• Avissar, R., 1992: Conceptual aspects of a statistical–dynamical approach to represent landscape subgrid-scale heterogeneities in atmospheric models. J. Geophys. Res., 97 , 27292742.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Avissar, R., , and Pielke R. A. , 1989: A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Mon. Wea. Rev., 117 , 21132136.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., 1996: A land surface model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: Technical description and user’s guide. NCAR Tech. Note NCAR/TN-417+STR, Boulder, CO, 150 pp.

  • Bonan, G. B., , Pollard D. , , and Thompson S. L. , 1993: Influence of subgrid-scale heterogeneity in leaf area index, stomatal resistance, and soil moisture on grid-scale land–atmosphere interactions. J. Climate, 6 , 18821897.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, M. X., , Dickinson R. E. , , Zeng X. , , and Hahmann A. N. , 1996: Comparison of precipitation observed over the continental United States to that simulated by a climate model. J. Climate, 9 , 22332249.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, M. X., , Zeng X. , , and Dickinson R. E. , 1998: Adjustment of GCM precipitation intensity over the United States. J. Appl. Meteor., 37 , 876887.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Collier, C. G., 1993: The application of a continental-scale radar database to hydrological process parameterization within atmospheric general circulation models. J. Hydrol., 142 , 301318.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Conner, M. D., , and Petty G. R. , 1998: Validation and intercomparison of SSM/I rain-rate retrieval methods over the continental Unites States. J. Appl. Meteor., 37 , 679700.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dai, A. G., 2001a: Global precipitation and thunderstorm frequencies. Part I: Seasonal and interannual variations. J. Climate, 14 , 10921111.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dai, A. G., 2001b: Global precipitation and thunderstorm frequencies. Part II: Diurnal variations. J. Climate, 14 , 11121123.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dai, Y. J., , and Zeng Q. C. , 1997: A land surface model (IAP94) for climate studies, Part I: Formation and validation in off-line experiments. Adv. Atmos. Sci., 14 , 433460.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dai, Y. J., and Coauthors, 2003: The Common Land Model. Bull. Amer. Meteor. Soc., 84 , 10131023.

  • Dickinson, R. E., , Henderson-Sellers A. , , and Kennedy P. J. , 1993: Biosphere–Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Model. NCAR Tech. Note NCAR/TN-387+STR, Boulder, CO, 72 pp.

  • Dinku, T., , and Anagnostou E. N. , 2005: Regional differences in overland rainfall estimation from PR-calibrated TMI algorithm. J. Appl. Meteor., 44 , 189205.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dolman, A. J., , and Gregory D. , 1992: The parameterization of rainfall interception in GCMs. Quart. J. Roy. Meteor. Soc., 118 , 455467.

  • Eltahir, E. A. B., , and Bras R. L. , 1993: Estimation of the fractional coverage of rainfall in climate models. J. Climate, 6 , 639644.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Entekhabi, D., , and Eagleson P. S. , 1989: Land surface hydrology parameterization for atmospheric general circulation models including subgrid scale spatial variability. J. Climate, 2 , 816831.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., , and Marks G. F. , 1995: The development of SSM/I rain-rate retrieval algorithms using ground-based radar measurements. J. Atmos. Oceanic Technol., 12 , 755770.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gao, X. G., , and Sorooshian S. , 1994: A stochastic precipitation disaggregation scheme for GCM application. J. Climate, 7 , 238247.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ghan, S. J., , Liljigren J. C. , , Shaw W. J. , , Hubbe J. H. , , and Dorman J. C. , 1997: Influence of subgrid variability on surface hydrology. J. Climate, 10 , 31573166.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gong, G., , Entekhabi D. , , and Salvucci G. D. , 1994: Regional and seasonal estimates of fractional storm coverage based on station precipitation observations. J. Climate, 7 , 14951505.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grecu, M., , and Anagnostou E. N. , 2001: Overland precipitation estimation from TRMM passive microwave observation. J. Appl. Meteor., 40 , 13671380.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grecu, M., , and Anagnostou E. N. , 2002: Use of passive microwave observations in a radar rainfall-profiling algorithm. J. Appl. Meteor., 41 , 702715.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hahmann, A. N., 2003: Representing spatial subgrid-scale precipitation variability in a GCM. J. Hydrometeor., 4 , 891900.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hahmann, A. N., , and Dickinson R. E. , 2001: A fine-mesh land approach for general circulation models and its impact on regional climate. J. Climate, 14 , 16341646.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Henderson-Sellers, A., , and Pitman A. J. , 1992: Land-surface schemes for future climate models: Specification, aggregation, and heterogeneity. J. Geophys. Res., 97 , 26872696.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Johnson, K. D., , Entekhabi D. , , and Eagleson P. S. , 1991: The implementation and validation of improved land surface hydrology in an atmospheric general circulation model. Ralph M. Parsons Laboratory Tech. Rep. 334, Massachusetts Institute of Technology, Cambridge, MA, 192 pp.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Koster, R. D., , and Suarez M. J. , 1992: Modeling the land surface boundary in climate models as a composite of independent vegetation stands. J. Geophys. Res., 97 , 26972715.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kummerow, C., and Coauthors, 2000: The status of the Tropical Rainfall Measuring Mission (TRMM) after two years in orbit. J. Appl. Meteor., 39 , 19651982.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Leung, L. R., , and Ghan S. J. , 1995: A subgrid parameterization of orographic precipitation. Theor. Appl. Climatol., 52 , 95118.

  • Liang, X., , Lettenmaier D. P. , , and Wood E. F. , 1996: One-dimensional statistical dynamic representation of subgrid spatial variability of precipitation in the two-layer variable infiltration capacity model. J. Geophys. Res., 101 , 2140321422.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCollum, J. R., , and Ferraro R. R. , 2003: Next generation of NOAA/NESDIS TMI, SSM/I, and AMSR-E microwave land rainfall algorithms. J. Geophys. Res., 108 .8382, doi:10.1029/2001JD001512.

    • Search Google Scholar
    • Export Citation
  • Mölders, N., , and Raabe A. , 1996: Numerical investigations on the influence of subgrid-scale surface heterogeneity on evapotranspiration and cloud processes. J. Appl. Meteor., 35 , 782795.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pitman, A. J., , Henderson-Sellers A. , , and Yang Z. L. , 1990: Sensitivity of regional climates to localized precipitation in global models. Nature, 346 , 734737.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pitman, A. J., , Yang Z. L. , , and Henderson-Sellers A. , 1993: Sub-grid scale precipitation in AGCMs: Reassessing the land surface sensitivity using a single column model. Climate Dyn., 9 , 3341.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ramírez, J. A., , and Senarath S. U. S. , 2000: A statistical–dynamical parameterization of interception and land surface–atmosphere interactions. J. Climate, 13 , 40504065.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Seth, A., , Giorgi F. , , and Dickinson R. E. , 1994: Simulating fluxes from heterogeneous land surfaces: Explicit subgrid method employing the Biosphere–Atmosphere Transfer Scheme (BATS). J. Geophys. Res., 99 , 1865118667.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shuttleworth, W. J., 1988: Macrohydrology—The new challenge for process hydrology. J. Hydrol., 100 , 3156.

  • Thomas, G., , and Henderson-Sellers A. , 1991: An evaluation of proposed representation of subgrid hydrological processes in climate models. J. Climate, 4 , 898910.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tian, Y., and Coauthors, 2004a: Comparison of seasonal and spatial variations of LAI/FPAR from MODIS and Common Land Model. J. Geophys. Res., 109 .D01103, doi:10.1029/2003JD003777.

    • Search Google Scholar
    • Export Citation
  • Tian, Y., , Dickinson R. E. , , Zhou L. , , Myneni R. B. , , Friedl M. , , Schaaf C. B. , , Carroll M. , , and Gao F. , 2004b: Land boundary conditions from MODIS data and consequences for the albedo of a climate model. Geophys. Res. Lett., 31 .L05504, doi:10.1029/2003GL019104.

