Journal of Geophysical Research: Biogeosciences

Representing a new MODIS consistent land surface in the Community Land Model (CLM 3.0)

Authors


Abstract

[1] Recently a number of studies have found significant differences between Moderate Resolution Imaging Spectroradiometer (MODIS) land surface mapping and the land surface parameters of the Community Land Model (CLM) of the Community Climate System Model (CCSM). To address these differences in land surface description, we have developed new CLM 3.0 land surface parameters that reproduce the physical properties described in the MODIS land surface data while maintaining the multiple Plant Functional Type (PFT) canopy and herbaceous layer representation used in CLM. These new parameters prescribe crop distributions directly from historical crop mapping allowing cropping to be described in CLM for any year from 1700 to current day. The new model parameters are calculated at 0.05 degrees resolution so they can be aggregated and used over a wider range of model grid resolutions globally. Compared to the current CLM 3.0 parameters, the new parameters have an increase in bare soil fraction of 10% which is realized through reduced tree, shrub, and crop cover. The new parameters also have area average increases of 10% for leaf area index (LAI) and stem area index (SAI) values, with the largest increases in tropical forests. The new land surface parameters have strong repeatable impacts on the climate simulated in CCSM 3.0 with large improvements in surface albedo compared to MODIS values. In many cases the improvements in surface albedo directly resulted in improved simulation of precipitation and near-surface air temperature; however, for the most part the existing biases of CCSM 3.0 remained with the new parameters. Further analysis of changes in surface hydrology revealed that the increased LAI of the new parameters resulted in lower overall evapotranspiration with reduced precipitation in CCSM 3.0. This was an unexpected result given that other research into the impacts of vegetation change suggests that the new parameters should have the opposite impact. This suggests that while the new parameters significantly improve the climate simulated in CLM 3.0 and CCSM 3.0, the new surface parameters have limited success in rectifying surface hydrology biases that result from the parameterizations within the CLM 3.0.

1. Introduction

[2] The Moderate Resolution Imaging Spectroradiometer (MODIS) instrument on NASA's Terra and Aqua satellites provides new global land surface data from imagery with spatial, temporal and spectral resolutions that was not previously available [Justice et al., 2002]. Tian et al. [2004a], Oleson et al. [2003], Wang et al. [2004] and Tian et al. [2004c] have all identified significant differences between the global land surface mapping products from MODIS, and the current land surface parameters of the Community Land Model (CLM) used with the Community Climate System Model (CCSM), and with the closely related Common Land Model (CoLM). Specifically these investigations identified substantial differences in Vegetation Cover, Leaf Area Index (LAI), Fraction of Photosynthetically Active Radiation (FPAR), and land surface albedo.

[3] To address these differences in land surface description between the CLM 3.0 and MODIS, we have developed new CLM 3.0 land surface parameters that reproduce the physical properties described in the new MODIS land surface data as closely as possible while maintaining the multiple Plant Functional Type (PFT) canopy and herbaceous layer representations of Bonan et al. [2002]. These new PFT mapping methods differ substantially from the new MODIS PFT mapping described by Tian et al. [2004a], with the new methods capturing multiple PFTs for each grid cell allowing the representation of complex vegetation through mixed canopy and herbaceous layers. This is essential for savannas, non intensive croplands, mixed forests and other mosaic landscapes that cannot be represented with a single PFT. The new model parameters also are calculated at the substantially finer resolution of 0.05 degrees so they can be aggregated and used over a wider range of model grid resolutions as well as for use in finer-scale land cover change experiments.

[4] The new parameters provide consistency with the MODIS Vegetation Continuous Fields data set from Hansen et al. [2003] for vegetation cover, while prescribing crop distributions directly from the historical cropping data set of Ramankutty and Foley [1999]. This direct prescription of cropping distributions has the added advantage of providing a consistent method for describing historical cropping in CLM for any year from 1700 to current day. The new parameters provide consistency with average monthly MODIS LAI through new monthly PFT LAI values that combine with the new PFT distributions to reproduce the average monthly MODIS LAI values at the 0.05 degree resolution. The new parameters also reduce differences in broadband surface albedo compared to average monthly MODIS values through new soil reflectance properties calculated directly from the MODIS data.

[5] Tian et al. [2004a, 2004b] have demonstrated that differences in land surface representation between the CLM and MODIS, especially land surface albedo, do significantly affect the climate simulated in the CLM and CCSM models. The importance in realistically describing land surface properties in CLM and CCSM is consistent with work by Sellers [1997], Pielke et al. [1998], Betts et al. [1996], Field and Avissar [1998] and others, that describe the complex relationships between vegetation cover, land surface properties, and atmospheric processes. This work has shown that these processes are highly sensitive to changes in the fluxes of radiation, momentum, energy, and moisture, between the soil, vegetation, and atmosphere. Further work on land surface and climate interactions by Feddema et al. [2005], Chase et al. [2001], Lawton et al. [2001], Zhao et al. [2001] and others, has demonstrated that the local changes in surface fluxes interact with larger-scale atmospheric conditions to impact weather and climate remotely through changes in global atmospheric circulation and moisture transport.

[6] To assess how the new MODIS consistent land surface parameters impact the surface energy balance, hydrology and atmospheric fluxes in CLM 3.0 and through that the larger-scale climate modeled in CCSM 3.0 we performed a number of global climate sensitivity experiments with the new parameters compared to the current CLM 3.0 parameters. The climate modeled in the sensitivity experiments was compared with globally observed climate data to evaluate the climate impacts of new land surface parameters relative to existing climate biases in the models with the current CLM 3.0 parameters.

[7] This paper provides an overview of the new data and methods used to generate the new finer-scale CLM 3.0 land surface parameters from MODIS land surface data as well as an assessment of the impacts of the new parameters on the climate simulated in CCSM 3.0. More detailed descriptions of the new data and methods as well as comparisons of the new land surface parameters to the current CLM 3.0 parameters are given by Lawrence and Chase [2006a]. This paper also contains a summary of the most important findings of the climate sensitivity experiments performed with the new land surface parameters and the current CLM 3.0 parameters. A more detailed assessment of these experiments is given by Lawrence and Chase [2006b].

2. Methods

2.1. NCAR Community Land Model

[8] The newest version of the NCAR Community Climate System Model (CCSM 3.0) includes a new standard land surface parameterization, the Community Land Model (CLM 3.0), described by Oleson et al. [2004]. The new model represents an improvement over previous land surface representations in the family of community climate models, as the CLM 3.0 simulates the land surface more realistically both in the description of physical processes and in the use of more representative land surface boundary data. Both of these changes helped to improve the overall climate simulation of the CLM and the CCSM [Bonan et al., 2002; Dickinson et al., 2006].

