• Abramopoulos, F., C. Rosenzweig, and B. Choudhury, 1988: Improved ground hydrology calculations for global climate models (GCMs): Soil water movement and evapotranspiration. J. Climate,1, 921–941.

  • Avissar, R., and R. Pielke, 1989: A parametrization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Mon. Wea. Rev.,117, 2113–2136.

  • Baumgartner, A., and E. Reichel, 1975: The World Water Balance. Elsevier, 179 pp.

  • Brubaker, K. L., D. Entekhabi, and P. S. Eagleson, 1993: Estimation of continental precipitation recycling. J. Climate,6, 1077–1089.

  • Budyko, M. I., 1978: The heat balance of the Earth. Climatic Change, J. Gribbin, Ed., Cambridge University Press, 85–113.

  • Choisnel, E. M., S. V. Jourdain, and C. J. Jacquart, 1995: Climatological evaluation of some fluxes of the surface energy and soil water balances over France. Ann. Geophys.,13, 666–674.

  • De Rosnay, P., and J. Polcher, 1998: Modeling root water uptake in a complex land surface scheme coupled to a GCM. Hydrol. Earth Syst. Sci.,2, 239–255.

  • Delworth, T. L., and S. Manabe, 1988: The influence of potential evaporation on the variabilities of simulated soil wetness and climate. J. Climate,1, 523–547.

  • ——, M. Bruen, and J. Polcher, 1998: Sensitivity of the surface fluxes to the number of layers in the soil model used in GCMs. Geophys. Res. Lett., in press.

  • Dickinson, R. E., 1984: Modeling evaporation for three-dimensional global climate models. Climate Processes and Climate Sensitivity, J. E. Hansen and T. Takahashi, Eds., Amer. Geophys. Union, 58–72.

  • ——, and A. Henderson-Sellers, 1988: Modelling tropical deforestation: A study of GCM land-surface parameterizations. Quart. J. Roy. Meteor. Soc.,114, 439–462.

  • Dirmeyer, P. A., 1997: The Global Soil Wetness Project: Initial results. GEWEX News,7, 3–6.

  • ——, and J. Shukla, 1993: Observational and modeling studies of the influence of soil moisture anomalies on atmospheric circulation (Review). Predictions of Interannual Climate Variations, J. Shukla, Ed., NATO ASI Series, Vol. 16, Springer-Verlag, 1–23.

  • ——, F. Zeng, A. Ducharne, J. Morrill, and R. Koster, 2000: The sensitivity of surface fluxes to soil water content in three land surface schemes. J. Hydrometeor.,1, 121–134.

  • Dolman, A., 1997: Workshop on Soil Vegetation Atmosphere Transfer (SVAT) Schemes/Land Surface Parameterizations (LSPs). GEWEX News,7, 13–14.

  • Dorman, J., and P. Sellers, 1989: A global climatology of albedo, roughness length and stomatal resistance for atmospheric general circulation models as represented by the Simple Biosphere Model (SIB). J. Appl. Meteor.,28, 833–855.

  • Ducharne, A., K. Laval, and J. Polcher, 1998: Sensitivity of the hydrological cycle to the parametrization of soil hydrology in a GCM. Climate Dyn.,14, 307–327.

  • ——, R. D. Koster, M. Suarez, and P. Kumar, 1999: A catchment-based land surface model for GCMS and the framework for its evaluation. Phys. Chem. Earth,24B, 769–773.

  • Ducoudré, N., K. Laval, and A. Perrier, 1993: SECHIBA, a new set of parametrizations of the hydrologic exchanges at the land–atmosphere interface within the LMD atmospheric general circulation model. J. Climate,6, 248–273.

  • Dunne, K. A., and C. J. Willmott, 1996: Global distribution of plant-extractable water capacity of soil. Int. J. Climatol.,16, 841–859.

  • Entekhabi, D., and P. Eagleson, 1989: Land surface hydrology parametrization for atmospheric general circulation models including subgrid scale spatial variability. J. Climate,2, 816–831.

  • ——, I. Rodriguez-Iturbe, and F. Castelli, 1996: Mutual interaction of soil moisture and atmospheric processes. J. Hydrol.,184, 3–17.

  • Fouquart, Y., and B. Bonnel, 1980: Computations of solar heating of the earth’s atmosphere: A new parametrisation. Beitr. Phys. Atmos.,53, 35–62.

  • Gadgil, S., and S. Sajani, 1998: Monsoon precipitation in the AMIP runs. Climate Dyn.,14, 659–689.

  • Garratt, J., 1993: Sensitivity of climate simulations to the land-surface and atmospheric boundary-layer treatments—A review. J. Climate,6, 419–449.

  • Gates, W. L., 1992: AMIP: The Atmospheric Model Intercomparison Project. Bull. Amer. Meteor. Soc.,73, 1962–1970.

  • Henning, D., 1989: Atlas of the Surface Heat Balance of the Continents. Gebrüder Borntraeger, 402 pp.

  • Hulme, M., 1992: A 1951–80 global land precipitation climatology for the evaluation of general circulation models. Climate Dyn.,7, 57–72.

  • Hutjes, R., and Coauthors, 1998: Biospheric aspects of the hydrological cycle—Preface. J. Hydrol.,213, 1–21.

  • Koster, R. D., and M. J. Suarez, 1992: Modeling the land surface boundary in climate models as a composite of independent vegetation stands. J. Geophys. Res.,97 (D3), 2697–2715.

  • ——, and P. C. D. Milly, 1997: The interplay between transpiration and runoff formulations in land surface schemes used with atmospheric models. J. Climate,10, 1578–1591.

  • Kuo, H., 1965: On formation and intensification of tropical cyclones through latent heat release by cumulus convection. J. Atmos. Sci.,22, 40–63.

  • Laval, K., 1988: Experience with surface processes at LMD. Proc. Parametrization of Fluxes over Land Surface, Reading, United Kingdom, European Centre for Medium-Range Weather Forecasts, 171–205.

  • ——, H. Le Treut, and R. Sadourny, 1981: Effects of cumulus parameterization on the dynamics of a general circulation model. Geophys. Astrophys. Fluid Dyn.,17, 113–127.

  • Legates, D. R., and C. J. Willmott, 1990: Mean seasonal and spatial variability in gauge-corrected, global precipitation. Int. J. Climatol.,10, 111–127.

  • Le Treut, H., and Z.-X. Li, 1991: Sensitivity of an atmospheric general circulation model to prescribed SST changes: Feedback effects associated with the simulation of cloud optical properties. Climate Dyn.,5, 175–187.

  • Liang, X., D. P. Lettenmaier, E. F. Wood, and S. J. Burges, 1994: A simple hydrologically based model of land surface water and energy fluxes for general circulation models. J. Geophys. Res.,99 (D7), 14 415–14 428.

  • Mahfouf, J.-F., and Coauthors, 1996: Analysis of results from the PILPS-RICE Workshop. Part 3: Transpiration. Global Planet. Change,13, 73–88.

  • Manabe, S., 1969: Climate and the ocean circulation 1. The atmospheric circulation and the hydrology of the earth’s surface. Mon. Wea. Rev.,97, 739–774.

  • ——, and R. Strickler, 1964: Thermal equilibrium of the atmosphere with a convective adjustment. J. Atmos. Sci.,21, 361–385.

  • Milly, P. C. D., 1994: Climate, soil water storage, and the average annual water balance. Water Resour. Res.,30, 2143–2156.

  • ——, and K. A. Dunne, 1994: Sensitivity of the global water cycle to the water-holding capacity of land. J. Climate,7, 506–526.

  • Mintz, Y., and Y. Serafini, 1992: A global monthly climatology of soil moisture and water balance. Climate Dyn.,8, 13–27.

  • Morcrette, J., 1991: Radiation and cloud radiative properties in the European Centre for Medium range Weather Forecasts forecasting system. J. Geophys. Res.,96, 9121–9132.

  • Patterson, K. A., 1990: Global distributions of total and total-available soil water holding capacities. M.S. thesis, Dept. of Geography, University of Delaware, 119 pp. [Available from Morris Library, University of Delaware, Newark, DE 19716.].

  • Polcher, J., 1995: Sensitivity of tropical convection to land surface processes. J. Atmos. Sci.,52, 3143–3161.

  • ——, and K. Laval, 1994: A statistical study of regional impact of deforestation on climate of the LMD-GCM. Climate Dyn.,10, 205–219.

  • Sadourny, R., and K. Laval, 1984: January and July performance of the LMD general circulation model. New Perspectives in Climate Modelling, A. L. Berger and C. Nicolis, Eds., Elsevier Science Publishers, 173–197.

  • Sellers, P., Y. Mintz, Y. Sud, and A. Dachler, 1986: A simple biopshere model (SiB) for use within general circulation models. J. Atmos. Sci.,43, 505–531.

  • Shukla, J., and Y. Mintz, 1982: Influence of land-surface evaporation on the earth’s climate. Science,215, 1498–1501.

  • Simmonds, I., and P. Hope, 1998: Seasonal and regional responses to changes in Australian soil moisture conditions. Int. J. Climatol.,10, 1105–1139.

  • Stamm, J. F., E. F. Wood, and D. P. Lettenmaier, 1994: Sensitivity of a GCM simulation of global climate to the representation of land-surface hydrology. J. Climate,7, 1218–1239.

  • Walker, J., and P. Rowntree, 1977: The effect of soil moisture on circulation and rainfall in a tropical model. Quart. J. Roy. Meteor. Soc.,103, 29–46.

  • Wallis, J. R., D. P. Lettenmaier, and E. F. Wood, 1991: A daily hydroclimatological data set for the continental United States. Water Resour. Res.,27, 1657–1663.

  • Wilson, M. F., and A. Henderson-Sellers, 1985: A global archive of land cover and soil data for use in general circulation climate models. J. Climatol.5, 119–143.

  • Zeng, X., Y.-J. Dai, R. Dickinson, and M. Shaikh, 1998: The role of root distribution for climate simulation over land. Geophys. Res. Lett.,25, 4533–4536.

  • Zobler, L., 1986: A world soil file for global climate modeling. National Aeronautics and Space Administration Tech. Memo. 87802, 32 pp.

  • View in gallery
    Fig. 1.

    Schematic diagram of the evolution of the two soil layers in SECHIBA

  • View in gallery
    Fig. 2.

