1. Introduction
The land surface influences weather and climate through surface fluxes of radiation, heat, moisture, and momentum. To simulate these fluxes, many land surface models discretize the earth’s land surface into large grid cells (typically about 3° latitude by 3° longitude for global climate models) with a single biome (e.g., needleleaf evergreen forest, savanna) assigned to each. Associated with each biome is a suite of physiological parameters, often based on field programs, that attempts to capture the most important characteristics of the vegetation determining energy, water, and momentum fluxes.
One disadvantage of this approach is that it ignores subgrid variability in land cover. Several studies have indicated that consideration of subgrid landscape variability is important for determining surface fluxes in weather/climate models (Henderson-Sellers and Pitman 1992; Koster and Suarez 1992; Bonan et al. 1993; Seth et al. 1994; Noilhan and Lacarrere 1995; Cooper et al. 1997; Shuttleworth et al. 1997). For example, the subgrid presence of wet surfaces (e.g., lakes, swamps, marshes) has been shown to significantly affect land surface fluxes and climate in a GCM (Bonan 1995a). Similarly, several studies argue that mesoscale circulations, induced by land surface inhomogeneities, should be parameterized in climate models (Avissar and Pielke 1989; Pielke et al. 1991, 1997; Vidale et al. 1997).
Another disadvantage of the biome approach is that it homogenizes a heterogeneous land surface into a few generalized biomes. However, biomes are not emergent landscape elements but are merely composed of plant species that coexist. Many of the required model parameters are leaf-level parameters that are inappropriate at the biome scale. A savanna, for example, does not have an inherent photosynthetic and stomatal response to light, but rather consists of grasses and trees that do have measurable leaf physiology. Moreover, landscapes, even at scales as small as tens of kilometers, are a mosaic of vegetated patches, with the structure and composition of each patch determined by site-specific microclimates, soils, geomorphology, and disturbance history (Bormann and Likens 1979; Shugart 1984). Modern biomes can be unlike past landscapes. For example, the needleleaf evergreen forest that dominated the southeastern United States 18 000 years ago at the height of the last glaciation is unlike the modern northern coniferous forest (Webb et al. 1993). The modern eastern deciduous forest did not form until 12 000 to 9000 years ago as the climate warmed and the ice sheet retreated northward. A more accurate ecological representation of land surfaces should represent landscapes as a shifting mosaic of coexisting physiologically distinct plants rather than as biomes with emergent properties.
The National Center for Atmospheric Research land surface model (NCAR LSM; Bonan 1996) partially achieves a subgrid ecological representation of the landscape by characterizing the land surface as a set of multiple plant functional types (PFTs) within each grid cell. However, the distribution of PFTs within a grid cell is specified based on one of 28 possible biomes. For example, a needleleaf evergreen forest is 75% needleleaf evergreen tree and 25% bare ground. Moreover, all needleleaf evergreen trees have a maximum leaf area index (LAI) of 5, a minimum LAI of 4, and a height of 17 m. The model, therefore, utilizes patches of PFTs but does not separate the physiology of the plant functional types (e.g., photosynthetic capacity) from their structure (e.g., LAI, canopy height).
The advent of high-resolution remote sensing data products such as land cover and LAI, two expected products from the National Aeronautics and Space Administration’s Earth Observing System (EOS) (Wharton and Myers 1997), rectifies this problem and allows for the separate specification of physiology and structure. Remote sensing is the only practical method by which land surface properties can be estimated at the temporal and spatial scales required by climate models (Roughgarden et al. 1991; Townshend et al. 1991; Wessman 1992; Sellers et al. 1995a). Studies in which satellite-derived LAI has been incorporated into land surface models have demonstrated its importance in climate simulations (Chase et al. 1996; Pitman et al. 1999; Bounoua et al. 2000).
In this paper, we report on the use of 1-km land-cover and LAI datasets developed from the Boreal Ecosystem–Atmosphere Study (BOREAS; Sellers et al. 1995b) to develop and to test a methodology for incorporating EOS-era data into a land surface model (the NCAR LSM) for use with a global climate model. The 1-km land-cover data were used to define the number and proportions of PFTs within each model grid cell in the BOREAS study area. The 1-km LAI data provided the necessary LAI for each PFT. We assessed the effects of incorporating these data both as separate and combined datasets by comparing simulated surface fluxes with those from a control experiment that utilized the model’s default dominant biome approach.
2. Methods
A model grid (referred to hereinafter as the hydromet grid) was defined as a domain extending from 52° to 57°N latitude and from 96° to 107°W longitude with a spacing of 10 min in longitude by 5 min in latitude (about 10 km by 10 km). The grid covers an area of roughly 360 000 km2 and encompasses the BOREAS northern and southern study areas. The dominant NCAR LSM surface and soil texture types for each hydromet grid cell were determined from 1-km maps. A control experiment (ECON) was run over the hydromet grid for 1994 using these dominant type maps. Next, an estimate of the relative proportions of PFTs within each hydromet grid cell was defined based on the 1-km data and incorporated into the model framework. A second experiment (EPFT) was run to assess the effects of these data on surface fluxes over the BOREAS region. Last, LAI was estimated for each subgrid PFT using a 1-km dataset and was incorporated into the model. A third experiment incorporating the combined PFT and LAI data sets (EPFTLAI) was then conducted, and its results were compared with the ECON and EPFT experiments.
a. Model
The NCAR LSM (version 1) has been described in detail by Bonan (1996). It was designed for coupling to atmospheric numerical models. Consequently, there is a compromise between computational efficiency and the complexity with which the necessary atmospheric, ecological, and hydrologic processes are parameterized. The model is meant not to be a detailed micrometeorological model but rather a simplified treatment of surface fluxes that reproduces at minimal computational cost the essential characteristics of land–atmosphere interactions important for climate simulations.
