1Department
of Geography, University of Maryland, College Park.|
1Department
of Geography, University of Maryland, College Park. |
The process of binning land surface properties into a small number of discrete groups defined by a classification scheme creates several weaknesses. First, the approach assumes that all vegetation within the area labeled as a given cover type functions similarly. For parameters like leaf area index [Prince, 1991b] or photosynthetic capacity [Chabot and Mooney, 1985], spatial and temporal variation within a traditional vegetation type can be twofold to 10-fold, with much of the variation reflecting disturbance as well as patterns in ecosystem characteristics such as limited water and nutrient availability. Second, discrete distributions of vegetation types described by land cover classification schemes do not realistically reflect gradients in vegetation types over the landscape or mixtures of vegetation types within a grid cell. Consequently, land surface properties that vary continuously over space or are discontinuous on a fine spatial scale are depicted in the models by discrete boundaries. Third, the parameter fields generated from discrete vegetation types are dependent on the classification scheme and the number of vegetation types included in the scheme.
As an example of the distortions that occur when characterizing vegetation gradients as discrete cover types, Plate 1 shows the location of a transect through central Asia that traverses 'from' north to south through tundra, coniferous forest, grassland, shrubs and bare ground, and bare ground. The land cover classification shown in Plate 1, derived from remotely sensed data [DeFries and Townshend, 1994a], portrays the vegetation along the transect as discrete cover types. However, summertime normalized difference vegetation index (NDVI) values; which are related to radiation absorption by photosynthetic tissue, vary gradually and not discretely across the transect (Figure 1). To the extent that NDVI describes an important aspect of vegetation function, parameter fields derive] from discrete cover types will not realistically reflect the true variations in ecosystem function.
Approaches that assign land surface parameters according to vegetation type implicitly assume that the classification scheme captures the functionally important variation in these parameters. Recent advances in remote sensing provide an alternative approach to estimating global distributions of these parameters. For example, NDVI provides a continuous distribution of a parameter related to surface primary productivity [Goward et al., 1985; Tucker et al., 1985b]. Some models [Sellers et al., 1994] use a combination of remotely sensed data and discrete vegetation types to calculate global fields of biophysical parameters. An alternative to approaches based on discrete types uses climate rather than vegetation as the basis for extrapolation. Examples of models based on this approach estimate net primary production [Pittick, 1986; Webb et al., 1983], ecosystem structure and function [Woodward et al. 1991], forest succession [Solomon and Bartlein, 1992], maximum photosynthetic rate [Woodward and Smith, 1993], and soil respiration [Raich and Schlesinger, 1992].
In this paper, we review several existing land surface classification schemes commonly used in atmosphere-biosphere models. We illustrate the difficult effects of using different vegetation classifications to derive global fields of land surface parameters: We then consider the functional properties of vegetation required to calculate parameter fields used in the models and suggest an approach for describing the land surface using continuous distributions of these functional properties.
This paper focuses on information about the vegetated land surface required
for models of water, energy, and carbon dioxide exchange between the atmosphere
and biosphere. We consider the land surface requirements for atmosphere-biosphere
models to be primarily related to vegetation characteristics. We do not
address other characteristics of the land surface, such as soil characteristics,
topography, and aspect. Moreover, it is likely that other types of global
change models including those dealing with exchanges of other trace gases
(e.g.; methane, nitrous oxide), vegetation and population dynamics, or
policy options will have different requirements for land surface information.
Global maps of vegetation created by these traditional means have a number of limitations that make comparisons between them difficult [Townshend et al., 1991]. First, they rely on compilations from numerous sources because ground observations using a single system of classification have never been attempted at a global scale. Integrating local studies into regional and, ultimately, global maps raises problems because the original maps are often produced at different times. Second, widely varying criteria are used to classify the vegetation, including several different physiognomic, floristic, and environmental properties [Fosberg, 1967; Kuchler, 1967; Mueller, Dumbois and Ellenberg, 1974]. Consequently, compilers of maps may have to force very different categories into a common scheme. Third, data reliability is inconsistent because the field-based large-scale maps are unavailable for much of the world. Without large-scale maps, extrapolation using indirect sources may be the only information available.
Digital versions of ground-based maps compiled to global scales are generally tailored to a particular application. For example, Hummel and Reck [1979] digitized the vegetation/ land use map in the Oxford Atlas [Cohen, 1973] for albedo studies and discriminated 12 vegetation types. Subsequently, Olson et al. [1983] updated the Hummel and Reck data set into 43 types to discriminate carbon densities of vegetation.
Matthews [1983] "translated" about 70 published vegetation maps into the Unesco [1973] hierarchical classification scheme and compiled the maps into a global digital data set at 1° x 1° resolution. The Matthews [1983] digital database identifies 178 types that can be grouped according to application. A commonly used version includes 32 vegetation types. As a companion to the vegetation database, Matthews also compiled a land use database specifying the cultivated fraction of each 1° cell. Wilson and Henderson-Sellers [1985] combined vegetation and land use information into a database of 53 dominant and secondary cover types.
There is considerable variation among the global land cover data sets
compiled from ground-based sources [Townshend et al., 1991]. Comparison
of three digital data sets [Matthews [1983] data set of 32 vegetation
types, the data set for primary land cover from Wilson and Henderson-Sellers,
1985], and the land cover data set of Olson et al. [1983]) reveals
that, after translating each scheme into a common scheme for the basis
of comparison, the three agree on only about 26% of the total land surface
[DeFries and Townshend, 1994b]. Two of the three data sets agree
on an additional 46% of the land surface, while none of the data sets agree
on 28% of the land surface (Plate 2). Some
of this lack of agreement is due to incompatible classification schemes
used in the three data sets, some due to the lack of agreement in geographical
extent of various cover types, and some due to the necessarily subjective
procedure for grouping the classes for the basis of comparison. Whatever
the reason, the comparison shows that there are large differences among
existing ground-based sources of information on global land cover.
Many combinations of climate parameters have been used to describe vegetation distributions. Although temperature and precipitation are the two climate parameters used in all the schemes, the schemes differ primarily in their respective emphasis on seasonal information. The Holdridge [1947] scheme probably because of its tropical origins, uses only annual mean values, while at the other extreme, the Troll and Paffen [1964] scheme uses detailed information on the seasonality of climate (e.g., maximum winter precipitation, moderately warm summers, less than 6 months of humidity). The scheme of Koppen [1936] is based on vegetation physiology constructed from the Grisebach [1872] vegetation map.