[9] In the prescribed configuration of CLM 3.0, the distribution and the seasonal cycles of vegetation and other land surface properties are directly specified through preprocessed land surface parameters. The CLM land surface model allows subgrid heterogeneity to be prescribed through fractional allocation of land cover to four or more Plant Functional Types (PFTs), and through prescribing the percentage of each grid cell occupied by ocean, lakes, wetlands, and glaciers. The properties of each of the subgrid land fractions are described through both monthly varying grid cell specific PFT parameters, as well as time and space invariant PFT and soil column parameters.

2.2. Experimental Design

[10] The new MODIS consistent land surface parameters were generated at 0.05 degrees and aggregated to the 0.5 degrees grid resolution of the land surface parameters distributed with CLM 3.0, and to the T42 CLM and CCSM model grid resolution used in our climate sensitivity experiments. The aggregation of the finescale data to the model resolutions used linear averaging of vegetated and bare grid cells for PFT and soil color data, and PFT weighted averaging for monthly PFT LAI, SAI and canopy height values based on the PFT percentages and PFT groupings in each 0.05 degree grid cell.

[11] To provide soil texture data at the 0.05 degree resolution, sand and clay percentages were directly prescribed from the global soil profile texture data sets of Reynolds et al. [1999]. These data sets were derived from the FAO/UNESCO Digital Soil Map of the World [Food and Agriculture Organization, 1995] using the depth and profile methods used by Batjes [1997]. This is the same base data set and methods used to generate the IGBP Global Soil Data Task 2000 used to generate the current CLM 3.0 parameters. The resulting finescale soil texture data set had the same spatial distribution and profile of sand and clay composition but at a finer resolution than the current CLM 3.0 T42 soil parameters [Lawrence and Chase, 2006a].

[12] The new MODIS parameters were assessed against the existing CLM 3.0 land surface parameters at the 0.5 degree resolution to determine the differences and similarities of the new land surface data and new parameter generation methods to the current parameters. Global maps of new and current CLM 3.0 parameters were produced for a range of parameters to show similarities and differences in global distributions between the two sets of parameters, and in the case of LAI against MODIS values. Changes in surface parameters over all land, and the three analysis regions of: Amazon, Boreal, and Sahara and Arabia were tabulated to assess the overall changes in parameter values.

[13] The climate impacts of the new methods on the CLM 3.0 and CCSM 3.0 models were investigated through sensitivity experiments with the new MODIS consistent land surface parameters compared to a control experiment with the current CLM 3.0 parameters. All experiments were performed at the T42 grid resolution using prescribed monthly climatology sea surface temperatures and sea ice distributions. The experiments were each run for a 15 year period, with the initial 5 years discarded as a spin up period. To assess the climate impacts of the new land surface parameters relative to climate variability that may be attributed to arbitrary starting conditions, a second new parameters experiment was conducted with different atmospheric and land surface initial conditions from the first experiment and the control experiment.

[14] The differences between the new parameters experiment and the control experiment were assessed relative to model climate biases of the CCSM 3.0, by comparing model results to observed climatology data from C. J. Willmott and K. Matsuura (Terrestrial air temperature and precipitation: Monthly and annual climatologies (Version 2.0.1), 2000, available online at http://climate.geog.udel.edu/∼climate/html_pages/download.html) (hereinafter referred to as Willmott and Matsuura, online data archives, 2000) for precipitation and near surface air temperature, and from MODIS for local noon Blacksky broadband surface albedo. The average monthly precipitation and air temperature data were compiled from 1970–1999 monthly values, aggregated from the 0.5 degree grid increment to the T42 grid increment. The monthly MODIS surface albedo was compiled from 2001–2003 Version 4 fortnightly albedo data. To ensure the MODIS albedo was directly comparable to the albedo calculated in CLM 3.0 the 0.05 degree MODIS data was aggregated to the T42 grid increment using the same MODIS land mask used in the generation of the new land surface parameters.

[15] Overall average changes in climate variables over all land, and the three analysis regions of: Amazon, Boreal, and Sahara and Arabia were tabulated for all seasons with the statistical strengths of these relationships assessed through multiple statistical tests. The statistical testing involved a Student's T test and a Wilcoxon Signed Rank test on each of the time series of seasonal variables to identify statistically significant differences in the experiments for all land and for the regions. The Wilcoxon Signed Rank test was used following von Storch and Zwiers [1999], as it is nonparametric and provides insurance against moderate departures from the distribution assumptions required for calculating T-test significance.

[16] In addition to the all land and regional analysis, the seasonal differences in a range of representative climate variables between the new parameters experiments and the CLM 3.0 control experiment were mapped globally. The mapped differences showed the differences in the mean climatological values for each season for each grid cell and for larger regions. Time series T tests were performed on each grid cell to identify statistically significant changes for the grid cell and for larger regions between experiments.

[17] Finally the robustness of the results of the first realizations of the new parameters experiments was assessed by comparing differences of the second realization of the new parameters experiment to the control experiment. The assessment involved spatial correlations of the differences of each of the new parameters experiments to the control experiment as well as mapping of differences of the second realization for comparison against the analysis of the first realization.

2.3. New Plant Functional Type (PFT) Parameter Mapping

[18] To incorporate MODIS land surface products wherever possible into the CLM Plant Functional Type parameters new methods were adapted from the original PFT parameter generation methods used for CLM 3.0 parameters and described by Bonan et al. [2002]. The fractions of land covered by tree, herbaceous and bare soil were determined directly from the MODIS Vegetation Continuous Fields data set from Hansen et al. [2003]. The further breakdown of the tree fraction to Needleleaf and Broadleaf, as well as Evergreen and Deciduous components was derived from AVHRR Continuous Fields Tree Cover Project data from Defries et al. [2000].

[19] The breakdown of the herbaceous fraction was derived directly from the global cropping of Ramankutty and Foley [1999] for crop fraction, with the remaining grass and shrub fractions determined from the MODIS global land cover mapping of Friedl et al. [2002]. All mapping and climate rules were applied at the 0.05 degree grid increment before aggregation to the CLM 3.0 model grid. A schematic of the new PFT mapping methods and source data sets is shown in Figure 1.

Figure 1.

Schematic of New Plant Functional Type (PFT) mapping from MODIS satellite data where possible and from historical cropping from Ramankutty and Foley [1999].

[20] The PFT physiology and climate rules of Nemani and Running [1996] were used to further split the tree, shrub and grass PFTs into tropical, temperate and boreal climate groupings and in the case of grasses C3/C4 photosynthetic pathways. These climate rules were assessed on the basis of climatological monthly surface air temperature and precipitation surfaces from Willmott and Matsuura (online data archives, 2000). To better represent the description of C3 and C4 grasses, the physiology and climate rules were modified to include the fractional C3/C4 mapping methods of Still et al. [2003]. The C3/C4 mapping methods used MODIS Version 4 average monthly LAI mapping to describe seasonal variations in vegetation cover as indicator for the C3/C4 growing season in the same manner as NDVI variation was used by Still et al. [2003].