The model characterizes the land surface through a combination of surface types and PFTs. Twelve PFTs are defined in the model that differ in leaf and stem areas, root profile, height, leaf dimension, optical properties, stomatal physiology, roughness length, displacement height, and biomass. The model allows for 28 different surface types, each composed of multiple PFTs and bare ground (see Table 1 for a list of the surface types present in the BOREAS region). For example, a mixed broadleaf deciduous and needleleaf evergreen forest consists of patches of broadleaf deciduous trees, needleleaf evergreen trees, and bare ground. Lakes and wetlands, if present, form additional patches. Soil effects are included by allowing thermal properties (heat capacity, thermal conductivity) and hydraulic properties (porosity, saturated hydraulic conductivity, saturated matric potential, slope of retention curve) to vary depending on percent sand and percent clay. Soils also differ in color, which affects soil albedos. Consequently, each grid cell in the domain of interest is assigned a surface type (which determines the PFTs and their cover), a fraction covered by lakes, a fraction covered by wetlands, a soil texture (percent sand, percent silt, percent clay), and a soil color.
Major features of the model are
prescribed time-varying leaf and stem areas;
absorption, reflection, and transmittance of solar radiation, accounting for the different optical properties of vegetation, soil, water, snow, and ice;
absorption and emission of longwave radiation allowing for emissivities less than 1;
sensible and latent heat fluxes, with partitioning of latent heat flux into canopy evaporation, soil evaporation, and transpiration from sunlit and shaded foliage;
turbulent transfer above and within plant canopies;
vegetation and ground temperatures that balance the surface energy budget (net radiation, sensible heat, latent heat, soil heat);
explicit linkage of photosynthesis and stomatal resistance in which stomatal physiology depends on the photosynthetic response to foliage temperature, absorbed photosynthetically active radiation, ambient carbon dioxide (CO2) concentration, vapor pressure deficit, soil water, and foliage nitrogen;
CO2 loss from plant respiration, which depends on temperature and varies among foliage, stem, and root biomass, and from microbial respiration, which depends on biomass, quality, soil water, and soil temperature;
interception, throughfall, and stemflow;
snow hydrological behavior;
infiltration and runoff;
temperatures for a six-layer soil column using a heat diffusion equation that accounts for phase change;
soil water for the same six-layer soil column using a one-dimensional conservation equation that accounts for infiltration input, gravitational drainage at the bottom of the column, evapotranspiration losses, and vertical water flow based on head gradients; and
temperatures for six-layer deep and shallow lakes accounting for eddy diffusion and convective mixing.
Bonan (1998) describes the climate that results when the model is coupled to the NCAR Community Climate Model. The model has been applied to study land–atmosphere CO2 exchange (Bonan 1995b; Craig et al. 1998), the effects of lakes and wetlands on climate (Bonan 1995a), the effect of vegetation and soil (Kutzbach et al. 1996) and lakes and wetlands (Coe and Bonan 1997) on the African monsoon in the middle Holocene, the effect of soil water on floods and droughts in the Mississippi River basin (Bonan and Stillwell-Soller 1998), the effect of tundra ecosystems on Arctic climate (Lynch et al. 1999), and the effects of temperate deforestation on climate (Bonan 1997, 1999). Bonan et al. (1997) evaluated the ability of the model to simulate fluxes from different surface types in the BOREAS study areas. Tower fluxes measured at the old jack pine and old aspen sites during the three 1994 intensive field campaigns were compared with simulated fluxes. The results suggested that the model is able to reproduce variability between vegetation types but not within vegetation types, thus motivating the current study in part.
b. Atmospheric forcings
A 1-h-resolution gridded atmospheric forcing dataset of air temperature, dewpoint temperature, pressure, wind speed, incident solar radiation, incident longwave radiation, and precipitation for the period 1 January to 31 December 1994 was assembled for the BOREAS hydromet grid (V. Pauwels 2000, personal communication). Data from numerous meteorological stations, rain gauges, Geostationary Operational Environmental Satellite estimates of solar radiation (Gu and Smith 1997), and Enterprise Electronics Corporation model WR100 rain radar (Schnur et al. 1997) were blended using a weighted-average interpolation scheme.
This dataset was used to force the model assuming all precipitation is large scale (uniformly distributed over the grid cell), 70% of the solar radiation is direct beam, 30% is diffuse, and solar radiation is evenly split into visible and near-infrared wavelengths. Observations of incident longwave were missing from 1 January to about 17 January 1994. These missing values were calculated from air temperature. Each hourly observation was repeated once to run the model at a 30-min time step. The model was run for four years, starting on 1 January 1994, by repeating the 1994 atmospheric forcing. The first three years were discarded for spinup from an arbitrary initialization of all temperatures to 10°C, soil water to 0.3 mm3 mm−3, and no snow.
c. Land surface data
A 1-km soil texture dataset was used to determine a dominant soil texture type for each hydromet grid cell (Shields et al. 1991). Thirty-eight soil texture classes were reduced to the sand, silt, and clay fractions for the 11 soil texture types of Cosby et al. (1984) and a 12th texture type defined as organic matter. A dominant soil texture type was chosen by computing the relative proportions of each of the 12 soil texture types within each grid cell. The model was modified to account for the effects of organic matter on soil hydraulic and thermal properties. Unfrozen and frozen thermal and hydraulic properties were determined from values for peat given by Dickinson (1997), Oke (1987), and Pielke (1984). Currently, the organic matter fraction is set to either 100% or 0% according to the soil texture map. Soil color was set to an intermediate value between light and dark for the region [color class 4; see Table 10 of Bonan (1996)]. These datasets were used to provide soil texture and color for all three experiments.