More recently, other researchers have developed data sets of potential
vegetation. Prentice [1990] compared several bioclimatic schemes
with the Matthews [1983] vegetation data set and modified the Holdridge
scheme to identify 29 bioclimatic zones. Prentice et al. [1992]
developed a global biome model that predicts global patterns in vegetation
based on primary driving variables such as mean coldest month temperature,
annual accumulated temperature, and a drought index. Box [1981]
developed a classification system with 90 plant life forms to describe
the major types of vegetation structure on the Earth and predicted the
occurrence of these life forms from climatic data.
At regional and global scales, phenologies of different vegetation types, as reflected in intrannual variations in the NDVI, provide a basis for mapping and monitoring vegetation. For the continent of Africa, Tucker et al. [1985a] used NOAA's Global Vegetation Index to produce a general land cover classification at a spatial resolution of 15-20 km based on mean annual NDVI and seasonality. Townshend et al. [1987] classified land cover for South America using multiple images from different dates throughout the year. In addition, Loveland et al. [1991] used AVHRR local area coverage data at approximately 1-km spatial resolution for an unsupervised classification that derived homogeneous land cover regions for the conterminous United States.
A few attempts have been made to use satellite data for deriving distributions of land cover types at a global scale. Koomanoff [1989] produced a global map of nine basic vegetation types using an integrated annual value of NDVI, and Shimoda et al. [1986] derived statistical estimates of global land cover types. DeFries and Townshend [1994a] produced a global distribution of major land cover types based on a global 1° x 1° NDVI data set [Los et al., 1994; Sellers et al., 1994]. Running et al. [1994] propose that a global vegetation classification for the purposes of biogeochemical models could be derived from a combination of primary attributes of plant-canopy structure. They also identify remotely sensed observations from multitemporal AVHRR data that could be used to classify vegetation according to these structural characteristics.
Like the maps of vegetation distributions compiled from field observations and modeled from bioclimatic classification schemes discussed above, vegetation distributions derived from satellite data have generally classified vegetation types into discrete units that do not necessarily reflect the gradients or mosaics of vegetation types across the landscape. Information from satellite data has not been fully exploited to provide continuous distributions of global land surface characteristics, although such information is inherent in the data. Deconvolution of landscape components to describe mixtures of cover types in terms of percentage cover, as has been carried out at local and regional scales using high-resolution data [Hanan et al., 1991; Smith et al., 1990a, b], is one approach that could provide information on continuous distributions of vegetation characteristics. Ustin et al. [1993] suggested using linear unmixing techniques to derive vegetation characteristics from satellite data that are independent of floristic descriptions. This approach has been applied to a variety of biome types, including semiarid [Smith et al., 1990a, b], coastal savanna [Gamon et al., 1992; Roberts et al., 1993], and tropical broadleaf forests [Adams et al., 1990].
Plate 3 is an example of a linear unmixing approach to obtain a continuous distribution of woodiness of the vegetation for a portion of the boreal zone in North America. For this example, woodiness was obtained from herbaceous/woody ratios from Friedlingstein et al. [1992] applied to fractional covers for tundra, trees, and grass obtained from a linear mixture model [Hanan et al., 1991] based on 1989 monthly NDVI and brightness values in a revised version of the Global Vegetation Index [Goward et al., 1994].
Another example of this approach that describes land cover as simultaneously measured, continuous fields of green foliage, woodiness/litter, soil cover, and shadow/shade is illustrated using a Landsat TM image covering the central California coast range (Figure 2). This image is 185 km x 185 km and covers an area that includes part of the north central California coast range to the west, most of the Sacramento Valley and the northern extent of the San Joaquin Valley, and into the Sierra Nevada range to the east. This image includes a wide range of vegetation types, including coastal mixed evergreen forests, chaparral, annual grasslands and savanna, oak woodlands, and mixed conifer forests. Figure 2 shows continuous variation in three landscape components that are derived from a linear unmixing of the pixel spectra: green foliage (Figure 2a), the wood/litter (Figure 2b), and the shade/shadow (Figure 2c). Several vegetation types can be distinguished because they exhibit consistent patterns in the relative proportions of these three spectral classes. For example, the distributions of deciduous oak woodlands, grasslands, Sierra Nevada and coastal evergreen forests, and agricultural lands can be readily identified on the basis of characteristic differences in the proportions of these spectral categories. Secondary classifications of the landscape components shown in Figure 2 have been used to distinguish plant communities from grasslands to tropical forests [Adams et al., 1990; Gamon et al., 1992; Roberts et al., 1993; Ustin et al., 1993, 1995].
The two examples shown in Plate 3 and Figure 2 use spectral unmixing to obtain continuous fields of vegetation characteristics. This technique, as well as additional approaches for extracting continuous distributions of land surface characteristics from satellite data, require further development, particularly at the global scale.
In addition to static descriptions of vegetation types, satellite data
have been used to detect changes in land cover over time, e.g., deforestation
in the Amazon basin [Skole and Tucker, 1993] and movements of the
desert margin in the Sahel [Tucker et al., 1986]. Such knowledge
is required if atmosphere-biosphere models are to predict realistically
the effects of both anthropogenically induced and natural changes in land
cover on exchanges of energy, water, and carbon dioxide between the atmosphere
and biosphere.
Based on a global data set of NDVI at 1° x 1° resolution [Los et al., 1994], the Matthews [1983] database with 178 vegetation types, the Matthews [1983] database regrouped into 32 vegetation types, the 29-vegetation types diagnosed from climate [Prentice, 1990], and 11 types deduced from seasonal variations in NDVI [DeFries and Townshend, 1994a], we attempted to reconstruct global NDVI fields based on vegetation distributions. The 1° x 1° NDVI distribution was treated as "perfect" observations, in the sense that global coverage does not depend on knowledge of vegetation type. The NDVI data set was used to calculate the mean NDVI of each vegetation type in each of the four vegetation distributions. Global NDVI distributions were reconstructed from each of the four data sets by extrapolating the mean NDVI value for each vegetation type to all grid cells identified in the data set as that type. Figure 3 shows the latitudinal variation of the zonally averaged NDVI using the full-resolution NDVI data set as well as for the three reconstructed distributions. The reconstructed distributions dearly depart from the original distribution. The difference is larger between the profiles calculated from the full-resolution NDVI data set and the reconstructed NDVI data set calculated from the 178 vegetation types than among those calculated from the 17&, 32, 29, and 11 vegetation types.