2.4. New Leaf Area Index (LAI) and Stem Area Index (SAI) Parameter Mapping

[21] New CLM PFT LAI parameters were derived from the MODIS Version 4 LAI data of Myneni et al. [2002] to address differences in monthly LAI between MODIS and the current CLM Parameters. The monthly MODIS LAI data was averaged for the 2001–2003 time period to produce climatological monthly LAI. Quality assurance flags were used to ensure that only highest-quality LAI values were included, with lower quality LAI values used only in the absence of high-quality data for the month for the three year period. The average monthly MODIS LAI values were divided to monthly PFT LAI values for each 0.05 degree grid cell based on the new PFT mapping, relative maximum PFT LAI values, and PFT leaf phenology rules.

[22] The initial breakdown of MODIS LAI to individual PFT LAI values was based on the annual maximum PFT LAI values prescribed by Bonan et al. [2002]. For summer green deciduous trees, the summer green leaf phenology of the Lund-Potsdam-Jena (LPJ) dynamic vegetation model, described by Sitch et al. [2003], was used to limit the initial maximum PFT LAI by leaf growing season. This method used the cumulative growing degree days from a base temperature to calculate a summer green phenology status. Growing degree days were calculated from average 1970–1999 monthly surface air temperature surfaces from Willmott and Matsuura (online data archives, 2000).

[23] A second breakdown of LAI was calculated for evergreen tree phenology following the methods and PFT parameters of Zeng et al. [2002]. This method ensured that the PFT LAI of evergreen PFTs could only reach a minimum fraction (0.7 Needleleaf and 0.8 Broadleaf) of the annual maximum PFT LAI calculated for the grid cell in the initial calculation of PFT LAI. The evergreen phenology calculation was performed to correct for MODIS LAI values for higher latitudes in winter months where snow cover and low sun angles prevented reasonable LAI retrieval from MODIS, and to differentiate the evergreen PFT LAI component of the MODIS LAI from the remaining non-evergreen PFT LAI. The final monthly PFT LAI values for non-evergreen PFTs were then calculated using the same calculation method as the initial PFT LAI breakdown, but with the evergreen PFT LAI components removed from the monthly MODIS LAI.

[24] Monthly PFT SAI values were calculated directly from the methods and PFT parameters of Zeng et al. [2002], using minimum PFT SAI values, the new PFT mapping, and the newly calculated monthly PFT LAI values. The monthly PFT SAI values represented the dead leaf component of the canopy, as well as the area of stems and branches associated with each PFT. The dead leaf component of the canopy was calculated from the difference in PFT LAI between the current month and the previous month added to a residual amount of dead leaf component from the previous month. A minimum PFT SAI value was calculated to account for the constant element of stem area not included in the monthly changes in PFT LAI and the residual SAI.

2.5. New Soil Color Parameter Mapping

[25] A new version of the soil color parameters used by CLM to define soil reflectance for use in calculating soil albedo was developed to reduce the land surface albedo differences found between the two stream radiation model of CLM and MODIS satellite observed surface albedo. The process of calculating the new soil colors involved fitting visible and near infrared soil reflectance values for each grid cell for each month, to find values that would reproduce the same average monthly snow free surface albedo in CLM as were observed by MODIS at local solar noon on the middle day of the month. The soil reflectance values were combined with climatological soil moisture from a control CCSM 3.0 run, to find the new soil color that would reproduce the monthly soil reflectance in CLM with the given soil moisture.

[26] The analysis of the CCSM 3.0 modeled climate identified that the range of soil reflectance values available from the soil color classes currently used in the model did not allow for the range of broad band surface albedo values observed over bare soils by MODIS. The highest broad band soil albedo values were over North African and Arabian deserts where visible soil albedo was up to 0.30 and near infrared soil albedo was up to 0.60. To allow the soil classes used in CLM 3.0 to reproduce the range of surface albedo values observed in the MODIS satellite data, the existing 8 soil colors were extended to the 20 soil colors with 11 soil colors of higher soil reflectance and a single soil color with lower soil reflectance added.

[27] The MODIS surface albedo was calculated for the 2001–2003 period, for the visible and near infrared spectrums over the Black Sky (Direct Beam) and White Sky (Diffuse Beam) components using the Ross-Li polynomial representation and the 16 day MODIS Version 4 isotropic, volumetric, and geometric scattering parameters described by Strahler et al. [1999] and Schaaf et al. [2002]. The MODIS 16 day Blacksky and Whitesky albedo values were then combined to produce 16 day actual or Bluesky albedo using the fraction of diffuse radiation reaching the surface at local solar noon for the middle day of each time period at each grid cell.

[28] The diffuse radiation fraction was calculated using the solar zenith angle at local noon and the SKYL lookup table provided with the MODIS Version 4 albedo parameters, available from C. B. Schaaf (Tools For MODIS BRDF/Albedo products, 2004, available at http://www-modis.bu.edu/brdf/userguide/tools.html). The optical depth of the sky was set at 0.2 to represent clear sky conditions and the atmospheric model was set to continental for all grid cells. The 16 day actual albedo time series were compiled to generate average monthly surface albedo for each spectrum, with only highest-quality cloud free data used. Where monthly cloud free data was not available in the three year period for a pixel, values were linearly extrapolated from the closest two months with cloud free data either side.

[29] The two stream radiation model of Sellers [1985] was used to calculate local noon broadband Bluesky albedo over a range of visible and near infrared soil reflectance values, with the new PFT, LAI and SAI mapping. For consistency with the MODIS albedo, the monthly local noon diffuse radiation fractions used with the two stream radiation model were the same as those used in the MODIS Bluesky albedo calculation. The soil reflectance values that produced the closest broadband albedo in the two stream radiation model were selected as the fitted values for the month.

[30] Soil color was then fitted with the climatological soil moisture for months that were snow free in the MODIS surface albedo data. The soil colors of the individual months were averaged over all snow free months to specify a representative soil color for the grid cell. In cases where there was no snow free surface albedo for the entire three year period, the soil color derived from snow affected albedo was used to give a representative soil color that included the effects of the permanent snow cover.

3. Results

3.1. New Plant Functional Type (PFT) Parameters

[31] The 0.05 degree PFT maps were aggregated to the 0.5 degree grid resolution of the CLM 3.0 raw data to compare with the current CLM 3.0 land surface parameters. The differences in PFT composition between the current CLM 3.0 parameters and the new parameters are shown for all land and the representative regions of Amazon, Boreal, and Sahara and Arabia in Table 1. Representative PFT distributions are globally mapped at the 0.5 degree grid resolution for the new parameters and the current parameters in Figures 2 and 3.