    • Search Google Scholar
    • Export Citation
  • Tian, Y., , Dickinson R. E. , , Zhou L. , , and Shaikh M. , 2004c: Impact of new land boundary conditions from Moderate Resolution Imaging Spectroradiometer (MODIS) data on the climatology of land surface variables. J. Geophys. Res., 109 .D20115, doi:10.1029/2003JD004499.

    • Search Google Scholar
    • Export Citation
  • Wang, G. L., , and Eltahir E. A. B. , 2000: Modeling the Biosphere–Atmosphere System: The impact of the subgrid variability in rainfall interception. J. Climate, 13 , 28872899.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wood, E. F., , Lettenmaier D. P. , , and Zartarian V. G. , 1992: A land-surface hydrology parameterization with subgrid variability for general circulation models. J. Geophys. Res., 97 , 27172728.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    Study region. On the right is the central Africa region, ranging from 15°S to 5°N and 10° to 30°E; on the left is the Amazon region ranging from 15°S to 5°N and 70° to 40°W.

  • View in gallery

    Comparison of rain rates between the theoretical pdfs and the histogram based on observation over (a) the central Africa and (b) the Amazon region. Lines with stars represent exponential pdf, and lines with circles represent lognormal pdf. The unit of rain rate is mm day−1.

  • View in gallery

    Water balance over central Africa in CTL simulation. All quantities as a fraction of precipitation: (a) runoff (including surface runoff and base runoff), (b) total evapotranspiration, (c) interception loss, (d) ground evaporation , and (e) plant transpiration.

  • View in gallery

    Same as Fig. 3, but for Africa EXP simulation.

  • View in gallery

    Heat fluxes over central Africa in CTL and EXP. (a) Sensible heat flux in CTL, (b) sensible heat flux in EXP, (c) latent heat flux in CTL, and (d) latent heat flux in EXP. The unit is W m−2.

  • View in gallery

    Central Africa: Soil water within the 10 layers (totaling 3.44 m) in (a) CTL and in (b) EXP. Soil water within the top 1 m in (c) CTL and in (d) EXP. The unit is kg m−2.

  • View in gallery

    Central Africa at a low-LAI (LAI = 1.8) site: Seasonal cycles of (a) soil water within the 10 layers (totaling 3.44 m), and (b) soil water within the top 1 m; and (c) transpiration and rainfall. (d), (e), and (f) Same as (a), (b), and (c), but at a high-LAI (LAI = 5.4) site. (Solid lines with circles: CTL; solid lines with triangles: EXP; solid lines: rainfall).

  • View in gallery

    Central Africa at a low-LAI (LAI = 1.8) site: Seasonal cycles of (a) LAI, (b) vegetation and ground temperatures, and (c) sensible and latent heat fluxes. (d), (e), and (f) Same as (a), (b), and (c), but at a high-LAI (LAI = 5.4) site. Dash–dot lines: vegetation temperature in CTL; solid lines: vegetation temperature in EXP; dash–dot lines with circles: ground temperature in CTL; solid lines with circles: ground temperature in EXP in (b) and (e). Dash–dot lines: sensible heat flux in CTL; solid lines: sensible heat flux in EXP; dash–dot lines with circles: latent heat flux in CTL; solid lines with circles: latent heat flux in EXP in (c) and (f).

  • View in gallery

    Central Africa: The impacts of LAI and rainfall coverage fraction on the EXP−CTL difference in interception loss ratio. (a) Difference vs LAI, (b) difference vs rainfall coverage fraction, (c) rainfall coverage fraction vs LAI, and (d) normalized difference (EXP–CTL difference in the interception loss ratio divided by LAI) vs rainfall coverage fraction.

  • View in gallery

    Water balance over the Amazon region. All quantities as a fraction of precipitation: (a) total evapotranspiration in CTL; (b) interception loss in CTL; (c) plant transpiration in CTL; (d) total evapotranspiration in EXP; (e) interception loss in EXP; and (f) plant transpiration in EXP.

  • View in gallery

    Same as Fig. 7, but for the Amazon region.

  • View in gallery

    Same as Fig. 9, but for the Amazon region.

  • View in gallery

    LAI over (a) the central Africa and (b) the Amazon region. Rainfall coverage fraction over (c) the central Africa and (d) the Amazon region

  • View in gallery

    Interception loss ratio under (a) exponential pdf, (b) lognormal pdf, and (c) spatially uniform rain rate within the rain-covered area over the central Africa region. (d), (e), and (f) Same as (a), (b), and (c) but for the Amazon region.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 13 13 1
PDF Downloads 9 9 1

Use of Satellite-Based Precipitation Observation in Improving the Parameterization of Canopy Hydrological Processes in Land Surface Models

View More View Less
  • 1 Department of Civil and Environmental Engineering, University of Connecticut, Storrs, Connecticut
© Get Permissions
Full access

Abstract

Precipitation exhibits significant spatial variability at scales much smaller than the typical size of climate model grid cells. Neglecting such subgrid-scale variability in climate models causes unrealistic representation of land–atmosphere flux exchanges. It is especially problematic over densely vegetated land. This paper addresses this issue by incorporating satellite-based precipitation observations into the representation of canopy interception processes in land surface models. Rainfall data derived from passive microwave (PM) observations are used to obtain realistic estimates of 1) conditional mean rain rates, which together with the modeled rain rate are used to estimate the rainfall coverage fraction at each model grid cell in this study, and 2) the probability density function (pdf) of rain rates within the rain-covered areas. Both of these properties significantly impact the land–atmosphere water vapor exchanges. Based on the above information, a statistical–dynamical approach is taken to incorporate the representation of precipitation subgrid variability into canopy interception processes in land surface models. The results reveal that incorporation of precipitation subgrid variability significantly alters the partitioning between runoff and total evapotranspiration as well as the partitioning among the three components of evapotranspiration (i.e., canopy interception loss, ground evaporation, and plant transpiration). This further influences soil water, surface temperature, and surface heat fluxes. It is shown that the choice of the rain-rate pdf within rain-covered areas has an effect on the model simulation of land–atmosphere flux exchanges. This study demonstrates that land surface and climate models can substantially benefit from the fine-resolution remotely sensed rainfall observations.

Corresponding author address: Dr. Guiling Wang, Dept. of Civil and Environmental Engineering, University of Connecticut, 261 Glenbrook Rd., Storrs, CT 06269-2037. Email: gwang@engr.uconn.edu

Abstract

Precipitation exhibits significant spatial variability at scales much smaller than the typical size of climate model grid cells. Neglecting such subgrid-scale variability in climate models causes unrealistic representation of land–atmosphere flux exchanges. It is especially problematic over densely vegetated land. This paper addresses this issue by incorporating satellite-based precipitation observations into the representation of canopy interception processes in land surface models. Rainfall data derived from passive microwave (PM) observations are used to obtain realistic estimates of 1) conditional mean rain rates, which together with the modeled rain rate are used to estimate the rainfall coverage fraction at each model grid cell in this study, and 2) the probability density function (pdf) of rain rates within the rain-covered areas. Both of these properties significantly impact the land–atmosphere water vapor exchanges. Based on the above information, a statistical–dynamical approach is taken to incorporate the representation of precipitation subgrid variability into canopy interception processes in land surface models. The results reveal that incorporation of precipitation subgrid variability significantly alters the partitioning between runoff and total evapotranspiration as well as the partitioning among the three components of evapotranspiration (i.e., canopy interception loss, ground evaporation, and plant transpiration). This further influences soil water, surface temperature, and surface heat fluxes. It is shown that the choice of the rain-rate pdf within rain-covered areas has an effect on the model simulation of land–atmosphere flux exchanges. This study demonstrates that land surface and climate models can substantially benefit from the fine-resolution remotely sensed rainfall observations.