    Reference distribution of WHC in SECHIBA: the WHC is 150 kg m−3 over dark regions, and 30 kg m−3 over the light-colored areas

  • View in gallery
    Fig. 3.

    Differences between the new WHC datasets and the reference WHC dataset (mm): (a) A-WHC–REF-WHC and (b) T-WHC)–REF-WHC. White areas over land indicate negative differences, with dashed contours at −50, − 100, and −200 mm. Positive differences appear in gray, with increasing shade between the solid contours at 0, 50, 100, 200 mm, and over

  • View in gallery
    Fig. 4.

    Evolution of global annual mean soil moisture (mm) in simulations REF, AWC, and TWC

  • View in gallery
    Fig. 5.

    July precipitation rates (mm day−1): (a) in REF (1980–88) and (b) in the climatology of Legates and Willmott (1990). Contours at 1, 3, 5, and 10 mm day−1

  • View in gallery
    Fig. 6.

    Annual mean differences in soil moisture (mm) between AWC and REF (AWC–REF). Contours at −100, −50, −10, 10, 50, and 100 mm

  • View in gallery
    Fig. 7.

    Zonal annual means over land, in REF and TWC, of soil moisture (mm), precipitation rate, evaporation rate, and moisture convergence (mm day−1)

  • View in gallery
    Fig. 8.

    Differences (TWC − REF) in Jul mean (a) evaporation and (b) precipitation (mm day−1), in the northern extratropics. Contours at −2, −1, −0.5, 0.5, 1, and 2 mm day−1

  • View in gallery
    Fig. 9.

    Tropical band: (a) annual mean precipitation in REF (mm day−1), and differences (TWC − REF) in annual mean (b) soil moisture (mm), (c) evaporation, (d) moisture convergence, and (e) precipitation (mm day−1). Contours at 1, 3, 5, and 10 mm day−1 in (a), −100, −50, −10, 10, 50, and 100 mm in (b), and −2, −1, −0.5, 0.5, 1, and 2 mm day−1 in (c), (d), and (e)

  • View in gallery
    Fig. 10.

    Significance of the annual mean differences in (a) evaporation and (b) precipitation, between TWC and REF. The dark areas correspond to statistically significant differences at the level α = 0.05; the spatial distribution of these differences is shown by the contours (−2, −1, −0.5, 0.5, 1, and 2, in mm day−1), with dotted lines for negative differences

  • View in gallery
    Fig. 11.

    Location of the seven areas selected to assess the statistical significance of the ITCZ changes

  • View in gallery
    Fig. 12.

    Sea level pressure: (a) annual mean distribution in REF, in hPa − 1000, contours at 7.5, 10, 15, and 20 hPa − 1000, (b) annual mean differences between TWC and REF: the contours (at −3, −2, −1, 1, 2, and 3 hPa − 1000, with dotted lines for negative values) correspond to TWC–REF, and the dark areas indicate the statistically significant differences at the level α = 0.05

  • View in gallery
    Fig. 13.

    Annual zonal means over land: differences between TWC and REF (TWC − REF) and AWC and REF (AWC − REF), for (a) soil moisture (mm), (b) evaporation, (c) precipitation, and (d) moisture convergence (mm day−1)

  • View in gallery
    Fig. 14.

    Mean seasonal cycles of soil moisture (mm), evaporation, and precipitation rates (mm day−1), simulated in REF, AWC, and TWC, in two tropical and northern midlatitude regions. The symbols on the AWC (TWC) curves indicate the statistical significance (at the level α = 0.05) of the difference between the monthly means: • indicates significant differences with both REF and TWC (AWC); * indicates a significant difference with REF only; and □ indicates a significant difference with TWC (AWC) only

  • View in gallery
    Fig. 15.

    Comparison of simulated (REF, AWC, and TWC) and observed (Wallis et al. 1991) seasonal cycles of (a) precipitation and (b) total runoff over the Mississippi basin (mm day−1)

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 305 84 26
PDF Downloads 102 36 5

Influence of the Realistic Description of Soil Water-Holding Capacity on the Global Water Cycle in a GCM

Agnès DucharneLaboratoire de Météorologie Dynamique du CNRS, Paris, France

Search for other papers by Agnès Ducharne in
Current site
Google Scholar
PubMed
Close
and
Katia LavalLaboratoire de Météorologie Dynamique du CNRS, Paris, France

Search for other papers by Katia Laval in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The sensitivity of the hydrological cycle to soil water-holding capacity (WHC) is investigated using the Laboratoire de Meteorologie Dynamique General Circulation Model (LMD GCM) coupled to a land surface model (LSM). A reference simulation (REF), with WHCs equal to 150 mm globally (except in deserts where it is set to 30 mm), is compared to two perturbation simulations using datasets with realistic WHC distributions:the “available WHC” (AWC) dataset is physically consistent with the definition of WHC in the LSM and has a global average close to 150 mm; the “total WHC” (TWC) dataset is used as a secondary reference for a large WHC increase (more than a doubling from 150 mm). The average impact over land of the increase in WHC (from REF to both AWC and TWC) is an increase in annual mean evaporation, split between increased annual precipitation and decreased annual mean moisture convergence. The regional responses, however, are more complex: precipitation increases in summer over the midlatitude landmasses through the recycling of increased evaporation; in the Tropics, moisture convergence and precipitation decrease in the intertropical convergence zone and precipitation increases in the surrounding areas, both behaviors being related to the sensitivity of tropical convection to surface energy fluxes in the LMD GCM.

Two important conclusions arise from these numerical results: first, the changes in the hydrological cycle are driven through evaporation by the WHC changes realized in the hydrologically active regions (continental midlatitude and tropical rainbelts); second, WHC increase of 10% to 20% in the rainbelts induces changes in the hydrologic cycle with similar patterns and almost the same amplitude as changes resulting from an increase greater than 100%. These results are strongly conditioned to the land–atmosphere feedbacks, which can only be allowed in a GCM environment.

Corresponding author address: Agnès Ducharne, UMR Sisyphe, UPMC, Boite 123, 4 Place Jussieu, 75252 Paris Cedex 05, France.

Email: ducharne@biogeodis.jussieu.fr

Abstract

The sensitivity of the hydrological cycle to soil water-holding capacity (WHC) is investigated using the Laboratoire de Meteorologie Dynamique General Circulation Model (LMD GCM) coupled to a land surface model (LSM). A reference simulation (REF), with WHCs equal to 150 mm globally (except in deserts where it is set to 30 mm), is compared to two perturbation simulations using datasets with realistic WHC distributions:the “available WHC” (AWC) dataset is physically consistent with the definition of WHC in the LSM and has a global average close to 150 mm; the “total WHC” (TWC) dataset is used as a secondary reference for a large WHC increase (more than a doubling from 150 mm). The average impact over land of the increase in WHC (from REF to both AWC and TWC) is an increase in annual mean evaporation, split between increased annual precipitation and decreased annual mean moisture convergence. The regional responses, however, are more complex: precipitation increases in summer over the midlatitude landmasses through the recycling of increased evaporation; in the Tropics, moisture convergence and precipitation decrease in the intertropical convergence zone and precipitation increases in the surrounding areas, both behaviors being related to the sensitivity of tropical convection to surface energy fluxes in the LMD GCM.

Two important conclusions arise from these numerical results: first, the changes in the hydrological cycle are driven through evaporation by the WHC changes realized in the hydrologically active regions (continental midlatitude and tropical rainbelts); second, WHC increase of 10% to 20% in the rainbelts induces changes in the hydrologic cycle with similar patterns and almost the same amplitude as changes resulting from an increase greater than 100%. These results are strongly conditioned to the land–atmosphere feedbacks, which can only be allowed in a GCM environment.

Corresponding author address: Agnès Ducharne, UMR Sisyphe, UPMC, Boite 123, 4 Place Jussieu, 75252 Paris Cedex 05, France.

Email: ducharne@biogeodis.jussieu.fr

1. Introduction

From an atmospheric point of view, evaporation is the key variable of surface hydrology since it links the energy budgets of the atmosphere and the surface, and more generally the energy and water cycles. In this atmospheric context, the main role of the land surface component is the partitioning of net incoming radiative energy into sensible and latent heat fluxes, which is strongly controlled by the availability of water for evaporation and transpiration, generally in the form of soil moisture. Numerous numerical studies have therefore been performed to study the influence of soil moisture on land–atmosphere interactions. Most of them made use of general circulation models (GCMs), the only numerical environment in which the feedbacks between the land surface, the overlying atmosphere, and the general circulation can be accounted for.

Three kinds of GCM soil moisture studies can be distinguished, as summarized in several review papers (Laval 1988; Garratt 1993; Entekhabi et al. 1996). Anomaly studies, inspired by the work of Walker and Rowntree (1977), focus on the atmospheric effects of initially prescribed soil moisture anomalies and their ability to sustain themselves through feedback processes (e.g., Dirmeyer and Shukla 1993; Simmonds and Hope 1998). A different kind of study, first carried out by Delworth and Manabe (1988), addresses the role of soil moisture, identified as a low-frequency component, on the timescales of atmospheric variability. The third kind consists of sensitivity studies. They include the work by Shukla and Mintz (1982), who showed an enhancement of land evaporation and the atmospheric water cycle when going globally from a totally dry to a saturated soil. Also relevant are sensitivity studies to soil hydrology parameterizations that influence soil moisture, such as drainage, soil water-holding capacity (WHC), its subgrid variability, or the vertical distribution of roots (e.g., Milly and Dunne 1994; Ducharne et al. 1998; Zeng et al. 1998).

These soil hydrology parameterizations have evolved along with the vegetation parameterizations in land surface models (LSMs) for GCMs, in response to studies showing the importance of land surface processes for the simulated climate. The first interactive LSM was the “bucket” model (Manabe 1969), representing the land surface as a one-layer soil reservoir, its water content controlling the evaporation rate. Systematic errors of this model in the short-term dynamics of evaporation have been explained by the absence of vegetation control over evaporation and by the lack of vertical discretization of the soil (Dickinson and Henderson-Sellers 1988; Mahfouf et al. 1996). This prompted the development of the Soil–Vegetation–Atmosphere Transfer (SVAT) models, which explicitly represent interception by the canopy and regulation of transpiration through a stomatal resistance. The latter depends on environmental stresses, including soil moisture, which is generally distributed among two to three connected soil layers (Dickinson 1984; Sellers et al. 1986; Abramopoulos et al. 1988). Later developments have focused on the subgrid variability of land surface processes (Entekhabi and Eagleson 1989; Liang et al. 1994; Ducharne et al. 1999), as well as on the vertical resolution, with an increase of the number of superposed soil layers (De Rosnay et al. 1998; J. Morrill et al. 1999, personal communication).