Figure 2.

Distribution of CLM tree and shrub plant functional types (PFTs) at the 0.5 degree grid resolution for (a, c, e, g) New MODIS parameters and (b, d, f, h) Current CLM 3.0 parameters.

Figure 3.

Distribution of CLM crop, grass, and bare plant functional types (PFTs) at the 0.5 degree grid resolution for (a, c, e, g) New MODIS parameters and (b,d, f, h) Current CLM 3.0 parameters.

Table 1. Average Global and Regional Land Surface Plant Functional Type Percentage Composition for Current CLM 3.0 Parameters and New MODIS Parametersa
PFTAll LandAmazonBorealSahara and Arabia
CLMNew (Diff)CLMNew (Diff)CLMNew (Diff)CLMNew (Diff)
  • a

    Plant functional type, PFT; Current CLM 3.0 parameters, CLM; New MODIS parameters, New. Differences between parameters are shown in brackets, with a dash signifying no change. Abbreviated plant functional types are specified in Figure 1 in the same numerical order.

Bare23.833.5 (+9.7)1.13.3 (+2.2)8.110.4 (+2.3)74.188.5 (+14.4)
Ndl Evg Tmp2.41.7 (−0.7)0.00.1 (+0.1)3.71.8 (−1.9)0.00.0 (−)
Ndl Evg Borl3.53.9 (+0.4)0.00.0 (−)17.718.1 (+0.4)0.00.0 (−)
Ndl Dec Borl0.90.7 (−0.2)0.00.0 (−)4.93.6 (−1.3)0.00.0 (−)
Brd Evg Trop6.06.1 (+0.1)39.438.5 (−0.9)0.00.0 (−)0.00.0 (−)
Brd Evg Tmp0.70.9 (+0.2)0.81.2 (+0.4)0.00.0 (−)0.00.0 (−)
Brd Dec Trop4.13.5 (−0.6)12.911.1 (−1.8)0.00.0 (−)0.40.3 (−0.1)
Brd Dec Tmp2.52.2 (−0.3)0.10.1 (−)1.31.1 (−0.2)0.00.0 (−)
Brd Dec Borl0.60.8 (+0.2)0.00.0 (−)2.22.3 (+0.1)0.00.0 (−)
Shr Evg Tmp0.20.2 (−)0.00.0 (−)0.00.0 (−)0.00.0 (−)
Shr Dec Tmp10.33.8 (−6.5)5.12.5 (−2.6)0.20.1 (−0.1)12.12.0 (−10.1)
Shr Dec Borl3.45.8 (+2.4)0.20.8 (+0.6)13.426.4 (+13.0)0.00.0 (−)
Grs C3 Arctic4.74.9 (+0.2)0.20.7 (+0.5)23.527.4 (−3.9)0.00.0 (−)
Grs C313.710.4 (−3.3)8.57.1 (−1.4)11.06.4 (−4.6)6.01.3 (−4.7)
Grs C48.011.1 (+3.1)17.430.1 (+12.7)0.10.0 (−0.1)5.16.1 (+1.0)
Crop15.310.5 (−4.8)14.44.7 (−9.7)14.110.3 (−3.8)2.21.6 (−0.6)

3.1.1. Bare

[32] The largest change in PFT distribution with the new parameters was the large increase in Bare PFT composition over all land compared to the current CLM 3.0 parameters (33.5% compared to 23.8%). Figures 3g and 3h show there are large differences in the distribution of Bare PFT between the parameters globally, with the largest differences in sparsely vegetated areas such as central Australia, southern Africa, southern South America, southwest North America, and central Asia. In these areas the new parameters specified substantially higher Bare PFT contribution than with the current CLM 3.0 parameters.

3.1.2. Needleleaf Trees

[33] There were relatively small changes in needle leaf tree contribution over all land, with the new parameters having lower contributions from Needleleaf Evergreen Temperate trees (1.7% compared to 2.4%), higher contribution from Needleleaf Evergreen Boreal trees (3.9% compared to 3.5%), and lower contribution from Needleleaf Deciduous Boreal trees (0.7% compared to 0.9%). Figures 2a and 2b show essentially the same distribution of needleleaf tree PFTs with both sets of parameters, with the new parameters having overall lower needleleaf tree composition.

3.1.3. Broadleaf Evergreen Trees

[34] The new parameters had higher contribution from broadleaf evergreen trees over all land, with a small increase in Broadleaf Evergreen Tropical trees (6.1% compared to 6.0%), and a relatively larger increase in Broadleaf Evergreen Temperate trees (0.9% compared to 0.7%). Figures 2c and 2d again show very similar distributions of broadleaf evergreen trees for both sets of parameters with some small regional variations.

3.1.4. Broadleaf Deciduous Trees

[35] The new parameters had overall lower contribution from broadleaf deciduous trees over all land, with decreases in Broadleaf Deciduous Tropical trees (3.5% compared to 4.1%) and Broadleaf Deciduous Temperate trees (2.2% compared to 2.5%), however there was a marginal increase in Broadleaf Deciduous Boreal trees (0.8% compared to 0.6%). Figures 2e and 2f show there were similar distributions in broadleaf deciduous trees, but with distinctly lower contributions from the new parameters. In tropical areas there were large decreases in broadleaf tree contributions in the savannas of the eastern South America, and central Africa. In temperate areas there were large decreases in the forest and agricultural lands of eastern North America, but with substantial increases in these same areas in Europe.

3.1.5. Shrubs

[36] The new parameters had mixed changes in contribution from shrubs over all land, with a large decrease in Broadleaf Deciduous Temperate shrubs (3.8% compared to 10.3%), a large increase in Broadleaf Deciduous Boreal shrubs (5.8% compared to 3.4%), and no change in Broadleaf Evergreen Temperate shrubs. Figures 2g and 2h show the large decreases in temperate shrubs corresponded to the large increases in bare soil in sparsely vegetated areas. The increase in boreal shrubs is shown to have been concentrated in North America, while there were decreases in shrub density over northern Russia.

3.1.6. Grasses

[37] The new parameters had similar overall contribution from grass, but with substantial changes in the partition to C3 and C4 grasses. Over all land the new parameters had a substantially lower contribution from C3 non-Arctic grass (10.4% compared to 13.7%), which was offset by a large increase in C4 grass (11.1% compared to 8.0%) and a smaller increase in C3 Arctic grass (4.9% compared to 4.7%). Figures 3c, 3d, 3e and 3f show that in tropical areas such as the Sahel in Africa, northern Australia, and south eastern Asia, the new parameters replaced equal mixes of C3 and C4 grasses with pure C4 grasses. In mid latitude areas such as the grasslands and the savannas of eastern South America, and southern Africa, the new parameters had increases in both C3 and C4 grasses. In Europe and in eastern North America there were increases in C3 grasses, but at higher latitudes there were substantially lower values for C3 grasses.