Corresponding author address: Dr. Guiling Wang, Dept. of Civil and Environmental Engineering, University of Connecticut, 261 Glenbrook Rd., Storrs, CT 06269-2037. Email: gwang@engr.uconn.edu

1. Introduction

Accurate estimation of land–atmosphere flux exchanges is an important aspect of climate modeling. However, because of the coarse resolution in climate models, land–atmosphere interactions are not well represented. For example, the typical size of an atmospheric general circulation model (AGCM) grid cell is 104–105 km2, while size of a typical convective rain cell (associated with the most intense rain rates) does not exceed an area of 100 km2 (e.g., Pitman et al. 1990). This scale mismatch causes climate models to predict very low rain rates, which represent grid averages instead of local precipitation intensity. As a result, climate models often overestimate the amount of water intercepted by vegetation canopy, which subsequently leads to unrealistic representation of other hydrological processes. In this paper, we investigate an approach for improving the representation of canopy hydrological processes using satellite-based precipitation observation.

There are three commonly used approaches to representing the subgrid variability of land–atmosphere processes. The mosaic approach (Avissar and Pielke 1989; Koster and Suarez 1992) divides each grid cell into tiles of different vegetation types without considering the physical location of each tile within the grid. Each tile interacts with the atmosphere independently. This approach is flexible in that the number of vegetation types can be arbitrarily set, but is limited in the description of subgrid atmospheric heterogeneity. The fine-mesh approach explicitly represents the surface spatial heterogeneities (Seth et al. 1994; Hahmann and Dickinson 2001) by breaking each atmosphere grid into subgrid cells at the land surface according to location. In simulating the land–atmosphere exchanges, this approach disaggregates the atmospheric forcing from the atmosphere grid to the land surface subgrid cells and aggregates surface fluxes from the subgrid cells to the atmosphere grid. Although it easily incorporates subgrid-scale parameterizations, the fine-mesh approach is expensive in computation. The statistical–dynamical approach is more commonly used for addressing this issue (Entekhabi and Eagleson 1989; Thomas and Henderson-Sellers 1991; Avissar 1992; Bonan et al. 1993; Mölders and Raabe 1996; Liang et al. 1996; Ghan et al. 1997; Ramírezj and Senarath 2000; Wang and Eltahir 2000). With this approach, land surface features are characterized by multiple values within a grid cell, typically represented by some statistical properties (e.g., probability density function) of relevant variables derived from historical data. It is easily implemented and computationally efficient. Large amounts of historical data may be required in order to derive the statistics needed.

The statistical–dynamical approach has been used to investigate the impact on hydrological processes of subgrid heterogeneity in soil moisture (Entekhabi and Eagleson 1989), precipitation (Entekhabi and Eagleson 1989; Thomas and Henderson-Sellers 1991), stomatal conductance (Avissar 1992), infiltration capacity (Wood et al. 1992), and topography (Leung and Ghan 1995). Among these factors, it is widely accepted that the subgrid variability of precipitation plays the dominant role in land surface–atmosphere interactions (Entekhabi and Eagleson 1989; Gao and Sorooshian 1994; Henderson-Sellers and Pitman 1992; Liang et al. 1996; Ghan et al. 1997). When applied to the subgrid variability of precipitation, the statistical–dynamical approach assumes that the model-generated precipitation only falls over a fraction μ of the model grid cell; within this fraction, precipitation is assumed to follow a certain statistical distribution. It has been shown that the choices of precipitation coverage fraction as well as the probability density function (pdf) used to model the precipitation variability within the rain-covered area greatly impact the hydrological processes (e.g., Pitman et al. 1990). It is therefore essential to achieve a realistic estimation of the fraction μ and the corresponding pdf type of rain rates.

Entekhabi and Eagleson (1989) estimated the rainfall coverage fraction using a relationship between the fraction area and the total storm depth. The U.K. Meteorological Office (UKMO) sets μ values according to rainfall types, for example, 0.5 for large-scale rainfall and 0.1 for convective rainfall. In reality, the coverage fraction for a specific rain type depends on such factors as location and season. Even in a single storm, μ varies during the life cycle of the storm (Eltahir and Bras 1993). Pitman et al. (1990) used different μ values (i.e., 0.1, 0.5, and 1.0) to investigate the effect of rainfall coverage fraction on the partitioning between evaporation and runoff over a tropical forest. Their simulation results demonstrated that changing the fraction from 1.0 to 0.1 changes an evaporation-dominated hydrological regime to a runoff-dominated one. Thomas and Henderson-Sellers (1991) also showed that the land surface hydrological processes were very sensitive to the rainfall coverage fraction value. For a greater physical realism, Eltahir and Bras (1993) presented an approach that estimates rainfall coverage fraction as the ratio of model-predicted rainfall intensity to intensity inferred from observations, which allows for temporal and spatial variations of the estimated rainfall coverage fraction. Wang and Eltahir (2000) applied this approach in a regional climate model over West Africa where ground observations of rainfall intensity were available from the Hydrology–Atmosphere Pilot Experiment in the Sahel (HAPEX Sahel). However, the broader application of this approach in climate models has been limited due to the lack of reliable and high-resolution global precipitation data.

Statistical characterization of rain-rate variability within rain-covered areas is another factor affecting the model simulation of hydrological processes. Exponential pdf is frequently assumed for this purpose (Shuttleworth 1988; Pitman et al. 1990; Thomas and Henderson-Sellers 1991). However, the spatial distribution of rain rate would vary with both rain regime and season. To acquire realistic estimates of the rain rates’ pdf, high-resolution global precipitation intensity data are needed.

Global high-resolution rainfall observations are currently available based on passive microwave sensors onboard a number of earth-orbiting platforms such as the Special Sensor Microwave Imager (SSM/I) instrument on the Defense Meteorological Satellite Program (DMSP) satellites, the Advance Microwave Scanning Radiometer (AMSR) aboard the Earth Observation Satellite (EOS) Aqua, and the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI). TRMM also carries the first space-based precipitation radar (PR) providing the most accurate overland rainfall estimates currently available from space (e.g., Kummerow et al. 2000; Grecu and Anagnostou 2002). The above sensors provide an opportunity to derive the conditional mean and pdf of rain rates to be used in climate models.

In this paper, a statistical–dynamical approach is described that incorporates multiyear (2001–03) high-resolution TMI rainfall estimates into the representation of canopy hydrological processes. The TMI rainfall data are used to obtain realistic estimates of μ and pdf. Two types of simulations, one with and one without representation of precipitation subgrid variability, are conducted using the Community Land Model (Dai et al. 2003). The effect of precipitation subgrid variability on hydrological processes is demonstrated through comparing results from the two types of simulations. In section 2, the study region and data used are discussed. Section 3 describes the approach used to improve the parameterization of canopy hydrological processes on the basis of satellite precipitation data. Experiment designs are described in section 4. Results are analyzed in section 5. In section 6 we present the conclusions and discussion.

2. Study region and data

Since canopy interception is the first of a sequence of hydrological processes affected by precipitation subgrid variability, we focus this study on two heavily vegetated tropical regimes: central Africa ranging from 15°S to 5°N, and 10° to 30°E, and the Amazon region ranging from 15°S to 5°N, and 70° to 40°W. The regions are shown in Fig. 1. High-resolution rainfall over these two regions is derived from 3 yr (2001–03) of TRMM satellite observations. TRMM carries a combination of active (PR) and multichannel passive microwave (TMI) sensors. Compared with PR, TMI provides larger swath width (760 km versus 220 km), which leads to more observation samples. TMI rain retrievals over land have been developed over the years (Ferraro and Marks 1995; Conner and Petty 1998; Grecu and Anagnostou 2001; McCollum and Ferraro 2003; Dinku and Anagnostou 2005). The most recent one by Dinku and Anagnostou (2005, hereafter DA05) was demonstrated to provide improved rainfall estimates compared with other techniques. The DA05 algorithm consists of 1) multichannel-based rain screening and convective/stratiform (C/S) classification schemes, and 2) nonlinear (linear) regressions for rain-rate retrieval of stratiform (convective) rain regimes. The technique parameters are determined using matched-resolution TMI brightness temperatures with PR-retrieved rainfall rates and precipitation classification.

