using the method employed here. Regarding radiation and wind components,
we used the splines for the sake of uniformity with other variables. Then,
we further controlled some unrealistic values (e.g., negative radiations):
we controlled the daily means (or cumulated values) from input (from RegCM3)
and output (for UTOPIA) of the interpolation to be
non-negative values.
In this study, we employed a single-model approach that has relatively larger
uncertainty: it is desirable to employ an ensemble approach, using multiple
models and/or initial conditions, to estimate the range of climate projections.
Our decision to employ the single-model approach is mainly due to limitations in
resources to perform multi-model ensemble simulations for both the RCM and land
surface model. Given such limitations, a high-resolution single model is
often an alternative choice, especially over a complex terrain.
made a fine-scale (20 km) single-model experiment using RegCM3 and found that
both the temperature and precipitation changes via RegCM3 were in line with the
CMIP3 and PRUDENCE ensemble results. Generally speaking, multi-model ensembles
tend to decrease the errors compared to an individual model; however, due to
the averaging operation (e.g., ensemble mean), the spatial and temporal
variability of the signal tends to decrease. Moreover, many previous studies
on various climate change impacts and projections had been performed using the
single-model approach e.g.,.
The multiple simulations performed for RC and FCs are presented in terms of the
temporal and spatial variability by displaying time series (annual cycles) and
two-dimensional maps, respectively, of the mean values of some variables. For time averaging,
suggested using monthly mean values for discussing the hydrologic budget
variations induced by climate change; however, we preferred a period of 10
days to better quantify time shifts of the physical variables. In this study,
the annual cycles are figured via the 10-day averages over the 30-year
simulation period, at each elevation-categorized grid-point set. Each month
has three 10-day periods: days 1 to 10, 11 to 20, and 21 to the end of the
month.
The analyzed variables include precipitation (PR), evapotranspiration (ET),
surface runoff (SR), and soil moisture (SM). We noticed that the general
trends of annual cycles are similar between RC and FCs. Therefore, in order
to accentuate the extent and direction of changes, the future variations in
the hydrologic budget components are shown as the differences between FCs and
RC; the PR difference (ΔPR) represents PRFC minus PRRC,
where FC is either FCB2 or FCA2 – similarly to ΔET and
ΔSR.
In this study, SM is defined as the quantity of water contained in soil that
is composed of solid particles, air, and water, and it is represented as
saturation ratio (S):
S=VwVw+Va=VwVv,
where Vw, Va, and Vv are the volumes of water, air, and voids,
respectively, in soil.
Results and discussion
In this section, we provide analyses on temporal variability and spatial
distribution of hydrologic budget components, making comparisons between RC
and FCs. The potential change in dryness (wetness) is also assessed through the
projection of the number of dry (wet) days. Finally, we compare our findings
with relevant previous studies, and discuss consistency and uniqueness of our
study.
Temporal variability of evapotranspiration, precipitation, runoff and soil moisture
Figure 2 compares the annual cycle of PR, ET, and SR in the plain area
(h ≤ 500 m a.s.l.). In the RC summer, ET exceeds PR from the end
of June (when ET peaks to about 22 mm) to the end of August (when SM is
minimal around 0.52 m3 m-3; see Fig. 3). PR shows its minimum
between mid-June and August, when it is lower than ET. In the RC winter, PR
is much higher than ET, and SR exceeds ET from October to March. In the
summers of FCs, ET exceeds PR for a longer period (in FCA2), and
both scenarios show larger water deficits in July and August, with the PR
minimum shifted to August in FCA2 (not shown). Furthermore, the
ET maxima shift towards July–August, in both FCA2 and
FCB2 (not shown), and the values increase by as much as 3–5 mm
(i.e., ΔETs).
Annual cycles of the 10-day average values of the surface hydrologic
budget components for the plain area for (a) RC,
(b) FCB2 - RC, and
(c) FCA2 - RC. Here, PR is precipitation, ET
evapotranspiration, and SR surface runoff. Units are millimeters (mm).