3.1.7. Crops

[38] The new parameters had a large overall decrease in Crops over all land (10.5% compared to 15.3%) as a result of prescribing cropping directly from Ramankutty and Foley [1999] rather than from land cover maps. The maps of Figures 3a and 3b show that the new parameters had small overall decreases in Crops in North America, Europe and Asia, but very large decreases in South America and southern areas of Australia. In most cases the decreases in Crop contribution corresponds with similar increases in grass PFTs.

3.2. New Leaf Area Index (LAI) and Stem Area Index (SAI) Parameters

[39] The 0.05 degree monthly PFT LAI and SAI parameters were aggregated to the 0.5 degree grid increment of the CLM 3.0 raw data to generate the new CLM LAI and SAI parameters. The equivalent total grid cell LAI and SAI values of the new parameters and the current CLM 3.0 parameters were assessed by combining the PFT composition values with the monthly PFT LAI and PFT SAI values. The differences between the new and current parameters for average seasonal LAI and SAI are shown in Tables 2 and 3, respectively, for all land and the representative regions of Amazon, Boreal, and Sahara and Arabia. The combined PFT LAI values of the new methods and the current CLM parameters are shown in Figure 4 against each other and against average monthly MODIS values. The differences in combined PFT SAI values for the new methods and the current CLM parameters are shown in Figure 5.

Figure 4.

Differences in CLM Leaf Area Index (a, b) between New MODIS parameters and Current CLM 3.0 parameters and (c, d) between New parameters and observed monthly MODIS values.

Figure 5.

Differences in CLM Stem Area Index (SAI) between the New MODIS parameters and Current CLM 3.0 parameters.

Table 2. Average Global and Regional Seasonal LAI for Current CLM 3.0 Parameters and New MODIS Parametersa
 All LandAmazonBorealSahara and Arabia
CLMNew (Diff)CLMNew (Diff)CLMNew (Diff)CLMNew (Diff)
  • a

    Current CLM 3.0 parameters, CLM; New MODIS parameters, New. Differences between parameters are shown in brackets, with a dash signifying no change. Composite LAI is calculated from parameters by summing PFT% * PFT Monthly LAI.

DJF0.880.94 (+0.06)2.602.98 (+0.38)0.610.67 (−0.06)0.020.06 (+0.04)
MAM0.981.13 (+0.15)2.803.17 (+0.37)0.850.87 (+0.02)0.020.05 (+0.03)
JJA1.331.46 (+0.13)2.763.18 (+0.42)2.152.22 (+0.07)0.050.08 (+0.03)
SON0.951.15 (+0.20)2.403.09 (+0.69)0.870.99 (+0.12)0.070.10 (+0.03)
Table 3. Average Global and Regional Seasonal SAI for Current CLM 3.0 Parameters and New MODIS Parametersa
 All LandAmazonBorealSahara and Arabia
CLMNew (Diff)CLMNew (Diff)CLMNew (Diff)CLMNew (Diff)
  • a

    Current CLM 3.0 parameters, CLM; New MODIS parameters, New. Differences between parameters are shown in brackets, with – signifying no change. Composite SAI is calculated from parameters by summing PFT% * PFT Monthly SAI.

DJF0.260.41 (+0.15)0.480.77 (+0.29)0.260.57 (+0.31)0.050.04 (−0.01)
MAM0.250.38 (+0.13)0.450.75 (+0.30)0.230.55 (+0.32)0.090.03 (−0.06)
JJA0.370.41 (+0.04)0.420.78 (+0.36)0.630.60 (−0.03)0.140.03 (−0.11)
SON0.320.52 (+0.20)0.390.76 (+0.37)0.521.03 (+0.51)0.090.04 (−0.05)

3.2.1. Leaf Area Index

[40] Over all land the new parameters had large increases in LAI for all seasons, with the largest increase in boreal autumn (1.15 compared to 0.95) and the smallest increase in boreal winter (0.94 compared to 0.88). Figure 4a shows the new parameters had mixed differences in DJF LAI over much of the northern hemisphere in the boreal winter, but had distinctly higher DJF LAI values for tropical forests, and distinctly lower LAI for savannas in the southern hemisphere during the austral summer. Figure 4b shows the new parameters had distinctly lower JJA LAI for eastern North America and Europe, and higher LAI for higher-latitude Arctic areas in the boreal summer. The new parameters also had distinctly higher JJA LAI for tropical forests, but mixed differences in southern hemisphere savannas and grasslands in the austral winter.

[41] Figure 4c shows the new parameters had no observable differences in DJF LAI compared to MODIS values, except for higher latitudes in the northern hemisphere. These areas corresponded with evergreen boreal forest of northern North America, Europe and Russia, which had substantially higher LAI values in the new parameters than could be retrieved by the MODIS satellite data for this time of year. Figure 4d shows there were practically no differences in JJA LAI between the new parameters and the values retrieved by the MODIS satellite data.

3.2.2. Stem Area Index

[42] Over all land the new parameters had relatively large increases in SAI for all seasons, with the largest increase in boreal autumn (0.52 compared to 0.32) and the smallest increase in boreal summer (0.41 compared to 0.37). Figure 5a shows the new parameters had increased DJF SAI over tropical forests, and over northern hemisphere temperate and boreal forests, but decreased DJF SAI over southern hemisphere savannas and grasslands. Figure 5b shows the new parameters again had increased JJA SAI for tropical forest, with slightly increased JJA SAI for northern hemisphere temperate and boreal forest and for southern hemisphere savanna and grasslands. The new parameter also had reduced JJA SAI for northern hemisphere grasslands and shrublands.

3.3. New Soil Color Parameters

[43] The 0.05 degree soil color maps generated from the two stream radiation model, monthly soil moisture, and MODIS albedo were aggregated to the 0.5 degree grid increment of the CLM 3.0 raw data to generate the new CLM soil color parameters. The differences in soil color and the associated differences in dry and saturated soil reflectance, between the current CLM 3.0 parameters converted to the new soil color classes and the new parameters are shown for all land and the representative regions of Amazon, Boreal, and Sahara and Arabia in Table 4. The soil color is globally mapped at the 0.5 degree grid increment for the new methods and the current parameters, converted to the new soil color classes, in Figure 6.

Figure 6.

CLM soil color class for (a) New MODIS parameters and (b) Control (CLM 3.0) parameters. Soil colors range from most to least reflective over the range 1 to 20.

Table 4. Average Global and Regional Soil Color and Visible and Near-Infrared Reflectance for Dry and Saturated Soil for Current CLM 3.0 Parameters and New MODIS Parametersa
 All LandAmazonBorealSahara and Arabia
CLMNew (Diff)CLMNew (Diff)CLMNew (Diff)CLMNew (Diff)
  • a

    Current CLM 3.0 parameters, CLM; New MODIS parameters, New. Differences between parameters are shown in brackets, with a dash signifying no change.