Global distributions of most land surface properties used in atmosphere-biosphere models are not available. Tabular data exist for some of the properties. For example, Whittaker and Likens [1975] provide widely used parameter values typical of major vegetation types. To obtain the global distribution, a common practice is to assign the parameter value to a vegetation type and extrapolate the value over the geographic distribution of the vegetation type using a digital vegetation distribution. This procedure requires a translation and matching of the vegetation definitions in the tabular and geographic databases. The results are illustrated in the following examples.
We extrapolated tabular data for three properties, biomass for 11 vegetation types [Reiners, 1973], leaf area index for 15 vegetation types [Lieth, 1975], and July-September albedo for 12 vegetation types [Hummer and Reck 1979], using three vegetation distributions. The vegetation distributions were 32 vegetation types compiled from maps [Matthews, 1983], 29 vegetation types diagnosed from climate [Prentice, 1990], and 11 types deduced from seasonal variations in NDVI [DeFries and Townshend, 1994a]. The three vegetation distributions were also used to map leaf area index (LAI) from a table of LAI values for 15 vegetation types and July-August-September albedos from a table of albedo values for 12 major vegetation types. The latitudinal profiles of the three resultant albedo, LAI, and biomass distributions are shown in Figure 4. The difference among the three estimates is as high as 50% of the mean. Albedo differences of approximately 5% (absolute value) in northern hemisphere midlatitudes are thought to be significant for simulations of the physical climate system [Dickinson and Hanson, 1984].
Clearly, the choice of land surface classification scheme can significantly influence the parameter fields for atmosphere-biosphere models. Discrepancies among the land cover data sets in geographic distributions of vegetation types add problems beyond those related to the loss of information resulting from binning a spatially varying quantity into discrete classes.
Ideally, either the fluxes of water, energy, and carbon dioxide or the fields of parameters that control them would be derived directly from remote sensing or other sources. Unfortunately, current remote sensing instruments and algorithms are not sufficient far this, at least for the broad range of parameters that could be included in a model. Given that we lack the ideal data, we are faced with the alternative of identifying land surface characteristics from which parameters that control fluxes can be derived. For example, a global mapping of LAI or roughness length requires knowledge of the vegetation's growth form and leaf type [Sellers et al., 1994]. Running et al. [1994] propose that a description of global vegetation can be derived from combinations of three primary attributes of plant-canopy structure: permanence of aboveground live biomass, leaf longevity, and leaf type.
The question becomes how to most accurately map these vegetation characteristics.
We have addressed some of the problems with the assumption underlying the
approach based on discrete classes. Now, we turn to the question of which
land surface characteristics control the fluxes and assess how these or
useful surrogates of them can be mapped globally. This is the first step
in an iterative process that involves (1) identifying the ideal land surface
characterization for a model, (2) identifying limitations on what can be
provided globally, and (3) modifying the models to utilize data that are
feasible to obtain [Ustin et al., 1993].
Models of atmosphere-biosphere exchanges of water, energy, and carbon require an overlapping suite of surface and ecosystem parameters. For energy and water exchange, important parameters include surface albedo, surface roughness, and canopy conductance, with the latter influenced by soil depth and soil moisture holding capacity. For carbon exchange, the information needed is CO2 fixation through photosynthesis and release through plant respiration and microbial decomposition, as well as the sizes and dynamics of carbon pools. Important parameters include photosynthetic capacity, stomata! conductance, LAI, intercepted solar radiation, partitioning of dry matter among leaves, stems, and roots, chemical composition of the plant biomass, and a number of soil properties related to depth, aeration, moisture holding capacity, and parent material. Individual models may require more or less detail, but the current understanding of the biophysical and biochemical controls on the processes encourages most models to work from similar parameters.
The following sections discuss the potential for a geographically referenced system for characterizing global vegetation as continuous distributions of a small number of functional properties. These properties are identified in terms of their end points, such as deciduous or evergreen vegetation. In reality, each grid cell is a combination of these end points. Describing the functional properties as mixtures of these end points would allow the calculation of continuous parameter fields for biosphere-atmosphere models. In this way, the parameter field would not depend on the number of vegetation types or the spatial resolution of the land cover data set as is currently the case with land cover data sets describing a discrete number of cover types.
It is not within the scope of this paper to address parameterization methods for all atmosphere-biosphere models. Rather, the notion of continuous fields to describe the land surface is proposed as an alternative approach to discrete classes to provide a richer and more accurate description of the land surface on which improved parameterization methods for atmosphere-biosphere models could be based.
In the following five sections, we consider the functional properties
required to calculate the most important parameters for atmosphere-biosphere
models, namely, absorbed radiation, energy exchange (albedo, canopy conductance,
and aerodynamic roughness), photosynthesis and transpiration, net primary
production, and carbon and nutrient dynamics.
Since FPAR summarizes aspects of the interactions between the terrestrial surface and the downwelling radiation, it is potentially amenable to quantification on the basis of satellite measurements of the solar radiation reflected from the Earth's surface. Satellite-based estimates of FPAR typically employ the large contrast between the red and infrared reflectances of leaves, not seen in the wood or soil background. This contrast is the basis for several spectral vegetation indices and has, in addition to FPAR, been related to green biomass [Gamon et al., 1992; Tucker et al., 1986] and gross and net primary production [Goward et al., 1985; Prince, 1991a].
The relationship between vegetation indices and FPAR depends on several factors, including the choice of vegetation index and spectral bands [Sellers, 1987], the color of the soil/ surface background [Asrar et al., 1992; Choudhury, 1987 Goward and Huemmrich, 1992; Sellers, 1985], solar and sensor geometry, and vegetation characteristics, including canopy architecture and spatial heterogeneity [Begue, 1993; Huemmrich and Goward, 1992; Myneni et al., 1992]. Errors resulting from each of these complicating factors can be corrected, but the corrections are facilitated with additional information, including some that depend on vegetation type, spatial distribution of canopies, and soil background properties.
The absence of a general theory for calculating FPAR from a vegetation
index encourages, at least at this point, an empirical approach at the
global level such as that developed by Sellers et al. [1994]. This
approach has revealed that vegetation characteristics do affect the relationships
between FPAR and vegetation indices calculated from remotely sensed data.
In general, the relationship differs according to whether the vegetation
is (1) tall, broadleaf evergreen vegetation including woodland and shrubs
with ground cover, (2) broadleaf deciduous vegetation, including mixed
needle and broadleaf, (3) tall, needleleaf vegetation, including deciduous
and evergreen classes, and (4) agriculture, including all short land cover
types with the exception of shrubs with ground cover. In conclusion, satellite-based
estimates of FPAR appear to require a land surface characterization scheme
that describes each cell in terms of percentage area covered by vegetation
with different seasonality (deciduous or evergreen) and leaf type (broadleaf
or needleleaf).