Whatever the complexity of soil hydrology in an LSM, total soil moisture remains an important prognostic variable, since it rules the length of time evaporation can be sustained in case of water input shortage. In this simplified context, the soil water-holding capacity, which defines the maximum soil moisture, is a critical LSM parameter because it controls the partitioning of evaporation and runoff, and the amount of water that can be stored during wet periods to be released in dry periods. The sensitivity of a GCM-simulated climate and water cycle to WHC was studied in particular by Milly and Dunne (1994). Their sensitivity study used the Geophysical Fluid Dynamics Laboratory (GFDL) GCM coupled to the bucket LSM, and consisted mainly of the comparison of two 10-yr integrations, with a globally constant WHC of 40 and 600 mm, respectively. The resulting increase in soil WHC induced a decrease in runoff and an increase in evaporation over land, which occurred mainly in the Tropics and northern midlatitude rainbelts. In both areas, the increase in continental evaporation was rather equally split between increased continental precipitation and decreased convergence of water vapor from ocean to land. The latter was linked to a weakening of the monsoon circulations. In the Tropics, however, the local moisture convergence was found to increase along the intertropical convergence zone (ITCZ). This was related to an intensification of the Hadley circulation, explained by the enhancement of moist convection in a moister atmosphere. Complementary experiments showed intermediate response to smaller increases of the WHC, and a saturation of the sensitivity for increases above 600 mm.

Stamm et al. (1994) also briefly examined the effect of soil WHC in the GFDL GCM, coupled this time to the Variable Infiltration Capacity (VIC) LSM, a SVAT model including a one-layer soil with a drainage term and a statistical treatment of the subgrid variability of soil WHC. They compared a 10-yr simulation with a globally constant WHC of 150 mm to a 10-yr simulation with a more realistic WHC, varying across the globe, with an average of 261 mm. Two 2-yr simulations with globally constant WHCs of 50 and 250 mm were also considered, although no statistical analysis could be performed on such short integrations. The authors found, “somewhat surprisingly,” that soil WHC had rather little influence on the simulated hydrological cycle, especially in the northern midlatitude land areas. The proposed explanation was their very effective drainage, limiting in all simulations the fraction of soil moisture that can be evaporated.

The above two studies, even though performed in a very similar GCM environment, exhibit very different sensitivities, which would be enough to motivate further investigation. Also, the scope of these studies presents some limitations that need to be assessed. Milly and Dunne (1994), in order to maximize the sensitivity of the hydrological cycle and allow better isolation of the relevant mechanisms, used unrealistic WHCs, both in value (40 and 600 mm) and distribution (globally constant). Stamm et al. (1994) considered much more realistic WHCs, but the main focus of their work was the validation of the VIC LSM. The sensitivity to WHC was examined as part of the validation process, and was not given strong attention. The influence of the WHC on the simulated climate might also have been altered by the simultaneous inclusion of a subgrid variability for this quantity.

The aim of this paper is to complement such studies, by investigating in detail the influence of realistic WHC distributions on soil moisture, the hydrological cycle, and climate. To this end, we devised a 10-yr sensitivity experiment in a GCM environment that accounts for the feedbacks between the land surface and the overlying atmosphere. The GCM and the coupled SVAT LSM are described in section 2. The design of the experiment, including a description of the realistic WHC distributions, is presented in section 3, and the hydrological cycle in the reference simulation is outlined in section 4. The results of the sensitivity experiment are then presented in section 5 and discussed in section 6, which also concludes the study.

2. Modeling environment

a. The atmospheric GCM

Our study used the atmospheric GCM of the Laboratoire de Météorologie Dynamique (LMD), the main features of which are described in Sadourny and Laval (1984) and Le Treut and Li (1991). This GCM is a finite difference, primitive equation model that uses a standard sigma coordinate in the vertical. In the horizontal, the grid points are distributed regularly in longitude and sine of latitude, defining grid cells of equal area across the globe. In the present study, we used 11 vertical levels and 64 × 50 grid points in the horizontal resulting in grid cells of approximately 160 000 km2 and a resolution in the Tropics of about 5.6° × 2.4°.

The radiation parameterization includes the solar radiation model proposed by Fouquart and Bonnel (1980) and the longwave radiation scheme of Morcrette (1991). The land surface snow-free albedo is prescribed monthly from the dataset collected by Dorman and Sellers (1989) and the ocean albedo is corrected in the presence of sea ice. Cloud cover and its radiative properties are defined according to the scheme developed by Le Treut and Li (1991), which is coupled to condensation parameterizations. These include a parameterization of large-scale condensation and two small-scale convection schemes:the moist convective adjustment technique from Manabe and Strickler (1964) and an adaptation (Laval et al. 1981) of the scheme proposed by Kuo (1965) for convection in unsaturated conditions. The turbulent fluxes in the planetary boundary layer (PBL) are computed with an implicit numerical scheme from the surface to the top of the PBL, where they vanish. The surface layer parameterization defines the drag coefficients as a function of the stability of the surface layer (Polcher and Laval 1994) and includes the land surface scheme Schématisation des Echanges Hydriques à l’Interface entre la Biosphère et l’Atmosphere (SECHIBA; Ducoudré et al. 1993), presented below.

b. Description of land surface hydrology

Land surface hydrology is calculated in the LMD GCM by the land surface scheme SECHIBA (Ducoudré et al. 1993). The soil moisture reservoir is characterized by its depth (1 m, representing a rooting depth) and a volumetric water holding capacity (WHC, in kg m−3) that is uniform across any given cell. The present sensitivity experiment addresses the influence of this WHC on the simulated climate, through the changes in its global distribution, which are described in section 3.

The evolution of soil moisture is computed with a two-layer soil model, based on Choisnel’s ideas (Ducoudré et al. 1993; Choisnel et al. 1995). The originality in this soil model is the variable depth of the upper soil layer, ranging from 0 to 10 cm. This layer is intimately related to the dynamics of rainfall and evaporation, and allows, in particular, evapotranspiration to occur at the potential rate after a storm. As shown in Fig. 1, the upper layer is created when it rains and the soil is composed of only one layer (usually after a dry spell). Additional rainfall can either fill this upper layer if it is not saturated, or deepen it otherwise, until the maximum depth (10 cm) is reached. Any water input beyond this point goes to the lower layer. As also shown in Fig. 1, evapotranspiration is extracted from the upper layer, when it exists, and from the lower one otherwise.

SECHIBA represents vegetation using the “mosaic” strategy (Koster and Suarez 1992; Avissar and Pielke 1989); the heterogeneous vegetation cover of a GCM grid cell is described by a set of homogeneous “tiles,” with each tile representing a different land surface type (bare soil or one of seven vegetation types). The total evaporation is computed as the weighted average of snow sublimation, bare soil evaporation, transpiration, and interception loss from each of the tiles in a grid cell. The general form for transpiration and bare soil evaporation is
i1520-0442-13-24-4393-e1
where ρ is air density, qsat(Tsurf) is the specific humidity of the surface at saturation, qair is near-surface air specific humidity, and ra is the aerodynamic resistance. The surface resistance rsurf is controlled by environmental stresses and parameterized as in Polcher and Laval (1994). The aridity coefficient β is defined as the maximum of the two aridity coefficients computed for the upper layer on one hand and the total soil column on the other hand. The aridity coefficient βi of a reservoir i is a function of its wetness θi (defined as the ratio of the water content to the WHC of the reservoir): βi = exp[−0.8(1 − θi)]. This coefficient is identical in all tiles in a grid cell since they share the same soil moisture.

Drainage, between the two soil layers and from the lower layer (base flow), is parameterized as in Ducharne et al. (1998). The drainage from a layer depends on its wetness θi in a nonlinear way; drainage is small until θi reaches 0.75, and it increases strongly beyond this threshold. Total runoff includes the base flow mentioned above and overflow runoff, simply defined as the excess water when the soil is saturated, like in the bucket model (Manabe 1969). One can notice that the drainage parameterization is the only difference between the present version of the LMD GCM and that used in Polcher and Laval (1994).

3. Design of the numerical experiment

a. Definitions

Two different WHCs are distinguished in this study. On one hand, the total WHC corresponds to the soil water capacity at saturation, when water fills the entire volume of pores. On the other hand, the available WHC is defined as the difference between the field capacity (in the absence of precipitation and water table, the water held between the total WHC and the field capacity is completely drained under the effect of gravity in two to three days) and the wilting point (the soil moisture below the wilting point cannot be extracted from the soil by transpiration or evaporation at normal conditions of temperature and pressure). The available WHC therefore corresponds to the maximum water amount available to sustain evapotranspiration in the long term (beyond three days); in SECHIBA, as in many LSMs for GCMs, the soil WHC is considered to be an available capacity.

The two main factors that determine all soil water retention characteristics (total WHC, field capacity, wilting point, and available WHC) are the soil structure and texture. Both of them depend on the pedogenesis, and therefore on the nature of bedrock, as well as on climate, topography, and vegetation. All these factors display a strong spatial variability at all scales, making it difficult to estimate representative values of total and available WHC for a GCM grid cell, as well as the global distribution of such estimates.

b. Reference WHC

In the first version of SECHIBA (Ducoudré et al. 1993), the WHC was equal to 150 kg m−3 in all continental grid cells. The first reason for such a simplification is historical; this value, in the range of typical values for available WHC, was used in the bucket model (Manabe 1969), which first attempted to parameterize land surface hydrology in a GCM. It has since been very commonly used in LSMs for GCMs. A detailed discussion on the estimation of available WHC and other soil water retention characteristics can be found in Mintz and Serafini (1992), who eventually proposed a very close value (146 kg m−3) to represent the average of available WHC for all textural types.