Soil Color15.316.6 (+1.3)16.017.8 (+1.8)16.018.1 (+2.1)13.18.9 (−4.2)
VIS Dry0.170.14 (−0.03)0.160.12 (−0.04)0.160.12 (−0.04)0.210.26 (+0.05)
VIS Sat0.090.07 (−0.02)0.080.06 (−0.02)0.080.06 (−0.02)0.110.15 (+0.04)
NIR Dry0.280.25 (−0.03)0.270.23 (−0.04)0.270.22 (−0.05)0.330.41 (+0.08)
NIR Sat0.170.15 (−0.02)0.160.12 (−0.04)0.160.12 (−0.04)0.220.30 (+0.08)

[44] Over all land the new parameters had substantially higher average soil color that resulted in substantially lower soil reflectance for both visible and near infrared dry and saturated soil reflectance. Figure 6 shows the differences in soil color between the new parameters and the current CLM 3.0 were not consistent globally. In general, the new parameters had higher soil color with lower soil reflectance for densely vegetated areas such as forests, and lower soil color with higher soil reflectance for sparsely vegetated areas such as the Sahara, Arabian Peninsula, and central Asia. There were some sparsely vegetated areas however, where the new parameters had substantially higher soil colors with lower soil reflectance, such as over Western Australia, the Kalahari desert of southern Africa and the Atacama desert of southern South America.

3.4. Changes in Broadband Land Surface Albedo Compared to MODIS

[45] The global and regional local noon broadband blacksky surface albedo of CLM 3.0 with the new MODIS consistent parameters and with the current CLM 3.0 parameters were compared to the equivalent average monthly MODIS values. The results of this analysis are shown for all land and the representative regions of Amazon, Boreal, and Sahara and Arabia in Table 5 and mapped globally in Figure 7. The albedo comparison investigated the effectiveness of the new parameters in reproducing the MODIS albedo in CLM 3.0. Table 5 shows that over all land the new parameters significantly reduced the CLM 3.0 surface albedo for all seasons. The decrease in surface albedo reduced the high albedo difference between the CLM 3.0 and MODIS albedo for all seasons over all land.

Figure 7.

Differences in average seasonal local noon broadband blacksky surface albedo (%) (a, b) between MODIS and the Current CLM 3.0 parameters experiment, and (c, d) with the New MODIS parameters experiment, and (e, f) differences between experiments. Differences with statistical significance greater than 95% are stippled.

Table 5. Average Global and Regional Local Noon Broadband Blacksky Surface Albedo for the Current CLM 3.0 Parameters, and the New MODIS Parameters Experimentsa
 All LandAmazonBorealSahara and Arabia
CLM (Diff)New (Diff)CLM (Diff)New (Diff)CLM (Diff)New (Diff)CLM (Diff)New (Diff)
  • a

    Current CLM 3.0 parameters, CLM; New MODIS parameters, New. Averages are given in percent. Differences between the experiments and MODIS are shown in brackets, with a dash signifying no difference. Statistical significance in differences between Current CLM 3.0 and New parameters experiments are shown as: n, neither test has significance > = 0.95; t, Student T Test has significance > = 0.95; w, Wilcoxon Signed Rank Test has significance > = 0.95; and b, both tests have significance > = 0.95.

DJF31 (+2)30 (+1) b13 (−)13 (−) n48 (+6)44 (+2) b25 (−8)31 (−2) b
MAM28 (+2)27 (+1) b13 (−)13 (−) b40 (+4)34 (−2) b24 (−8)31 (−1) b
JJA21 (+3)20 (+2) b14 (+1)14 (+1) b17 (+5)14 (+2) b24 (−8)29 (−3) b
SON26 (+2)25 (+1) b13 (−)13 (−) b27 (−)23 (−4) b24 (−8)30 (−2) b

[46] In the Amazon the new parameters had little change in broadband surface albedo, with all changes smaller than 1%. In the Boreal region the new parameters did significantly decrease the surface albedo of CLM 3.0, bringing the model albedo closer to MODIS for all seasons except SON. In DJF, MAM and JJA the high model albedo was reduced to values that were only slightly higher or lower than MODIS, while in SON the model albedo was changed to substantially lower albedo with the new parameters.

[47] In the Sahara and Arabia region the new parameters significantly increased the modeled surface albedo of CLM 3.0 for all seasons, with broadband surface albedo changed from substantially lower (−8%) to slightly lower than MODIS (−1 to −3%). The new soil color parameters removed much of the difference in surface albedo between CLM 3.0 and MODIS for this region, however small albedo differences remained for all seasons.

[48] The maps of Figure 7 show that the changes in soil color with the new parameters had the largest impact in areas with sparse vegetation that were not affected by snow, such as the Sahara, Arabia, central Asia and Australia. In more densely vegetated areas such as the Amazon, central Africa and south east Asia, small differences between CLM 3.0 and MODIS persisted. In snow affected areas such as Eurasia, the Tibetan plateau, and northern North America, albedo in CLM 3.0 was highly constrained by snow cover, with the differences in DJF surface albedo between CLM 3.0 and MODIS persistent with both sets of parameters, independent of changes in soil color, PFT composition or seasonal LAI and SAI differences.

3.5. Changes in Precipitation Compared to Observation

[49] Table 6 shows that over all land the new parameters significantly reduced the positive precipitation bias modeled in CCSM 3.0 compared to Willmott and Matsuura's (online data archives, 2000) findings for all seasons. In the Amazon region, there were no significant changes with the new parameters. In the Boreal region the new parameters significantly reduced the high precipitation biases for JJA and SON. In the Sahara and Arabia region the new parameters changed the wet bias of MAM to a dry bias, and had large significant reductions in the wet precipitation biases of JJA and SON.

Table 6. Average Global and Regional Seasonal Precipitation for the Current CLM 3.0 Parameters, and With the New MODIS Parameters Experimentsa
 All LandAmazonBorealSahara and Arabia
CLM (Diff)New (Diff)CLM (Diff)New (Diff)CLM (Diff)New (Diff)CLM (Diff)New (Diff)
  • a

    Current CLM 3.0 parameters, CLM; New MODIS parameters, New. Precipitation given in mm/day. Differences between the experiments and 1970–1999 observed climatological values from Willmott and Matsuura (online data archives, 2000) are shown in brackets, with a dash signifying no difference. Statistical significance in differences between Current CLM 3.0 and New parameters experiments are shown as: n, neither test has significance > = 0.95; t, Student T Test has significance > = 0.95; w, Wilcoxon Signed Rank Test has significance > = 0.95; and b, both tests have significance > = 0.95.