It is conspicuous that the summer PR decreases in the future – between the end
of May and the beginning of September in FCB2 (Fig. 2b), and
between July and September in FCA2 (Fig. 2c). On the contrary, PR
generally increases in winter, between December and February, in both FCs. In
autumn, ΔPRs show large variations in short periods: for instance, in
FCB2, it varies as much as -6 mm in mid-September, +10 mm in late
September, -12 mm in late October, +15 mm in mid-November, and -7 mm
in late November. Regarding ΔET, there are almost no variations in
cold months, while there is a small increment (up to 3 mm) between April and
September in FCB2, and a larger increment in the same period in
FCA2, with the largest value in August (∼ 5 mm). This
large variation in PR is partly due to the orographic effect. As reported by ,
in winter the southwesterly flow increases across
the Alps, and causes a maximum of precipitation increase over the southern
Alps; in autumn the main circulation change is in the easterly and
southeasterly direction.
Figure 3 shows the 10-day mean values of SM for the plain area, expressed as
saturation ratio – see Eq. (1). Variations of SM in plains are almost
negligible in a colder period (late November–mid-May), but are large
during a warmer period (late May–mid-November): the driest points are
antedated by ∼ 10 days in FCs, still being in August, and their values
decrease by ∼ 0.1 m3 m-3. The decrease begins already in spring
(from late May) and continues till late October (FCB2) or early November
(FCA2), with the largest depletion in early August (FCB2) and in
early to mid-August (FCA2). Moreover, the period that future SM values
are lower than the lowest SM of RC (i.e., ∼ 0.52 m3 m-3 in
mid-August) extends from early July to early September in FCB2 and to
mid-September in FCA2. In the driest periods of FCs, several grid points
in the plains go below their permanent wilting points (PWPs), which vary
according to soil type, or remain below PWP for an excessive duration by
about 1 month. Our results regarding the future changes in SM in the warm
period – an increase in days of SM lower than the lowest SM of RC, and a
surplus of period below PWP – signify that, if the land use of the grid
points is pasture, we need appropriate countermeasures to ensure an adequate
productivity. During the cold period in plains, SM shows the highest values
(∼ 0.73 m3 m-3) in both RC and FCs; the SM values of FCs slightly
exceed those in RC, due to the small increments of PR in this period (see
Fig. 2b and c).
Annual cycles of the 10-day average values of SM, expressed as
saturation ratio (in m3 m-3), at the soil surface layer (a depth of
0.05 m) in RC, FCB2, and FCA2 for the plain area.
Figure 4 shows the annual cycle of hydrologic budget components over the
high-mountain area (h > 2000 m a.s.l.). In both RC and FCs, PR does not
exceed ET, while the gap between the two variables narrows in the FC summers,
due to an increase in ET and a decrease in PR. In RC, ET peaks in mid-July
while PR peaks in late June. The peak of SR, between May and June, is out of
phase because it is also affected by the concurrent snowmelt. It is
noteworthy that PRs in summer and autumn generally decrease in FCs (i.e.,
ΔPR < 0) from mid-June to mid-November: except for short terms in
early July, from mid- to late August, and from late September to late October
in FCB2, and except only from early October to early November in
FCA2. On the contrary, in winter and spring, PRs generally increase in
FCs from mid-January to early June except for short-term decreases in
mid-April and mid-May. Regarding ΔET, there are almost no variations
in cold months, as expected (due to snow cover), whereas there is a large
increment (∼ 10 mm) between May and June, and a low-to-moderate increment
(∼ 2–6 mm) between July and October in FCs.