In our reference run, the WHC was kept to 150 kg m−3, except in desert areas, where the value was lowered to 30 kg m−3. The aim of this modification was to decrease unrealistically high evaporation rates, recycled into unrealistically high precipitation rates, and its success (although not quantified) is the first evidence of the impact of soil WHC on the hydrological cycle in the LMD GCM coupled to SECHIBA. The resulting WHC dataset, displayed in Fig. 2, will be referred to as the reference WHC dataset, or REF-WHC.

c. Realistic distributions of WHC

Another reason why SECHIBA and many other LSMs for GCMs did not use a realistic global distribution of WHC was, for a long time, the lack of good quality data, apart from what could be derived from soil texture datasets (Wilson and Henderson-Sellers 1985; Zobler 1986). Patterson (1990) and Dunne and Willmott (1996) have carefully derived global soil WHC datasets from the Food and Agriculture Organization/United Nations Educational, Scientific and Cultural Organization (1971–81) soil types and the corresponding representative soil profiles, as well as from rooting depth and optionally organic matter content. For the purpose of our sensitivity experiment, two global WHC datasets proposed by Patterson (1990) at the 0.5° × 0.5° resolution were selected and interpolated to the LMD GCM resolution to replace the reference dataset. The first dataset, A-WHC, gives the available WHC of the top first meter and is physically consistent with the definition of the WHC in SECHIBA. The second dataset, T-WHC, corresponds to a total WHC and was chosen to examine the impact of a larger change in WHC. Both datasets ignore the organic matter, and give vertically integrated WHCs (in kg m−2), over a constant 1-m soil depth (independent of the vegetation cover). These values are equivalent to the volumetric WHC values of the top first meter of soil and can then be used directly in SECHIBA. In the remainder of this paper, all WHCs and soil moistures will be expressed in mm, equivalent to kg m−2, since the soil depth is 1 m in SECHIBA.

The WHC average over land is 126.7 mm in REF-WHC, 139.1 mm in A-WHC, and 266.1 mm in T-WHC. Figure 3 shows the areas where WHC increases and decreases when the former is replaced by the two latter. It confirms that T-WHC has much higher WHC compared to REF-WHC. The only spots where T-WHC is smaller than REF-WHC are rather dry areas of tundra (subarctic areas and Tibetan Plateau) and they have a small extent. The areas where A-WHC is smaller than REF-WHC have a larger extent, and the increases in WHC from REF-WHC to A-WHC have a much smaller amplitude than from REF-WHC to T-WHC. These changes are the largest in the desert areas, where REF-WHC had been decreased to 30 mm. Overall, T-WHC can be considered as a large increase and A-WHC as a moderate (and less generalized) increase of WHC compared to REF-WHC.

d. Description of the simulations

Three 10-yr simulations were performed with the previously described version of the LMD GCM, using interannually varying sea surface temperatures (SST) from the 1979–88 Atmospheric Model Intercomparison Project dataset (Gates 1992). The only differences between the three simulations are related to the WHC distributions. REF is the reference simulation, using the reference WHC dataset REF-WHC. This WHC is replaced with the available WHC from A-WHC in AWC, and with the total WHC from T-WHC in TWC.

In order to decrease the number of years needed to overcome the effects of initial conditions (spin-up time), the initial soil moisture for AWC and TWC was modified from the initial soil moisture in REF. For both soil layers, the new moisture initial state is proportional to the one in REF; the proportionality factor in any land grid cell is the ratio of the new to the reference WHC. Over land, on average, the initial soil moisture was 75.7 mm in REF and was modified to 77.2 mm in AWC and 144.6 mm in TWC. There was no significant trend in the year to year variations of soil moisture during the last nine years of integration, as illustrated in Fig. 4, for the global average of soil moisture. Only the first year of integration was consequently excluded from the analysis period.

4. Overview of the reference hydrological cycle

Table 1 gives several estimates of the world water balance from different recent sources. There is a wide confidence margin on these estimates, because of insufficient spatial sampling of observed data, the biases in measurements, and the assumptions of the assimilation methods (Henning 1989). The table, however, shows a large overestimation of continental precipitation in REF. Total runoff is parameterized in SECHIBA as the residual of the continental water budget. It is therefore more sensitive to overestimation of continental precipitation than continental evaporation, which is rather realistically simulated in the annual average over land. A consequence is the inversion, between the climatological estimates and the reference simulation, of the relative positions of total runoff and evaporation. The marked overestimation of total runoff is also associated with unrealistically large continental moisture convergence (approximated by PcEc, as the interannual variations of soil moisture are ignored), related to excessive moisture divergence over the oceans by means of water balance (clear when one considers the volumetric fluxes, by accounting for the respective areas of the land and ocean masses, given in Table 2). Such an enhanced hydrological cycle is a systematic feature of the LMD GCM, but it is especially pronounced in the present version.

Figure 5 compares the average July precipitation field simulated in REF to the July precipitation climatology from Legates and Willmott (1990). In the Tropics, the simulated precipitation presents a realistic zonal structure, dynamically related to the Hadley–Walker circulation: a band with very high precipitation rates, the ITCZ, is surrounded in the subtropics by areas with very low precipitation. These areas, however, do not extend far enough westward over the oceans, and the precipitation rates related to the ITCZ are too high over land, especially in Africa. These features result from the previously mentioned overestimation of moisture convergence over land and moisture divergence over the oceans. The main flaw of the simulated tropical precipitation in July, however, is found over southeastern Asia and the western Pacific, where an unrealistic zonal structure is found between 80° and 115°E: two high-precipitation bands at 30° and 5°N surround a relatively low-precipitation region around 15°N, where the climatological precipitation has, in contrast, its local maximum. This indicates that the Indian summer monsoon circulation is not well simulated. This is the case in many GCMs (Gadgil and Sajani 1998) but is especially marked in the present version of the LMD GCM, where the convective systems associated with this circulation not only do not migrate sufficiently northward, but do not even reach the Indian subcontinent. The Northern Hemisphere rainbelts, localized over Europe and eastern North America, are characterized in REF by overestimated precipitation rates along the east coast of North America, and over the relief (Rocky Mountains, Alps), and by a marked underestimation over the U.S. Great Plains and Europe. Evaporation rates are also underestimated by the GCM in these regions in summer (not shown).

In January, the precipitation distribution in REF is also realistic, but the continental precipitation associated with the midlatitude rainbelts and the ITCZ are too high. The Amazon basin is a notable exception; most precipitation occurs over the southeastern coast of Brazil instead of the Amazon basin, where precipitation rates are strongly underestimated in the LMD GCM. One can refer for further details to Polcher and Laval (1994), who compare the January precipitation field in a version of the LMD GCM similar to REF with the Hulme (1992) climatology.

5. Results

a. Annual water budgets

Table 2 summarizes the differences between REF, AWC, and TWC, over land and oceans. As already mentioned, the soil water-holding capacity is on the average very similar in REF and AWC, and roughly two times higher in TWC. The differences in mean soil moisture tend to follow the above differences in WHC; the mean soil moisture, similar in REF and AWC, almost doubles when going to TWC. The relative positions of REF and AWC are reversed, however, when one considers soil moisture (AWC < REF). The reason is that the largest WHC increases from REF to AWC occur in desert areas (Fig. 3), where the annual mean soil moisture is not notably changed from REF to AWC (Fig. 6), because no rainfall can fill the increased WHC. Subsequently, soil moisture variations over land between REF and AWC are related to the WHC variations in nondesert areas, which are dominated by the strong decreases over Asia, following WHC (Fig. 3a).

Soil water fluxes are also significantly sensitive to WHC changes. The annual mean drainage decreases from REF to both AWC and TWC, due to a decrease in the wetness of the lower soil layer (not shown). The annual mean surface runoff also decreases. This occurs in areas where the WHC increases from REF to AWC and TWC, leading to fewer instances of saturation. The largest change over land by far is the massive increase in annual mean evaporation from REF to both AWC and TWC. It is surprising, however, that the annual mean evaporation over land is almost as high in AWC as in TWC, when both its mean soil moisture and WHC are twice as small, and this raises the question of the origin of these evaporation increases.

Table 2 further shows that all other variables in AWC, over both land and ocean, are significantly closer to their values in TWC than in REF (the statistical significance of the difference in annual means was assessed with Student’s t-test). This behavior follows that of continental evaporation and contrasts with soil moisture and WHC. This indicates that the induced changes in continental evaporation are the main intermediary influencing the hydrological cycle and climate, as has been indicated in several studies (e.g., Simmonds and Hope 1998). The increase in continental evaporation from REF to AWC and TWC is split between increased continental precipitation (39% and 45% in AWC and TWC, respectively) and decreased moisture convergence over land (61% and 55%, respectively). Over the oceans, the mean evaporation rate is extremely similar in all simulations because of the prescribed sea surface temperature, and the reduction in mean moisture divergence in AWC and TWC is directly translated into an increase in mean precipitation. Finally, the changes in continental evaporation are also directly related to changes in the energy budget, as illustrated in Table 2 by the mean sensible heat flux and surface temperature over land.

The strong differences between TWC and REF will be analyzed first (section 5b), to identify the mechanisms linking evaporation and the atmospheric hydrological cycle in the LMD GCM. In a second step (section 5c), AWC and TWC will be compared to investigate the striking similarities between the two simulations, with a special emphasis on the interactions between soil moisture and evaporation.

b. Sensitivity to TWC

Figure 7 compares the annual zonal means over land from simulations REF and TWC, for soil moisture and three important terms in the atmospheric water budget:evaporation, precipitation, and moisture convergence. The zonal means south of 40°S are discarded, because they include either less than three land points, or land-ice points (Antarctica), where the land surface processes related to soil moisture are not enabled.

In agreement with Table 2, Fig. 7 shows a strong increase in soil moisture. This increase occurs at all latitudes and is associated with an increase in evaporation, by means of soil moisture stress reduction. North of 60°N, however, the evaporation increase is more modest than could be expected from the soil moisture increase, since potential evaporation is reduced because of low incoming radiation and the presence of snow. The interpretation of the subsequent differences in precipitation and moisture convergence between the two simulations calls for a distinction between the Tropics and the Northern Hemisphere extratropics, where different processes are relevant.