DJF2.6 (+0.4)2.5 (+0.3) b5.7 (−0.3)5.6 (−0.4) n1.6 (+0.5)1.6 (+0.5) n0.1 (−0.2)0.1 (−0.2) n
MAM2.4 (+0.2)2.4 (+0.2) b4.8 (−1.3)4.7 (−1.4) n1.6 (+0.6)1.6 (+0.6) n0.3 (+0.1)0.1 (−0.1) b
JJA2.5 (+0.1)2.4 (−) b3.0 (−0.5)3.0 (−0.5) n2.2 (+0.2)2.0 (−) b2.6 (+2.1)2.3 (+1.8) b
SON2.6 (+0.3)2.6 (+0.3) b4.8 (+0.7)4.6 (+0.5) n2.2 (+0.6)2.1 (+0.5) b1.3 (+1.1)1.1 (+0.9) b

[50] The maps of Figure 8 show that while the new parameters effectively reduced CCSM 3.0 precipitation biases in many areas, large precipitation biases remained with the same general patterns as with the original CLM 3.0 parameters. Comparing the DJF precipitation biases of Figures 8a and 8c with the precipitation differences between the two parameter experiments in Figure 8e reveals that the biggest reductions in DJF precipitation bias was in central Africa, while the largest increase in bias was in western Australia. Over the rest of the world the changes in precipitation were mixed with little distinctive pattern or significance.

Figure 8.

Differences in average seasonal precipitation (a, b) between Willmott and Matsuura (online data archives, 2000) 1970–1999 climatology and Current CLM 3.0 parameters experiment and (c, d) with New MODIS parameters experiment, and (e. f) differences between experiments. Differences with statistical significance greater than 95% are stippled.

[51] Comparing the JJA precipitation biases of Figures 8b and 8d with the differences between the two experiments in Figure 8f reveals the largest reduction in precipitation bias with the new parameters were over the Arabian peninsula and over northern and central Africa, where the large wet biases of the current CLM 3.0 parameters experiment were substantially reduced. There also were increases in both wet and dry JJA biases in North America, Europe, Russia, eastern China and south eastern Asia. Interestingly the areas that had the largest reductions in precipitation bias also were the areas with the largest reductions in albedo differences to MODIS.

3.6 Changes in Near Surface Air Temperature Compared to Observation

[52] Table 7 shows that over all land the new parameters had significant increases in the warm biases of near surface air temperature for MAM, JJA and SON compared to Willmott and Matsuura's (online data archive, 2000) findings. In the Amazon region the new parameters had a significant reduction in the JJA warm bias, with changes in all other seasons nonsignificant. For the Boreal region the new parameters significantly increased the warm bias of MAM and changed the cool bias to a warm bias for JJA. For the Sahara and Arabia region the new parameters significantly reduced warm biases for DJF and MAM, reduced the cool bias for JJA, and changed the warm bias to a cool bias for SON.

Table 7. Average Global and Regional Near-Surface Air Temperature for the Current CLM 3.0 Parameters, and With the New MODIS Parameters Experimentsa
 All LandAmazonBorealSahara and Arabia
CLM (Diff)New (Diff)CLM (Diff)New (Diff)CLM (Diff)New (Diff)CLM (Diff)New (Diff)
  • a

    Current CLM 3.0 parameters, CLM; New MODIS parameters, New. Temperatures are given in °C. Differences between the experiments and 1970–1999 observed climatological values from Willmott and Matsuura (online data archives, 2000) are shown in brackets, with a dash signifying no change. Statistical significance in differences between Current CLM 3.0 and New parameters experiments are shown as: n, neither test has significance > = 0.95; t, Student T Test has significance > = 0.95; w, Wilcoxon Signed Rank Test has significance > = 0.95; and b, both tests have significance > = 0.95.

DJF6.2 (+1.3)6.2 (+1.3) n24.1 (+0.2)24.1 (+0.2) n−15.4 (+3.2)−15.6 (+3.0)n17.8 (+1.5)16.9 (+0.6) b
MAM9.7 (+0.6)10.0 (+0.9) b24.4 (+0.8)24.5 (+0.9) n−4.0 (−0.1)−3.2 (+0.9) t25.6 (+1.5)24.9 (+0.8) b
JJA14.2 (+0.5)14.9 (+1.2) b24.6 (+2.0)24.3 (+1.7) b10.1 (−1.4)12.5 (+1.0) b29.1 (−0.8)29.4 (−0.5) b
SON10.9 (+1.0)11.1 (+1.2) b25.6 (+1.5)25.4 (+1.3) n−0.9 (+1.5)−0.6 (+1.8) n24.8 (+0.1)24.3 (−0.4) b

[53] The maps of Figure 9 show that while the new parameters changed reference height temperatures relative to observation, the same general patterns of temperature bias between the modeled values and observation were repeated with both sets of parameters. Despite these similarities in biases, there was a cooling over the DJF warm bias of the Sahara, the Arabian Peninsula, Europe and central Asia. There also was warming of the JJA cool bias over northern latitudes, but additional warming of the JJA warm bias over the rest of North America, Europe and central Asia.

Figure 9.

Differences in average seasonal near-surface air temperature (a, b) between Willmott and Matsuura (online data archives, 2000) 1970–1999 climatology and Current CLM 3.0 parameters experiment, and (c, d) with New MODIS parameters experiment. Differences between experiments and climatology with statistical significance greater than 95% are stippled.

3.7. Changes in CLM Surface Hydrology

[54] Further analysis detailed by Lawrence and Chase [2006b], found that the decrease in precipitation over all land with the new MODIS consistent parameters corresponded with increased canopy interception, surface runoff and deep soil drainage, resulting in significant decreases in soil moisture through the entire soil profile. The increase in canopy interception corresponded with increases in LAI and SAI for all seasons in the new parameters. There also were small but statistically significant increases in transpiration, as well as larger significant increases in canopy evaporation over all land for all seasons with the new parameters. The increases in transpiration and canopy evaporation, however, were completely offset by large decreases in soil evaporation, resulting in a net decrease in evapotranspiration with the new parameters. The net decreases in evapotranspiration and precipitation were unexpected given the net increase in LAI with the new MODIS parameters, an issue we will come back to in the discussion.

3.8. Dependence of Climatology on Model Initial Conditions

[55] The dependence of initial conditions on the changes in climate simulated in the CLM 3.0 and the CCSM 3.0 with the new MODIS consistent parameters was assessed through a second realization of the new parameters experiment with different initial conditions [Lawrence and Chase, 2006b]. The two realization experiments had very high spatial correlation for seasonal differences with the control experiment for broadband land surface albedo (0.96–0.98), precipitation (0.65–0.77), near surface air temperature (0.54–0.94) as well as a range of surface hydrology and surface flux parameters (0.79–0.97). In general, the highest spatial correlations were found in JJA while the lowest correlations were in DJF. The seasonal differences in spatial correlation were most probably associated with the variability of snow cover which has largest extent in DJF and smallest extent in JJA. The snow cover extent is highly dependent on atmospheric conditions which have large interannual variability and are sensitive to initial conditions, where as snow free conditions are more directly influenced by the prescribed land surface parameters.

