Improvements in the evaluation of land surface models would
translate into more reliable predictions of future climate changes,
as significant uncertainties persist in the quantification and
representation of the relative contributions of soil and vegetation
to the water and energy cycles. In this paper, we investigate the
usefulness of water stable isotopes in land surface models studying
land surface processes. To achieve this, we implemented
The response of the simulated
This study's results confirm that the use of water stable isotopes measurements helps constrain the representation of key land surface processes in land surface models.
Uncertainties in the representation of land surface processes in the land surface component of climate models may contribute significantly to the spread in climate projections (Pitman et al., 2009; Boé and Terray, 2008; Crossley et al., 2000). Several studies investigated in land surface models, both in stand-alone mode or coupled to atmospheric General Circulation Models (GCMs), the impact of key parameters on the soil hydrology and found that the contribution of the terrestrial biosphere to the water cycle was highly uncertain (Cheruy et al., 2013; Koster et al., 2006; Polcher, 1996). Improvements in the evaluation of such models would translate into more reliable predictions of future changes.
Water molecules carry hydrogen and oxygen isotopes in different
proportions. The most common and stable species are hydrogen
Our main objective is here to confirm the usefulness of water isotopes in investigating land surface processes. Using similar approaches as in other isotope-enabled land surface models (Haese et al., 2013; Aleinov and Schmidt, 2006; Yoshimura et al., 2006; Cuntz et al., 2003), we implemented for this study water isotopes in the most recent version of the land surface model ORCHIDEE (following Risi, 2009), hereafter referred to as ORCHIDEE-iso.
Due to its unique hydrological features (presence of wetlands, permafrost), Siberia has been in recent years the focus of several climate change studies (e.g. Bulygina et al., 2015, 2011). In particular, studies focus on the sensitivity of Siberian hydrological and biogeochemical cycles to climate perturbations, especially in view of a warming trend (e.g. Walter et al., 2006). Of particular concern is a possible deepening of the soil seasonal melting layer, the Active Layer Thickness, which could bear effects on the hydrological cycle as well as on the greenhouse gas budget, considering the large amount of carbon stored in deep permafrost (Callaghan et al., 2011).
Being able to predict water oxygen isotopes in soil and leaf water
pools is also a prerequisite to understand the global atmospheric
budget of
In this study we thoroughly tested ORCHIDEE-iso's performance at instrumented stations located in a boreal region of Western Siberia on the left bank of the river Ob and investigated the added value of water isotopes in constraining the description of key hydrological processes in the model. For this, ORCHIDEE-iso stand-alone (i.e. not coupled to an atmospheric model) was set up and run at these four locations. The results of the simulations were evaluated against local vertical profiles of water isotopes in soil water.
As hydrogen and oxygen stable isotopes in water molecules are closely
associated, isotopic ratios and fractionations of the two elements are
usually discussed together. In this study, following common practice,
we refer to the isotopic enrichment by the
Expressed in ‰,
Vertical profiles of
The measurement stations, delimiting a triangular 30
The heterogeneity of the study area in Labytnangi is reflected in the
depth of the permafrost. The shallowest seasonal melting layer is
found for station 3 at 40
At each site, three soil cores were collected and divided into
10
ORCHIDEE-iso (Risi, 2009) is based on the land surface model ORCHIDEE (Krinner et al., 2005) developed at the Institut Pierre-Simon Laplace (IPSL), in France. The setup used for this study is the natural evolution of the one developed and extensively tested by Risi (2009) and has been accordingly evaluated against isotopic water measurement at stations, representative of different climates, belonging to the MIBA (Moisture Isotopes in Biosphere and Atmosphere) network (Wingate et al., 2010; Raz-Yaseef et al., 2010; Knohl et al., 2007). Differences between the two ORCHIDEE-iso setups solely depend on differences between the version of ORCHIDEE isotopes were originally implemented in and a more recent version of ORCHIDEE this study is based on.
ORCHIDEE includes three separate modules: SECHIBA resolves water and energy exchanges between land surface and atmosphere (de Rosnay, 1999; Ducoudré et al., 1993); STOMATE accounts for vegetation phenology and growth and for the carbon cycle (Krinner et al., 2005); LPJ (Sitch, 2003) simulates the dynamic evolution of the vegetation cover through competition and fire. The model considers different land cover types in boreal, temperate, and tropical regions: within each grid cell bare soil, deciduous or evergreen forests, C3 and C4 grasslands may co-exist, up to a maximum of 12 different plant functional types (characterized by leaf area index, albedo, root depth, height, among others), for which water fluxes and reservoirs are computed independently. In this study the STOMATE and LPJ modules were not activated; prescribed land cover maps and related parameters were used instead, with vegetation phenology parameterized according to a simple growing degree day (GDD) model.