DA05 provides extensive evaluation of the algorithm performance as well as comparative statistics with the formal TRMM overland TMI retrieval. Validation results are shown in Figs. 7 and 8 of DA05 for the period of the experiment (2001–03) over central Africa and the Amazon region, respectively. Evaluation statistics show an adequate statistical agreement between DA05 rain estimates and the more definitive PR rain rates.

3. Model description

The Community Land Model (CLM) (Dai et al. 2003) is used in this paper to investigate the effects of precipitation subgrid variability on canopy interception and subsequently on other land surface hydrological processes. The model was developed based on three existing models: the Land Surface Model (LSM) (Bonan 1996), the Biosphere–Atmosphere Transfer Scheme (Dickinson et al. 1993), and the 1994 version of the Chinese Academy of Sciences Institute of Atmospheric Physics Land Surface Model (Dai and Zeng 1997). In addition, it allows other land surface parameterizations to be incorporated into it easily. A more detailed description of the model can be found in Dai et al. (2003). The version used here is otherwise the same as the released version 3.0 except for two modifications. The first concerns the representation of subsurface runoff, for which substantial improvement is needed. To avoid the uncertainty related to the parameterization for subsurface lateral runoff, we assume that all subsurface runoff comes from drainage at the bottom of the soil model. The second modification is directly related to the topic of this study, that is, rainfall interception by vegetation canopy, and is explained in the following.

a. Default interception scheme in the CLM

Canopy water storage is determined by canopy interception, canopy drip, and loss via evaporation:
i1525-7541-6-5-745-e1
where S is the canopy water storage, Ic is the canopy interception rate, Dr is the canopy drip rate, and Ew is the evaporation rate from wet foliage. In the default CLM, the canopy interception is considered as an exponential function of canopy density:
i1525-7541-6-5-745-e2
where Pm is the grid-averaged precipitation intensity (simulated by atmosphere models or derived from reanalysis data), LAI is one-sided leaf area index, and SAI is one-sided stem area index. The canopy drip occurs when canopy water exceeds the water storing capacity:
i1525-7541-6-5-745-e3
where C is the canopy water storing capacity, and Δt is the model time step. The canopy water storing capacity is related to leaf and stem area index of canopy:
i1525-7541-6-5-745-e4
where A is a tunable parameter. In this study, we follow the default CLM and set the parameter A to 0.1 mm, but will modify the canopy interception representation using the methodology described below.

b. Interception scheme with precipitation subgrid variability

To incorporate precipitation subgrid variability into the interception scheme in CLM, we use a statistical–dynamical approach to improve the representation of the canopy interception processes facilitated by high-resolution TMI precipitation data. Following Shuttleworth (1988), for a given rainfall coverage fraction μ and a given pdf of rainfall f (P), the canopy interception at a grid cell (i.e., grid-averaged interception) can be expressed as
i1525-7541-6-5-745-e5
where μ is the rainfall coverage fraction, Imax is the maximum canopy interception rate [Imax= (CS)/Δt], Pm is the grid-averaged unconditional rain rate (simulated by atmosphere models or derived from reanalysis data), Po is the conditional mean rain rate derived from observations, Δt is the model time step, and f (P) is the probability density function of rain rate P.

In our study, Eq. (2) is replaced by Eq. (5) to form a new canopy interception scheme. High-resolution observations of precipitation intensity are required to get realistic estimates of rainfall coverage fraction μ and pdf of rain rate P within the rain-covered area. Supported by the passive microwave rain observations, two commonly used pdf types, the exponential and lognormal, are used to characterize the statistical rain-rate variability over central Africa and the Amazon region. The equations of these pdf types are shown in the appendix. If precipitation follows the exponential pdf, an analytical solution on canopy interception can be derived as shown in the appendix. If precipitation follows the lognormal pdf, numerical integration will be used to solve Eq. (5).

c. Rainfall coverage fraction

As mentioned earlier, the rainfall coverage fraction μ and pdf of rain rate within the rain-covered area are two factors that control canopy interception in our representation of precipitation subgrid variability. Rainfall coverage fraction, though, is expected to play a more significant role compared with how rain rates distribute within rainfall-covered areas (Liang et al. 1996).

In previous studies, μ was often specified as a constant over the globe and through all seasons, for example, 1.0 or 0.5 for large-scale rainfall and 0.1 or 0.3 for convective rainfall (Dolman and Gregory 1992). However, such choice of the parameter value is rather arbitrary, and it has been shown that a model’s climate at the land surface is very sensitive to the choice of μ (Pitman et al. 1990; Johnson et al. 1991; Thomas and Henderson-Sellers 1991). Actually, precipitation features significant spatial and seasonal variability (Chen et al. 1996, 1998; Dai 2001a, b), causing the rainfall coverage fraction to vary spatially and with season. More importantly, the fraction in climate models often depends on the model resolution. A more realistic approach is to estimate the rainfall coverage fraction as the ratio of the model-predicted rainfall intensity to the conditional mean rainfall intensity inferred from observations: μ = Pm/Po (Eltahir and Bras 1993; Collier 1993; Gong et al. 1994; Wang and Eltahir 2000). Here we use this approach to estimate μ and its temporal and spatial variations.

To use this approach, high-resolution observations of rainfall intensity are needed. Here, we used multiyear (2001–03) overland rain estimates at 0.1° resolution derived by the DA05 algorithm. The conditional (rain rate > 0.0) mean values of rain rate were calculated for each month and 1° grid cells of two study areas. The rainfall coverage fraction μ was then calculated taking the ratio of modeled grid-averaged rainfall intensity to the corresponding observation-based conditional mean rain rate. This approach ensures that surface rainfall predicted by the model is delivered to the surface in a rain rate statistically the same as observed. Because the conditional mean rainfall intensities are obtained for the different months and each grid cell, μ values from this approach captures the spatial and seasonal variability of precipitation.

d. Precipitation distribution

There have been contradictory findings regarding the type of precipitation distribution within the rain-covered area and the model sensitivity to different types of distributions. For example, Pitman et al. (1990, 1993) found a high sensitivity of model climate to precipitation spatial distribution, but Liang et al. (1996) documented little difference in their model results between experiments assuming uniformly distributed precipitation and those assuming exponential distribution. This may indicate that the sensitivity to the spatial distribution is model dependent. In our study, two commonly used pdf types (i.e., exponential and lognormal) are used as candidates to represent the experimental pdfs derived from TMI rain rates. Here, the model grids are grouped into two climate regimes: Africa and Amazon. As shown in Table 1, χ2 values for the exponential pdf are lower than those for the lognormal pdf over central Africa (with the exception of the month of February), while χ2 values for the lognormal pdf are smaller for all months at the Amazon region. This is also evident from the comparison between the theoretical and experimental (histograms) pdfs shown in Fig. 2, which takes the month of March as an example over both central Africa and the Amazon region. In February over central Africa, the χ2 value for the lognormal pdf is slightly smaller than that for the exponential pdf. However, since this difference is small, we chose to use the exponential pdf to characterize the rain-rate distribution over central Africa for all seasons.