Finally, for ΔSR in high mountains, there is a weak increase (< 5 mm)
between late November and late March, a stronger increment (∼ 10 mm) in April,
especially in FCB2, a strong decrease (up to -25 to -31 mm) between
May and June, and a general weak decrease in summer between July and September (see
Fig. 4b and c). As a result, the maxima of SR in FCs significantly decrease and
their occurrence dates shift ahead to May for FCB2 and between April
and May for FCA2 because snowmelt occurs nearly 30–40 days earlier
see
– see also the analysis on frost frequency in .
also reported that, regarding the 75th percentile in the Alpine areas, the snowmelt-driven
runoff timing moves earlier by about 35 days – due to the largest decrease in snow
cover between April and June, sustaining the spring runoff maximum. Those variations
in our result are in line with the changes in snowpack in FCs, which starts to melt
earlier, between late April and early May. We should consider that changes in snow
cover affect the surface energy budget through the snow–albedo feedback mechanism
;
that is, a reduction of snow cover decreases the surface albedo and thus increases
the absorption of solar radiation at the surface, resulting in warming. Moreover, soil
temperature starts rising earlier in the year in the snowless areas. Our results also
agree with other studies, carried out using RCMs over the Alpine areas: for example,
for PR and ET, and
for snow. Note that SR in RC is almost null between mid-December and March
while ΔSRs in FCs in the same period are positive: this indicates the
presence of rainfall and/or snowmelt over at least some parts of the
high-mountain grid points, even in the coldest periods.
Same as in Fig. 2 but for the high-mountain area. Grey bars at the
lower portion in (a) represent the snow cover (in m) in RC varying
from 0 m (null) to 1 m (max); in (b) and (c) the snow
cover difference (in m) between the corresponding FC and RC varies from
-1 m (Neg) to 1 m (Pos). The periods of snow ablation (late spring) and
accumulation (mid or late autumn) are well identified.
Figure 5 shows SM at the high-mountain grid points and demonstrates the
effects
of hydrologic budget components on surface SM. We note that the behaviors of SM
in high mountains are substantially different from those over plains (cf. Fig. 3).
In RC, the highest SM (∼ 0.65 m3 m-3) occurs in early June while the
lowest SM (∼ 0.51 m3 m-3) arises in early to mid-March. The
increase in SM from late March to early June is related to snowmelt due
to an increase in net radiation.
Surface SM in RC starts to decrease as the cold season starts in early
November, reaching the minimum in mid-March. Note that SMs during the same
cold period in FCs are larger than SM in RC, evidencing a larger amount of
liquid precipitation in FCs: in other words, winter rainfall will be more
frequent in the future. The peak of SM in spring is advanced by 10–20 days
in FCs, occurring in early May. The magnitude of maximum SM in FCs is a bit
lower than that in RC but the spread is larger, implying that snow ablation
starts much earlier and lasts longer. In addition, the occurrence of the
minimum SR shifts from mid-March in RC to summer in FCs: in both February and
early August (i.e., two minima) in FCB2, and late August in FCA2.
This shifting is mainly caused by the enhancement of ET.
Same as in Fig. 3 but for the high-mountain area.
Spatial distribution of evapotranspiration, precipitation, runoff and soil moisture
Our analyses illustrate that the differences in the SM behaviors between RC
and FC, both over plains (Fig. 3) and in high mountains (Fig. 5), are
strongly linked to the variations in the hydrologic budget components.
In this section, to understand such linkage more clearly, we perform analyses on the
spatial distribution of hydrologic variables (i.e., PR, ET, SR, and surface SM) along
with discussions on the associated energy variables (i.e., net radiation,
NR, and surface soil temperature, ST), during summer when such variables generally
show their largest values. Details in the analyses of NR and ST are referred
to .
Figure 6 shows the variables averaged in the month of July, in which PR and
surface SM are close to their annual minima while ET is close to its annual
maximum. Here, we discuss the variables in terms of anomalies of
FCA2 only because of similar patterns to but larger variations
than those of FCB2. Variables in Fig. 6 are anomalies of
hydrologic budget components: ΔET, ΔPR, ΔSR, and
ΔSM where, e.g., ΔET represents
ETFCA2 - ETRC.