The 1-km Advanced Very High Resolution Radiometer (AVHRR) BOREAS land-cover map of Steyaert et al. (1997) was used to assign a single dominant NCAR LSM surface type to each hydromet grid cell for the ECON experiment. The land-cover classes defined at 1-km resolution were first translated to model surface types using the provided land-cover type descriptions. A single surface type was then determined for each grid cell by reprojecting the 1-km land-cover data to the hydromet grid and selecting the dominant type by area. This surface type was used in the model to determine the PFTs in the grid cell and their LAI from prescribed lookup tables (e.g., Table 1). Fractions of each grid cell occupied by water and wetlands were also determined.
In the EPFT and EPFTLAI experiments, the vegetation composition (i.e., the number of PFTs and their abundance) and structure (i.e., LAI) of each grid cell varied based on high-resolution data. The 1-km Steyaert et al. land-cover data were used to calculate the fractional areas of lakes, wetlands, and PFTs occurring within each grid cell. Each pixel was assigned an NCAR LSM surface type, which provided the PFTs and their relative abundance for that pixel (Table 1). The data were then aggregated to the hydromet grid. The model was modified to accommodate up to six PFTs within each grid cell (the maximum number occurring in any grid cell) plus lake and wetland fractions for a total of up to eight subgrid patches.
In the EPFT experiment, LAI for each PFT was prescribed by the model. In the EPFTLAI experiment, a 1-km LAI dataset derived from AVHRR data (Cihlar et al. 1997) was used to provide LAI for each PFT as a replacement for the model’s prescribed LAI. Three LAI images derived from 10-day composites for the BOREAS region were available (21–31 May, 21–31 July, and 1–10 September 1994). The LAI obtained from remote sensing is a quantity averaged over the field of view of the sensor (i.e., a 1-km pixel) and thus incorporates any bare ground within the field of view. In contrast, the LAI specified in the model is for the vegetated fraction of the grid cell only. Therefore, we first calculated an average LAI for each surface type in each grid cell from the 1-km data. Next, the average surface type LAI was divided by the model’s total vegetation fraction (Table 1) to obtain a PFT LAI. In cases where two PFTs are associated with a single surface type (e.g., cool grassland consists of C3 and C4 grasses), we assigned the same LAI value to both PFTs. In the hydromet region, this assumption was necessary for cool grassland and mixed woodland. This assumption is not unrealistic for cool grassland, because C3 and C4 grasses are likely to have the same phenological characteristics. For cool mixed woodland, this assumption is less valid, particularly for the May and September images, because broadleaf deciduous trees may have lower LAI than do needleleaf evergreen trees in these months. However, information required to unmix the relative LAI signals of these two PFTs was unavailable for this region. Further study is required to solve this problem for future applications involving seasonal datasets.
A monthly value of LAI for each PFT throughout the year was required to conduct the 3-yr spinup. LAI for months between observations was obtained from linear interpolation. For other months, the model’s PFT phenological characteristics (specified from Dorman and Sellers 1989) were merged with the remotely sensed LAI by adjusting the model LAI for the difference between the two datasets in July. Some additional adjustments were made to ensure that unrealistic temporal profiles did not occur (e.g., negative LAI, decreases in LAI during greenup, increases in LAI during senescence). To ensure that the stem area index remained in proportion to the remotely sensed LAI, these were adjusted similarly.
3. Results
a. Surface data
When viewed in terms of surface types, the study region is dominated by needleleaf evergreen forest (Fig. 1). Other areas dominated by broadleaf deciduous forest occur in the southwest, as do cropland and some grassland. Mixed woodlands (consisting of needleleaf evergreen and broadleaf deciduous trees) are interspersed throughout the region. Also notable are several large regions classified as recent burns/bare ground. Pixels identified as sparse vegetation, rock outcrops, bare ground, and recent burns in the Steyaert et al. classification were lumped into this surface type because of a lack of information on the PFTs, if any, in these areas.
In contrast, the inclusion of the 1-km PFT data indicates a more varied range of PFT fractions in the study region (Fig. 2). Some of the largest differences are located in the northeast, where grid cells classified as 100% bare ground in the dominant classification show substantial fractions of vegetation (primarily needleleaf evergreen). Similarly, grid cells dominantly classified as needleleaf evergreen have larger fractions of bare ground. In the northeast, the fractional cover of the broadleaf deciduous PFT is generally more prevalent than the fractional cover prescribed by the mixed woodland surface type. In the southwest, where the PFTs are not affected by large changes in the bare ground fraction, the relative proportions of the needleleaf evergreen, broadleaf deciduous, cropland, and C3 grass PFTs also exhibit large differences. The differences in C4 grass are very small because this PFT is only present as a small subgrid patch in the cool grassland surface type (Table 1). Up to six PFTs coexist within the grid cells as opposed to the two to three allowed for in the dominant type scheme.
Figure 3 shows the differences in LAI determined from the remotely sensed LAI and the model’s PFT-prescribed LAI (EPFTLAI − EPFT) for June–August. We only discuss results from these three months because they are the most constrained by the observed LAI. In general, the remotely sensed LAI is lower than the PFT-prescribed LAI in the north and east and higher in the south and west. There is also a stronger seasonal signal in the remotely sensed LAI, particularly in the south and west. In the south and west, the areas of higher remotely sensed LAI are primarily associated with areas dominated by broadleaf deciduous or mixed woodland. Some areas of cropland also have higher remotely sensed values. In the north and east, areas dominated by needleleaf evergreen have lower remotely sensed LAI (∼20%).
b. Influence of remotely sensed data on surface fluxes
In this section, we examine the significance of the differences in the PFT and LAI datasets in terms of surface fluxes predicted by the model. Surface flux differences calculated as EPFTLAI − ECON describe the changes caused by incorporation of the combined datasets; differences calculated as EPFTLAI − EPFT and EPFT − ECON describe changes due to LAI and PFTs alone, respectively.
