Miba Spezial 93 Pdf Converter. ACIER POUR TREFILAGE. ACIER POUR TUBE. ACIER POUR USAGE GENERAL. ACIER POUR VERIN HYDRAULIQUE. The database recognizes 1,746,000 software titles and delivers updates for your software.

Figure 3: Evaluation of hydrological and isotopic variables simulated by ORCHIDEE on different MIBA or Carbo-Europe sites. A, d, g, j, m: latent (green) and sensible (red) heat fluxes observed locally when available (circles), simulated in the ERA-Interim reanalyses (stars) and simulated by ORCHIDEE (lines). B, e, h, k, n: normalized soil moisture content (SWC, without unit) observed locally (circles) and simulated by ORCHIDEE (lines). C, f, i, l, o: δ 18O of the surface soil (brown) and stems (green) simulated by ORCHIDEE in the control offline simulations (thin curves) and observed (circles). Observed δ 18O in precipitation (thick dashed red) and vapor (thick dashed blue) used as forcing are also shown. A-c: Le Bray, d-f: Yatir, g-i: Morgan-Monroe, j-l: Donaldson Forest, m-o: Anchorage.

The normalized SWC (soil water content) is calculated. Figure 4: Same as Figure 3 but for Mitra (a-c), Bily Kriz (d-f), Brloh (g-i), Hainich (j-l: Donaldson Forest), and Tharandt (m-o) The soil moisture seasonality is very well simulated at all sites where data is available ( Figures 3 and 4), except for a two-month offset at Yatir ( Figure 3f). Water isotopes in the soil water: The evaluation of the isotopic composition of soil water is crucial before using ORCHIDEE to investigate the sensitivity to the evapo-transpiration partitioning or to infiltration processes, or in the future to simulate the isotopic composition of paleo-proxies such as speleothems [].

In observations, at all sites, δ 18O s remains close to δ 18O p, within the relatively large month-to-month noise and spatial heterogeneities ( Figures 3 and 4) At most sites (Le Bray, Donaldson Forest, Anchorage, Bily Kriz and Hainich), observed δ 18O s exhibits no clear seasonal variations distinguishable from month-to-month noise. At Morgan- Monroe and Mitra, and to a lesser extent at Brloh and Tharandt, δ 18O s progressively increases throughout the spring, summer and early fall, by up to 5‰ at Morgan-Monroe. The increase in δ 18O s in spring can be due to the increase in δ 18O p. The increase in δ 18O s in late summer and early fall, while δ 18O p starts to decrease, is probably due to the enriching effect of bare soil evaporation.

At Yatir, δ 18O s increases by 10‰ from January to June, probably due to the strong evaporative enrichment on this dry site. Then, the δ 18O s starts to decline again in July. This could be due to the diffusion of depleted atmospheric water vapor in the very dry soil.

ORCHIDEE captures the order of magnitude of annual-mean δ 18O s on most sites, and captures the fact that it remains close to δ 18O p. ORCHIDEE captures the typical δ 18O s seasonality, with an increase in δ 18O s in spring-summer at Morgan-Monroe, Donaldson Forest, Mitra and Bily Kriz.

However, the sites with a spring-summer enrichment in ORCHIDEE are not necessarily those with a spring-summer enrichment in observations. This means that ORCHIDEE misses what controls the inter-site variations in the amplitude of the δ 18O s seasonality. The seasonality is not well simulated at Yatir. This could be due to the missed seasonality in soil moisture and evapo-transpiration. This could be due also to the fact that at Yatir ORCHIDEE underestimates the proportion of bare soil evaporation to total evapo-transpiration: less than 10% in ORCHIDEE versus 38% observed [], which could explain why the spring enrichment is underestimated. Besides, ORCHIDEE does not represent the diffusion of water vapor in the soil, which could explain why the observed δ 18O s decrease at Yatir in fall is missed.

When comparing the different sites, annual-mean δ 18O s follows annual-mean δ 18O p, with an inter-site correlation of 0.99 in observations. Therefore, it is easy for ORCHIDEE to capture the intersite variations in annual-mean δ 18O s. A more stringent test is whether ORCHIDEE is able to capture the inter-site variations in annual-mean δ 18O s - δ 18O p. This is the case, with a correlation of 0.85 ( Figure 5a) between ORCHIDEE and observations. In ORCHIDEE (and probably in observations), spatial variations in δ 18O s - δ 18O p are associated with the relative importance of bare soil evaporation. Figure 5: a) Relationship between simulated and observed annual-mean δ 18O in the soil water (red), stem water (blue) and leaf water (green), to which the precipitation-weighted annual-mean precipitation δ 18O is subtracted. In the case of perfect model-data agreement, markers should fall on the y=x line.

Kings Court Card Game Instructions more. B) Relationship between the annual-mean δ 18O in the soil water and in stem water, to which the precipitation-weighted annual-mean precipitation δ 18O is subtracted, for both ORCHIDEE (magenta) and observations (cyan). When soil and stem water share the same δ 18O, they fall on the y=x line. Water isotopes in the stem water: In observations, observed δ 18O stem exhibits no seasonal variations distinguishable from month-tomonth noise ( Figures 3 and 4). At Le Bray, Yatir, Mitra, Brloh, Hainich, observed δ 18O stem is more depleted than the surface soil water.

It likely corresponds to the δ 18O values in deeper soil layers, suggesting that the rooting system is quite deep. For example, at Mitra, the root system reaches least 6 m deep, and could at some places reach as deep as 13 m where it could use depleted ground water.