ORCHIDEE can run at different spatial scales, from point-wise
simulations up to global experiments. The standard temporal resolution
is of 30 min. The model represents all fluxes and pools relevant for
the soil water budget. Throughfall rain, snowmelt and throughfall dew
reach the soil surface. In the ORCHIDEE setup (Choisnel et al., 1995)
used for this study the soil is 2
Water stable isotopes
In spite of the unresolved vertical dimension of soil water pool of
the ORCHIDEE setup it is based on, we endowed ORCHIDEE-iso with
a diagnostic representation of the vertical distribution of water
stable isotopes in the soil column. The approach, originally
implemented by Risi (2009) ensures water budget conservation along the
total soil column (and thus it does not interfere with the soil water
budget of ORCHIDEE). Soil water is vertically distributed over several
layers, defined, based on the available water soil water, from the top
to the bottom of the soil column. At each time step, water sources and
sinks are budgeted and, if necessary, the discretization revised; the
isotopic composition is updated accordingly. The uppermost layer
contains the water pool subject to bare-soil evaporation and has
a water equivalent height:
In this study ORCHIDEE-iso was run in stand-alone mode, point-wise, at four stations in Labytnangi (cf. 2.2) using ORCHIDEE's standard time resolution.
The model spin-up consisted in running ORCHIDEE iteratively five times over the same meteorological year (2012), at each location. A further year of simulation was then performed using initial state variable values from the spin-up simulations at each location.
In the following, we describe the meteorological forcing, the isotopic forcing, and the characterization of the stations in ORCHIDEE-iso based on the available soil and vegetation data at the measurement sites.
ORCHIDEE-iso was forced for all experiments by 6 hourly time series
of locally-measured surface air temperature and humidity, wind speed,
precipitation, shortwave and longwave downward radiation. All these
meteorological fields, except precipitation and radiation, are
routinely recorded at the WMO meteorological station of Salekhard, 20 km
southeast of Labytnangi. For the missing variables,
ERA-INTERIM reanalysis were used to complete the dataset. Due to the
spatial proximity (
Besides pure meteorological forcing, ORCHIDEE-iso requires the isotope composition of precipitation and water vapor as additional input variables as the relative composition of all land surface water reservoirs is, at the first order, driven by the composition of precipitation.
The
The four experimental stations were characterized in ORCHIDEE-iso by attributing different types of soil and plant functional types (Table 1). At this stage in the model it is not possible to represent more than one soil type at a time in each grid cell and for the whole soil column. The soils were thus classified according to the predominant grain size (sand, silt or clay, as defined by USDA, the United States Department of Agriculture) at each station, using data collected in summer 2012 by the University of Yekaterinburg (Valdayskikh et al., 2013, and summarized in Sect. 2.2).
Plant function types were characterized by a predominance of C3 grass
at all stations, with a maximum fraction, 99 %, at station
3. Boreal broad-leaf summer-green trees were, although in different,
limited proportions, represented at all sites. Station 1 and 2,
located in the tundra, also host deciduous needleleaf species
(
In this section we first present the experimental soil water isotope
profiles and then we proceed in evaluating the ability of ORCHIDEE-iso
to reproduce
For each station, three profiles of
We assume that meteorological conditions do not significantly differ from station to station and that, therefore, differences in isotopic enrichment among the sites mostly arose from different characteristic of the soils and of the vegetation cover. At short temporal scales, it cannot be nevertheless excluded that episodic weather phenomena influenced differently the stations far from one another.
We show the average profiles of
All profiles of Fig. 3 show an isotopic enrichment near the
surface. This is due to the preferred evaporation of the lighter
isotopes from bare soil (Barnes and Allison, 1998). At the surface
(first 10
Proceeding downwards, the
In Fig. 4 we show d-excess profiles calculated from observed
All profiles show a minimum near the surface. This descends from the
fact that evaporation of HDO is favored during bare soil evaporation
compared to evaporation of
The territory of Labytnangi, extremely heterogeneous in its vegetation cover, is characterized by marked horizontal and vertical variability in morphology, soil, and processes. The horizontal inter-site variability has been addressed by repeating the model tuning (below) for each of the study sites. The vertical heterogeneity in soil types at a given station is, on the other hand, not accounted for in ORCHIDEE at this stage, nor are vertical processes as cryoturbation (e.g. Bockheim and Tarnocai, 1998) that may alter the physical properties of the soil (e.g. in gley tundra) and/or determine rearrangements of the soil vertical structure, in turn affecting the isotopic composition of soil moisture.