The changes in regionally averaged monthly fluxes were small (<3 W m−2) (Table 2). However, the standard deviations indicate some substantial variability in the responses of the individual grid cells. In general, the simultaneous incorporation of the PFT and LAI data decreased net radiation, resulting in decreases in both sensible and latent heat flux. Increases in ground evaporation were balanced somewhat by decreases in transpiration and canopy evaporation. This result is consistent with the fact that the regionally averaged remotely sensed LAI was smaller than the model LAI (not shown). Changes in ground evaporation and transpiration were highly variable in comparison with canopy evaporation. July was somewhat atypical in that the change in latent heat flux due to the incorporation of the combined datasets was positive. An increase in ground evaporation in response to a relatively high rainfall total in this month was responsible (the average precipitation over the region in July was 82 mm as compared with 44 and 15 mm in June and August, respectively).
In general, the changes in regional averages of net radiation, and sensible and latent heat flux due to the separate incorporation of the two datasets had the same sign. The sensible heat changes in June were the only exception to this result. LAI generally had less of an effect on the average net radiation than the PFT data did and caused much less spatial variability. The influence of LAI on the sensible and latent heat fluxes was comparable in magnitude to the PFT data but was less variable than the PFT data for sensible heat. The LAI data generally had a slightly larger influence on the components of latent heat than did the PFT data. The variability caused by LAI was comparable to that induced by the PFT data.
Because the downward longwave and shortwave radiation were prescribed in these experiments, changes in net radiation were primarily due to differences between PFT albedo, the influence of LAI on albedo, and upward longwave radiation changes attributable to surface temperature variations (Fig. 4). LAI had a minor influence on net radiation for the most part. Significant changes occurred only for those grid cells in which the LAI was sufficiently low as to expose the soil. The higher soil albedo decreased the net radiation for these grid cells. Increases in longwave radiation loss due to higher surface soil temperatures accentuated the decrease in net radiation.
Grid cells with slightly increased net radiation were generally associated with higher remotely sensed LAI, which lowered the gridcell albedo. Some increases in net radiation also occurred for grid cells that had lower LAI. In these cases, the increase in net radiation was due to interactions between the effects of LAI on vegetation albedo and temperature, and soil moisture on soil albedo. Lower LAI increased the vegetation temperature and albedo, thus resulting in lower net radiation. On the other hand, lower LAI allowed more precipitation to reach the soil, thereby increasing the surface soil water, lowering the soil albedo, and increasing the net radiation. This latter effect was most important for grid cells with a dry soil surface in the ECON experiment. However, these changes were relatively small.
In the north and east, the largest increases (decreases) in net radiation attributable to the combined datasets were due to decreases (increases) in the bare ground fraction because of the strong contrast between soil and vegetation albedos. The changes in net radiation were accentuated somewhat by changes in outgoing longwave radiation caused by surface temperature differences. In the south and west, where changes in bare ground fraction were small, net radiation was affected by changes in the relative proportions of cropland/grassland and forest. Increases (decreases) in net radiation were correlated with negative (positive) changes in cropland/grassland fractional cover. The model calculates higher albedos for cropland/grassland than for forested PFTs.
Ground evaporation and transpiration were strongly affected by the changes in both PFT and LAI, particularly in July (Table 2). The largest effects of changes in PFTs on ground evaporation and transpiration were associated with changes in the bare ground fraction (Fig. 5). Increases (decreases) in ground evaporation were highly correlated with increases (decreases) in bare ground. Transpiration was increased (decreased) for grid cells with decreases (increases) in bare ground. Changes in the fractions of other PFTs had smaller effects on the latent heat flux components. In the southwest, where changes in bare ground fraction were minimal, the distributions of the PFTs had some effects on transpiration. Grid cells with increases in the broadleaf deciduous PFT (and corresponding decreases in grassland and cropland PFTs) generally had higher transpiration because of larger LAI assigned to these PFTs in this month. The opposite was also true. However, in some cases, grid cells with decreased needleleaf evergreen and increased cropland fractions had increased transpiration despite the reduction in LAI. For identical atmospheric conditions, the model prescribes a higher photosynthetic rate, lower stomatal resistance, and thus higher transpiration for the cropland PFT than for the needleleaf evergreen PFT. These changes in the physiological parameters more than offset the effects of reduced leaf area.
Over much of the region, the changes in ground evaporation and transpiration due to PFT changes were somewhat compensating, so that latent heat displayed much less spatial variability than did either component alone. Grid cells that had positive latent heat flux changes generally had large reductions in bare ground fraction, decreased ground evaporation, and increased transpiration. However, because the surface area of the vegetation canopy was greater than the ground area (for LAI > 1 m2 m−2), transpiration increased by more than the corresponding decrease in ground evaporation. The grid cells exhibiting decreases in latent heat flux had increases in bare ground fraction, modest increases in ground evaporation, and reductions in transpiration. Here, increased ground evaporation was not sufficient to offset the reductions in transpiration.
The spatial variability in the latent heat flux components induced by the remotely sensed LAI was similar to that generated by incorporation of the PFT data (Table 2 and Fig. 6). The largest changes in ground evaporation and transpiration occurred for grid cells with low remotely sensed LAI. Ground evaporation increased and transpiration decreased for these grid cells. On the other hand, increases in LAI caused only small increases in transpiration (and decreases in ground evaporation). Grid cells dominated by cropland had a slightly enhanced response for the reasons mentioned above. As with the PFT data, variability in the latent heat flux components cancelled out somewhat so that the latent heat flux response was much less variable.