At Donaldson Forest, Morgan-Monroe, Anchorage and Tharandt, δ 18O stem is very close to δ 18O s, maybe reflecting small vertical variations in isotopic composition within the soil or shallow root profiles. At Bily Kriz, observed δ 18O stem is surprisingly more enriched than surface soil water. Several hypotheses could explain this result: (1) the surface soil water could be depleted by dew or frost at this mountainous, foggy site; (2) spruce has shallow roots and therefore sample soil water that is not so depleted; (3) the twigs that were sampled were relatively young so that evaporation from their surface could have occurred when they were still at tree; (4) twigs were sampled in sun-exposed part of the spruce crowns during sunny conditions, which could favor some evaporative enrichment. Additional measurements show a lower Deuterium excess in the stem water compared to the soil water, supporting evaporative enrichment of stems. ORCHIDEE captures the fact that δ 18O stem is nearly uniform throughout the year.

As for soil water, it is easy for ORCHIDEE to capture the inter-site variations in annual-mean δ 18O stem (inter-site correlation between ORCHIDEE and observations of 0.90). ORCHIDEE is able to capture some of the inter-site variations in annual-mean δ 18O stem - δ 18O p, with a inter-site correlation between ORCHIDEE and observations of 0.60. However, ORCHIDEE simulates δ 18O stem values that are very close to δ 18O s values ( Figure 5b).

It is not able to capture δ 18O stem values that are either more enriched or more depleted than δ 18O s. This could be due to the fact that ORCHIDEE underestimates vertical variations in soil isotopic composition. Also, ORCHIDEE is not designed to represent deep ground water sources or photosynthesizing twigs. Vertical profiles of soil water isotope composition: At Le Bray, we compare our offline simulation for 2007 with soil profiles collected from 1993 to 1997 and in 2007 ( Figure 6a-6b).

The year mismatch adds a source of uncertainty to the comparison. In summer (profiles of August 1993 and September 1997), the data exhibits an isotopic enrichment at the soil surface of about 2.5‰ compared to the soil at 1 m depth ( Figure 6a), likely due to surface evaporation [].

Then, by the end of September 1994, the surface becomes depleted, likely due to the input of depleted rainfall. Previously enriched water remains between 20 and 60 cm below the ground, suggesting an infiltration through piston-flow []. ORCHIDEE predicts the summer isotopic enrichment at the surface, but slightly later in the season (maximum in September rather than August) and underestimates it compared to the data (1.5‰ enrichment compared to 2.5‰ observed, Figure 6b). The model also captures the surface depletion observed after the summer, as well as the imprint of the previous summer enrichment at depth. However, ORCHIDEE simulates the surface depletion in December, whereas the surface depletion can be observed sooner in the data, at the end of September 1994. Figure 6: Vertical profiles of soil δ 18O measured (a,c) and simulated by ORCHIDEE for the control offline simulations (b,d) on the Bray site (a,b) and the Yatir sites (b,d).

Beware that the y-scales for observations and simulations are different. This is because the representation of the soil water content is very rudimentary in the ORCHIDEE model, preventing any quantitative comparison of measured and simulated soil depth. The horizontal black dashed line represents the bottom of the observed profiles.

Model outputs are sampled at the same time as the data. For the Yatir sites, frequent soil sampling for the same year allowed us plot representative bi-monthly averages for both measured and simulated profiles. This could not be the case for Le Bray. Some soil profiles were observed at Le Bray in 2007, but we do not show them because they are limited to the top 24 cm of the soil only.

At Yatir, observed profiles exhibit a strong isotopic enrichment from deep to shallow soil layers in May-June by up to 10‰ ( Figure 6c). As for Le Bray, the model captures but underestimates this isotopic enrichment in spring and summer by about 3‰ ( Figure 6d). This discrepancy could be the result of underestimated bare soil evaporation.

Observed profiles also feature a depletion at the surface in winter that the model does not reproduce. This depletion could be due to back-diffusion of depleted vapor in dry soils [,-], a process that is not represented in ORCHIDEE but likely to be significant in this region. Soil evaporation fluxes measured with a soil chamber at Yatir shows that when soils are dry, there is adsorption of vapor from the atmosphere to the dry soil pores before sunrise and after sunset []. Water isotopes in leaf water: It is important to evaluate the simulation of the isotopic composition of leaf water by ORCHIDEE if we want to use this model in the future for the simulation of paleoclimate proxies such tree-ring cellulose [,], for the simulation of the isotopic composition of atmospheric CO 2 which may be used to partition CO 2 fluxes into respiration from vegetation and soil [,] or for the simulation of the isotopic composition of atmospheric O 2 which may be used to infer biological productivity [,].

In the observations, δ 18O leaf exhibits a large temporal variability reflecting a response to changes in environmental conditions (e.g., relative humidity and the isotopic composition of atmospheric water vapor). At all sites except at Yatir, δ 18O leaf is most enriched in summer than in winter, by up to 15‰ ( Figures 3 and 4). This is because the evaporative enrichment is maximum in summer due to drier and warmer conditions. ORCHIDEE captures the maximum enrichment in summer. However, ORCHIDEE underestimates the annual-mean δ 18O leaf at most sites ( Figure 5).

This could be due to the fact that most leaf samples were collected during the day, when the evaporative enrichment is at its maximum, while for ORCHIDEE we plot the daily-mean δ 18O leaf. At Le Bray, if we sample the simulated δ 18O leaf during the correct days and hours, simulated δ 18O leaf increases by 4‰ in winter and by 10‰ in summer. Such an effect can thus quantitatively explain the model-data mismatch.

After taking this effect into account, simulated δ 18O leaf may even become more enriched than observed. This is the case at Le Bray, especially in summer. The overestimation of summer δ 18O leaf could be due to neglecting diffusion in leaves or non-steady state effects.

Again, Yatir is a particular case. Minimum δ 18O leaf occurs in spring-summer while the soil evaporative enrichment is maximum. In arid regions and seasons, leaves may close stomata during the most stressful periods of the day, inhibiting transpiration, and thus retain the depleted isotopic signal associated with the moister conditions of the morning [,]. ORCHIDEE does not represent this process and thus simulates too enriched δ 18O leaf. Summary: Overall, ORCHIDEE is able to reproduce the main features of the seasonal and vertical variations in soil water isotope content, and seasonal variations in stem and leaf water content.