Hydrologic budget components: differences between FCA2
and RC (i.e., FCA2 - RC) of the mean values of
(a) ET (in mm), (b) PR (in mm), (c) SR (in mm), and
(d) surface SM (in m3 m-3). The mean is calculated over the
month of July.
Compared to RC, we notice a large increment of NR everywhere in
FCA2 (not shown), with the exception of a few grid points located
in the central and western Alps. Regarding ΔET (Fig. 6a), plains along
the Po River and the northern off-alpine regions (i.e., middle-slope and/or
foot) show the largest increments, well correlated to ΔNR, implying
that most
of the available energy excess is used for evaporative processes. In contrast,
in the Apennines and central Alps, ΔETs are almost null or slightly
negative while ΔNRs are insignificantly positive. ΔPR (Fig. 6b)
and ΔSR (Fig. 6c) show similar signals, with a general deficit,
especially in the eastern and western Alpine areas. In particular, consistent
with ,
ΔPR depicts a dipolar pattern, especially in the eastern part of the
Italian Alps, with positive values over the Alps and its north and negative
values over south of the Alps. Surface ΔSM (Fig. 6d) shows a general
reduction, larger in the zones at latitudes lower than 45∘ N, whereas
surface ΔST (not shown) is almost uniformly larger in the considered
domain. As ETs increase (i.e., ΔET > 0), SMs generally decrease;
however, both decrease over some regions where ΔSMs are strongly
negative – on the western mountainous Emilia-Romagna region and Tuscany,
and along the Po River and in central and southern Piedmont as well (cf.
Fig. 6a and d). When SM decreases below the wilting point, evaporation
generally ceases because there is no available water for further ET, and the
ET anomaly (i.e., ΔET) can be negative. Considering that most of those
areas are important for agricultural production see, e.g.,a study
on grapevine in Piedmont region,
our results constitute a threatening challenge for future agricultural
productivity.
It is evident that ΔET and ΔPR do not show a linear correlation
(cf. Fig. 6a and b). ΔETs are generally positive, whereas ΔPRs
are distributed around null with some positive peaks on the Apennines and
northwestern Italy and large negative peaks on some Alpine locations. This
disparity brings about and/or enhances the nonlinear interactions among
temperature, evaporation, soil moisture, etc. Noting that nonlinearity can
develop even with small perturbations e.g.,,
our results elucidate that similar investigations can only be conducted using
models that are able to give a correct estimation of energy and hydrologic
processes.
Number of dry and wet days in the future climate
The availability of the SM estimations enables us to evaluate the occurrence
of dry and wet days, instead of using atmospheric relative humidity as usual,
in a similar way to figure the warm and cold days via the ST estimations.
We employ SM to assess the dry and wet days in FCs because we consider it to
be a more valuable indicator of the soil hydrologic conditions, directly
reflecting the hydrologic status of the soil water, e.g., that used by plants.
Here, we limit the analysis to the surface SM (i.e., in the top soil layer
with a depth of 5 cm), due to its significant impact on several agricultural
productions.
The anomalies in number of (a) dry days and
(b) wet days in FCA2.
Actually, for the short grass vegetation category considered in our simulations,
the root layer is only 5 cm deep, as the grass is only 10 cm high. Despite this
value seeming too low, it represents the typical height for the landscapes of Po
Valley (at least in its portion occupied by natural vegetation). Furthermore,
the upper soil layer represents the greatest effect of the atmosphere–land
surface–soil interactions. Given that we are interested in the present versus
future hydrologic budget components, it is appropriate to focus on the top soil
layer, where the most dynamic interactions with atmosphere and land surface occur.