In August, a very dry month in comparison with July, the spatial variability of ground evaporation and transpiration was greatly reduced (Table 2). The latent heat flux in the control experiment was generally low for this month because of the lack of available soil water. Increases in ground evaporation due to reductions in vegetation cover associated with the PFT and LAI datasets only occurred for grid cells that had sufficient precipitation to keep the surface soil layer adequately wet (not shown). Decreases in ground evaporation due to increased vegetation cover were minor because ground evaporation was limited more by the available surface soil moisture. Transpiration was similarly limited by the rooting zone soil moisture. Thus, changes in PFTs and LAI only had a significant effect on grid cells that had adequate water in the rooting zone. The latent heat flux was slightly more variable in August because of the imbalance created by these water-limiting effects on ground evaporation and transpiration. For example, decreases in transpiration due to lower LAI were frequently not compensated by higher ground evaporation.
Changes in sensible heat flux due to PFTs (Fig. 5) were more spatially variable than were changes due to LAI (Fig. 6). Because of the relative spatial uniformity of the latent heat response, much of the change in net radiation (Fig. 4) was implemented as changes in sensible heat flux. Positive (negative) changes in net radiation created positive (negative) changes in sensible heat flux. As discussed earlier, the changes in latent and sensible heat due to the separate datasets were additive when the datasets were combined, both in terms of the regional average and spatial variability (Fig. 7). Sensible heat was more spatially variable than latent heat was because of the offsetting effects of ground evaporation and transpiration. Also, the experiments were somewhat water limited, even in July, thus limiting the response of the latent heat flux. The sensible heat and net radiation changes were highly correlated (correlation coefficient r = 0.86).
We note that the averaging of the diurnal cycle in the above analyses somewhat underestimates the spatial variability in flux differences. The frequency distributions in Fig. 8 show that the effects of incorporating remotely sensed PFTs and LAI were much more pronounced in daytime than at nighttime. The differences in daytime fluxes and surface temperature were of larger magnitude and were more spatially variable. The diurnal range in surface temperature was slightly smaller for the EPFTLAI simulation, with warmer nighttime temperatures and cooler daytime temperatures on average. These differences in the simulation of the diurnal cycle may have implications for the evolution of the planetary boundary layer and regional climate in a coupled model implementation of this approach.
4. Discussion and conclusions
We have described and tested a method for incorporating remotely sensed subgrid plant functional type and leaf area index data into a land surface model for use with a climate model. Large differences in PFT fractional cover between the subgrid and dominant type representations were found for the BOREAS region. Changes in the needleleaf evergreen PFT fraction were the most extensive both in terms of spatial distribution and magnitude (up to ±40%). Large differences in LAI between the remotely sensed and PFT-prescribed representations were also found (up to ±3 m2 m−2). Although the response of the model to these differences was somewhat small in terms of the average change in surface fluxes over the BOREAS region, the spatial variability of the model response was substantial. The PFT and LAI data were generally of equal importance in modifying the surface fluxes.
The relatively small regionally averaged changes may be due in part to the similar physiological parameters associated with each PFT (Bonan 1996). In addition, the model does not distinguish between wet and dry needleleaf evergreen trees (e.g., black spruce and jack pine). These are distinctly different physiological types, with black spruce adapted to cold, wet, nutrient-poor soils and jack pine growing on warm, dry sites (Bonan and Shugart 1989). As a result, they have very different surface energy fluxes (Moore et al. 2000; Baldocchi et al. 1997; Jarvis et al. 1997; Goulden et al. 1997). However, the ability to represent physiological variants within a class of plants (e.g., needleleaf evergreen tree) is limited by necessary datasets to characterize the distribution of those variants across the landscape. For the hydromet region, wet and dry conifer could be inferred from the 1-km land-cover data; however, Potter et al. (1999) point out that there is a high degree of undifferentiated pixel mixing among wet and dry conifer in this dataset. For global applications of the model such data do not exist (e.g., DeFries et al. 1999).
Inclusion of wet and dry needleleaf evergreen trees would necessitate subgrid variability in edaphic factors. Spatial variability in the hydrological properties of the soil has been shown to be as important as that induced by heterogeneity in vegetation (Cooper et al. 1997; Boone and Wetzel 1999). We did not consider subgrid edaphic variability because it would require assumptions about how soils, topography, and vegetation covary and would greatly increase the number of subgrid patches in the model.
The methodology used to adapt these datasets to the model framework can be improved in several ways. First, it was assumed that each 1-km pixel contained a single “pure” PFT. In reality, any given pixel may be heterogeneous. Unmixing techniques applied to EOS-era data hold promise for extracting subpixel PFT information that can be incorporated easily into the scheme described here to calculate PFT fractions with improved accuracy. These techniques may also provide PFT information for sparse and early regrowth vegetation that was not available in the datasets used here. Unmixing may also provide a means for estimating the individual PFT LAI of mixed pixels (e.g., the case of broadleaf deciduous and needleleaf evergreen mixtures encountered here). Substantial progress has been made such that prototype data are now available (e.g., DeFries et al. 1999). Second, improved estimates of bare ground fractional cover, whether fire-induced or not, are required. Two prototype datasets have emerged that show promise in meeting this requirement (Gutman and Ignatov 1998; Zeng et al. 2000). However, the methodology used to derive bare ground fraction must be consistent with the assumptions made in estimating LAI to use them simultaneously in a land surface model (Gutman and Ignatov 1998). Third, the temporal variability in LAI was not evaluated because of the limited datasets available. A full seasonal cycle of remotely sensed LAI would be useful for evaluating the temporal profiles used by the model. A recent study using satellite data has pointed out the need to update the model’s vegetation phenology (R. Myneni 2000, personal communication).