Canopy conductance varies widely among vegetation types. Schulze et al. [1994] conclude that maximum canopy conductance varies in parallel to maximum leaf conductance and that maximum leaf conductance is highly correlated with maximum photosynthetic capacity. Sellers et al. [1992] showed how maximum leaf conductance and FPAR can be combined to estimate total (area-averaged) canopy conductance. Thus the determinants of photosynthetic capacity provide a useful starting point for estimating maximum canopy conductance. The most important exception involves the distinction between C3 and C4 plants, with the latter typically operating at lower conductances than the former. Among the ecosystems considered by Schulze et al. [1994], agricultural ecosystems tended to have substantially higher maximum conductances than any of the natural ecosystems, suggesting that a land surface characterization indicating extent or intensity of agriculture would be useful, especially if the associated soils information indicated the amount of fertilizer used.
Aerodynamic roughness is mainly a function of plant height, architecture,
and spacing and a weak function of LAI [Grace, 1977]. The most important
information is whether the land surface is forested. Because conifers typically
differ from broad-leaved species in the height at which the projected horizontal
wind velocity drops to zero, a land surface characterization indicating
the relative abundances of these two leaf types would be a major asset
for modeling. Additional useful information would come from distinctions
between grasslands and shrublands and between even aged and uneven aged
forests.
The most important biochemical determinant of photosynthetic capacity is the leaf nitrogen content [Field and Mooney, 1986]. Nitrogen is one of the nutrients most limiting to plant growth [Vitousek and Howarth, 1991]. Nitrogen availability can control photosynthesis at many levels, including biochemical capacity of individual leaves, canopy development, and the suite of species successful on a site [Chapin et al., 1986]. While it may ultimately be practical to quantify leaf nitrogen with remote sensing [Wickland, 1991], the best currently available predictors are combinations of measures that quantify canopy development (e.g., NDVI) and growth potential. The important characteristics tend to be whether the vegetation is woody or nonwoody, annual or perennial, and the photosynthetic pathway (C3 or C4) [Chapin, 1993].
Biochemical capacity for photosynthesis is highly correlated with maximum relative growth rate [Schulze and Chapin, 1987], with the highest levels of both in herbaceous plants characteristic of disturbed sites and the lowest in woody plants found in resource-poor sites in which low light, low nutrient availability, low or very high water availability, or low temperature strongly limit growth [Chapin, 1993]. Plants characteristic of resource poor sites appear to utilize low potential growth and photosynthesis as a mechanism for tolerating chronic resource shortages [Chapin, 1980]. In addition, such plants conserve resources, especially nutrients, through slow tissue turnover and prolonged persistence of leaves, roots, and stems. Stress tolerance minimizes sensitivity to chronic and acute resource shortages, but with the consequence that potential growth, photosynthesis, and transpiration are all constrained [Chapin, 1991].
Woodiness is the most characteristic feature of stress-resistant plants, especially in sites with low biomass. The ranges of maximum relative growth rates, photosynthesis, and conductance overlap between woody and nonwoody plants, but the overlap is modest. Therefore percentage of woody and nonwoody plants in each cell should be the first feature used to characterize the land surface for quantifying photosynthesis and transpiration.
Among woody plants, long-lived species tend to be slower growing than short-lived species, and the same pattern is generally true for herbaceous species [Grime and Campbell, 1991]. Short-lived species are successful only in habitats where they can gather enough resources to grow and reproduce in a short time. The mechanistic reason that long life implies slow growth is not clear, but the pattern is present. The pattern is perhaps most clearly expressed in the contrast between annual and perennial herbaceous plants (essentially all woody plants are perennials). Annuals tend to grow fast, construct short-lived, high-nitrogen tissues, and have high photosynthetic rates. Percentage of annuals and perennials within a cell therefore needs to be the second level of a land surface characterization.
A land surface characterization for calculating photosynthesis and transpiration
should also include photosynthetic pathway. Though C3 and C4
species overlap in essentially all physiological functions, differences
in temperature dependence, and nitrogen and water requirements are too
consistent to omit from a land surface characterization scheme [Pearcy
and Ehleringer, 1984].
A scheme to characterize vegetation to support NPP models should be similar to a scheme designed for photosynthesis and transpiration (above) because photosynthesis and growth are interdependent [Chapin, 1993]. Information on growth form (tree, shrub, herb, grass) is additionally important for NPP because carbon loss through respiration varies as a function of both plant growth and plant biomass [Amthor, 1989].
Several global models utilize a formulation in which NPP is calculated from a product of IPAR and a radiation use efficiency [Field et al., 1994]. Measured radiation use efficiencies are typically higher in crops with the C4 than the C3 photosynthesis pathway [Sinclair and Horie, 1989], and higher in annual crops than in natural grasslands or forests [Ruimy et al., 1994], emphasizing the C3-C4 and annual-perennial distinctions discussed in relation to photosynthesis and transpiration.
The role of woody biomass as a control on NPP is difficult to assess at the global scale. Forest NPP tends to be higher than grassland NPP [Chapin, 1993], though this may reflect the frequent dominance of woody plants on richer sites more than inherent differences in growth potential. As for photosynthesis, extended tissue longevity and, especially, evergreen plants are more common in low-resource habitats and represent adaptations for coping with resource limitation [Chapin, 1980]. The possibility that slow growth is an adaptation to resource limitation [Chapin, 1980] also argues that evergreen vegetation should be associated with low NPP.
The primary vegetation characteristics that need to be captured for
global NPP models are woody versus nonwoody, evergreen versus deciduous,
C3 versus C4 photosynthetic pathway, and crop versus
noncrop. These characteristics will also facilitate models of APAR, nutrient
dynamics, and photosynthesis and transpiration, which often play a role
in NPP calculations.
For woody vegetation the second level of information required in a land surface characterization is disturbance history. Within a region, the amount of carbon and nutrients in biomass varies widely as a function of disturbance. Storing large amounts of carbon and nutrients in structural material requires a substantial amount of time following disturbance. Natural disturbance (windstorms, wildfire) usually occurs on a relatively fine scale (from <1 to 1000 km2), and its influence can be considered to be in a quasi-steady state over sufficiently coarse scales. In contrast, human-caused disturbance (forest harvesting, land clearing, anthropogenic fires) can cause a significant decrease in carbon in woody vegetation on continental scales [Detwiler and Hall, 1988; Skole and Tucker, 1993]. This distinction is also important because of the systematic decline in soil carbon with cultivation [Post and Mann, 1990; Schlesinger, 1986] and because the intensity of human use in cultivated areas affects nutrient transformations as well as nutrient gains, losses, and storage. Large regions that are altered by human activity should therefore be distinguished from areas with little human disturbance.