More specifically, the water content of the soil layer that represents the
largest variations of moisture is subject to direct evaporation; to the
transpiration from vegetation roots; to the gravitational drainage to the
second soil layer; to the capillary suck of moisture from the second soil
layer; and finally to the eventual precipitation, eventual vegetation drainage,
and eventual snow runoff.
In order to find the absolute thresholds for SM, we have selected two
parameters: PWP and the field capacity. PWP is the SM level below which the
osmotic pressure of the plant roots is insufficient to extract water from the
soil, and is usually considered as an indicator of a serious water deficit
for agricultural practices. The field capacity represents the SM level above
which the gravitational drainage, due to soil hydraulic conductivity, causes
a rapid removal of the excess water through percolation into deeper
layers; thus, it is considered to be a threshold above which soil is very wet, as
in the cases of very intense precipitation, sometimes causing floods. Since
these two values change according to the soil type and texture, we define a
non-dimensional index, QI, which is independent from soil type, as follows:
QI=q1-qwiqfc-qwi,
where q1 is the moisture of the top soil layer, qwi is PWP, and
qfc is the field capacity. All the values are expressed as a
soil saturation ratio. In this way, the soil wetness is categorized in terms
of QI as extremely dry soil for QI ≤ 0, and extremely wet soil
for QI ≥ 1. In this study, we define the thresholds for dry soil and
wet soil as QI=0 and QI=0.8, respectively. Note that it is quite
rare to see the cases with QI=1 because the 3-hourly precipitation data
from RegCM3 are interpolated to hourly data by keeping the constant rain
rate, to be used as input for UTOPIA. Therefore, we have arbitrarily defined
the threshold for wet soil as QI=0.8.
Figure 7 shows the anomalies of dry and wet days in FCA2. The
number of dry days generally increases in most of the domain except the Alpine
high-mountain areas (Fig. 7a). A higher number of dry days (e.g., 30–50 days)
occur over the regions of extreme soil dryness – the coastal areas as well
as the off-alpine regions of the Alps and the Apennines (cf. Fig. 6d). The
interannual variability of the dry-day occurrence also decreases (not shown),
implying that our results are relatively robust and that we may experience
drought over the non-high-mountain areas in almost every year.
The number of wet days, on the other hand, is almost stationary over plains
but increases by 10–15 days in some localized regions close to the Alps in
the Italian side (especially in the Lombardy region), and by even more than
20 days at the feet of the Alps in Switzerland, France, and Austria (Fig. 7b).
The interannual variability is generally stationary, but increases in
the areas with the largest numbers of wet days (not shown). Therefore, in
FCA2, we can have more occasions of reaching high values of surface
SM, and
hence a potentially higher risk of floods. This also implicates a corresponding
higher possibility of hydrogeological instability over the same areas of
higher flood risk.
Overall, in the plain areas including the Po Valley, ΔET is positive
while ΔPR is weakly negative and ΔSM is moderately negative
(especially during summer, as in Figs. 2 and 3). With more significant overall
increases in NR over plains, the combined effect will bring about larger
evaporation and lower soil moisture, and thus an overall increase in the number of
dry days, mostly attributed to a much drier climate in summer. Meanwhile, over
the high-mountain areas, PR, SR, and SM increase while ET shows little
variation in spring and winter (see Figs. 4 and 5). As SM is large over high
mountains, we have more source of atmospheric moisture through evaporation
there. Then, through the combined effect of terrain-induced convective
motion, increase in NR (though less significant), and pre-existing snow, we
can have more snowmelt (during spring) and more liquid precipitation
(especially during winter), resulting in more wet days, again mostly
attributed to much a wetter climate in winter.
Comparative discussion on previous works
The Mediterranean basin is recognized as one of the climatic
hotspots around the world .
The Alps and their adjacent areas, including the Po River basin in
Italy, have been a target region of many climate projection studies, using
either a single RCM or an ensemble of GCMs and/or RCMs e.g., to mention
just a few,.