As noted in the introduction, future work should also consider the contribution of mesocale heat fluxes to the total flux. Research has indicated that the mosaic or patch approach to modeling land surface heterogeneity may only be appropriate for scales much smaller than the depth of the boundary layer (e.g., Pielke et al. 1991). The BOREAS region in particular is very heterogeneous, with strong potential for the development of mesoscale circulations resulting from contrasts between vegetation types and water/wetland bodies, and gradients in orography (Vidale et al. 1997). Because these circulations cannot be explicitly modeled at current global climate model resolutions, they should be parameterized. Pielke et al. (1997) provide one such parameterization appropriate for large-scale models.
Acknowledgments
We thank the BOREAS information system staff for processing the original soils data from Agriculture Canada, V. Pauwels for making the atmospheric forcing dataset available, and three anonymous reviewers for constructive comments. This research was funded under NASA Interagency Agreement W-19046.
REFERENCES
Avissar, R., and R. A. Pielke, 1989: A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meterology. Mon. Wea. Rev.,117, 2113–2136.
Baldocchi, D. D., C. A. Vogel, and B. Hall, 1997: Seasonal variation of energy and water vapor exchange rates above and below a boreal jack pine forest canopy. J. Geophys. Res.,102, 28 939–28 951.
Bonan, G. B., 1995a: Sensitivity of a GCM simulation to inclusion of inland water surfaces. J. Climate,8, 2691–2704.
——, 1995b: Land–atmosphere CO2 exchange simulated by a land surface process model coupled to an atmospheric general circulation model. J. Geophys. Res.,100, 2817–2831.
——, 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, 150 pp. [Available from UCAR Communications, P.O. Box 3000, Boulder, CO 80307-3000.].
——, 1997: Effects of land use on the climate of the United States. Climatic Change,37, 449–486.
——, 1998: The land surface climatology of the NCAR land surface model coupled to the NCAR Community Climate Model. J. Climate,11, 1307–1326.
——, 1999: Frost followed the plow: Impacts of deforestation on the climate of the United States. Ecol. Appl.,9, 1305–1315.
——, and H. H. Shugart, 1989: Environmental factors and ecological processes in boreal forests. Annu. Rev. Ecol. Syst.,20, 1–28.
——, and L. M. Stillwell-Soller, 1998: Soil water and the persistence of floods and droughts in the Mississippi River Basin. Water Resour. Res.,34, 2693–2701.
——, D. Pollard, and S. L. Thompson, 1993: Influence of subgrid-scale heterogeneity in leaf area index, stomatal resistance, and soil moisture on grid-scale land–atmosphere interactions. J. Climate,6, 1882–1897.
——, K. J. Davis, D. Baldocchi, D. Fitzjarrald, and H. Neumann, 1997: Comparison of the NCAR LSM1 land surface model with BOREAS aspen and jack pine tower fluxes. J. Geophys. Res.,102, 29 065–29 075.
Boone, A., and P. J. Wetzel, 1999: A simple scheme for modeling sub-grid soil texture variability for use in an atmospheric climate model. J. Meteor. Soc. Japan, 77 (1B), 317–333.
Bormann, F. H., and G. E. Likens, 1979: Pattern and Process in a Forested Ecosystem. Springer-Verlag, 253 pp.
Bounoua, L., G. J. Collatz, S. O. Los, P. J. Sellers, D. A. Dazlich, C. J. Tucker, and D. A. Randall, 2000: Sensitivity of climate to changes in NDVI. J. Climate,13, 2277–2292.
Chase, T. N., R. A. Pielke, T. G. F. Kittel, R. Nemani, and S. W. Running, 1996: Sensitivity of a general circulation model to global changes in leaf area index. J. Geophys. Res.,101, 7393–7408.
Cihlar, J., J. Chen, and Z. Li, 1997: Seasonal AVHRR multichannel data sets and products for studies of surface–atmosphere interactions. J. Geophys. Res.,102, 29 625–29 640.
Coe, M. T., and G. B. Bonan, 1997: Feedbacks between climate and surface water in northern Africa during the middle Holocene. J. Geophys. Res.,102, 11 087–11 101.
Cooper, H. J., E. A. Smith, and J. Gu, 1997: Modeling the impact of averaging on aggregation of surface fluxes over BOREAS. J. Geophys. Res.,102, 29 235–29 253.
Cosby, B. J., G. M. Hornberger, R. B. Clapp, and T. R. Ginn, 1984:A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soil. Water Resour. Res.,20, 682–690.
Craig, S. G., K. J. Holmén, G. B. Bonan, and P. J. Rasch, 1998: Atmospheric CO2 simulated by the National Center for Atmospheric Research Community Climate Model 1. Mean fields and seasonal cycles. J. Geophys. Res.,103, 13 213–13 235.
DeFries, R. S., J. G. R. Townshend, and M. C. Hansen, 1999: Continuous fields of vegetation characteristics at the global scale at 1-km resolution. J. Geophys. Res.,104, 16 911–16 923.
Dickinson, R. E., 1997: Incorporation of boreal forest ecosystem description into climate modeling framework. NASA Final Report, Project NAG-5-2310, 78 pp. [Institute of Atmospheric Physics, University of Arizona, Tucson, AZ 85721.].
Dorman, J. L., and P. J. Sellers, 1989: A global climatology of albedo, roughness length and stomatal resistance for atmospheric general circulation models as represented by the Simple Biosphere Model (SiB). J. Appl. Meteor.,28, 834–855.
Goulden, M. L., B. C. Daube, S.-M. Fan, D. J. Sutton, A. Bazzaz, J. W. Munger, and S. C. Wofsy, 1997: Physiological responses of a black spruce forest to weather. J. Geophys. Res.,102, 28 987–28 996.
Gu, J., and E. A. Smith, 1997: High-resolution estimates of total solar and PAR surface fluxes over large-scale BOREAS study area from GOES measurements. J. Geophys. Res.,102, 29 685–29 705.