1) The Northern Hemisphere extratropics

In land areas north of 30°N, annual mean precipitation increases by an amount equivalent to the increase in evaporation, and moisture convergence is not significantly changed (Fig. 7). Most of the increase in evaporation in the Northern Hemisphere extratropics occurs 1) in the midlatitude rainbelts, and 2) in summer, when radiation drives high evaporation rates. This supports the assumption by Milly (1994) that WHC is important for the annual water balance because it allows interseasonal storage of water from the periods of water supply (winters) to the periods of water demand (summers). The importance of winter precipitation for evaporation increase is illustrated by the area to the east of the Aral Sea; the increase in WHC from REF-WHC to T-WHC is one of the highest in the extratropics (Fig. 3b), but the change in mean July evaporation is smaller than 0.5 mm day−1, because no rainfall is available to increase soil moisture in this very dry spot.

The increase in precipitation between REF and TWC is also maximum in summer in the Northern Hemisphere extratropics, and there is a strong spatial correlation in July between the areas of evaporation increase and precipitation increase in TWC relative to REF (Fig. 8). The statistical significance of the above changes was assessed in every grid point using Student’s t-test to compare the July means in REF and TWC from 1980 to 1988. Most evaporation increases greater than 1 mm day−1 were found to be significant at the level α = 0.05, and so were the precipitation increases, although to a lesser extent. The recycling of evaporation is recognized as an important source of precipitation over land, especially in the midlatitude rainbelts in summer (Brubaker et al. 1993) and can explain, in part, the simulated increase in precipitation. However, an increase in moisture convergence, related to dynamic feedbacks, is needed in areas where the increase in precipitation exceeds the increase in evaporation (such as over eastern Asia).

2) Sensitivity of the atmospheric water cycle to evaporation in the Tropics

In the Tropics, precipitation and moisture convergence exhibit a totally different response to the increased evaporation from REF to TWC. Figure 7 shows that the zonal annual mean precipitation barely changes, and the increase in evaporation is mainly reflected in a decrease in moisture convergence, centered on the ITCZ (characterized by the maximum of precipitation rates and moisture convergence). Figure 9 shows the spatial distribution of the annual mean differences in soil moisture, evaporation, moisture convergence, and precipitation between TWC and REF in the tropical band. The upper panel (Fig. 9a) displays the annual mean precipitation in REF, as a reference for the position of the ITCZ in the model. Figure 9b shows that the increase in soil moisture from REF to TWC is a general feature in hydrologically active areas (precipitation greater than 1–3 mm day−1 in the annual average). In these areas, rainfall permits the potential of increased WHC to lead effectively to an increase in soil moisture. Conversely, where precipitation is low, as in the western part of the Amazon basin, the increase in WHC, although larger than 200 mm there (Fig. 3), is not enough in itself to induce a meaningful increase in soil moisture.

Figure 9c shows that evaporation changes closely follow soil moisture changes, leading to a general increase in evaporation in the tropical rainbelts. The changes in moisture convergence over land (Fig. 9d) mainly consist of a decrease, which occurs in the areas of maximum precipitation (ITCZ). This decrease in moisture convergence over land is related to an increase in moisture convergence (or decrease in divergence) over the oceans, also located along the ITCZ. The changes in precipitation over land (Fig. 9e) can be detailed in reference to the above changes in evaporation and moisture convergence. The marked increase in precipitation from REF to TWC on the edges of the ITCZ, where the moisture convergence is not significantly changed, suggests a positive feedback between evaporation and precipitation. In the areas of strongest convection and precipitation, precipitation tends, in contrast, to decrease, following to a lesser extent the decrease in moisture convergence. Finally, the changes of precipitation over the oceans, where evaporation is basically unchanged, are directly related to the changes in moisture convergence.

The statistical significance of the above changes was assessed in every grid point using Student’s t-test to compare the annual means in REF and TWC from 1980 to 1988. Figure 10a indicates that the increases in evaporation located around the mean position of the ITCZ over land are, for the most part, statistically significant at the level α = 0.05. In the areas of moderate convection, most increases in precipitation are also statistically significant (Fig. 10b).

This analysis identified two different responses to increased evaporation in the Tropics over land: the first one favors a decrease in moisture convergence in the ITCZ, while the second one favors an increase in precipitation along the edges of the ITCZ. Although the focus was on annual mean behavior, the two different kinds of responses were also observed at monthly timescales. Table 3 details the annual mean differences in surface water and energy budgets, as well as their statistical significance, in the seven tropical areas shown in Fig. 11. The first six areas in Table 3, with high precipitation rates and moisture convergence, are representative of the ITCZ (see also Fig. 9a). In all these areas, evaporation increases from REF to TWC, whereas moisture convergence and precipitation decrease, in agreement with Figs. 9c and 9d. These changes are statistically significant, except for precipitation in one area. The sensible heat flux is also significantly decreased in TWC, like surface temperature. This behavior can be explained in terms of the sensitivity of tropical convection to surface energy fluxes in the LMD GCM, which was studied in detail by Polcher (1995), for a deforestation experiment. In particular, he showed that the triggering of deep convection, strongly related to large-scale moisture convergence, was more sensitive to sensible heat flux than to latent heat flux.

However, Table 3 does not show any relationship in the six sample regions between the magnitude of the decrease in sensible heat flux and the decrease in moisture convergence and precipitation. Spearman’s rank correlation coefficient (SRCC) was computed to quantify the correlation between the six regional means of precipitation difference and the six regional means for each other variable in Table 3. The closer the values to 1 or −1, the stronger the rank correlation, positive or negative, in the six sample regions. As mentioned above, the differences in precipitation are not correlated to the differences in sensible heat flux. They are significantly correlated to the differences in moisture convergence because of the strong physical link between moisture convergence and precipitation in areas of intense convection. The only other variables with significant correlations to the precipitation differences are annual mean precipitation and moisture convergence in REF. The proposed explanation is that these two quantities are related to the duration of intense convection and to the number of intense convective events, which are the most sensitive to sensible heat flux changes in the LMD GCM (Polcher 1995). The change in sensible heat flux can then be seen as a triggering factor to changes in convective activity, their magnitude being controlled by the intensity and duration of convection.

Area 2 on the northern coast of South America (Venezuela and surroundings) is given as a representative example of the behavior observed along the edges of the ITCZ. It is characterized by moderate convection, as indicated by the moderate precipitation rates and a low annual mean moisture convergence in REF. This area experiences a significant and important increase in precipitation (close to a doubling), which originates for the most part in the significant increase in evaporation. This response of precipitation to increased evaporation, which contrasts with that of highly convective areas, is also consistent with the sensitivity of tropical convection to surface energy fluxes as analyzed by Polcher (1995) in the LMD GCM. In areas of moderate convection, precipitation is not as strongly related to large-scale moisture convergence and is more directly controlled by the latent heat flux, which is associated with a release of water vapor in the atmosphere. This sensitivity is similar to that exhibited in the northern midlatitudes, where increased precipitation is related to increased evaporation through recycling.

3) Relationship with atmospheric circulation

The global distribution of sea level pressure (SLP) is examined to illustrate the influence of a change in WHC on atmospheric circulation. Figure 12a shows the reference global distribution of annual mean SLP in REF, and Fig. 12b shows large-scale increases in annual mean SLP over land from REF to TWC. It is noteworthy that these increases in SLP are related to decreases in surface temperature, which have very similar patterns and statistical significance to that of the SLP increase. This indicates that the main origin of the SLP changes is the increase in evaporation, which induces the decrease in surface temperature.

In the northern midlatitudes, the increases of SLP over land are specifically located in the areas of lowest SLP in REF and were found to be statistically significant at the level α = 0.05, using Student’s t-test. In contrast, the SLP decreases significantly over the oceans in areas of high SLP in REF. This results in a decrease in the low-level pressure gradients between land and oceans, inducing a weakening of the planetary waves and related zonal circulations. In the Tropics, SLP also increases over land, in areas that are characterized in REF by low annual mean SLP and correspond approximately to the areas affected by the ITCZ. The increase in SLP in these continental areas is related to a decrease in the large-scale mass circulation between land and ocean. This is consistent with the simulated decrease in moisture convergence in the ITCZ (Fig. 9), since the latter is strongly tied to the Hadley–Walker circulation.

c. Comparison of AWC and TWC

1) Meridional distributions over land

Figure 13 compares the annual zonal means over land of the differences TWC–REF and AWC–REF in soil moisture, evaporation, precipitation, and moisture convergence. The evaporation rates are very similar in AWC and TWC, whereas their soil moisture is markedly different. This shows that the strong nonlinearity between soil moisture and evaporation rates from REF to AWC to TWC, previously pointed out on average over land (Table 2), holds at all latitudes. The zonal differences in precipitation rate and moisture convergence are also extremely similar in TWC and AWC compared to REF, showing that the same processes explain the response of the hydrologic cycle to increased evaporation from REF to the two perturbation simulations (AWC and TWC). In the Northern Hemisphere extratropics, the increases in both evaporation and precipitation are smaller in AWC than in TWC. This is consistent with the importance of recycling at these latitudes, indicated by the very small amplitude of the moisture convergence changes. TWC and AWC exhibit even more similar hydrological behavior in the Tropics, with a strong decrease in both moisture convergence and precipitation at the equator (mean zonal location of the ITCZ, see Fig. 7) and an increase in precipitation rate in the surrounding areas. These changes in AWC and TWC compared to REF are directly related to the sensitivity of tropical convection to surface energy fluxes in the LMD GCM, which is different in areas of intense and moderate convection as previously shown by comparing TWC and REF in section 5b(2).

2) Regional analysis

The remaining question in understanding the similarities between AWC and TWC is the following: why are the evaporation rates so close in AWC and TWC despite the large discrepancies in their soil moisture? It was previously shown that evaporation is increased from REF to TWC in the hydrologically active areas, that is, the midlatitude and tropical rainbelts (Figs. 8, 9). This is also the case from REF to AWC (not shown). Figure 14 shows the annual cycles of soil moisture, evaporation and precipitation in two selected regions amid these rainbelts. The first region, located in the midlatitudes, is composed of six grid cells in western Europe (excluding Great Britain). The second region, in the Tropics, is area 2 in Fig. 11. The mean WHC in these two regions for the three simulations is given in Table 4. In reference to REF, the WHC is increased in both regions in AWC, but to a much smaller extent than in TWC; it is increased by 10% to 20% in AWC and by more than 100% in TWC.