4. Experiment designs

To investigate the effects of precipitation subgrid variability on land surface hydrological processes, two types of model simulations (control and experiment) are designed. Control simulations use the default canopy interception scheme of CLM, which estimates the canopy interception according to Eq. (2). With this scheme, the canopy interception is not related to the subgrid precipitation variability. It is solely controlled by the grid-averaged precipitation intensity. Experiment simulations account for the impact of precipitation subgrid variability on canopy interception and use Eq. (5) to calculate the canopy interception. As such, the interception depends on both grid-average modeled precipitation intensity and observation-based subgrid variability characterized by rainfall coverage fractions and rain-rate pdf varying for each specific area and season.

Following the data analysis of section 3, the exponential (lognormal) pdf was used in the experiments over the central Africa (Amazon) region. Further, to study the effect of rain-rate pdf type on the hydrological processes, we expanded our experiment to include two pdf types (i.e., exponential, the lognormal) and spatially uniform rain rate within the rain-covered area for each region.

CLM in its stand-alone mode is used in this study to simulate land surface hydrological processes. National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data (Kalnay et al. 1996) is used to force the land surface model. The very issue of unrealistic representation of canopy interception arises from the coarse resolution of the atmosphere model in coupled land–atmosphere simulations. In using the stand-alone version of CLM, the reanalysis data are used as a surrogate for model-simulated atmosphere forcing. Vegetation properties (e.g., LAI) in the default CLM were derived from Advanced Very High Resolution Radiometer (AVHRR) normalized difference vegetation index (NDVI) data. The quality of AVHRR data has been questioned because of factors such as satellite drift and changeover. In our study, we use the LAI derived from Moderate Resolution Imaging Spectroradiometer (MODIS) observations (Tian et al. 2004a, b,c) instead, which is of superior quality compared with AVHRR data.

In both the control and experiment simulations, the time step is set to 1 h. To obtain the initial values of model-state variable (e.g., vegetation temperature, soil temperature, soil moisture), 5 yr of spinup is conducted before running the model. To make the initial conditions consistent with model simulations, we used the default interception scheme and the modified interception scheme in the spinups of the control and the experiment runs, respectively. The simulation period is from 1 January 2001 to 31 December 2001.

5. Results

To explore the effect of the precipitation subgrid variability on hydrological processes, we compare results from the control run (labeled as CTL, which does not allow for the precipitation subgrid variability) with results from the experiment run (labeled as EXP, which includes the precipitation subgrid variability). Our results analyses focus on water balance, soil water conditions, surface temperature, and heat fluxes over both central Africa and the Amazon region. Similarities and differences between the two regions are identified and analyzed.

a. Over central Africa

1) Water and energy balance

Figures 3 and 4 present the runoff, the total evapotranspiration (ET) and the three components of ET in CTL and EXP, respectively. All the variables we present are annual averages. Rainfall is intercepted by vegetation canopy before reaching the ground. Under uniform precipitation distribution over the whole model grid cell, the grid-averaged rain rate (NCEP–NCAR reanalysis precipitation) is much smaller than the observed rain rate. Given the same canopy water storing capacity as in EXP, much more water is intercepted by the canopy in CTL and consequently less becomes throughfall. Substantial underestimation of the amount of water that reaches the ground results in significant underestimation of runoff if the subgrid variability of precipitation is neglected (Fig. 3a versus Fig. 4a). As a result, the total evapotranspiration is overestimated (Fig. 3b versus Fig. 4b). Corresponding to the overestimation of total evapotranspiration, the latent heat flux is overestimated under spatially uniform rain rate (Fig. 5c versus Fig. 5d), and the sensible heat flux is underestimated (Fig. 5a versus Fig. 5b), especially around the equator where the vegetation is very dense.

Precipitation subgrid variability significantly impacts the partitioning between runoff and total ET as well as the partitioning of ET among interception loss, ground evaporation, and plant transpiration, especially over heavily vegetated areas. Comparing Figs. 3c–e with Figs. 4c–e reveals a dramatic difference in the three components of ET between CTL and EXP. Under spatially uniform rain rate over the whole model grid cell, substantial overestimation of vegetation evaporation arises from overestimation of canopy interception. Underestimation of ground evaporation results primarily from the reduction in water reaching the ground, causing less water storage in the top soil layer. Transpiration is determined by many factors such as soil water stress, solar radiation, humidity deficit between vegetation and canopy air, and the wetted fraction of foliage. As explained later, underestimation of transpiration in CTL is caused by smaller humidity deficit and larger fraction of wetted leaf compared with those in EXP. Table 2 presents quantitative comparison between CTL and EXP of the domain averages of hydrological variables. For example, including the precipitation subgrid variability changes the runoff averaged over the whole domain from 7.7 × 10−6 to 19.8 × 10−6 mm s−1.

Although much less water reaches the ground when precipitation subgrid variability is not considered, the water within the 10 soil layers (totaling 3.44 m) is only slightly underestimated (Figs. 6a and 6b). If we take the first meter of soil as plant root zone, the soil water amount in the top 1 m reflects plant water availability. There is no big difference in the root zone soil moisture between CTL and EXP either, as shown in Figs. 6c and 6d.

To better understand the difference in soil water, two grid points were chosen as examples to illustrate how the soil water changes from CTL to EXP, one with a low LAI of 1.8 at 13.5°S, 21.5°E and one with a large LAI of 5.4 at 2.5°S, 23.5°E. Figure 7 presents seasonal cycles of soil water and transpiration at these two sites. For the low-LAI site, the water within the total 10 soil layers from EXP is slightly larger than that from CTL (Fig. 7a). The top 1-m soil water from the two runs is almost the same (Fig. 7b). Small but considerable difference in soil water is simulated for the high-LAI site (Figs. 7d and 7e), as high as approximately 10%–15% relative to the soil water in CTL. Soil water is one of the factors that determine transpiration. Since the difference in soil water between CTL and EXP is small, water stress in the soil is not the primary cause for the transpiration difference between the two runs shown in Figs. 7c and 7f. Besides water stress, humidity deficit between vegetation and canopy air as well as the leaf area available for photosynthesis also control the transpiration. In CTL (i.e., under spatially uniform rain rate), the lower vegetation temperature (Figs. 8b and 8e) results in lower humidity of the air within stomatas, and more intercepted rainfall causes moister canopy air, which leads to smaller humidity deficit compared with EXP. At the same time, the larger wetted leaf area leaves less dry leaf area to photosynthesize. The transpiration difference between CTL and EXP in our study is primarily attributed to differences in humidity deficit and wetted leaf area.

2) Seasonal cycles of temperature and heat fluxes

The precipitation subgrid variability does not impact the seasonal cycles of vegetation temperature and ground temperature, even though it changes the temperature value of each month. This statement holds for both sparse (13.5°S, 21.5°E) and dense canopy (2.5°S, 23.5°E) sites, as shown in Figs. 8b and 8e. However, there are big differences in seasonal cycles of sensible and latent heat fluxes between CTL and EXP. For the sparse canopy site (Fig. 8c), the minimum latent heat flux occurs in July in both CTL and EXP, while the maximum occurs in different months in the two runs. Both the minimum and maximum values of sensible heat flux vary between the two runs. For the dense canopy site (Fig. 8f), there is no obvious seasonal cycle in either CTL or EXP. However, including precipitation subgrid variability changes the magnitude of heat fluxes. For the two kinds of sites, the seasonal cycle of latent heat flux is closely related to the leaf index area seasonality. As a result, the stronger seasonality of LAI at the sparse canopy site results in the stronger seasonality of heat fluxes, compared with the dense canopy site.