Gutman, G., and A. Ignatov, 1998: The derivation of the green vegetation fraction from NOAA/AVHRR data for use in numerical weather prediction models. Int. J. Remote Sens.,19, 1533–1543.
Henderson-Sellers, A., and A. J. Pitman, 1992: Land-surface schemes for future climate models: Specification, aggregation, and heterogeneity. J. Geophys. Res.,97, 2687–2696.
Jarvis, P. G., J. M. Massheder, S. E. Hale, J. B. Moncrieff, M. Rayment, and S. L. Scott, 1997: Seasonal variation of carbon dioxide, water vapor, and energy exchanges of a boreal black spruce forest. J. Geophys. Res.,102, 28 953–28 966.
Koster, R. D., and M. J. Suarez, 1992: Modeling the land surface boundary in climate models as a composite of independent vegetation stands. J. Geophys. Res.,97, 2697–2715.
Kutzbach, J., G. Bonan, J. Foley, and S. P. Harrison, 1996: Vegetation and soil feedbacks on the response of the African monsoon to orbital forcing in the early to middle Holocene. Nature,384, 623–626.
Lynch, A. H., G. B. Bonan, F. S. Chapin III, and W. Wu, 1999: The impact of tundra ecosystems on the surface energy budget and climate of Alaska. J. Geophys. Res.,104, 6647–6660.
Moore, K. E., D. R. Fitzjarrald, R. K. Sakai, and J. M. Freedman, 2000: Growing season water balance at a boreal jack pine forest. Water Resour. Res.,36, 483–493.
Noilhan, J., and P. Lacarrere, 1995: GCM grid-scale evaporation from mesoscale modeling. J. Climate,8, 206–223.
Oke, T. R., 1987: Boundary Layer Climates. Routledge, 435 pp.
Pielke, R. A., 1984: Mesoscale Meteorological Modeling. Academic Press, 612 pp.
——, G. A. Dalu, J. S. Snook, T. J. Lee, and T. G. F. Kittel, 1991: Nonlinear influence of mesoscale land use on weather and climate. J. Climate,4, 1053–1069.
——, X. Zeng, T. J. Lee, and G. A. Dalu, 1997: Mesoscale fluxes over heterogeneous flat landscapes for use in larger scale models. J. Hydrol.,190, 317–336.
Pitman, A. J., M. Zhao, and C. E. Desborough, 1999: Investigating the sensitivity of a land surface scheme’s simulation of soil wetness and evaporation to spatial and temporal leaf area index variability within the Global Soil Wetness Project. J. Meteor. Soc. Japan,77 (1B), 281–290.
Pokrant, H., 1991: Land cover map of Canada derived from AVHRR images. Manitoba Remote Sensing Centre, Winnipeg, Manitoba, Canada. [Available from Manitoba Remote Sensing Centre, Land Information Division, Department of Natural Resources, 1007 Century Street, Winnipeg, MB R3H0W4, Canada.].
Potter, C. S., J. C. Coughlan, and V. Brooks, 1999: Investigations of BOREAS spatial data in support of regional ecosystem modeling. J. Geophys. Res.,104, 27 771–27 788.
Roughgarden, J., S. W. Running, and P. A. Matson, 1991: What does remote sensing do for ecology? Ecology,72, 1918–1922.
Schnur, R., T. W. Krauss, F. J. Eley, and D. P. Lettenmaier, 1997: Spatiotemporal analysis of radar-estimated precipitation during the BOREAS summer 1994 field campaigns. J. Geophys. Res.,102, 29 417–29 427.
Sellers, P. J., and Coauthors, 1995a: Remote sensing of the land surface for studies of global change: Models–algorithms–experiments. Remote Sens. Environ.,51, 3–26.
——, and Coauthors, 1995b: The boreal ecosystem–atmosphere study (BOREAS): An overview and early results from the 1994 field year. Bull. Amer. Meteor. Soc.,76, 1549–1577.
Seth, A., F. Giorgi, and R. E. Dickinson, 1994: Simulating fluxes from heterogeneous land surfaces: Explicit subgrid method employing the Biosphere–Atmosphere Transfer Scheme (BATS). J. Geophys. Res.,99, 18 651–18 668.
Shields, J. A., C. Tarnocai, K. W. G. Valentine, and K. B. MacDonald, 1991: Soil landscapes of Canada—procedures manual and user’s handbook. Publication 1868/E, LRRC Contribution Number 88-29, Land Resource and Research Centre, Agriculture Canada, Ottawa, Canada, 74 pp. [Available online at http://res.agr.ca/CANSIS/NSDB/SLC/__overview.html].
Shugart, H. H., 1984: A Theory of Forest Dynamics: The Ecological Implications of Forest Succession Models. Springer-Verlag, 278 pp.
Shuttleworth, W. J., Z.-L. Yang, and A. M. Arain, 1997: Aggregation rules for surface parameters in global models. Hydrol. Earth Syst. Sci.,1, 217–226.
Steyaert, L. T., F. G. Hall, and T. R. Loveland, 1997: Land cover mapping, fire regeneration, and scaling studies in the Canadian boreal forest with 1 km AVHRR and Landsat TM data. J. Geophys. Res.,102, 29 581–29 598.
Townshend, J. G., C. Justice, W. Li, C. Gurney, and J. McManus, 1991: Global land cover classification by remote sensing: Present capabilities and future possibilities. Remote Sens. Environ.,35, 243–255.
Vidale, P. L., R. A. Pielke, L. T. Steyaert, and A. Barr, 1997: Case study modeling of turbulent fluxes and mesoscale fluxes over the BOREAS region. J. Geophys. Res.,102, 29 167–29 188.