Discrepancies can be explained by some sampling protocols, by shortcomings in the hydrological simulation or by neglected processes in ORCHIDEE (e.g., fractionation in the vapor phase). The strong spatial heterogeneity of the land surface at small scales does not prevent ORCHIDEE from performing reasonably well. This suggests that in spite of some small-scale spatial heterogeneities at each site, local isotope measurements contain large-scale information and are relevant for the evaluation of large-scale LSMs. Sensitivity analysis Sensitivity to evapo-transpiration partitioning: Several studies have attempted to partition evapo-transpiration into the transpiration and bare soil evaporation terms at the local scale [-,]. Estimating E/ET, where E is the bare soil evaporation and ET is the evapo-transpiration, requires measuring the isotopic composition of soil water, stem water and of the evapo-transpiration flux.

The isotopic composition of the evapo-transpiration can be estimated through “Keeling plots” approach [], but this is costly [] and the assumptions underlying this approach are not always valid []. Considering a simple soil water budget at steady state and with vertically-uniform isotopic distribution, we show that although estimating E/ET requires measuring the isotopic composition of the evapo-transpiration flux, estimating E/I (where I is the precipitation that infiltrates into the soil) requires measuring temperature, relative humidity (h) and the isotopic composition of the soil water (δ 18O s), water vapor (δ 18O v) and precipitation (δ 18O p) only. Such variables are available from several MIBA and Carbo-Europe sites. More specifically, E/I is proportional to δ 18O p - δ 18O s: (1) where α eq and α k are the equilibrium and kinetic fractionation coefficients respectively. Below, we show that this equation can apply to annual-mean quantities, neglecting effects associated with daily or monthly covariations between different variables. We investigate to what extent this equation allows us to estimate the magnitude of E/I at local sites. At the Yatir site, all the necessary data for equation 1 is available.

An independent study has estimated E/I =38% []. Download Free Flash Template Interactive Design. Using annually averaged observed values (δ 18O p=-5.1‰ and δ 18O s=-3.7‰ in the the surface soil), we obtain E/I =46%. However, in ORCHIDEE, the annually averaged surface δ 18O s is 0.8 lower when sampled at the same days as in the data.

When correcting for this bias, we obtain E/I =28%. Observed E/I lies between these two estimates. This shows the applicability of this estimation method, keeping in mind that estimating E/I is the most accurate where E/I is lower.

When we perform sensitivity tests to ORCHIDEE parameters at the various sites, the main factor controlling δ 18O s is the E/I fraction. This is illustrated as an example at Le Bray and Mitra sites ( Figure 7). Sensitivity tests to parameters as diverse as the rooting depth or the stomatal resistance lead to changes in δ 18O s - δ 18O p and in E/I that are very well correlated, as qualitatively predicted by equation 13. This means that whatever the reason for a change in E/I, the effect on δ 18O s - δ 18O p is very robust.

Figure 7: Isotopic difference between soil water and precipitation (δ 18O s- δ 18O p) as a function of E/I (fraction of the infiltrated water that evaporates at the bare soil surface), for different sensitivity tests in ORCHIDEE. A) at Le Bray and b) at Mitra. All values are annual means. The horizontal dashed line represents the observed values for δ 18O s- δ 18O p. The orange dashed line shows the best linear fit between the different sensitivity tests. Quantitatively, the slope of δ 18O s - δ 18O p as a function of E/I among the ORCHIDEE tests is of 0.78‰/% (r=0.94, n=6) at Le Bray and of 0.25‰/% (r=0.999, n=5) at Mitra, compared to about 0.25-0.3‰/% predicted by equation 13.

The agreement is thus very good at Mitra. The better agreement at Mitra is because it is a dry site where E/I varies greatly depending on sensitivity tests. In contrast, Le Bray is a moist site where E/I values remains small for all the sensitivity tests, so numerous effects other than E/I and neglected in equation 13 can impact δ 18O s - δ 18O p. To summarize, local observations of δ 18O s - δ 18O p could help constrain the simulation of E/I in models. This would be useful since the evapo-transpiration partitioning has a strong impact on how an LSMs represents land-atmosphere interactions []. Sensitivity to soil infiltration processes: Partitioning between evapo-transpiration, surface and drainage depends critically on how precipitation water infiltrates the soil [,,], which is a key uncertainty even in multi-layer soil models where infiltration processes are represented explicitly []. It has been suggested that observed isotopic profiles could help understand infiltration processes at the local scale [].

The capacity of ORCHIDEE to simulate soil profile allows us to investigate whether measured isotope profiles in the soil could help evaluate the representation of these processes also in largescale LSMs. With this aim, we performed sensitivity tests at Le Bray. The simulated profiles are sensitive to vertical water fluxes in the soil. When the diffusivity of water in the soil column is decreased by a factor 10 from 0.1 to 0.01 compared to the control simulation, the deep soil layer becomes more depleted by about 0.7‰ ( Figure 8) and the isotopic gradient from soil bottom to top becomes 30% steeper in summer, because the enriched soil water diffuses slower through the soil column.

Figure 8: Sensitivity of simulated δ 18O s profiles to the parameterization of infiltration processes in the soil at Le Bray. July (a) and December (b) are shown for three different parameterizations in offline simulations: control simulation (solid red), a simulation in which the soil water diffusivity was divided by 10 (dashed blue) and a simulation is which the water infiltrates the soil uniformly in the vertical (crude representation of preferential pathways, dash-dotted green) rather than in a piston-like way as is the case for other simulations.