For nonwoody vegetation a land surface characterization should distinguish between annual and perennial-dominated areas. Perennial plants (most important, perennial grasses) are not readily distinguished from annuals in their peak season APAR, but they function wholly differently in the ways that they allocate carbon and nutrients and in their responses to natural and anthropogenic disturbance.
Many of the critical controls on carbon and nutrient dynamics involve
below ground processes, including nutrient uptake, root production and
turnover, and organic matter decomposition. All of these interact strongly
with soils characteristics. The properties of soils, especially vegetated
soils, are relatively inaccessible through remote sensing. There is thus
a need for improved characterization of global soils as well as for global
vegetation.
Spatially explicit information about continuous gradients and mixtures of vegetation characterized by its functional properties would improve descriptions of the spatially heterogeneous vegetation over the Earth's land surface. This approach would improve the accuracy of the land surface characterization over the traditional approach of assigning grid cells to discrete land cover types. A small suite of continuous descriptors potentially identifies more combinations of ecosystem states than even the most complex topological schemes. With continuous descriptors, there is no fixed number of vegetation classes, so the number of vegetation classes can be adjusted according to requirements of different models. However, obtaining some of these land surface characteristics will require fundamentally different approaches to collecting information, including combinations of remote sensing and ground-based efforts.
The information about the vegetated land surface required for calculations of FPAR, albedo, canopy conductance, roughness, photosynthesis and transpiration, net primary production, and carbon and nutrient dynamics could largely be met by geographically referenced information on continuous distributions of a relatively small number of vegetation characteristics (Table 2). The most important vegetation characteristics are the spatial extent of woody plants (gradations between bare ground, herbaceous plants, and trees and/or shrubs), seasonality of photosynthetic activity of woody vegetation (deciduous, evergreen), leaf type (broadleaf, coniferous), photosynthetic pathway of nonwoody vegetation, plant longevity of nonwoody vegetation (annual, perennial), and disturbance type (e.g., cultivation, fire history). Many of these distinctions are also called for by Running et al. [1994]. In continuous fields describing these characteristics, each grid cell would be characterized by percentage of land surface area occupied by vegetation with the characteristic.
Ideally, these required land surface characteristics would be obtained from remote sensing, so that the parameter fields would be internally consistent. Some of the characteristics, such as a distinction between deciduous and evergreen vegetation, woody and nonwoody vegetation, and to some extent, broadleaf and needleleaf vegetation, can be remotely sensed based on temporal dynamics in the NDVI (e.g., mean NDVI, amplitude in NDVI temporal profile, greenness-duration), red and near-infrared reflectances, and land surface temperature [DeFries et al., 1995; Nemani et al., 1993; Running et al., 1994]. Other parameters, such as a distinction between C3 and C4 vegetation, cannot be directly measured with remote sensing but can be derived from climatic parameters (e.g., grassland locations with warm, wet climates are favorable to C4 physiology) (G. J. Collatz and J. A. Berry, manuscript in preparation, 1995). Still other distinctions, such as whether nonwoody vegetation is annual or perennial, would probably need to rely on ground-based information (Table 2). Methodologies to develop continuous fields for each of the land surface characteristics are likely to be unique to the particular characteristic.
In order to use remote sensing to obtain continuous fields, substantial
methodological development is required. This paper provides examples of
one possible approach, linear unmixing that has been applied at local and
regional scales to obtain continuous fields of vegetation attributes (Plate
3 and Figure 2). In order for this approach
to be globally applicable, globally valid "pure end-members" for each vegetation
characteristic need to be identified. Other methods in addition to linear
unmixing methods need to be investigated as well.
Amthor, J. S., Respiration and Crop Productivity, 215 pp., Springer-Verlag, New York, 1989.
Asrar, G., R. B. Myneni, and B. J. Choudhury, Spatial heterogeneity in vegetation canopies and remote sensing of absorbed photosynthetically active radiation: A modeling study, Remote Sens. Environ., 22, 209-233, 1992.
Badwhar, G. D., Use of Landsat-derived profile features for spring small-grains classification, Int. J. Remote Sens., 5, 783-797, 1984.
Begue, A., Leaf area index, intercepted photosynthetically active radiation, and spectral vegetation indices: A sensitivity analysis for regular dumped canopies, Remote Sens. Environ., 46, 45-59, 1993.
Box, E. O., Macroclimate and Plant Forms: An Introduction to Predictive Modeling in Phytogeography, Dr. W. Junk, Norwell, Mass., 1981.
Chabot, B. F., and H. A. Mooney, Physiological Ecology of North American Plant Communities, Chapman and Hall, New York, 1985.
Chapin, F. S., III, The mineral nutrition of wild plants, Annul Rev. Ecol. Syst., 11, 233-260, 1980.
Chapin, F. S., III, Integrated responses of plants to stress, Bioscience, 41, 29-36, 1991.
Chapin, F. S., III, The functional role of growth forms in ecosystem and global processes, in Scaling Physiological Processes: Leaf to Globe, edited by J. R. Ehleringer and C. B. Field, pp. 287-312 Academic, San Diego, Calif, 1993.
Chapin, F. S., III, P. M. Vitousek, and K. Van Cleve, The nature of nutrient limitation in plant communities, Am. Natur., 127, 48-58, 1986.
Choudhury, B. J., Relationships between vegetation indices, radiation absorption and net photosynthesis evaluated by a sensitivity analysis, Remote Sens. Environ., 22, 209-233,' 1987.
Cohen, S. B., The Oxford World Atlas, 190 p p., Oxford Univ. Press, New York, 1973.
Conese, C., G. Maracchi, F. Miglietta, F. Maselli, and V. M. Sacco Forest classification by principal component analyses of TM data, Int. J. Remote Sens., 9, 1597-1612, 1988.
DeFries, R. S., and J. G. R. Townshend, NDVI-derived land cover classification at global scales, Int. J. Remote Sensing, 15(17), 35673586, 1994a.
DeFries, R. S., and J. R. G. Townshend, Global land-cover: Comparison of ground-based data sets to classifications with AVHRR data, in Environmental Remote Sensing From Regional to Global Scales, edited by G. Foody and P. Curran, pp. 84-110, John Wiley, New York, 1994b.