These increases in WHC explain the soil moisture increases in both regions during the rainy season (defined from the precipitation cycle in REF in the lower panels). An important difference between the two regions is found in the relative phasing of their evaporation and precipitation cycles. In the midlatitudes, the strongest control on evaporation is the incoming surface radiation. In the selected Northern Hemisphere region, evaporation therefore occurs in spring and summer, and the increase in soil moisture occuring in the winter rainy season is not used to enhance evaporation until this period. Conversely, in the Tropics, incoming surface radiation is high all year long. However, as the selected tropical area is mostly located in the Northern Hemisphere, radiation displays a maximum in boreal summer. This coincides with the precipitation maximum, thus maximizing evaporation and evaporation sensitivity to soil moisture in boreal summer.

In the two areas, the large and moderate increases in soil moisture, from REF to TWC and AWC, respectively, both allow significant increases in evaporation, which have similar amplitude during the period of increasing evaporation (from March to June in the midlatitude region and from March to August in the tropical region). During this period, precipitation also increases, by an amount that is very close to the increase in evaporation. This suggests that the new rates of evaporation and precipitation in AWC and TWC are linked through the process of local/regional recycling. The similar behavior of AWC and TWC during this period of increasing evaporation results from both the positive feedbacks related to recycling and the original two-layer soil model in SECHIBA (section 2b). The precipitation rates control the depth and wetness of this top soil layer, which drives the evaporation rate when it is present. This is the case in the two studied regions during the period of increasing evaporation (when the radiation stress is stronger than the soil moisture stress), so that the evaporation rates can be comparably high in AWC and TWC.

The evaporation rates are no longer similar in these two simulations during the period of “decreasing evaporation” (from July to September in the midlatitude region and in September and October in the tropical region). By this period, soil moisture has already been depleted to a large extent by evaporation. In AWC, it becomes limiting (as in REF), which explains the sharp drop in evaporation rate in the two regions. In TWC, the large WHC is related to seasonal variations of soil moisture around a higher average than in AWC and REF. In particular, soil moisture is greater than in these two simulations during the period of decreasing evaporation, allowing higher evaporation rates to be sustained.

During this period, the response of precipitation to evaporation rates in TWC is different in the two selected regions. In the midlatitude region, the higher evaporation rates in TWC in late summer are related, by means of recycling, to significantly higher precipitation rates compared to the other two simulations. This is not the case in the tropical region, where the largest differences between TWC and REF occur from June to August for precipitation, and in September and October for evaporation. The seasonal cycle of precipitation is defined in the Tropics in relation to the seasonal shift of the ITCZ and the structure of the Hadley circulation; a rainy season occurs when the ITCZ (center or edges) covers the region, and when the ITCZ migrates zonally, the region becomes subsident and divergent, and this strongly inhibits convection and precipitation. As a result, the phase of the precipitation seasonal cycle is not markedly changed between the three simulations, and the main difference is the amplitude of this cycle, which depends on the evaporation rate during the rainy season.

The remaining hydrologically active zone is the ITCZ. Comparison of Tables 5 and 3 shows that the water and energy budgets at the surface are very similar in AWC and TWC, in the six regions selected as representative of the ITCZ on the annual average (areas 1 and 3–7 in Fig. 11). In addition, the precipitation differences between AWC and REF, as between TWC and REF, are significantly correlated to the annual averages of precipitation and moisture convergence in REF (ρ = 0.88, in both cases). This indicates that the relationship between increased evaporation and decreased moisture convergence and precipitation in the ITCZ is explained in AWC and TWC by the same processes: the increased evaporation is related through the surface energy balance to a decreased sensible heat flux, which reduces the number of deep convective events, thus moisture convergence and precipitation (Polcher 1995).

d. Comparison with observational data

1) Global water balance

The above studies showed that, in the LMD GCM, the hydrological cycle, and more generally the simulated climate, is highly sensitive to the definition of soil water-holding capacity. It is therefore important to understand how the latter can influence the realism of the simulated climate. Comparison of Tables 1 and 2 shows that the excessive moisture convergence over land (and divergence over ocean) in REF (section 4) is partly reduced in simulations AWC and TWC. This reduction arises from increases in both evaporation and precipitation over land, which thus deviate further from their climatological estimates. However, this allows the ratio of evaporation to precipitation to exceed the ratio of runoff to precipitation in AWC and TWC, which is more realistic than the opposite situation in REF.

2) Spatial distributions

The increases in WHC occuring in both AWC and TWC also allow some qualitative improvements of the precipitation and evaporation fields over land. In the Tropics, the precipitation rates in the ITCZ, which are too high in REF compared to the climatology from Legates and Willmott (1990), are significantly reduced in the two perturbation simulations. The increase in precipitation occuring on the edges of the ITCZ tends to improve the precipitation in the Amazon basin in January, which is strongly underestimated in REF (section 4). But this improvement is small compared to the extent of the underestimation and does not result from a fundamental change in the regional circulation, which does not penetrate far enough north into the Amazon basin (Polcher and Laval 1994). The summer monsoon over southeastern Asia is not notably improved in AWC and TWC compared to REF, either.

A major improvement related to the WHC increase is the significant increase in both evaporation and precipitation in the midlatitude rainbelts in summer (Fig. 8). In REF, the evaporation drop in late summer is particularly unrealistic and leads to an excessive warming of the continental surface. However, this drop is only delayed in TWC (Fig. 14), based on a total WHC (T-WHC) that is not physically consistent with the available WHC required in SECHIBA (sections 3a and 3b). In addition, precipitation is noticeably increased, in summer and winter, over the midlatitude mountain ranges, where orographic lift favors convection and condensation and precipitation is already overestimated in REF.

Decreases in WHC can also bring about some improvements in the simulated water cycle, as is the case on the eastern sides of the Northern Hemisphere continents (eastern Siberia, and Quebec and Labrador in Canada). The July mean precipitation is overestimated in REF in these areas (Fig. 5), and it is decreased in both TWC (Fig. 8b) and AWC (not shown), following the decrease in WHC (Fig. 3). This highlights the importance of a realistic large-scale distribution of the WHC.

3) Case study of the Mississippi basin

Figure 15 displays the simulated and observed (Wallis et al. 1991) cycles of precipitation in the Mississippi basin. The use of a larger WHC (326 mm in TWC and 177 mm in AWC, compared to 150 mm in REF) increases precipitation all year long, and the annual mean precipitation is closer to the observational estimate in REF than in AWC and TWC, because the latter two simulations strongly overestimate the precipitation maximum, in May and June. However, the summer precipitation, greatly underestimated in REF, is significantly increased in AWC and TWC, even if it remains too small compared to the observational data. Figure 15 also compares the simulated and observed total runoff (Wallis et al. 1991). The latter is approximated as the ratio of measured streamflow to the contributing area at the gauging station. On a monthly basis, it is a sensible estimate of total runoff at the GCM grid scale, because of the many gauging stations in the Mississippi basin, chosen to be as free as possible from the effects of water management (Wallis et al. 1991). The simulated runoff is quantitatively improved in winter in AWC and TWC compared to REF, because the higher WHC reduces the production of saturation excess runoff. In contrast, total runoff is not improved (increased) in summer, despite higher precipitation rates in AWC and TWC than in REF. The reason is the simultaneous increase in evaporation, which prevents water being available for runoff.

6. Conclusions

a. Summary

The aim of this work was to assess the sensitivity of the GCM simulated climate, and particularly of the hydrological cycle, to the definition of soil water-holding capacity. This question has already been addressed, by Milly and Dunne (1994), especially. These authors compared the climate simulated with the GFDL GCM coupled to a bucket LSM (one-layer soil and no explicit vegetation control on evapotranspiration), in the case of two extremely different values of globally constant WHC: 40 and 600 mm. In order to understand how more realistic changes in WHC, both in distribution and in range of values, can influence the water cycle and climate, we compared three 10-yr simulations with the LMD GCM coupled to a two-layer-soil SVAT model. The reference simulation uses WHCs (almost) globally equal to 150 mm, which is a “traditional” value for available WHC, whereas the two perturbation simulations use realistically distributed WHC datasets established by Patterson (1990). The available WHC dataset (in AWC) is physically consistent with the WHC definition in SECHIBA and has a global average close to 150 mm; the total WHC dataset (in TWC) is used as an example of large WHC increase (more than a doubling from 150 mm). These datasets have been improved (Dunne and Willmott 1996) since the initiation of our study. The main difference between these two versions of WHC datasets lies in how the soil and rooting depths are evaluated. We believe, however, that the resulting differences in WHC are not crucial in a sensitivity study such as ours.

The average impact over land of the increase in WHC (from REF to both AWC and TWC) is an increase in annual mean evaporation, which is split between increased annual precipitation and decreased annual mean moisture convergence. This annual mean impact, already pointed out by many studies (e.g, Shukla and Mintz 1992; Milly and Dunne 1994; Ducharne et al. 1998) is the simple result of water conservation in a steady simulation. These changes occur mainly in the continental rainbelts. In the midlatitude rainbelts, precipitation increases mostly during the evaporative period, through the recycling of increased evaporation, allowed by an increased interseasonal storage of soil moisture. In the Tropics, both moisture convergence and precipitation strongly decrease in the ITCZ, and precipitation increases in the surroundings areas. These two behaviors are both related to the sensitivity of the tropical convection to surface energy fluxes in the LMD GCM, which is a function of the intensity of convection (Polcher 1995).

Finally, the studied changes in WHC have some positive influence on the realism of the simulated hydrologic cycle, although some problems remain at all scales. The increase in evaporation and precipitation where they are underestimated in REF (midlatitude rainbelts over land) is a major improvement. However, this increase is generally too small, as shown in the Mississippi basin. Many studies (e.g., De Rosnay and Polcher 1998) did previously show that, in many GCMs, an important cause of the excessive dryness of the Northern Hemisphere landmasses in late summer is excessive incoming surface radiation. It is possible that the correction of this systematic error combined with an increase in WHC may improve precipitation (and evaporation and temperature) in the continental midlatitude summer.

b. Discussion

Two main results emerge in terms of sensitivities: first, the above changes in the hydrological cycle are driven by evaporation changes, and therefore WHC changes, that occur in the hydrologically active areas of the continental midlatitude and tropical rainbelts; second, a 10%–20% increase of WHC in the rainbelts (AWC) induces changes in the hydrological cycle that have similar patterns and almost the same amplitude as the changes induced by a WHC increase of more than 100% in the rainbelts (TWC). It is important to keep in mind that these results are model dependent, but this is, of course, true whichever GCM is considered.