Simulated profiles are also sensitive to the way precipitation infiltrates the soil. When precipitation is added only to the top layer (piston-flow infiltration) the summer enrichment is reduced by mixing of the surface soil water with rainfall, and it propagates more easily to lower layers during fall and winter. Conversely, when rainfall is evenly spread throughout the soil column (a crude representation of preferential pathway infiltration), the surface enrichment is slightly more pronounced and the deep soil water is more depleted by up to 0.8‰ in winter ( Figure 8). However, the observed surface depletion occurs in February with preferential pathways, compared to December in the piston-like in infiltration. The quick surface depletion observed after the summer suggests that infiltration is dominated by the pistonlike mechanisms. To summarize, we show that vertical and seasonal variations of δ 18O s are very sensitive to infiltration processes, and are a powerful tool to evaluate the representation of these processes in LSMs. Global-scale Simulations Using the Coupled LMDZORCHIDEE Model Simulation set-up To compare with global datasets, we performed LMDZ-ORCHIDEE coupled simulations.

In all our experiments, LMDZ three-dimensional fields of horizontal winds are nudged towards ECMWF (European Center for Medium range Weather Forecast) reanalyses []. This ensures a realistic simulation of the large-scale atmospheric circulation and allows us to perform a day-to-day comparison with field campaign data [,]. At each time step, the simulated horizontal wind field is relaxed towards the reanalysis following this equation: where is the reanalysis horizontal wind field, is the effect of all simulated dynamical and physical processes on, and τ is a time constant set to 1 h in our simulations []. To compare with global datasets, LMDZ-ORCHIDEE simulations are performed for the year 2006, chosen arbitrarily. We are not interested in inter-annual variations and focus on signals that are much larger. To ensure that the water balance is closed at the annual scale, we performed iteratively 10 times the year 2006 as spin-up.

In these simulations, the Peclet and non-steady state effects are de-activated. To compare with field campaign observations in 2002 and 2005, we use simulations performed for these specific years, initialized from the 2006 simulation. In these simulations, we test activating or deactivating the Peclet effect. In all LMDZ-ORCHIDEE simulations, canopy-interception was de-activated (consistent with simulations that our modeling group performed for the Fourth Assessment Report). Evaluation of water isotopes in leaf water at the diel scale during campaign cases Daily data from field campaigns: Two field campaigns are used to evaluate the representation of δ 18O leaf diurnal variability.

The first campaign covers six diurnal cycles in May and July 2002 in a grassland prairie in Kansas (39.20 ° N 96.58 ° W []). The second campaign covers four diurnal cycles in June 2005 in a pine plantation in Hartheim, Germany (7.93 ° N, 7.60 ° E []). Because meteorological and isotopic forcing are not available for the entire year, we prefer to compare these measurements with LMDZORCHIDEE simulations. At both sites, the simulated δ 18O v and δ 18O stem are consistent with those observed (model-data mean difference lower than 1.4‰ in Kansas and 0.4‰ at Hartheim), allowing us to focus on the evaluation of leaf processes. Evaluation results: At the Kansas grassland site, δ 18O leaf exhibits a diel cycle with an amplitude of about 10‰ []. LMDZ-ORCHIDEE captures this diel variability, both in terms of phasing and amplitude ( Figure 9).

The model systematically overestimates δ 18O leaf by about 4‰, in spite of the underestimation of the stem water by 1.4‰ on average. This may be due to a bias in the simulated relative humidity (LMDZ is on average 13% too dry at the surface, which translates into an expected enrichment bias of 3.9‰ on the leaf water assuming steady state based on Equation 7) or to uncertainties in the kinetic fractionation during leaf water evaporation. Figure 9: δ 18O of stem and grass leaves measured during two series of 3 diurnal cycles in May and July 2002 over the plains of Kansas [120] and simulated by LMDZ-ORCHIDEE for the same year in the grid box containing the observation site. Δ 18O of vapor (blue), pine leaves (pink and red) and stems (green) measured during four diurnal cycles in June 2005 in Hartheim, Germany [121] and simulated by LMDZ-ORCHIDEE for the same year in the grid box containing the observation site.

Simulated values are dashed, observed values solid. Two kinds of leaves were sampled during this campaign: one-year-old leaves (solid pink) and current-year leaves (solid brown). Two leaf water diagnostics were computed for in LMDZ-ORCHIDEE: stationary state at the evaporative site (dashed red, equation 7) or non-stationary state in the lamina, taking into account the Peclet effect (dashed brown, equation 9, using an effective length scale of 25 mm). At the Hartheim pine plantation, δ 18O leaf is on average 8‰ more depleted for current-year needles than for 1-year-old needles. Also, the observed diel amplitude is weaker for current-year needles (5 to 8‰) than for 1-year-old needles (10 to 15‰). These observations are consistent with a longer diffusion length for current-year needles (15 cm) than for 1-year-old needles (5 cm) [] and with a larger transpiration rate, leading to a stronger Peclet effect. When neglecting Peclet and non-steady state effects, ORCHIDEE simulates an average δ 18O leaf close to that of 1-year-old needles, consistent with the small diffusion length and evaporation rate of these leaves.

ORCHIDEE captures the phasing of the diurnal cycle, but underestimates the diel amplitude by about 4‰. This is probably due to the underestimate of the simulated diel amplitude of relative humidity by 20%. Accounting for Peclet and non-steady state effects strongly reduces both the average δ 18O leaf and its diel amplitude ( Figure 9), in closer agreement with current-year needles. To summarize, ORCHIDEE simulates well the leaf water isotopic composition. The leaf water isotope calculation based on Craig et al. [] simulates the right phasing and amplitude for leaves that have short diffusive lengths or low transpiration rates.

Non-steady state and diffusion effects need to be considered in other cases. By activating or de-activating these effects, ORCHIDEE can simulate all cases. Evaluation of water isotopes in precipitation Precipitation datasets: To evaluate the spatial distribution of precipitation isotopic composition simulated by the LMDZORCHIDEE coupled model, we use data from the Global Network for Isotopes in Precipitation (GNIP []), further complemented by data from Antarctica [] and Greenland []. We also use this network to construct isotopic forcing at sites where the precipitation was not sampled, complemented with the USNIP (United States Network for Isotopes in Precipitation []) network. Evaluation results: At the global scale, the LMDZ-ORCHIDEE coupled model reproduces the annual mean distribution in δ 18O p and d p observed by the GNIP network reasonably well ( Figure 10), with correlations of 0.98 and 0.46 and Root Mean Square Errors (RMSE) of 3.3‰ and 3.5‰ respectively. Figure 10: a) Annual mean δ 18O p from GNIP [122], Antarctica [123] and Greenland [124] data.