3) Effects of LAI and rainfall coverage fraction

The differences of hydrological variables between CTL and EXP depend on LAI and rainfall coverage fraction. In the following, we take the interception loss ratio (i.e., the ratio of interception loss to precipitation) as an example to investigate the effects of LAI and rainfall coverage fraction on the differences. Figure 9a shows the relationship between the difference in the interception loss ratio and the LAI, and Fig. 9b shows the relationship between the difference and the rainfall coverage fraction. To partly eliminate the compensatory impact, we divided μ into three groups when investigating the LAI effect and divided LAI into three groups when investigating the μ effect. It is evident that the LAI and the rainfall coverage fraction affect the difference in the interception loss ratio differently. The larger the LAI is, the larger the interception loss ratio is in both CTL and EXP. However, the interception loss ratio in CTL increases with LAI faster than that in EXP, leading to the increase of the difference between CTL and EXP with LAI. The rainfall coverage fraction μ affects the interception loss ratio in EXP. Large μ value favors large interception loss ratio in EXP, causing the difference between CTL and EXP to decrease with μ.

However, the LAI and the rainfall coverage fraction μ are not independent. It is therefore necessary to examine the relationship between the LAI and the rainfall coverage fraction in order to understand their respective impact on the EXP–CTL difference in the interception loss ratio. It is apparent from Fig. 9c that μ increases with increase of LAI. As a result, the LAI effect can offset the effect of rainfall coverage fraction in Fig. 9b, even though LAI was divided into three different groups. To further demonstrate the effect of the rainfall coverage fraction on the difference, we normalized the difference in the interception loss ratio by LAI, thus eliminating the effect of LAI. It is obvious that there is a negative relationship between the normalized difference in interception loss ratio and the rainfall coverage fraction (Fig. 9d). The difference in interception loss ratio is more sensitive to changes in the rainfall coverage fraction when the rainfall coverage fraction is small, and this sensitivity decreases as μ becomes larger.

b. Over the Amazon region

The comparisons of water balance, surface temperatures seasonality and heat fluxes seasonality between CTL and EXP over the Amazon region are similar to those over central Africa. For example, Fig. 10 shows the comparison between CTL and EXP in the evapotranspiration ratio (the ratio of total evapotranspiration to precipitation), the interception loss ratio (the ratio of interception loss to precipitation), and the plant transpiration ratio (the ratio of transpiration to precipitation). Similar to the results for central Africa, neglecting the precipitation subgrid variability leads to overestimated interception loss and total evapotranspiration as well as underestimated plant transpiration. The interception loss ratio over the forest region is ∼50% in the CTL and is ∼20% in the EXP. The latter is much closer to observations (Shuttleworth 1988). Water within the 10 soil layers (totaling 3.44 m) and top 1-m soil are almost the same in EXP as in CTL over the sparse canopy site (Figs. 11a and 11b). At the dense canopy site, differences in soil water are very small (Figs. 11d and 11e). Figure 12 presents the effects of LAI and rainfall coverage fraction μ on the difference between CTL and EXP in the interception loss ratio over the Amazon region. Comparing Fig. 9 with Fig. 12 indicates that the difference varies with LAI and μ in the Amazon region in a similar way to that in central Africa. We notice however that the relationship between the normalized difference in interception loss ratio and the rainfall coverage fraction over the Amazon region is more scattered than that over central Africa (Fig. 9d versus Fig. 12d).

The differences between CTL and EXP in water balance, temperatures, and heat fluxes are smaller over the Amazon region than over central Africa. Table 2 shows a more quantitative comparison averaged over the two model domains. These differences between central Africa and the Amazon region can be attributed to their differences in LAI and in rainfall coverage fraction. Figure 13 presents the yearly mean LAI and rainfall coverage fraction over the two regions. In general, the LAI over the Amazon region is slightly larger than that over central Africa (Fig. 13a versus Fig. 13b). However, the rainfall coverage fraction over the Amazon region is much larger than the fraction over central Africa (Fig. 13c versus Fig. 13d). As analyzed before, the larger the LAI is, the larger the difference in hydrological variable between CTL and EXP. In contrast, the larger the rainfall coverage fraction is, the smaller the difference between CTL and EXP. When the difference in LAI is small, the rainfall coverage fraction plays the dominant role in determining the difference between the two runs. It is therefore apparent that differences in hydrological variables are much smaller over the Amazon region than central Africa.

c. Impact of rain-rate subgrid distribution

The rainfall coverage fraction and the subgrid distribution of rain rate within the fraction are the two major factors that impact the hydrological processes when precipitation subgrid variability is considered. Previous studies demonstrated that the impact of rainfall coverage fraction is dominant, while the impact of subgrid distribution is a secondary factor. To investigate the effect of rain-rate subgrid distribution on the hydrological processes, we conducted two additional simulations for each region: one with a lognormal pdf and one with a spatially uniform rain rate within the rain-covered area to compare with the simulation using exponential pdf over central Africa; one with an exponential pdf and one with a spatially uniform rain rate within the rain-covered area to compare with the simulation using lognormal pdf over the Amazon region. Note that the spatially uniform rain rate here is only applied to the rain-covered fraction of the grid cell, while in the control simulations (sections 5a and 5b) rainfall is uniformly distributed over the whole grid cell.

Comparing Fig. 14a with Fig. 14b indicates that the interception loss ratio from the simulation with the exponential pdf is similar to that with the lognormal pdf in both the spatial pattern and the magnitude over central Africa. Interception loss ratio from the simulation with the spatially uniform rain rate is significantly larger than that associated with the exponential and lognormal pdfs (Fig. 14a versus Fig. 14c and Fig. 14b versus Fig. 14c), even though the spatial patterns of the ratio are similar. Similar conclusions can be drawn over the Amazon region (Figs. 14d–f).

6. Discussion and conclusions

In this paper, we used a multiyear high-resolution (0.1°) rainfall dataset to examine the implication of representing rain-rate subgrid variability in the land surface component of climate models using a statistical–dynamical approach. The rain rates are retrieved from Tropical Rainfall Measuring Mission PR-calibrated TMI observations. The study is conducted over two densely vegetated tropical regions (central Africa and Amazon). The impact of rain-rate subgrid variability on the hydrological cycle starts with the canopy interception processes, which subsequently influences the partitioning between runoff and total evapotranspiration as well as the partitioning among the three components of evapotranspiration. Further, it impacts soil moisture, surface temperature, and surface heat fluxes.

Realistic estimation of rainfall coverage fraction is an important aspect in climate model simulations when precipitation subgrid variability is included. We estimated the fractional coverage as the ratio of model-predicted rain intensity (for which the NCEP–NCAR reanalysis precipitation is used as a surrogate) to intensity derived from TMI observations. The rainfall coverage fraction here is merely a scaling factor for rainfall intensity (scaling it from grid average to local intensity). It is comparable with the observed fraction only if the model-predicted rainfall amount over a whole grid cell is close to the observed rainfall amount. This approach guarantees that model-simulated rainfall is delivered to the land surface in a realistic intensity regardless of the total amount.

By comparing results from the control runs (which treat rainfall as uniformly distributed over the entire grid cell) with those from the experiment runs (which include the rainfall subgrid variability), we found that including rainfall subgrid variability significantly alters the surface water balance, favoring more transpiration (and ground evaporation) and less interception loss. The difference in interception loss ratio between the control and experiment runs is the largest among all hydrological variables. The impact of rainfall subgrid variability on land surface hydrological processes is consistent between central Africa and the Amazon region. However, differences between the two regions in the magnitude of the impact do exist. These differences may have risen from differences in LAI and rain coverage fraction.

Because of the lack of interception loss data, model simulations were not compared with observations. However, we do note that, over the Amazon forest region, the interception loss ratio in the simulation with subgrid variability is approximately 20%–25%. This magnitude is more comparable with published site data [∼12.5% in Shuttleworth (1988)] than the 50% in the simulation without subgrid precipitation variability. Our modification of the canopy hydrological representation based on satellite observations is therefore considered an improvement of the model parameterization.