With the goal of reproducing the observed vertical isotopic profiles,
we investigated the sensitivity of the model output to various
parameters influencing the infiltration, the bare-soil evaporation,
and the vertical diffusion. Once identified the value of each
parameter leading to the best fit to the observed profiles for each
station, we checked its consistency both with the a priori knowledge
of the sites (Valdayskikh et al., 2013) and with the information from
the analysis of the observed isotopic profiles. In Fig. 3 and Fig. 4
(green curves) we show the outcome of the best simulation (i.e. the
profiles closest to the observed values) of
Figure 3 shows the best-simulated
For station 3, we observe a good match in the surface enrichment and
in the transition down to 25
The model is able to capture for each site the recent summer
enrichment, but not the memory of the previous summer, i.e. the
secondary peak we observe 45
Figure 4 shows the simulated d-excess profiles obtained using for the
same stations the same values of the key parameters that led to the
best fit of
The d-excess forcing might be affected by a large error, due to
(
Profiles of isotopes in soil moisture may show steep vertical gradients (our study, Wingate et al., 2009; Gazis and Geng, 2004), resulting from a combination of surface evaporation and small-scale vertical hydrological processes, such as vertical infiltration. Vertical processes are described in ORCHIDEE-iso in a rather simplified and idealized way as compared to state-of-the-art LSMs, but even in vertically resolved modeling approaches, where they are explicitly described (de Rosnay, 1999), infiltration processes are still difficult to simulate (de Rosnay et al., 2002). With the aim of evaluating such processes in ORCHIDEE-iso, we investigated the sensitivity of the simulated isotopic profiles to vertical diffusivity, infiltration pathway of the precipitation (Gazis and Geng, 2004), which feed back on the relative proportions of evapotranspiration, surface runoff, and drainage (Boone et al., 2010; Ducharne et al., 1998), and to surface evaporation.
For each set of tests we considered as reference run the best fit
obtained for
In Fig. 5 we show for each station the profiles corresponding to the
best fit and those obtained with different values of
The extent to which water infiltrates in soil depends in first place on the soil hydraulic conductivity, a function of the soil texture (e.g. percentage of sand, silt and clay) and structure. The coarser the soils, the larger the pores, and the faster the vertical transfer. A variety of factors (flow through shrinkage cracks, passages created by roots and earthworms) may though, in finer soils, result in the formation of highly water-conductive pathways. This feature may drastically raise the overall infiltration rate of such soils, otherwise characteristically slow.
The mechanism of water progressively penetrating in the soil with time, or piston flow, has been described by some conceptual models (e.g. Green and Ampt, 1911). In ORCHIDEE-iso, this pathway is represented by having the rainfall collected in first place only by the model upper soil layer and, then, having it propagate vertically with time. Preferential pathway is, on the other hand, represented in our model by evenly distributing the precipitation input throughout the soil column (i.e. assuming instantaneous vertical flow).
Figure 6 shows the impact of the chosen infiltration pathway on the
simulated
The infiltration pathway does not appear to significantly influence the surface enrichment, apart from a slight effect (higher by 0.5 ‰ for the piston case) in case of station 3.
The infiltration pathway best fitting the observation profiles is of
preferential flow type for all stations except station 1. Piston-flow
reproduces best the profile at this station, as it is characterized by
thinner soil. In case of station 3, for which the seasonal signal is
clear in the observations (Fig. 3), the curve obtained with
piston-like transfer does not capture properly the 40
In our ORCHIDEE setup, each plant functional type is associated with
a seasonally varying portion of bare soil. This variation is
calculated monthly according to the following equation:
In Fig. 7 we explore the impact of varying ext_c, i.e. modifying the
relative fraction of bare soil, on the isotopic profiles. Increasing
ext_c leads to lower values of bare_soil. Reduced amount of
evaporated water translates, in turn, into lower isotopic surface
enrichment. It is important here to stress the fact that the four
sites are not directly comparable, as they were a priori differently
characterized in terms of vegetation cover in the model (only the two
tundra stations share the same fractions of PFTs, Table 1), and that
the different bare soil fractions tested here are relative to the PFTs
of this initial characterization. For station 2–4, for which the
infiltration follows a preferential pathway, the effect of ext_c is
remarkable, particularly in the first 20
With the aim of explaining the differences among different sites, we chose two among the four experimental sites: station 1 (silty tundra) and 4 (floodplain). The corresponding profiles have contrasting shapes and are relatively well captured by the model. At these two sites the best fits have been obtained with different values of the three tunable parameters, namely infiltration pathway, bare soil fraction, and vegetation cover. We tried to decompose the difference between the simulated profiles into the single contribution of each of those three parameters.