The data is gridded over a coarse 7.5 × 6.5° grid for visualization purposes. B) Same as a) but for annual mean d p. C) Annual mean δ 18O p simulated by coupled LMDZ-ORCHIDEE model for the control simulation. D) same as c) but for annual mean d p. This good model-data agreement can be obtained even when we deactivate ORCHIDEE. When we use LMDZ in a stand-alone mode, in which the isotope fractionation at the land surface is neglected [], the model-data agreement is as good as when we use LMDZ-ORCHIDEE.

Therefore, fractionating processes at the land surface have a second order effect on precipitation isotopic composition, consistent with [,-]. To quantify in more detail the effect of fractionation at the land surface, we performed additional coupled simulations with LMDZORCHIDEE. We compare the control simulation described above (ctrl) to a simulation in which fractionation at the land surface was deactivated (nofrac) ( Figure 11). In nofrac, the composition of bare soil evaporation equals that of soil water. Even when restricting the analysis to continental regions, the spatial correlations between the ctrl and nofrac simulations are 0.999 and 0.95 for δ 18O p and d p respectively, and the root mean square differences are 0.27‰ and 1.1‰ for δ 18O p and d p respectively.

This confirms that fractionation at the land surface has a second-order effect on precipitation isotopic composition compared to the strong impact of atmospheric processes. Figure 11: a) Annual-mean δ 18O p in the ctrl simulation (LMDZ-ORCHIDEE) minus annual mean δ 18O p in the nofrac simulation (LMDZ-ORCHIDEE in which the isotopic fractionation was de-activated during bare soil evaporation). This shows the effect of isotopic fractionation at the soil surface on δ 18O p. B) Same as a) but for d p. However, to second order, a detailed representation of fractionation at the land surface lead to a slight improvement in the simulation of δ 18O p and to a significant improvement in that of d p.

In ctrl, δ 18O p is lower by up to 1.5‰ and d p higher by up to 5‰ than in nofrac over boreal continental regions such as Siberia, Canada and central Asia, consistent with the expected effect of fractionation at surface evaporation []. Taking into account fractionation at the land surface leads to a better agreement with the GNIP data over these regions, where δ 18O p is overestimated by about 4‰ and d p underestimated by 4 to 7‰ when neglecting fractionation at the land surface. The effect of fractionation is maximal over these boreal regions because (1) the fraction of bare soil evaporation is maximal, (2) a significant proportion of evaporativelyenriched soil water is lost by drainage and (3) a larger proportion of the moisture comes from land surface recycling [,,].

Similar results were obtained with other models []. To summarize, LMDZ-ORCHIDEE simulates well the spatial distribution of precipitation isotopic composition, but this distribution is not a very stringent test for the representation of land surface processes in ORCHIDEE.

In the next section, we argue that the distribution of river isotopic composition is a more stringent test. Evaluation of water isotopes in river water: Large rivers integrate a wide range of hydrological processes at the scale of GCM grid boxes [,,-]. Here we evaluate the isotopic composition of river water simulated by ORCHIDEE using data collected by the Global Network for isotopes in Rivers (GNIR [,]). Observed annual mean δ 18O river follows to first order the isotopic composition of precipitation [], and is thus also well simulated by LMDZ-ORCHIDEE ( Figure 12a and 12b), with a spatial correlation between measured and simulated δ 18O river of 0.80 and a RMSE of 3.2‰ over the 149 LMDZ grid boxes containing data. Regionally however, the δ 18O difference between precipitation and river water (δ 18O river - δ 18O p) can be substantial and provides a stronger constraint for the model.

Over South America, Europe and some parts of the US, the river water is typically 1‰ to 4‰ more depleted than the precipitation ( Figure 12a), because precipitation contributes more to rivers during seasons when it is the most depleted []. In contrast, over central Asia or northern America, river water is more enriched than precipitation, due to evaporative enrichment of soil water [,,].

This is further confirmed by a simulation where fractionation at the land surface was neglected (not shown), for which the river water is in global average 5‰ more depleted. Figure 12: (a) Annual mean δ 18O p in rivers (δ 18O river) measured from the GNIR database. (b) Simulated by LMDZ-ORCHIDEE for the control simulation.

(c) Annual mean δ 18O river-δ 18O p observed from the GNIR and GNIP databases. (d) Simulated by LMDZ-ORCHIDEE for the control simulation. On sub-plots d and f the United States, where the GNIR network is the densest, are enlarged for better readability.

ORCHIDEE reproduces moderately well the magnitude and patterns of δ 18O river - δ 18O p, with a spatial correlation of 0.39 and a RMSE of 2.7‰ over the 22 LMDZ grid boxes that contain δ 18O river observations. It simulates the negative values over the western US, Europe and South America and the positive value over Mongolia.

However, the model does not capture the positive δ 18O river - δ 18O p in Eastern US, though positive values are simulated further North. This suggests that such a diagnostic may help identify biases in the representation of the soil water budget, as discussed in the following section. Sensitivity to the representation of pathways from precipitation to rivers At the local scale, water isotopes have already been used to partition river discharge peaks into the contributions from recent rainfall and soil water [-]. Given the property of rivers to integrate hydrological processes at the basin scales [,,-], we now explore to what extent δ 18O river could help evaluate pathways from precipitation to rivers in LSMs. We illustrate this using seasonal variations in δ 18O river on two well established GNIR and GNIP stations in Vienna (Danube river) and Manaus (the Amazon) ( Figure 13). The seasonal cycle in δ 18O river is attenuated compared to that in δ 18O p, and δ 18O river lags δ 18O p (by 5 month at Vienna and 1-3 months at Manaus).