In general, those studies showed good agreements with our results and produced
consistent results of climate projections at the end of 21st century over the
study region. However, none of them studied hydroclimate projections of the full
water cycle by assessing all hydrologic components – precipitation, evapotranspiration,
runoff, and soil moisture – as in our study. Most them focused on just some specific
component(s) of the water cycle, e.g., precipitation and/or surface runoff. For instance,
studied climate change projections for the Mediterranean region, focusing on precipitation and temperature;
studied the impact of climate change on the Po basin, addressing discharge; and
carried out ensemble RCM projections over the Alps, focusing on
precipitation.
Nevertheless, it is meaningful to compare our findings, on overall hydrologic
components, with other studies over the same study area. Basically, most of
the previous studies showed consistent results with ours, as shown in the following.
illustrated the positive anomaly of precipitation in the future climate over the southern
Alps from autumn through spring, and the negative anomaly in summer over the highest
peaks of the Alps.
remarked on the peculiar behavior of the Alpine region, compared to the Mediterranean basin,
with moderate drying during warmer seasons and an increase in precipitation in winter, and
a large increment of interannual variability, which can lead to an increase in extreme
events such as droughts and floods.
discussed a surrogate climate change simulation over the Alpine region and found that
the winter precipitation increased with a significant dependence on elevation while
the summer precipitation decreased over the Alpine mountain chain, due to a local
surface–atmosphere feedback mechanism involving reduced snow cover and soil moisture
at the beginning of summer.
showed that future precipitation decreased during summer and increased during winter
and spring over the Alps;
also noted the increment of winter precipitation in the future climate in the Alpine
region, due to a negative correlation with decreasing pressure patterns.
addressed the larger changes in the precipitation regime in the higher-elevation
Alpine catchments.
examined the variation in the discharge maxima of the Po River in the future climate,
and concluded that the winter–spring maximum would increase and the summer–autumn
maximum will decrease.
confirmed that future precipitation would increase (decrease) over northern
(southern) Europe, with most of the Alpine region exhibiting a positive (negative)
precipitation change in the winter (summer).
found a robust signal of decreasing snowfall amounts, from September to May,
over most parts of the Alps, with relative changes in mean snowfall being strongly
dependent on elevation. In a review paper based on the existing literature
and additional analyses on climate change in the Alps,
concluded that warming induces a seasonal precipitation change – increase
in winter and decrease in summer – and a drastic decrease in snow cover
below 1500–2000 m in the Alps.
Compared to most other studies, which focused on the subcomponent(s) of hydrologic
cycle, our study is quite exhaustive and has its own uniqueness: our study provides
more complete analyses on all hydrologic components, including soil moisture, for both
the reference climate and future projections. Furthermore, along with a study on the land
surface energy balance ,
we provide discussions on the linkages between the hydrologic and energy
components to complete the full description of hydroclimatic changes. These
enable us to better quantify some significant variations in the frame of
changing climate in the Alpine and adjacent areas, in which the climatic
change shows a larger variability.
Conclusions
In this study, we investigated the characteristic changes in hydrologic budget
components and soil moisture, over the Alpine areas and northern Italy, under the
projected conditions of the future climate (FC; 2071–2100), compared to the reference
climate (RC). We employed the University of TOrino model of land Processes Interaction
with Atmosphere in offline simulations. The meteorological input data in FCs
are provided by the Regional Climate Model version 3, based on the A2 and B2
scenarios from the Intergovernment Panel on Climate Change Special Report on
Emissions Scenarios.