DeFries, R. S., and J. R. G. Townshend, An initial coarse resolution NDVI-derived global land cover classification, in ISLSCP Initiative 1: Global Data Sets for Land-Atmosphere Models 1987-1988, CD Rom, vols. 1-5, edited by B. W. Meeson et al., NASA, Greenbelt, Md., 1995.
DeFries, R., M. Hansen, and J. Townshend, Global discrimination of land cover types from metrics derived from AVHRR Pathfinder data, Remote Sens. Environ., in press, 1995.
Demetriades-Shah, T. H., M. Fuchs, E. T. Kanemasu and I. Flitcroft A note of caution concerning the relationship between cumulated intercepted solar radiation and crop growth, Agric. Forest Meteorol., 58, 193-207, 1992.
Desjardins, R. L, P. H. Schuepp, J. I. MacPherson, and D. J. Buckley, Spatial and temporal variations of the fluxes of carbon dioxide and sensible and latent heat over the FIFE site, J. Geophys. Res., 97, 18,467-18,475, 1992.
Detwiler, R. P., and C. A. S. Hall, Tropical forest and the global carbon budget, Science, 239, 42-47, 1988.
Dickinson, R. E., Land processes in climate models, Remote Sens. Environ., 51, 27-38, 1995.
Dickinson, R. E., and B. Hanson, Vegetation-albedo feedbacks, in Climate Processes and Climate Sensitivity, Geophys. Monogr. Ser., vol. 29, edited by J. Hansen and T. Takahashi, pp. 180-186, AGU Washington, D. C., 1984.
Field, C., and H. A. Mooney, The photosynthesis-nitrogen relationship in wild plants, in On the Economy of Plant Form and Function, edited by T. J. Givnish, pp. 25-55, Cambridge Univ. Press, New York, 1986.
Field, C. B., J. T. Randerson, C. M. Malmstrom, and C. S. Potter, Ecological controls on net primary production: Lessons from global models, in Scaling Up, edited by P. van Gardingen, G. Foody, and P. Curran, Br. Soc. for Exp. Biol., London, England, 1994.
Fosberg, F. R., A classification of vegetation for general purposes, in Guide to the Checksheet for IBP Areas, IBP Handb. 4, edited by G. F. Peterken, pp. 73-120, Blackwell, Cambridge, Mass., 1967.
Friedlingstein, P., C. Delire, J. F. Muller, and J. C. Gerard, The climate induced variation of the continental biosphere: A model simulation of the last glacial maximum, Geophys. Res. Lett., 19, 897-900, 1992.
Fuller, R. M., R. J. Parsell, M. Oliver, and G. Wyatt, Visual and computer classifications of remotely sensed images: A case study of grasslands in Cambridgeshire, Int. J. Remote Sens., 10, 193-210 1989.
Gamon, J. A., C. B. Field, D. A. Roberts, S. L. Ustin, and R. Valentini, Functional patterns in an annual grassland during and AVIRIS overflight, Remote Sens. Environ., 44, 239-253, 1992.
Gates, D. M., Biophysical Ecology, 611 pp., Springer-Verlag, New York, 1980.
Goward, S. N., and K. F. Huemmrich, Vegetation canopy PAR absorptance and the normalized difference vegetation index: An assessment using the SAIL model, Remote Sens. Environ., 39, 119-140, 1992.
Goward, S. N., C. J. Tucker, and D. G. Dye, North American vegetation patterns observed with NOAA 7 advanced very high resolution radiometer, Vegetation, 64, 3-14, 1985.
Goward, S. N., S. Turner, D. G. Dye, and S. Liang, The University of Maryland improved Global Vegetation Index product, Int. J. Remote Sens., 15(17), 3365-3396,1994.
Grace, J., Plant Responses to Wind, Academic, San Diego, Calif., 1977.
Grime, J. P., and B. D. Campbell, Growth rate, habitat productivity and plant strategy as predictors of stress response, in Response of Plants to Multiple Stresses, edited by H. A. Mooney, W. E. Winner, and E. J. Pell, pp., Academic, San Diego, Calif., 1991
Grisebach, A., Die Vegetatia der Erde nach ihrer klimatischen Anordnung, vols. 1 and 2, Wilhelm Engelmann, Leipzig, 1872.
Hanan, N. P., S. D. Prince, and P. H. Y. Hiernaux, Spectral modelling of multicomponent landscapes in the Sahel, Int. J. Remote Sens., 12 1243-1258, 1991.
Harmon, M. E., et al., Ecology of coarse woody debris in temperate ecosystems, Adv. Ecol. Res., 15, 133-302, 1986.
Hill, J., and J. Megier, Regional land cover and agricultural area statistics in the Departement Ardeche, France, by use of thematic mapper data, Int. J. Remote Sens., 9, 1573-1596, 1988
Holdridge, L. R., Determination of world formations from simple climatic data, Science, 105, 367-368, 1947.
Huemmrich, K. F., and S. N. Goward, 1992, Spectral vegetation indexes and the remote sensing of biophysical parameters, in Proceedings of the International Geoscience and Remote Sensing Symposium, pp. 1017-1019, Inst. of Elec. and Electron. Eng., New York, 1992.
Hummel, J., and R. Reck, A global surface albedo model, J. Appl. Meteorol., 18, 244, 1979.
Kimes, D. S., Radiative transfer in homogeneous and heterogeneous vegetation canopies, in Photon-Vegetation Interactions, edited by R. B. Myneni and J. Ross, pp. 339-388, Springer-Verlag, New York,
Koomanoff, V. A., Analysis of global vegetation patterns: A comparison between remotely sensed data and a conventional map, Master's thesis, Dep. of Geogr., Univ. of Md., 1989.
Koppen, W., Das geographische system der klimate, in Handbuch der Klimatotogie 1(C), edited by W. Koppen and R. Geiger, Gebr, Borntraeger, Berlin, 1936.
Korner, C. H., J. A. Scheel, and H. Bauer, Maximum leaf diffusive conductance in vascular plants, Photosynthetica, 13, 45-82, 1979.
Kuchler, A. W., Vegetation Mapping, John Wiley, New York, 1967.
Kumar, M., and J. L. Monteith, Remote sensing of crop growth, in Plants and the Daylight Spectrum, edited by H. Smith, pp. 133-144, Academic, New York, 1981.
Li, X., and A. H. Strahler, Geometric-optical bidirectional reflectance modeling of a conifer forest canopy, IEEE Trans. Geosci. Remote Sens., GE-24, 906-919, 1985.