Figure 13: Seasonal variations in δ 18O p (a,b) and δ 18O river (c,d) observed (solid black) and simulated for the control LMDZ-ORCHIDEE simulation (dashed black) for (a,c) the Danube river in Vienna and (b,d) the Amazon river in the Manaus region (average over the 8 ° S-3 ° S-56 ° W 63 ° W domain). Also shown are δ 18O river for simulations where the total runoff is partitioned into surface runoff only without drainage (dash-dotted blue) and where we multiplied by two the time residence in the reservoir collecting drainage in the routing scheme (dash-dotted red). Beware that the y-scale is different on the two sites. The difference in the annual-mean values between the two sites reflect the difference in the annual-mean δ 18O p. LMDZ-ORCHIDEE (control simulation) simulates qualitatively well the amplitude and the phasing observed in δ 18O p and δ 18O river.

To understand better what determines the attenuation and lag of the seasonality in δ 18O river compared to that in δ 18O p, we perform sensitivity tests to ORCHIDEE parameters. Parameters tested include the partitioning of excess rainfall into surface runoff and drainage and the residence time scale of different reservoirs (slow, fast and stream) in the routing scheme. River discharge is extremely sensitive to these parameters []. If all the runoff occurs as surface runoff ( Figure 13), then the seasonal cycle of δ 18O river is similar to that of δ 18O p. This shows that the attenuation and lag of the seasonality in δ 18O river compared to that in δ 18O p are caused by the storage of water into the slow reservoir, which accumulates drainage water. When the residence time scale of the slow reservoir is multiplied by 2 (i.e., the water from the slow reservoir is poured twice faster into the streams, Figure 12), the simulated lag of δ 18O river at Vienna increases from 4 to 5 months (in closer agreement with the data). In contrast, the seasonal cycle in δ 18O river is not sensitive to residence time scales in the and fast reservoirs, which are too short to have any impact at the seasonal scale.

To summarize, ORCHIDEE performs well in simulating the seasonal variations in δ 18O river. In turn, δ 18O river observations could help estimate the proportion of surface runoff versus drainage and calibrate empirical residence time constants in the routing scheme, offering a mean to enhance model performance. Evapo-transpiration partitioning In this section, we generalize at the global scale our results on evapo-transpiration partitioning estimates. We apply equation 1 to annual-mean outputs from a LMDZORCHIDEE. We compare E/I estimated from Equation 1 to E/I directly simulated by LMDZ-ORCHIDEE. The spatial pattern of E/I is remarkably well estimated by Equation 1 ( Figure 14). The equation captures the maximum over the Sahara, Southern South America, Australia, central Asia, Siberia and Northern America.

The isotope-derived spatial distribution of E/I correlates well with the simulated distribution (r=0.91). Average errors are lower than 50% of the standard deviation at the global scale. This confirms that covariation between the different variables at sub-annual time scales has a negligible effect, so that the equation can be applied to annual-mean quantities. Generally, E/I estimates are best where E/I is relatively small. Figure 14: a) annual mean E/I (proportion of infiltrating water recycled back to the atmosphere as bare soil evaporation) simulated by LMDZ-ORCHIDEE for the control simulation. B) E/I estimated from water isotopes measurements. We perform the estimations only on grid points where the denominators in the equation are different from 0 and where the soil water contents and the water fluxes whose compositions we need are strictly positive.

Grid points where estimations cannot be performed are left white. To test the effect of the assumption that the soil water isotopic composition is vertically constant, we applied Equation 1 using δ 18O s - δ 18O p from a simulation with soil profiles activated. This assumption is a significant source of uncertainty on estimating E/I ( Table 4). We also analyzed the effect of potential measurement errors in δ 18O s, δ 18O p, δ 18O v temperature or relative humidity on the E/I reconstruction. Results are relatively insensitive to small errors in these measurements ( Table 4). However, results are sensitive to the choice of the n exponent in the calculation of the kinetic fractionation α k ( Table 4): knowing the n exponent with an accuracy of 0.07 (e.g., estimated n ranges from 0.63 to 0.70) is necessary to estimate E/I with an absolute precision of 2%. Absolute or relative error RMS absolute error on r E/I RMS relative error on r E/I, when r E/I>4% (37% of total land aread) soil profiles 12% 50% Δ T =1°C 0.2% 1% Δ rh =1% 0.5% 1% Δδ p=1 3% 35% Δδ v=1 1% 8% Δδ s=1 5% 49% Δn = 0.5 14% 52% Table 4: Uncertainties in the estimation of E/I related to measurement errors and assumptions necessary in the simple conceptual model.

Values give absolute (in ratio) and relative variations (in%) in estimated E/I when temperature T is modified by 1 ° C (line 4), when relative humidity rh is modified by 1% (line 5), when δ 18O v, δ 18O p and δ 18O s are modified by 1, when n in the kinetic fractionation is varied from 0.5 to 1, and when the soil δ 18O is not homogeneous vertically. The resulting variations in estimated E/I are averaged over all land grid points where the estimation could be performed. Finally, estimating E/I using equation 1 bears additional sources of uncertainty in that we cannot estimate using the ORCHIDEE model. These are related to all processes that ORCHIDEE does not simulate. For example, ORCHIDEE underestimates or mis-represents the vertical isotopic gradients in soil water at some sites and does not represent the effect of water vapor diffusion in the soil.