In FCs based on the A2 and B2 scenarios (FCA2 and FCB2,
respectively), the most significant changes are the increment of
evapotranspiration (ET) and the subsequent depletion of soil moisture (SM),
more remarkably in FCA2. Precipitation (PR) shows the lowest values
while ET depicts the highest values in the future summer (in particular,
July), when SMs are the lowest in many grid points. In the plain area, the
minimum SM in FC occurs about 20–30 days earlier than in RC, and remains low
for the successive months up to November. In the high-mountain area, the
surface runoff (SR) coming from the snowmelt keeps the soil water amount
sufficiently high to maintain the ET levels high from May to October,
especially in FCA2; thus, ET (or latent heat flux) always
exceeds sensible heat flux (SHF). Over plains, the period in which ET exceeds PR
elongates by about 1 month, mainly in spring. Moreover, SM decreases also
for one more month in summer, falling below the wilting point threshold in
the surface soil layer. In high mountains, due to the earlier occurrence of
snowmelt, the land surface becomes snowless for an additional month.
We found that these changes in the hydrologic budget components are strongly
related to the variations of net radiation (NR), which generally increase in
the Alpine area, causing the warming of both the top soil layer and the soil
surface – the former through an enhanced SHF, and the latter due to the highest
soil heat flux see, e.g.,.
Under the future conditions of increasing NR and soil temperature along with
decreasing SM, we expect two climatic feedbacks to take place: (1) a drier
soil brings about higher albedo, and (2) a warmer soil emits more longwave
radiations. Both feedbacks act to decrease NR eventually – i.e., negative
feedbacks. However, there are coincident increments of SHF to the atmosphere
as well as longwave radiation emitted by the warmer atmosphere. The overall
outcome cannot be generalized because it depends on the intensity of
individual component of the energy and hydrologic budgets. This confirms that
the climate system is quite complex and that, to evaluate well the surface
conditions, it is essential to calculate the energy and hydrologic budget
components in detail.
The values presented in this study refer only to the average conditions; however,
considering the large interannual variability of hydrologic variables registered
over those areas in RC, we expect to have more frequent and more intense occurrences
of longer dry spells (hence severe droughts) and heat waves in FCs, especially
in middle summers. As most agricultural products intensively grow in summer
(e.g., wheat, rice, maize, and grapevine, as well as other typical products in the Po
valley), the potential conditions of elongated drought will significantly exert
unfavorable impacts on agricultural production .
Other activities related to water supply (e.g., industry, hydroelectric power
production, etc.) can also suffer serious problems, consequently exerting
harmful impacts on economy and human health in local regions.
On the contrary, during winter, PRs generally increase in FCs, with a larger
number of the liquid precipitation events at high elevations. Furthermore, in
spring, snowmelt occurs earlier by about 1 month, thus resulting in
precedence of the SR peak by about 20–30 days. In winter, the SR amount
generally increases. By taking into account the large interannual variability
of PR, this runoff increases the occurrence and/or duration of wet periods
(e.g., heavy rainfall and floods) during winter and spring in FCs.
We also examined potential changes in the number of dry and wet days in
FCA2 by analyzing surface SMs. Our results report a higher possibility
of having SMs below the wilting point in the plain and coastal areas, and a
probability of slightly increasing wet days, particularly in the off-alpine
areas.
We note that the numerical values of all variables are dependent on the performance
of the model employed.
Noting that our study is based on a single-model approach, uncertainties in
the projected changes related to model bias and ensemble variability can be
large; thus, our results should be interpreted with caution. In this context,
further research is needed to obtain more robust results from an ensemble
approach.
Recent studies demonstrate that the accuracy of land surface processes
diagnosed by land surface models can be further improved by considering
various aspects of vegetation effects in the subgrid-scale parameterizations
e.g.,.
Moreover, the model uncertainties can be significantly reduced by optimal
estimation of the parameter values in the schemes e.g.,
and/or seeking for an optimized set among multiple-physics optional schemes e.g.,.
By applying these methods, the details of model-generated spatial and temporal
changes in the future energy and hydrologic budgets can be different from the
current results; however, we believe that the general trends are not
significantly disparate. Overall, our findings can provide a useful guideline
to plan the managements of water resources, floods, irrigation, forestry,
hydropower, and many other activities relevant for human life.