Li, X., and A. H. Strahler, Geometric-optical modeling of a conifer forest canopy, IEEE Trans. Geosci. Remote Sens., GEE-23, 508-514, 1986.
Lieth, H., Primary production of major vegetation units of the world, in Primary Productivity of the Biosphere, edited by H. Lieth and R. H. Whittaker, pp. 203-215, Springer-Verlag, New York, 1975.
Los, S., C. Justice, and C. J. Tucker, A global 1 X 1 degree NDVI data set for climate studies derived from the GIMMS continental NDVI data, Int. J. Remote Sens., 15(17), 3493-3518, 1994.
Loveland, T. R., J. W. Merchant, D. O. Ohlen, and J. F. Brown, Development of a land-cover characteristics database for the conterminous U.S., Photogramm. Eng. Remote Sens., 57, 1453-1463 1991.
Matthews, E., Global vegetation and land use: New high resolution data bases for climate studies, J. Clim. Appl. Meteorol., 22, 474-487,
Monteith, J. L., Climate and the efficiency of crop production in Britain, Philos. Trans. R. Soc. London B. 281, 277-294, 1977.
Mooney, H. A., and S. L. Gulmon, Environmental and evolutionary constraints on photosynthetic characteristics of higher plants, in Topics in Plant Population Biology, edited by O. T. Solbrig et al., pp 316-337, Columbia Univ. Press, New York, 1979.
Mueller-Dombois, D., Classification and mapping of plant communities: A review with emphasis on tropical vegetation, in SCOPE 23: The Role of Terrestrial Vegetation in the Global Carbon Cycle, edited by G. M. Woodwell, pp. 21-90, John Wiley, New York, 1984.
Mueller-Dumbois, D., and H. Ellenberg, Aims and Methods of Vegetation Ecology, John Wiley, New York, 1974.
Myneni, R. B., G. Asrar, D. Tanre, and B. J. Choudhury, Remote sensing of solar radiation absorbed and reflected by vegetated land surfaces, IEEE Trans. Geosci. Remote Sens., 30(2), 302-314, 1992.
Nemani, R., L Pierce, S. Running, and S. Goward, Developing satellite-derived estimates of surface moisture status, J. Appl. Meteorol., 32, 548-557, 1993.
Olson, J. S., J. Watts, and L Allison, Carbon in live vegetation of major world ecosystems, Rep. W-7405-ENG26, U.S. Dep. of Energy, Oak Ridge Natl. Lab., Oak Ridge, Tenn., 1983.
Pearcy, R. W., and J. Ehleringer, Comparative ecophysiology of C3 and C4 plants, Plant Cell Environ., 7, 1-13, 1984.
Pittick, A. B., The effect of changing climate on Australian biomass production--A preliminary study, Clim. Change, 8, 243-255, 1986.
Post, W. M., and L. K Mann, Changes in soil organic carbon and nitrogen as a result of cultivation, in Soils and the Greenhouse Effect, edited by A. F. Bouwman, pp. 401-406, John Wiley, New York, 1990.
Prentice, I. C., W. Cramer, S. P. Harrison, R. Leemans, R. A. Monserud, and A. M. Soloman, A global biome model based on plant physiology and dominance, soil properties and climate, J. Biogeogr., 19, 117-134, 1992.
Prentice, K. C., Bioclimatic distribution of vegetation for general circulation model studies, J. Geophys. Res., 95, 11,811-11,830, 1990.
Prince, S. D., A model of regional primary production for use with coarse resolution satellite data, Int. J. Remote Sens., 12, 1313-1330, 1991a.
Prince, S. D., Satellite remote sensing of primary production: Comparison of results from Sahelian grasslands 1981-1988, Int. J. Remote Sens., 12, 1301-1331, 1991b.
Raich, 1. W., and W. H. Schlesinger, The global carbon dioxide flux in soil respiration and its relationship to vegetation and climate, Tellus, 44B, 81-99, 1992.
Reiners, W. A., Terrestrial detritus and the carbon cycle, Brookhaven Symp. Biol., 25, 303-327, 1973.
Reiners, W. A., Achievements and challenges in forest energetics, in Concepts of Ecosystem Ecology, edited by L. R. Pomeroy and J. J. Alberts, pp. 75-114, Springer-Verlag, New York, 1988.
Roberts, D. A., R. O. Green, D. E. Sabol, and J. B. Adams, Temporal changes in endmember abundances, liquid water, and water vapor over vegetation at Jasper Ridge: Summaries of the Fourth Annual JPL Airborne Geoscience Workshop, pp. 153-156, Jet Propul. Lab., Pasadena, Calif. Oct. 25-29,1993.
Ruimy, A., G. Dedieu, and B. Saugier, Methodology for the estimation of terrestrial net primary production from remotely sensed data, J. Geophys. Res., 99, 5263-5284, 1994.
Running, S. W., and E. R. Hunt, Generalization of a forest ecosystem process model for other biomes, BIOME-BGC, and an application for global-scale models, in Scaling Physiological Processes: Leaf to Globe, edited by J. R. Ehleringer and C. B. Field, pp. 141-158, Academic, San Diego, Calif., 1993.
Running, S. W., T. R. Loveland, and L. L. Pierce, A vegetation classification logic based on remote sensing for use in global biogeochemical models, Ambio, 23, 77-81, 1994.
Running, S. W., T. W. Loveland, L. L. Pierce, R. R. Nemani, and E. R. Hunt Jr., A remote sensing based vegetation classification logic for global land cover analysis, Remote Sens. Environ., 51, 39-48, 1995.
Russell, G., P. G. Jarvis, and J. L. Monteith, Absorption of radiation by canopies and stand growth, in Plant Canopies: Their Growth, Form and Function, edited by G. Russell, B. Marshall, and P. G. Jarvis, pp. 21-39, Cambridge Univ. Press, New York, 1989.
Schlesinger, W. H., Changes in soil carbon storage and associated properties with disturbance and recovery, in The Changing Carbon Cycle: A Global Analysis, edited by J. R. Trabalka and D. E. Reichle, pp. 194-200, Springer-Verlag, New York, 1986.
Schulze, E.-D., Carbon dioxide and water vapor exchange in response to drought in the atmosphere and in the soil, Annul Rev. Plant Physiol., 37, 247-274, 1986.
Schulze, E.-D., and F. S. Chapin III, Plant specialization to environments of different resource availability, in Potentials and Limitations in Ecosystem Analysis, edited by E.-D. Schulze and H. Zwolfer, pp. 120-148, Springer-Verlag, New York, 1987.