These effects may disturb the proportionality between E/I and δ 18O s - δ 18O p in practical applications. To summarize, co-located isotope measurements in precipitation, vapor and soil water could provide an accurate constrain on the proportion of bare soil evaporation to precipitation infiltration. Conclusion and Perspectives The ORCHIDEE LSM, in which we have implemented water stable isotopes, reproduces the isotopic compositions of the different water pools of the land surface reasonably well compared to local data from MIBA and Carbo-Europe and to global observations from the GNIP and GNIR networks. Despite the scale mismatch between local measurements and a GCM grid box, and despite the strong spatial heterogeneity in the land surface, the capacity of ORCHIDEE to reproduce the seasonal and vertical variations in the soil isotope composition suggests that even local measurements can yield relevant information to evaluate LSMs at the large scale. We show that the simulated isotope soil profiles are sensitive to infiltration pathways and diffusion rates in the soil. The spatial and seasonal distribution of the isotope composition of rivers is sensitive to the partitioning of total runoff into surface runoff and drainage and to the residence time scales in underground reservoirs. The isotopic composition of soil water is strongly tied to the fraction of infiltrated water that evaporates through the bare soil.

These sensitivity tests suggest that isotope measurements, combined with more conventional measurements, could help evaluate the parameterization of infiltration processes, runoff parameterizations and the representation of surface water budgets in LSMs. Evaluating an isotopic LSM requires co-located observations of the isotope composition in precipitation, vapor and soil at least at the monthly scale. However, such co-located measurements are still very scarce, and most MIBA and Carbo-Europe sites are missing one of the components. Therefore, for LSM evaluation purpose, we advocate for the development of co-located isotope measurements in the different water pools at each site, together with variables. Our results suggest that isotope measurements are spatially relatively well representative and that even monthly values are already valuable to identify model bias or to estimate soil water budgets. Therefore, in the perspective of LSM evaluation, if a compromise should be made with sampling frequency and spatial coverage, we favor co-located measurements of all the different water pools at the monthly scale on a few sites representative of different climatic conditions, rather than multiplying sites where water pools are not all sampled.

Additionally, at each observation site, collecting different soil samples a few meters apart is helpful to check that they are spatial representative. In the future, development in laser technology [,] will allow the generalization of water vapor isotope monitoring at the different sampling sites, which has long been a very tedious activity []. From the modeling point of view, kinetic fractionation processes during bare soil evaporation are a source of uncertainty, and a better understanding and quantification of this fractionation is necessary [,]. In addition, the accuracy of isotopic simulations by LSM is expected to improve as the representation of hydrological processes improves. In particular, given the importance of vertical water exchanges for the isotopic simulation, implementing water isotopes in a multi-layer hydrological parameterization with sufficient vertical resolution [] is crucial. In the future, we plan to implement water isotopes in the latest version of ORCHIDEE, which is multi-layer and more sophisticated [-].

Finally, latest findings largely based on water isotopic measurements suggest that different water pools coexist within a soil column and that evaporation, transpiration, runoff and drainage tap from these different pools [,,]. These effects are not yet represented explicitly in global LSMs.

These effects were mainly evidenced based on isotope measurements, and in turn, their representation expected to significantly impact isotopic simulations. Such feedbacks between isotopic research and hydrological parameterization improvements should lead to LSM improvements in the future. With this in mind, LSM inter-comparison projects would strongly benefit from including water isotopes as part of their diagnostics, in the lines of iPILSP (isotope counterpart of the Project for Intercomparison of Land-surface Parameterization Schemes []). Representation of isotope fractionation during evaporation from land surface water pools Processes for which we neglect fractionation: Snow sublimation is associated with a slight fractionation due to exchanges between snow and vapor in snow pores [,,]. However, we assume that these effects are small enough to be neglected, as in other GCMs [].

Water uptake by roots has been shown to be a non-fractionating process [,], but fractionation at the leaf surface during transpiration impacts the composition of transpired fluxes at scales shorter than daily [,]. As the application of ORCHIDEE in the context of our study focuses mainly on time scales of a month or longer, we assume here that the transpiration and stem water have the composition of soil water extracted by the roots. Evaporation from bare soils and canopy-intercepted water: We represent isotope fractionation during evaporation of soil and canopyintercepted water using the model of Craig []: at any time t, the isotopic composition of evaporation R E is given by: (2) where R l and R v are the isotopic compositions of liquid water at the evaporative site and of water vapor respectively, h is the relative humidity normalized to surface temperature, α eq is the isotopic fractionation during liquid-vapor equilibrium [] and α k is the kinetic fractionation during water vapor diffusion. The kinetic fractionation during soil evaporation is still very uncertain [,]. We use the very widespread formulation of [,]: (3) where D and D i are the molecular diffusivities of light and heavy water vapor in air, respectively, and n is an exponent that depends on the flow regime (0.5, 0.67 and 1 for turbulent, laminar and stagnant regimes respectively) but remains difficult to estimate [,]. In this study, we take n =0.67 for both evaporation of soil and canopyintercepted water, corresponding to moist conditions in the case of soils [].

However, we also tried 0.5 and 1.0 to estimate the range of uncertainty related to this parameter. The isotopic composition of precipitation is only slightly sensitive to the formulation of the kinetic fractionation: when n varies from 0.5 to 1, significant changes in δ 18O p and d p are restricted to areas where bare soil covers more than 70%. Even in those case, changes in δ 18O p and d p never exceed 2‰ and 7‰ respectively. The impact is slightly stronger on soils. Varying n from 0.5 to 1 leads to δ 18O s variations of 2‰ in offline simulations on the Bray site, of the order of the observed average difference between two samples collected on the same day (2.2‰). In coupled simulations, the impact on δ 18O s and ds reaches 8‰ and 20‰ respectively on very arid regions such as the Sahara. To calculate the temporal mean isotopic composition of evaporation over the time step Δt,, we assume R v and h are constant throughout each time step.

On the other hand, we allow the isotopic ratio of liquid water to vary over the simulation time step Δt following []. While assuming constant R l is a valid assumption for models with very short time steps [], it is not the case in ORCHIDEE (Δt =30min). We then calculate as: (4) where R l0 is the initial isotopic ratio of liquid water, f is the remaining liquid fraction in the water reservoir affected by isotopic enrichment, and β and γ are parameters defined by Stewart []: and For canopy-intercepted water, the water reservoir is sufficiently small to assume that the water reservoir affected by isotopic enrichment is the total canopy-intercepted water. For soil evaporation on the other hand, we assume that the depth of the water reservoir affected by isotopic enrichment equals the average distance traveled by water molecules in the soil: (5) where K D is the effective self-diffusivity of liquid water in the soil column.