Figure 8 shows the differences between
The difference in the experimental data profiles (green, with
associated standard deviation in yellow) shows a higher enrichment
(negative values) at the surface for station 4. ORCHIDEE-iso, although
qualitatively capturing such enrichment (blue curve), overestimates
it. In the model world such enrichment might be due to the higher
fractional bare soil in the vegetation coverage (Table 1) prescribed
for station 4. This effect overcompensates the lower ext_c
value. Below 30
The previous section has highlighted the impact of the bare soil
fraction on the surface enrichment. Here we investigate whether water
isotope observations allow us to estimate the contribution of bare
soil evaporation to the evapotranspiration (defined here as sum of
bare soil evaporation and transpiration), i.e. the
The measured soil isotopic composition
The
This paper shows the results of simulations of vertical profiles of
ORCHIDEE-iso has been tuned at each station with respect to bare soil
cover, infiltration pathway, and vertical diffusivity; it reproduces
qualitatively and quantitatively well the surface enrichment in
observed vertical profiles of
Sensitivity tests on the impacts of vertical soil water diffusivity
and infiltration on the isotopic vertical composition simulated by
ORCHIDEE-iso show that the choice of the infiltration pathway is
essential in determining the shape of the profile and most relevant
among the two parameters. A piston-like flow does not affect the
surface enrichment, but highlights the memory of the past season below
20
Our approach combining model results and observations allowed
estimating the ratio
A model set up with a basic a priori knowledge of a site could deliver further information on the site itself. We confirmed that interactions between model and measurements is not unidirectional, but mutual. If on one hand isotopes help in best tuning the model, simulation results help in tracing back information about conditions and processes at the experimental sites.
The approach followed in this study appears to be even more promising in view of the current developments of ORCHIDEE toward a version for high latitudes. On one hand the model is transitioning to a setup with eleven vertical layers (de Rosnay et al., 2002), on the other hand, in progress is also the description of the dynamics of permafrost and of its interaction with snow and vegetation (Gouttevin et al., 2012), coupled with a vertically resolved snow model (Wang et al., 2013). This higher resolution in processes would confer a much higher degree of realism to the simulations.
Furthermore, to simulate the water cycle in modern climate scenarios and, among other things, address persistent and systematic biases affecting the models (Meehl et al., 2007), running simulations coupling ORCHIDEE-iso with an atmospheric general circulation model (Risi et al., 2013, 2010b) would be valuable. Improvements in the description of the partitioning of evapotranspiration into evaporation and transpiration, which could benefit from isotopic measurements, may moreover help in reducing the uncertainties in the land surface hydrological response to climate change.
The observation data used in this study were the minimum required to test a land surface model. Ideally, measurements in all water reservoirs (soil, precipitation, water vapor – in situ or remotely detected –, leaf and stem water) and time-series (a full seasonal cycle, at monthly scale) would be desirable. Moreover, co-located temperature and soil moisture profiles should be taken, accompanied by complete meteorological forcing, to capture the variability at small-scale.
This work was partly supported by the WSibIso project with a grant of Ministry of Education and Science of the Russian Federation under the contract No. 11.G34.31.0064. The authors are thankful to E. Rakov and V. Gryazin (Ural Federal University) for fruitful discussions.
Bare soil, vegetation coverage (fraction of the grid cell), and soil type (predominant USDA grain size) in ORCHIDEE for the four simulation sites. Vegetation is expressed in terms of plant functional types.
Values of the tunable parameters for which the simulated profiles best
fit the observations in
Identifying the differences between simulated
The territory of Labytnangi with the measurement sites indicated. Image from Google Earth, map data: ©2015 CNES/Astrium, Landsat, ©2015 DigitalGlobe, ©2015 Google.
A schematic description of ORCHIDEE-iso. In red the processes for which isotopic fractionation occurs.
d-excess calculated from in situ
Sensitivity of the simulated
Sensitivity of the simulated
Sensitivity of the simulated
Differences between stations 1 (silty tundra) and 4
(floodplain) in terms of observed
Red lines: difference between