Schulze, E.-D., F. M. Kelliher, C. Korner, J. Looyd, and R. Leuning, Relationship among maximum stomata! conductance, ecosystem surface conductance, carbon assimilation rate, and plant nitrogen nutrition: A global ecology scaling exercise, Annul Rev. Ecol. Syst., 25, 629-660, 1994.
Sellers, P. J., Canopy reflectance, photosynthesis and transpiration, Int. J. Remote Sens., 6, 1335-1372, 1985.
Sellers, P.J., Canopy reflectance, photosynthesis and transpiration, II, The role of biophysics in the linearity of their interdependence, Remote Sens. Environ., 21,143-183, 1987.
Sellers, P.J., J.A. Berry, G.J. Collatz, C.B. Field, and F.G. Hall, Canopy reflectance, photosynthesis and transpiration, III, A reanalysis using improved leaf models and a new canopy integration scheme, Remote Sens. Environ., 42,187-216, 1992.
Sellers, P.J., C.J. Tucker, G.J. Collatz, S.O. Los, C.O. Justice, D. A. Dazlich, and D. A. Randall, A global 1 degree by 1 degree data set for climate studies, Part 2; The generation of global fields of terrestrial biophysical parameters from the NDVI, Int. J. Remote Sens., 15(17) 3519-3545, 1994.
Shimoda H., K. Fukue, T. Hosomura, and T. Sakata, Global vegetation monitoring using NOAA GAC data, pp. 505-508, Remote Sens. for Resour. Dev. and Environ. Manage., Enschede, The Netherlands, 1986.
Sinclair, T. R., and T. Horie, Leaf nitrogen, photosynthesis, and crop radiation use efficiency: A review, Crop Sci., 29, 90-98, 1989.
Skole, D., and C. Tucker, Tropical deforestation and habitat fragmentation in the Amazon: Satellite data from 1978 to 1988, Science, 260, 1905-1910, 1993.
Smith, M. O., S. L. Ustin, J. B. Adams, and A. R. Gillespie, Vegetation in deserts: I, A regional measure of abundance from multispectral images, Remote Sens. Environ., 31,1-26, 1990a.
Smith, M. O., S. L. Ustin, J. B. Adams, and A. R. Gillespie, Vegetation in deserts: II, Environmental influences on regional abundance, Remote Sens. Environ., 31, 27-52, 1990b.
Solomon, A. M., and P. J. Bartlein, Past and future climate change: Response by mixed deciduous-coniferous forest ecosystems in northern Michigan, Can. J. For. Res., 22, 1727-1738, 1992.
Strahler, A. H., and D. L B. Jupp, Geometric-optical modeling of forests as remotely sensed scenes composed of three-dimensional, discrete objects, in Photon-Vegetation Interactions, edited by R. B. Myneni and J. Ross, pp. 415-440, Springer-Verlag, New York, 1991.
Townshend, J. R. G., C. O. Justice, and V. T. Kalb, Characterization and classification of South American land cover types using satellite data, Int. J. Remote Sens., 8,1189-1207, 1987.
Townshend, J. R. G., C. O. Justice, W. Li, C. Gurney, and J. McManus, Global land cover classification by remote sensing: Present capabilities and future possibilities, Remote Sens. Environ., 35, 243-256, 1991.
Troll, C., and K. Paffen, The seasonal climates of the earth (in German with English summary), Erkunde, 18, 5-28, 1964.
Tucker C. J., J. R. G. Townshend, and T. E. Goff, African land-cover classification using satellite data, Science, 227, 369-375, 1985a.
Tucker, C. J., C. L. Vanpraet, M. J. Sharman, and G. Van Ittersum, Satellite remote sensing of total herbaceous biomass production in the Senegalese Sahel: 1980-1984, Remote Sens. Environ., 17, 233249, 1985b.
Tucker, C. J., C. O. Justice, and S. D. Prince, Monitoring the grasslands of the Sahel, Int. J. Remote Sens., 7,1571-1582, 1986.
Unesco, International classification and mapping of-vegetation, Ecol. Conserv., 6,15-16, 1973.
Ustin, S. L., M. O. Smith, and J. B. Adams, Remote sensing of ecological processes: A strategy for developing and testing ecological models using spectral mixture analysis, in Scaling Physiological Processes: Leaf to Globe, edited by J. Ehrlinger and C. B. Field, pp. 339-357, Academic, San Diego, Calif., 1993.
Ustin, S. L., L. Duan, Q. J. Hart, and R. Haxo, Seasonal changes observed in AVIRIS images of Jasper Ridge, California, paper presented at the International Geoscience and Remote Sensing Symposium (IGARSS-94), Calif. Inst. of Technol., Pasadena, 1995.
Valentini, R., J. A. Gamon, and C. B. Field, Ecosystem gas exchange in a California serpentine grassland: Seasonal patterns and implications for scaling, Ecology, in press, 1995.
Vitousek, P. M., and R. W. Howarth, Nitrogen limitation on land and in the sea: How can it occur?, Biogeochemistry, 13, 87-115, 1991.
Waring, R. H., and W. H. Schlesinger, Forest Ecosystems: Concepts and Management, 340 pp., Academic, San Diego, Calif., 1985.
Webb, W. L., W. K. Laurenroth, S. R. Szarek, and R. S. Kinerson, Primary production and abiotic controls on forests, grasslands, and desert ecosystems in the United States, Ecology, 64, 134-151, 1983.
Whittaker, R. H., and G. E. Likens, The biosphere and man, in Primary Productivity of the Biosphere, edited by H. Lieth and R: H. Whittaker, pp. 281-302, Springer-Verlag, New York, 1975.
Wickland, D. E., Mission to planet earth: The ecological perspective, Ecology, 72,1923-1933, 1991.
Wilson, M. F., and A. Henderson-Sellers, A global archive of land cover and soils data for use in general circulation models, J. Climatol., 5, 119-143, 1985.
Woodward, F. I., and T. M. Smith, Predictions and measurements of the maximum photosynthetic rate, Amax at the global scale, in Ecophysiology and Photosynthesis, Ecol. Stud., vol. 100, edited by E. D Schulze and M. M. Caldwell, pp. Springer-Verlag, New York,
Woodward, F. I., G. B. Thompson, and I. F. McKee, The effects of elevated concentrations of carbon dioxide on individual plants, populations, communities and ecosystems, Ann. Bot., 67, suppl. 1, 23 28,
Yoder, B. J., Photosynthesis of conifers: Influential factors and potentials for remote sensing, Ph.D. dissertation, Oreg. State Univ., Corvallis, 1992.