Neglecting the dispersion term, K D is given by Munnich et al. [,,-]: (6) where is the molecular liquid water selfdiffusivity [,], τ is the soil tortuosity and θ l is the volumetric soil water content. In the control simulation, we assume θ l.τ =0.1 leading to L =0.67 mm. This choice is consistent with a τ of 0.67 [] and an average θ l of about 15%. At the Bray, measurements along profiles show θ l varying from about 5 to 30%.

Since these values are difficult to constrain observationally and very variable spatially and temporally, sensitivity tests to θ l.τ are performed and described. We neglect the vapor phase in the soil and associated fractionation and diffusion processes []. Dew formation: We assume fractionation during dew and frost formation following a Rayleigh distillation of the vapor in the lowest 10 hPa ( ~ 80 m) of the atmosphere. Since the atmospheric water vapor condenses in small proportion during frost and dew, this choice of the depth of atmosphere involved in the condensation has almost no impact on the composition of the dew and frost formed. Following common practice, we use equilibrium fractionation coefficient from Merlivat et al. [,,] and the kinetic fractionation formation of [] with λ =0.004, whose choice has very little impact on the results.

Leaf water evaporation: At isotopic steady state, the composition of water transpired by the vegetation is equal to that of the soil water extracted by the roots. In default simulations, we assume that isotopic steady state for plant water is established at any time and we diagnose the composition of the leaf water at the evaporation site, SS e R, by inverting the Craig and Gordon equation []: (7) where R s and R v are the isotopic ratio in soil water and water vapor respectively, h is the relative humidity normalized to surface temperature, α eq is the isotopic fractionation during liquid-vapor equilibrium [] and α k is the kinetic fractionation during water vapor diffusion. We take the same kinetic fractionation formulation as for the soil evaporation [], with n =0.67 [,].

Leaf water compositions are significantly sensitive to parameter n, with variations of the order of 10‰ as n varies from 0.5 to 1. We assume that the leaf temperature used to calculate α eq is equal to the soil temperature, but results are very little sensitive to this assumption. The isotopic composition of leaf water has been the subject of many observational and numerical modeling studies [,-].

Several studies have shown that the composition of the leaves is affected by mixing with xylem water and by non-stationary effects [,]. Nonsteady state effects are also incorporated in ORCHIDEE following [].

The isotopic ratio in the leaf mesophyll R L SS is the result of the mixing between leaf water at the evaporative site and xylem water (Peclet effect): (8) where f is a coefficient decreasing as the Peclet effect increases: and p is the Peclet parameter [,]: E is the transpiration rate per leaf area, L eff is the effective diffusion length and W is the leaf water content per leaf volume (assumed equal to 10 3 kg/m 3, order of magnitude in []). The Peclet number P can be tuned by changing L eff, that depends on leaf geometry and drought intensity (e.g., 7 to 12 mm in Cuntz et al. [], 50 to 150 mm in Barnard et al. We take L eff =8 mm to optimize our simulation on Hartheim. For some simulations, we account for the effect of water storage in leaves (leading to some memory in the leaf water isotopic composition) following Dongmann []. Assuming that W is constant, we calculate the leaf lamina composition R L as Farquhar []: (9) where and g is the sum of the total (stomatic and boundary layer) conductances.

The isotopic composition of transpiration is then calculated so as to conserve isotope mass. Representation of the vertical distribution of soil water isotopic composition Principle: In control simulations, we assume that the isotopic composition of soil water is homogeneous vertically and equals the weighted average of the two soil layers. In addition, to test this assumption, we implemented a representation of the vertical distribution of the soil water isotopic composition: the soil water is spread vertically between several layers. The first layer contains a water height, where K D is the diffusivity of water molecules in water and Δt is the time step of the simulation, and the other layers contain a water height resol.L. The parameter resol can be tuned to find a compromise between vertical resolution and computational time. Layers are created from the top to bottom until all layers are full with water except the deepest one that contains the remaining soil water.

For example, with L =0.67 mm, up to 16 layers can thus be created if the soil is saturated. Bare soil evaporation is extracted from the first layer. Transpiration is extracted from the different layers following a root extraction profile that reflects the sensitivity of transpiration to soil moisture []. Drainage takes water from the deepest layer. In the control simulation, rain and snow melt are added to the first layer (piston-like flow). In a sensitivity test, that can also be homogeneously distributed in the different layers, to crudely represent preferential pathways through fractures or pores in the soil.

At each time step, the soil water isotopic composition in each layer is re-calculated by taking into account the sources and sinks for each layer and ensuring that each layer remains full except the deepest one. Isotopic diffusion between adjacent layers is applied at each time step (Equation 6). The water budget of the total soil remains exactly the same as without vertical discretization. Evaluation for an idealized case: The module representing vertical distribution of water isotopes in the soil is first evaluated for an idealized case when it is not yet embedded into ORCHIDEE. First, we use a case in which the soil column evaporates at its top and is permanently refilled at the bottom by a water with δ 18O of -8‰ [].

The soil remains saturated, and we focus on the steady state reached after a few hundreds of days []. An analytical solution is available for this case [,]. The analytical solution and a much more sophisticated model of soil water isotopes (MuSICA []) yield very similar results ( Figure 15a): the bottom of the soil is at -8‰ while the top of the soil is enriched up to 15‰. The soil module of ORCHIDEE is able to reproduce these results when the value of θ l.τ is set to be very low (0.001) and when the vertical resolution is sufficiently high (layers of 0.75 mm). Whatever the value for θ l.τ, ORCHIDEE results become less sensitive to the vertical discretization when layers are thinner than about 2 mm.