We would like to point out the importance of land surface–atmosphere feedbacks in the presented results. In the framework of the Global Soil Wetness Project (Dirmeyer 1997), the influence of soil moisture on land surface fluxes was recently studied offline (Dirmeyer et al. 2000); the involved LSMs are driven by observational meteorological forcing, with no feedback allowed from the land surface to the atmosphere. In this offline study, the land surface fluxes, including evaporation rates, show the highest sensitivity to soil moisture in dry areas, where soil moisture stress has a strong effect on evaporation. Conversely, in a GCM environment, the highest sensitivity to soil moisture changes is found in the continental rainbelts, as shown in the present study, and also in Milly and Dunne (1994), and in Shukla and Mintz (1982), to cite only a few of the relevant studies. This strong difference in patterns between offline and GCM studies is directly related to the land–atmosphere feedbacks, which allow enhancement of the atmospheric hydrological cycle only in GCM studies. The strong similarity in this study between the evaporation rates in AWC and TWC, despite very different WHC and soil moisture, is also related to land–atmosphere feedbacks:positive feedback between the sensible heat flux and intense convection in the ITCZ; recycling in the midlatitudes and in the tropical areas of moderate convection. The intensity of the latter feedback has been shown to depend on the original two-layer structure of the soil in SECHIBA (section 5c).

In a recent review of the Core Project Biosphere Aspects of the Hydrological Cycle of the International Geosphere Biosphere Programme (Hutjes et al. 1998), the following concern is at the crux of one of the future key themes: “the required accuracy of simulated soil moisture (in terms of quantity and spatial pattern) in climate models is unclear.” Our study argues that soil water-holding capacity needs to be measured with a high level of accuracy in the continental rainbelts. There, a small change of available WHC (approximately +15%), from the uniform distribution REF-WHC to the realistic distribution A-WHC, resulted in significant changes of the hydrological cycle and climate. These changes were as important as those induced by a much larger increase in WHC (approximately +100%), when REF-WHC was replaced by T-WHC, which is not an available but a total WHC and, therefore, is not physically consistent with the LSM SECHIBA. This result is of particular interest in the context of LSM development, since it points out the importance of physically based parameter estimates. This is particularly striking if one compares the magnitude of the changes induced in two similar versions of the LMD GCM by a moderate increase in WHC on one hand (AWC, in this paper) and by direct changes of soil hydrology parameterizations (Ducharne et al. 1998) on the other; the first ones are by far larger than the second ones. The determination of WHC data for GCMs, however, is a complex issue, for numerous reasons. The estimation of the soil water retention characteristics, including WHC, are subject to many uncertainties at the field scale; the best way to aggregate (or, conversely, to distribute) such estimates at the scale of the GCM cell remains an area of intensive research, in relation to the recognition of lateral subgrid variability as a fundamental aspect of LSM developments (Dolman 1997). But the most important issue in the determination of WHC estimates for GCMs may be the lack of consensus inside the GCM modelers community about the definition of WHC and even soil moisture (Koster and Milly 1997).

The strong sensitivity of climate to soil WHC demonstrated by our study in a present day climate also has important implications in the context of climate change and/or land use change experiments. Both climate and land use changes are likely to change 1) the volumetric water capacity through the organic matter content and 2) the rooting depth, resulting in changes in areal WHC. Because of the nonlinear sensitivity of the hydrological cycle to WHC, the uncertainties in such possible WHC changes must be carefully assessed with regard to climate prediction error.

The last, though by no means the least, consequence of the extreme sensitivity of climate to soil WHC in our experiment is the need to ascertain its validity. In particular, Milly and Dunne (1994) examined a wide range of globally constant WHCs: 4, 150, 300, 600, 1200, and 2400 mm, and they showed an increase of about 70 mm of annual mean evaporation over land for each doubling of WHC until almost 600 mm. This is notably different from our results, which show a larger increase in evaporation, not only from a doubling in WHC, but also from a 15% increase in the continental rainbelts. There could be many explanations for these discrepancies, including the differences in the studied WHC range and in the numerical environment. This calls for further work, and it would be of particularly great interest to compare the sensitivities of evaporation and the hydrological cycle to WHC in a bucket and a SVAT model coupled to the same GCM, in order to extract the responses that are caused by the LSM itself from the responses that are explained by the GCM and its physical parameterizations (convection, condensation, etc.).

Acknowledgments

We would like to thank Krista Dunne (née Patterson) for providing us with the WHC datasets used in this study, as well as Dennis P. Lettenmaier for the observed data over the Mississippi basin. We are also very grateful to Randal Koster, Jan Polcher, and Jennifer Crossley for their helpful comments. This work was supported, in terms of computational facilities, by the Institut du Développement et des Ressources en Informatique Scientifique.

REFERENCES

  • Abramopoulos, F., C. Rosenzweig, and B. Choudhury, 1988: Improved ground hydrology calculations for global climate models (GCMs): Soil water movement and evapotranspiration. J. Climate,1, 921–941.

  • Avissar, R., and R. Pielke, 1989: A parametrization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Mon. Wea. Rev.,117, 2113–2136.

  • Baumgartner, A., and E. Reichel, 1975: The World Water Balance. Elsevier, 179 pp.

  • Brubaker, K. L., D. Entekhabi, and P. S. Eagleson, 1993: Estimation of continental precipitation recycling. J. Climate,6, 1077–1089.

  • Budyko, M. I., 1978: The heat balance of the Earth. Climatic Change, J. Gribbin, Ed., Cambridge University Press, 85–113.

  • Choisnel, E. M., S. V. Jourdain, and C. J. Jacquart, 1995: Climatological evaluation of some fluxes of the surface energy and soil water balances over France. Ann. Geophys.,13, 666–674.

  • De Rosnay, P., and J. Polcher, 1998: Modeling root water uptake in a complex land surface scheme coupled to a GCM. Hydrol. Earth Syst. Sci.,2, 239–255.

  • Delworth, T. L., and S. Manabe, 1988: The influence of potential evaporation on the variabilities of simulated soil wetness and climate. J. Climate,1, 523–547.

  • ——, M. Bruen, and J. Polcher, 1998: Sensitivity of the surface fluxes to the number of layers in the soil model used in GCMs. Geophys. Res. Lett., in press.

  • Dickinson, R. E., 1984: Modeling evaporation for three-dimensional global climate models. Climate Processes and Climate Sensitivity, J. E. Hansen and T. Takahashi, Eds., Amer. Geophys. Union, 58–72.

  • ——, and A. Henderson-Sellers, 1988: Modelling tropical deforestation: A study of GCM land-surface parameterizations. Quart. J. Roy. Meteor. Soc.,114, 439–462.

  • Dirmeyer, P. A., 1997: The Global Soil Wetness Project: Initial results. GEWEX News,7, 3–6.

  • ——, and J. Shukla, 1993: Observational and modeling studies of the influence of soil moisture anomalies on atmospheric circulation (Review). Predictions of Interannual Climate Variations, J. Shukla, Ed., NATO ASI Series, Vol. 16, Springer-Verlag, 1–23.

  • ——, F. Zeng, A. Ducharne, J. Morrill, and R. Koster, 2000: The sensitivity of surface fluxes to soil water content in three land surface schemes. J. Hydrometeor.,1, 121–134.

  • Dolman, A., 1997: Workshop on Soil Vegetation Atmosphere Transfer (SVAT) Schemes/Land Surface Parameterizations (LSPs). GEWEX News,7, 13–14.

  • Dorman, J., and P. Sellers, 1989: A global climatology of albedo, roughness length and stomatal resistance for atmospheric general circulation models as represented by the Simple Biosphere Model (SIB). J. Appl. Meteor.,28, 833–855.

  • Ducharne, A., K. Laval, and J. Polcher, 1998: Sensitivity of the hydrological cycle to the parametrization of soil hydrology in a GCM. Climate Dyn.,14, 307–327.

  • ——, R. D. Koster, M. Suarez, and P. Kumar, 1999: A catchment-based land surface model for GCMS and the framework for its evaluation. Phys. Chem. Earth,24B, 769–773.

  • Ducoudré, N., K. Laval, and A. Perrier, 1993: SECHIBA, a new set of parametrizations of the hydrologic exchanges at the land–atmosphere interface within the LMD atmospheric general circulation model. J. Climate,6, 248–273.

  • Dunne, K. A., and C. J. Willmott, 1996: Global distribution of plant-extractable water capacity of soil. Int. J. Climatol.,16, 841–859.

  • Entekhabi, D., and P. Eagleson, 1989: Land surface hydrology parametrization for atmospheric general circulation models including subgrid scale spatial variability. J. Climate,2, 816–831.

  • ——, I. Rodriguez-Iturbe, and F. Castelli, 1996: Mutual interaction of soil moisture and atmospheric processes. J. Hydrol.,184, 3–17.

  • Fouquart, Y., and B. Bonnel, 1980: Computations of solar heating of the earth’s atmosphere: A new parametrisation. Beitr. Phys. Atmos.,53, 35–62.

  • Gadgil, S., and S. Sajani, 1998: Monsoon precipitation in the AMIP runs. Climate Dyn.,14, 659–689.

  • Garratt, J., 1993: Sensitivity of climate simulations to the land-surface and atmospheric boundary-layer treatments—A review. J. Climate,6, 419–449.

  • Gates, W. L., 1992: AMIP: The Atmospheric Model Intercomparison Project. Bull. Amer. Meteor. Soc.,73, 1962–1970.

  • Henning, D., 1989: Atlas of the Surface Heat Balance of the Continents. Gebrüder Borntraeger, 402 pp.

  • Hulme, M., 1992: A 1951–80 global land precipitation climatology for the evaluation of general circulation models. Climate Dyn.,7, 57–72.

  • Hutjes, R., and Coauthors, 1998: Biospheric aspects of the hydrological cycle—Preface. J. Hydrol.,213, 1–21.

  • Koster, R. D., and M. J. Suarez, 1992: Modeling the land surface boundary in climate models as a composite of independent vegetation stands. J. Geophys. Res.,97 (D3), 2697–2715.