4. Discussion

[56] The comparison of the broadband surface albedo modeled in CLM 3.0 with the new parameters and MODIS surface albedo, showed there were large improvements in the modeled albedo in sparsely vegetated areas such as the Sahara, Arabia, central Asia and Australia, but substantial differences in albedo remained in densely vegetated areas. Investigation into the differences between CLM 2.0 modeled albedo and MODIS by Tian et al. [2004a] found that the large differences in albedo in vegetated areas could be traced to the PFT contribution of grasses and crops, and their leaf scattering albedos which were substantially higher than broadleaf trees and shrubs. Very large differences also remained between the modeled albedo and MODIS in areas affected by snow. In these areas changes in PFT contribution, LAI and SAI did have some impacts, however the distribution of snow modeled in CCSM 3.0 had a larger impact than changes in the surface parameters.

[57] The comparison of the climate modeled in CLM 3.0 and CCSM 3.0 with the current CLM 3.0 parameters and the new MODIS consistent land surface parameters, also found the new parameters resulted in an average climate that had year round reduced precipitation, drier soils and reduced evapotranspiration. This relationship was very robust for densely vegetated regions and dominated other changes in surface hydrology associated with changes in precipitation. It appears that this reduction in precipitation, along with drier soils and reduced evapotranspiration was the main driver for the large increase in JJA near surface air temperature over land at mid to high latitudes in the northern hemisphere. For sparsely vegetated areas, however, the changes in precipitation had a more direct impact on soil evaporation and therefore on evapotranspiration.

[58] Modeling investigations into the climate impacts of tropical deforestation in South America, Africa, and South East Asia, by Zhang et al. [1995], Polcher and Laval [1994], Bounoua et al. [2002] and others, have found the main impact of removing tropical forests is reduced evapotranspiration through reduced transpiration. In their research they found the reduction in atmospheric moisture combined with the increase in sensible heat flux, results in reduced cloud formation and precipitation. The reduced cloud cover significantly increases solar radiation flux which offsets any increase in albedo associated with the deforestation, resulting in further increases in sensible heat flux and further reduction in precipitation and cloud cover.

[59] Our investigations found that this relationship is reversed in CLM 3.0 with reduced evapotranspiration the product of large decreases in soil evaporation that more than offset the increases in canopy evaporation and transpiration associated with increased LAI. The absolute magnitudes of transpiration compared to soil and canopy evaporation in both experiments indicate that plants in CLM 3.0 transpire at only a fraction of the evaporation rate of bare soils and canopy intercepted precipitation. Further analysis detailed by Lawrence and Chase [2006b], found the warmer and drier reference surface conditions impacted overlying atmospheric conditions, resulting in higher planetary boundary layer heights with reduced convection having positive feedbacks through reduced precipitation and cloud cover. The reduced precipitation further influenced the surface hydrology making land surface drier, while the reduced cloud cover increased shortwave radiation resulting in increased surface warming.

[60] The numerous statistical tests demonstrated that changes in land surface parameters do have highly robust impacts on the climate simulated in CCSM 3.0. The Wilcoxon Signed Rank test supported the significance found with the T test for all cases except MAM differences in near surface air temperature in the Boreal region. The very high spatial correlations of the differences between our second realization of the new parameters experiment and the control experiment, compared to the differences with the first realization, further demonstrate that the changes in climate found with the new parameters are repeatable, independent of initial conditions. This makes the statistical tests more meaningful as the areas identified as significant are replicated in the second realization. This demonstrates the importance of realistically simulating the land surface and the surface fluxes from the land to the atmosphere in CLM 3.0 for the larger-scale climate simulation in CCSM 3.0.

[61] While not directly investigated in this research, there are other benefits to providing a more realistic finescale representation of vegetation within the CLM 3.0. The new land surface data sets are being used in a wide range of applications from global fine mesh land surface simulations at 0.05 degrees resolution, as well as in sensitivity experiments to assess how the new representation impacts the uptake of atmospheric CO2 through changes in photosynthesis, and for changes in the emission of volatile organic compounds such as isoprene. Following this the new land surface parameters have been adopted by the CCSM Land Working Group as the new standard prescribed land surface parameters to be used in future versions of the model.

5. Conclusions

[62] We have developed new methods for generating finer resolution (0.05 degrees) global CLM 3.0 land surface parameters that are consistent with MODIS land surface data, while maintaining the multiple PFT canopy and herbaceous layers of Bonan et al. [2002]. The parameters prescribe crop distributions directly from the historical cropping data set of Ramankutty and Foley [1999], providing a consistent method for describing historical cropping in CLM for any year from 1700 to current day. The finer resolution of the new parameters also allows them to be used over a wider range of model grid resolutions globally than were previously available. The new parameters result in increased bare soil fraction of 10% which is realized through reduced tree, shrub and crop cover. The new parameters also have area average increases of 10% for LAI and SAI, with the largest increases in tropical forests.

[63] The sensitivity experiments performed with the new MODIS consistent land surface parameters demonstrate that the CCSM 3.0 does have strong robust responses to changes in land surface parameters through changes in surface radiation budgets, energy balances and hydrology, and through that the fluxes of energy, moisture and momentum from the surface to the atmosphere. In many cases improvements in surface albedo directly resulted in improved simulation of precipitation and near surface air temperature, however for the most part large biases remained with the new parameters.

[64] Our analysis of changes in surface hydrology revealed that the increased LAI of the new parameters resulted in lower overall evapotranspiration with reduced precipitation. This was an unexpected result given that other research into the impacts of vegetation change indicates that the new parameters should have the opposite impact. This suggests that providing improved descriptions of land surface parameters from MODIS land surface data is important for improving climate simulated in CLM 3.0 and CCSM 3.0, however, the new surface parameters have limited success in rectifying surface hydrology biases that result from the parameterizations within CLM 3.0. As a result further work is required to investigate how changing these parameterizations would impact the surface hydrology and therefore the surface fluxes and the climate simulated in CLM 3.0 and CCSM 3.0.

Acknowledgments

[65] The use of the computing time for the model experiments was supplied through a grant from the National Center for Atmospheric Research, Community Climate System Model (CCSM) Land Working Group, which is sponsored by the National Science Foundation. Funding for this research also was supported by National Science Foundation grants ATM0001476 and ATM0437538. MODIS land surface products were provided by the MODIS Land Science Team. Historical cropping data were provided by the Center for Sustainability and the Global Environment, University of Wisconsin.

Ancillary