The rain-rate pdf type within the rain-covered area is of importance in determining the hydrological processes. Unsuitable pdf models may lead to unrealistic representation of subgrid variability, causing inaccurate simulation of hydrological processes. If long-term observations of rain rate are available, statistical analysis of these data is the best way to derive climatological pdfs.

Because including the precipitation subgrid variability significantly alters the surface water balance, this study has hydrological implications for river flow forecast and agriculture. Flows from large rivers can be forecasted using land surface models coupled with river routing models. Neglecting the precipitation subgrid variability gives rise to underestimation of surface runoff, and therefore, underestimated river flow. In addition, over regions where strong seasonality of precipitation necessitates irrigation during the dry season, the underestimated river flow leads to underestimated water storage in reservoirs. As a result, although soil moisture within plant root zone is similar between the simulations with and without representation of subgrid variability (Figs. 6, 7 and 11), water availability for irrigation is not. Neglecting rainfall subgrid variability underestimates irrigation-available water, causing false alarms in water resources planning and management.

In this study, we examined the impact of including the rain-rate subgrid variability on land surface hydrological processes using a stand-alone land surface model. However, as the land surface feeds back to the atmosphere, such impact will further propagate into the atmospheric hydrological cycle (Wang and Eltahir 2000; Hahmann 2003), which would further influence the surface hydrological processes. We are currently working on investigating this issue using coupled land surface–atmosphere models. Such study will help us assess the impact of rain-rate subgrid variability on the full water cycle through interactions between land surface and the overlying atmosphere.

Acknowledgments

This study was supported by the NASA Global Water and Energy Cycle Program (NAG511527) (to Dr. Anagnostou) and by the NOAA GEWEX Americas Prediction Project (NA030AR4310080) (to Dr. Wang). Mr. Tufa Dinku conducted the calibration/validation of the TMI rain retrievals. TRMM data were provided by NASA DAAC.

REFERENCES

  • Avissar, R., 1992: Conceptual aspects of a statistical–dynamical approach to represent landscape subgrid-scale heterogeneities in atmospheric models. J. Geophys. Res., 97 , 27292742.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Avissar, R., , and Pielke R. A. , 1989: A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Mon. Wea. Rev., 117 , 21132136.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., 1996: A land surface model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: Technical description and user’s guide. NCAR Tech. Note NCAR/TN-417+STR, Boulder, CO, 150 pp.

  • Bonan, G. B., , Pollard D. , , and Thompson S. L. , 1993: Influence of subgrid-scale heterogeneity in leaf area index, stomatal resistance, and soil moisture on grid-scale land–atmosphere interactions. J. Climate, 6 , 18821897.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, M. X., , Dickinson R. E. , , Zeng X. , , and Hahmann A. N. , 1996: Comparison of precipitation observed over the continental United States to that simulated by a climate model. J. Climate, 9 , 22332249.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chen, M. X., , Zeng X. , , and Dickinson R. E. , 1998: Adjustment of GCM precipitation intensity over the United States. J. Appl. Meteor., 37 , 876887.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Collier, C. G., 1993: The application of a continental-scale radar database to hydrological process parameterization within atmospheric general circulation models. J. Hydrol., 142 , 301318.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Conner, M. D., , and Petty G. R. , 1998: Validation and intercomparison of SSM/I rain-rate retrieval methods over the continental Unites States. J. Appl. Meteor., 37 , 679700.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dai, A. G., 2001a: Global precipitation and thunderstorm frequencies. Part I: Seasonal and interannual variations. J. Climate, 14 , 10921111.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dai, A. G., 2001b: Global precipitation and thunderstorm frequencies. Part II: Diurnal variations. J. Climate, 14 , 11121123.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dai, Y. J., , and Zeng Q. C. , 1997: A land surface model (IAP94) for climate studies, Part I: Formation and validation in off-line experiments. Adv. Atmos. Sci., 14 , 433460.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dai, Y. J., and Coauthors, 2003: The Common Land Model. Bull. Amer. Meteor. Soc., 84 , 10131023.

  • Dickinson, R. E., , Henderson-Sellers A. , , and Kennedy P. J. , 1993: Biosphere–Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Model. NCAR Tech. Note NCAR/TN-387+STR, Boulder, CO, 72 pp.

  • Dinku, T., , and Anagnostou E. N. , 2005: Regional differences in overland rainfall estimation from PR-calibrated TMI algorithm. J. Appl. Meteor., 44 , 189205.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dolman, A. J., , and Gregory D. , 1992: The parameterization of rainfall interception in GCMs. Quart. J. Roy. Meteor. Soc., 118 , 455467.

  • Eltahir, E. A. B., , and Bras R. L. , 1993: Estimation of the fractional coverage of rainfall in climate models. J. Climate, 6 , 639644.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Entekhabi, D., , and Eagleson P. S. , 1989: Land surface hydrology parameterization for atmospheric general circulation models including subgrid scale spatial variability. J. Climate, 2 , 816831.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ferraro, R. R., , and Marks G. F. , 1995: The development of SSM/I rain-rate retrieval algorithms using ground-based radar measurements. J. Atmos. Oceanic Technol., 12 , 755770.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gao, X. G., , and Sorooshian S. , 1994: A stochastic precipitation disaggregation scheme for GCM application. J. Climate, 7 , 238247.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ghan, S. J., , Liljigren J. C. , , Shaw W. J. , , Hubbe J. H. , , and Dorman J. C. , 1997: Influence of subgrid variability on surface hydrology. J. Climate, 10 , 31573166.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gong, G., , Entekhabi D. , , and Salvucci G. D. , 1994: Regional and seasonal estimates of fractional storm coverage based on station precipitation observations. J. Climate, 7 , 14951505.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grecu, M., , and Anagnostou E. N. , 2001: Overland precipitation estimation from TRMM passive microwave observation. J. Appl. Meteor., 40 , 13671380.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grecu, M., , and Anagnostou E. N. , 2002: Use of passive microwave observations in a radar rainfall-profiling algorithm. J. Appl. Meteor., 41 , 702715.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hahmann, A. N., 2003: Representing spatial subgrid-scale precipitation variability in a GCM. J. Hydrometeor., 4 , 891900.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hahmann, A. N., , and Dickinson R. E. , 2001: A fine-mesh land approach for general circulation models and its impact on regional climate. J. Climate, 14 , 16341646.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Henderson-Sellers, A., , and Pitman A. J. , 1992: Land-surface schemes for future climate models: Specification, aggregation, and heterogeneity. J. Geophys. Res., 97 , 26872696.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Johnson, K. D., , Entekhabi D. , , and Eagleson P. S. , 1991: The implementation and validation of improved land surface hydrology in an atmospheric general circulation model. Ralph M. Parsons Laboratory Tech. Rep. 334, Massachusetts Institute of Technology, Cambridge, MA, 192 pp.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Koster, R. D., , and Suarez M. J. , 1992: Modeling the land surface boundary in climate models as a composite of independent vegetation stands. J. Geophys. Res., 97 , 26972715.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kummerow, C., and Coauthors, 2000: The status of the Tropical Rainfall Measuring Mission (TRMM) after two years in orbit. J. Appl. Meteor., 39 , 19651982.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Leung, L. R., , and Ghan S. J. , 1995: A subgrid parameterization of orographic precipitation. Theor. Appl. Climatol., 52 , 95118.

  • Liang, X., , Lettenmaier D. P. , , and Wood E. F. , 1996: One-dimensional statistical dynamic representation of subgrid spatial variability of precipitation in the two-layer variable infiltration capacity model. J. Geophys. Res., 101 , 2140321422.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McCollum, J. R., , and Ferraro R. R. , 2003: Next generation of NOAA/NESDIS TMI, SSM/I, and AMSR-E microwave land rainfall algorithms. J. Geophys. Res., 108 .8382, doi:10.1029/2001JD001512.