  • ——, and P. C. D. Milly, 1997: The interplay between transpiration and runoff formulations in land surface schemes used with atmospheric models. J. Climate,10, 1578–1591.

  • Kuo, H., 1965: On formation and intensification of tropical cyclones through latent heat release by cumulus convection. J. Atmos. Sci.,22, 40–63.

  • Laval, K., 1988: Experience with surface processes at LMD. Proc. Parametrization of Fluxes over Land Surface, Reading, United Kingdom, European Centre for Medium-Range Weather Forecasts, 171–205.

  • ——, H. Le Treut, and R. Sadourny, 1981: Effects of cumulus parameterization on the dynamics of a general circulation model. Geophys. Astrophys. Fluid Dyn.,17, 113–127.

  • Legates, D. R., and C. J. Willmott, 1990: Mean seasonal and spatial variability in gauge-corrected, global precipitation. Int. J. Climatol.,10, 111–127.

  • Le Treut, H., and Z.-X. Li, 1991: Sensitivity of an atmospheric general circulation model to prescribed SST changes: Feedback effects associated with the simulation of cloud optical properties. Climate Dyn.,5, 175–187.

  • Liang, X., D. P. Lettenmaier, E. F. Wood, and S. J. Burges, 1994: A simple hydrologically based model of land surface water and energy fluxes for general circulation models. J. Geophys. Res.,99 (D7), 14 415–14 428.

  • Mahfouf, J.-F., and Coauthors, 1996: Analysis of results from the PILPS-RICE Workshop. Part 3: Transpiration. Global Planet. Change,13, 73–88.

  • Manabe, S., 1969: Climate and the ocean circulation 1. The atmospheric circulation and the hydrology of the earth’s surface. Mon. Wea. Rev.,97, 739–774.

  • ——, and R. Strickler, 1964: Thermal equilibrium of the atmosphere with a convective adjustment. J. Atmos. Sci.,21, 361–385.

  • Milly, P. C. D., 1994: Climate, soil water storage, and the average annual water balance. Water Resour. Res.,30, 2143–2156.

  • ——, and K. A. Dunne, 1994: Sensitivity of the global water cycle to the water-holding capacity of land. J. Climate,7, 506–526.

  • Mintz, Y., and Y. Serafini, 1992: A global monthly climatology of soil moisture and water balance. Climate Dyn.,8, 13–27.

  • Morcrette, J., 1991: Radiation and cloud radiative properties in the European Centre for Medium range Weather Forecasts forecasting system. J. Geophys. Res.,96, 9121–9132.

  • Patterson, K. A., 1990: Global distributions of total and total-available soil water holding capacities. M.S. thesis, Dept. of Geography, University of Delaware, 119 pp. [Available from Morris Library, University of Delaware, Newark, DE 19716.].

  • Polcher, J., 1995: Sensitivity of tropical convection to land surface processes. J. Atmos. Sci.,52, 3143–3161.

  • ——, and K. Laval, 1994: A statistical study of regional impact of deforestation on climate of the LMD-GCM. Climate Dyn.,10, 205–219.

  • Sadourny, R., and K. Laval, 1984: January and July performance of the LMD general circulation model. New Perspectives in Climate Modelling, A. L. Berger and C. Nicolis, Eds., Elsevier Science Publishers, 173–197.

  • Sellers, P., Y. Mintz, Y. Sud, and A. Dachler, 1986: A simple biopshere model (SiB) for use within general circulation models. J. Atmos. Sci.,43, 505–531.

  • Shukla, J., and Y. Mintz, 1982: Influence of land-surface evaporation on the earth’s climate. Science,215, 1498–1501.

  • Simmonds, I., and P. Hope, 1998: Seasonal and regional responses to changes in Australian soil moisture conditions. Int. J. Climatol.,10, 1105–1139.

  • Stamm, J. F., E. F. Wood, and D. P. Lettenmaier, 1994: Sensitivity of a GCM simulation of global climate to the representation of land-surface hydrology. J. Climate,7, 1218–1239.

  • Walker, J., and P. Rowntree, 1977: The effect of soil moisture on circulation and rainfall in a tropical model. Quart. J. Roy. Meteor. Soc.,103, 29–46.

  • Wallis, J. R., D. P. Lettenmaier, and E. F. Wood, 1991: A daily hydroclimatological data set for the continental United States. Water Resour. Res.,27, 1657–1663.

  • Wilson, M. F., and A. Henderson-Sellers, 1985: A global archive of land cover and soil data for use in general circulation climate models. J. Climatol.5, 119–143.

  • Zeng, X., Y.-J. Dai, R. Dickinson, and M. Shaikh, 1998: The role of root distribution for climate simulation over land. Geophys. Res. Lett.,25, 4533–4536.

  • Zobler, L., 1986: A world soil file for global climate modeling. National Aeronautics and Space Administration Tech. Memo. 87802, 32 pp.

Fig. 1.
Fig. 1.

Schematic diagram of the evolution of the two soil layers in SECHIBA

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 2.
Fig. 2.

Reference distribution of WHC in SECHIBA: the WHC is 150 kg m−3 over dark regions, and 30 kg m−3 over the light-colored areas

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 3.
Fig. 3.

Differences between the new WHC datasets and the reference WHC dataset (mm): (a) A-WHC–REF-WHC and (b) T-WHC)–REF-WHC. White areas over land indicate negative differences, with dashed contours at −50, − 100, and −200 mm. Positive differences appear in gray, with increasing shade between the solid contours at 0, 50, 100, 200 mm, and over

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 4.
Fig. 4.

Evolution of global annual mean soil moisture (mm) in simulations REF, AWC, and TWC

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 5.
Fig. 5.

July precipitation rates (mm day−1): (a) in REF (1980–88) and (b) in the climatology of Legates and Willmott (1990). Contours at 1, 3, 5, and 10 mm day−1

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 6.
Fig. 6.

Annual mean differences in soil moisture (mm) between AWC and REF (AWC–REF). Contours at −100, −50, −10, 10, 50, and 100 mm

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 7.
Fig. 7.

Zonal annual means over land, in REF and TWC, of soil moisture (mm), precipitation rate, evaporation rate, and moisture convergence (mm day−1)

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 8.
Fig. 8.

Differences (TWC − REF) in Jul mean (a) evaporation and (b) precipitation (mm day−1), in the northern extratropics. Contours at −2, −1, −0.5, 0.5, 1, and 2 mm day−1

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 9.
Fig. 9.

Tropical band: (a) annual mean precipitation in REF (mm day−1), and differences (TWC − REF) in annual mean (b) soil moisture (mm), (c) evaporation, (d) moisture convergence, and (e) precipitation (mm day−1). Contours at 1, 3, 5, and 10 mm day−1 in (a), −100, −50, −10, 10, 50, and 100 mm in (b), and −2, −1, −0.5, 0.5, 1, and 2 mm day−1 in (c), (d), and (e)

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 10.
Fig. 10.

Significance of the annual mean differences in (a) evaporation and (b) precipitation, between TWC and REF. The dark areas correspond to statistically significant differences at the level α = 0.05; the spatial distribution of these differences is shown by the contours (−2, −1, −0.5, 0.5, 1, and 2, in mm day−1), with dotted lines for negative differences

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 11.
Fig. 11.

Location of the seven areas selected to assess the statistical significance of the ITCZ changes

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 12.
Fig. 12.

Sea level pressure: (a) annual mean distribution in REF, in hPa − 1000, contours at 7.5, 10, 15, and 20 hPa − 1000, (b) annual mean differences between TWC and REF: the contours (at −3, −2, −1, 1, 2, and 3 hPa − 1000, with dotted lines for negative values) correspond to TWC–REF, and the dark areas indicate the statistically significant differences at the level α = 0.05

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 13.
Fig. 13.

Annual zonal means over land: differences between TWC and REF (TWC − REF) and AWC and REF (AWC − REF), for (a) soil moisture (mm), (b) evaporation, (c) precipitation, and (d) moisture convergence (mm day−1)

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 14.
Fig. 14.

Mean seasonal cycles of soil moisture (mm), evaporation, and precipitation rates (mm day−1), simulated in REF, AWC, and TWC, in two tropical and northern midlatitude regions. The symbols on the AWC (TWC) curves indicate the statistical significance (at the level α = 0.05) of the difference between the monthly means: • indicates significant differences with both REF and TWC (AWC); * indicates a significant difference with REF only; and □ indicates a significant difference with TWC (AWC) only

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Fig. 15.
Fig. 15.

Comparison of simulated (REF, AWC, and TWC) and observed (Wallis et al. 1991) seasonal cycles of (a) precipitation and (b) total runoff over the Mississippi basin (mm day−1)

Citation: Journal of Climate 13, 24; 10.1175/1520-0442(2000)013<4393:IOTRDO>2.0.CO;2

Table 1.

World water budget, per unit area (mm yr−1): comparison of estimates from different sources with annual averages from 1980–88 in REF. The letters P, E, and Y denote precipitation, evaporation, and total runoff, respectively, and the subscripts c and o denote the continents and oceans, respectively

Table 1.
Table 2.

Comparison of the simulations REF, AWC, and TWC: annual means over land and ocean areas. A * (†) indicates a statistically significant difference at the level α = 0.05 (α = 0.01). The values in parentheses are values of moisture convergence in 1012 m3 yr−1

Table 2.
Table 3.

Summary of the annual mean differences in climate between REF and TWC, in the areas located in Fig. 11. All differences in means are statistically significant at the level α = 0.05, except if denoted NS. The notations are Ngp for number of grid points in the region, E for evaporation (mm day−1), P for precipitation (mm day−1), PE for moisture convergence (mm day−1), H for sensible heat flux (W m−2), Ts for surface temperature (C), and Δ is the difference operator TWC − REF. SRCC denotes Spearman’s rank correlation coefficientof ΔP and the other variables in the six ITCZ regions (6, 7, 1, 3, 4, and 5); significant correlation at the level α = 0.05 is denoted with a ★

Table 3.
Table 4.

Mean WHC (mm) in the two midlatitude and tropical regions analyzed in Fig. 14

Table 4.
Table 5.

Summary of the annual mean differences in climate between REF and AWC, in the areas located in Fig. 11. All differences in means are statistically significant at the level α = 0.05, except those denoted NS. The notations are the same as in Table 3, but Δ denotes the difference operator AWC–REF

Table 5.
Save