Webb, T., III, P. J. Bartlein, S. P. Harrison, and K. H. Anderson, 1993:Vegetation, lake levels, and climate in eastern North America for the past 18,000 years. Global Climates since the Last Glacial Maximum, H. E. Wright et al., Eds., University of Minnesota Press, 415–467.
Wessman, C. A., 1992: Spatial scales and global change: Bridging the gap from plots to GCM grid cells. Annu. Rev. Ecol. Syst.,23, 175–200.
Wharton, S. W., and M. F. Myers, Eds., 1997: MTPE/EOS Data Products Handbook. NASA, Goddard Space Flight Center, Greenbelt, MD, 266 pp. [Available from NASA Goddard Space Flight Center, Code 902, Greenbelt, MD 20771.].
Zeng, X., R. E. Dickinson, A. Walker, M. Shaikh, R. S. DeFries, and J. Qi, 2000: Derivation and evaluation of global 1-km fractional vegetation cover data for land modeling. J. Appl. Meteor.,39, 826–839.
Dominant surface type in the BOREAS region. In the ECON simulation, the surface type determined the PFTs, their abundance, and their LAI for a grid cell. See Table 1 for a description of the PFTs associated with each surface type.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Dominant surface type in the BOREAS region. In the ECON simulation, the surface type determined the PFTs, their abundance, and their LAI for a grid cell. See Table 1 for a description of the PFTs associated with each surface type.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Dominant surface type in the BOREAS region. In the ECON simulation, the surface type determined the PFTs, their abundance, and their LAI for a grid cell. See Table 1 for a description of the PFTs associated with each surface type.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Difference between remotely sensed and model-prescribed PFT fractional cover (EPFT − ECON) (%). The model-prescribed PFTs were based on the dominant surface type (Fig. 1, Table 1). In the EPFT simulation, PFT fractional cover was based on 1-km satellite-derived land cover.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Difference between remotely sensed and model-prescribed PFT fractional cover (EPFT − ECON) (%). The model-prescribed PFTs were based on the dominant surface type (Fig. 1, Table 1). In the EPFT simulation, PFT fractional cover was based on 1-km satellite-derived land cover.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Difference between remotely sensed and model-prescribed PFT fractional cover (EPFT − ECON) (%). The model-prescribed PFTs were based on the dominant surface type (Fig. 1, Table 1). In the EPFT simulation, PFT fractional cover was based on 1-km satellite-derived land cover.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Difference between remotely sensed LAI and model PFT-prescribed LAI (EPFTLAI − EPFT) (m2 m−2). In the EPFT simulation, LAI was prescribed for each PFT. In the EPFTLAI simulation, LAI was based on 1-km satellite-derived data.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Difference between remotely sensed LAI and model PFT-prescribed LAI (EPFTLAI − EPFT) (m2 m−2). In the EPFT simulation, LAI was prescribed for each PFT. In the EPFTLAI simulation, LAI was based on 1-km satellite-derived data.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Difference between remotely sensed LAI and model PFT-prescribed LAI (EPFTLAI − EPFT) (m2 m−2). In the EPFT simulation, LAI was prescribed for each PFT. In the EPFTLAI simulation, LAI was based on 1-km satellite-derived data.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
(top) Effect of remotely sensed LAI plus PFT (EPFTLAI − ECON) on net radiation for Jul 1994 (W m−2). (bottom) Effect of remotely sensed LAI (EPFTLAI − EPFT) on net radiation.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
(top) Effect of remotely sensed LAI plus PFT (EPFTLAI − ECON) on net radiation for Jul 1994 (W m−2). (bottom) Effect of remotely sensed LAI (EPFTLAI − EPFT) on net radiation.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
(top) Effect of remotely sensed LAI plus PFT (EPFTLAI − ECON) on net radiation for Jul 1994 (W m−2). (bottom) Effect of remotely sensed LAI (EPFTLAI − EPFT) on net radiation.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed PFT (EPFT − ECON) on latent heat, ground evaporation, transpiration, and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed PFT (EPFT − ECON) on latent heat, ground evaporation, transpiration, and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed PFT (EPFT − ECON) on latent heat, ground evaporation, transpiration, and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed LAI (EPFTLAI − EPFT) on latent heat, ground evaporation, transpiration, and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed LAI (EPFTLAI − EPFT) on latent heat, ground evaporation, transpiration, and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed LAI (EPFTLAI − EPFT) on latent heat, ground evaporation, transpiration, and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed LAI plus PFT (EPFTLAI − ECON) on latent and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed LAI plus PFT (EPFTLAI − ECON) on latent and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Effect of remotely sensed LAI plus PFT (EPFTLAI − ECON) on latent and sensible heat for Jul 1994 (W m−2).
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Histograms of daytime (solid line, 1900 UTC) and nighttime (dashed line, 0900 UTC) flux and surface temperature differences (EPFTLAI − ECON) determined from the average diurnal cycle for Jul.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Histograms of daytime (solid line, 1900 UTC) and nighttime (dashed line, 0900 UTC) flux and surface temperature differences (EPFTLAI − ECON) determined from the average diurnal cycle for Jul.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Histograms of daytime (solid line, 1900 UTC) and nighttime (dashed line, 0900 UTC) flux and surface temperature differences (EPFTLAI − ECON) determined from the average diurnal cycle for Jul.
Citation: Journal of Hydrometeorology 1, 5; 10.1175/1525-7541(2000)001<0431:TEORSP>2.0.CO;2
Plant functional type (PFT) fractional cover for each surface type in the BOREAS region.
Monthly differences (± two standard deviations), with the percent change from the control simulation in parentheses, in simulated net radiation (Q*), sensible (QH) and latent (QE) heat, ground evaporation (Eg), transpiration (Et), and canopy evaporation (Ec) averaged over the hydromet grid (W m−2).