    • Search Google Scholar
    • Export Citation
  • Mölders, N., , and Raabe A. , 1996: Numerical investigations on the influence of subgrid-scale surface heterogeneity on evapotranspiration and cloud processes. J. Appl. Meteor., 35 , 782795.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pitman, A. J., , Henderson-Sellers A. , , and Yang Z. L. , 1990: Sensitivity of regional climates to localized precipitation in global models. Nature, 346 , 734737.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pitman, A. J., , Yang Z. L. , , and Henderson-Sellers A. , 1993: Sub-grid scale precipitation in AGCMs: Reassessing the land surface sensitivity using a single column model. Climate Dyn., 9 , 3341.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ramírez, J. A., , and Senarath S. U. S. , 2000: A statistical–dynamical parameterization of interception and land surface–atmosphere interactions. J. Climate, 13 , 40504065.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Seth, A., , Giorgi F. , , and Dickinson R. E. , 1994: Simulating fluxes from heterogeneous land surfaces: Explicit subgrid method employing the Biosphere–Atmosphere Transfer Scheme (BATS). J. Geophys. Res., 99 , 1865118667.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shuttleworth, W. J., 1988: Macrohydrology—The new challenge for process hydrology. J. Hydrol., 100 , 3156.

  • Thomas, G., , and Henderson-Sellers A. , 1991: An evaluation of proposed representation of subgrid hydrological processes in climate models. J. Climate, 4 , 898910.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Tian, Y., and Coauthors, 2004a: Comparison of seasonal and spatial variations of LAI/FPAR from MODIS and Common Land Model. J. Geophys. Res., 109 .D01103, doi:10.1029/2003JD003777.

    • Search Google Scholar
    • Export Citation
  • Tian, Y., , Dickinson R. E. , , Zhou L. , , Myneni R. B. , , Friedl M. , , Schaaf C. B. , , Carroll M. , , and Gao F. , 2004b: Land boundary conditions from MODIS data and consequences for the albedo of a climate model. Geophys. Res. Lett., 31 .L05504, doi:10.1029/2003GL019104.

    • Search Google Scholar
    • Export Citation
  • Tian, Y., , Dickinson R. E. , , Zhou L. , , and Shaikh M. , 2004c: Impact of new land boundary conditions from Moderate Resolution Imaging Spectroradiometer (MODIS) data on the climatology of land surface variables. J. Geophys. Res., 109 .D20115, doi:10.1029/2003JD004499.

    • Search Google Scholar
    • Export Citation
  • Wang, G. L., , and Eltahir E. A. B. , 2000: Modeling the Biosphere–Atmosphere System: The impact of the subgrid variability in rainfall interception. J. Climate, 13 , 28872899.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wood, E. F., , Lettenmaier D. P. , , and Zartarian V. G. , 1992: A land-surface hydrology parameterization with subgrid variability for general circulation models. J. Geophys. Res., 97 , 27172728.

    • Crossref
    • Search Google Scholar
    • Export Citation

APPENDIX

Pdfs of Precipitation

The pdf of the precipitation P within the rain-covered area that follows the exponential distribution can be expressed by
i1525-7541-6-5-745-ea1
where P is the precipitation intensity derived from observation, Pm is the precipitation intensity from atmosphere model, μ is the rain coverage fraction. Given a exponential distribution f (P), the grid-averaged rate of canopy interception can be derived (Shuttleworth 1988) as
i1525-7541-6-5-745-ea2
The pdf of the precipitation P within the rain-covered area that follows the lognormal distribution can be expressed by
i1525-7541-6-5-745-ea3
where
i1525-7541-6-5-745-eqa1
P is averaged precipitation P, and σp is the standard deviation of precipitation P.

Fig. 1.
Fig. 1.

Study region. On the right is the central Africa region, ranging from 15°S to 5°N and 10° to 30°E; on the left is the Amazon region ranging from 15°S to 5°N and 70° to 40°W.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 2.
Fig. 2.

Comparison of rain rates between the theoretical pdfs and the histogram based on observation over (a) the central Africa and (b) the Amazon region. Lines with stars represent exponential pdf, and lines with circles represent lognormal pdf. The unit of rain rate is mm day−1.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 3.
Fig. 3.

Water balance over central Africa in CTL simulation. All quantities as a fraction of precipitation: (a) runoff (including surface runoff and base runoff), (b) total evapotranspiration, (c) interception loss, (d) ground evaporation , and (e) plant transpiration.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 4.
Fig. 4.

Same as Fig. 3, but for Africa EXP simulation.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 5.
Fig. 5.

Heat fluxes over central Africa in CTL and EXP. (a) Sensible heat flux in CTL, (b) sensible heat flux in EXP, (c) latent heat flux in CTL, and (d) latent heat flux in EXP. The unit is W m−2.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 6.
Fig. 6.

Central Africa: Soil water within the 10 layers (totaling 3.44 m) in (a) CTL and in (b) EXP. Soil water within the top 1 m in (c) CTL and in (d) EXP. The unit is kg m−2.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 7.
Fig. 7.

Central Africa at a low-LAI (LAI = 1.8) site: Seasonal cycles of (a) soil water within the 10 layers (totaling 3.44 m), and (b) soil water within the top 1 m; and (c) transpiration and rainfall. (d), (e), and (f) Same as (a), (b), and (c), but at a high-LAI (LAI = 5.4) site. (Solid lines with circles: CTL; solid lines with triangles: EXP; solid lines: rainfall).

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 8.
Fig. 8.

Central Africa at a low-LAI (LAI = 1.8) site: Seasonal cycles of (a) LAI, (b) vegetation and ground temperatures, and (c) sensible and latent heat fluxes. (d), (e), and (f) Same as (a), (b), and (c), but at a high-LAI (LAI = 5.4) site. Dash–dot lines: vegetation temperature in CTL; solid lines: vegetation temperature in EXP; dash–dot lines with circles: ground temperature in CTL; solid lines with circles: ground temperature in EXP in (b) and (e). Dash–dot lines: sensible heat flux in CTL; solid lines: sensible heat flux in EXP; dash–dot lines with circles: latent heat flux in CTL; solid lines with circles: latent heat flux in EXP in (c) and (f).

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 9.
Fig. 9.

Central Africa: The impacts of LAI and rainfall coverage fraction on the EXP−CTL difference in interception loss ratio. (a) Difference vs LAI, (b) difference vs rainfall coverage fraction, (c) rainfall coverage fraction vs LAI, and (d) normalized difference (EXP–CTL difference in the interception loss ratio divided by LAI) vs rainfall coverage fraction.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 10.
Fig. 10.

Water balance over the Amazon region. All quantities as a fraction of precipitation: (a) total evapotranspiration in CTL; (b) interception loss in CTL; (c) plant transpiration in CTL; (d) total evapotranspiration in EXP; (e) interception loss in EXP; and (f) plant transpiration in EXP.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 11.
Fig. 11.

Same as Fig. 7, but for the Amazon region.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 12.
Fig. 12.

Same as Fig. 9, but for the Amazon region.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 13.
Fig. 13.

LAI over (a) the central Africa and (b) the Amazon region. Rainfall coverage fraction over (c) the central Africa and (d) the Amazon region

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Fig. 14.
Fig. 14.

Interception loss ratio under (a) exponential pdf, (b) lognormal pdf, and (c) spatially uniform rain rate within the rain-covered area over the central Africa region. (d), (e), and (f) Same as (a), (b), and (c) but for the Amazon region.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM438.1

Table 1.

The χ2 test values of fit goodness for each month over both the central Africa and the Amazon region. The numbers of degree of freedom are 18 and 17 for the exponential pdf and the lognormal pdf, respectively

Table 1.
Table 2.

Comparison between the control simulation and the experiment simulation in water and energy balance over the central Africa region. All the variables are domain averages. CTL is the control simulation. EXP is the experiment simulation.

Table 2.
Save