2. BIOSPHERE

2.1 Introduction

This chapter is divided into two parts. After defining the fundamental relationships between climate and the biosphere, it examines the interactions between the vegetation, soils, land cover/land use and the climate system. It then discusses the nature of explanatory and predictive models of these climate and ecosystem interactions. Each part highlights some (but not all) of the reasons for measurement of specific environmental variables.

2.2 Understanding Climate Impact on the Biosphere and Feedbacks to Climate

On the scale of regions to continents (e.g., >10⁵ km²), climate is the dominant arbiter of the productivity and distribution of vegetation. For example, minimum winter temperatures constrain broad-leaved deciduous trees from invading and excluding higher latitude boreal evergreen conifers at their mutual border, and from being invaded by more productive subtropical species at their lower latitude border (Prentice, et al., 1992). Soil moisture determines whether forest or steppe occupies a region. Its seasonal availability can determine whether forests are dominated by needle-leaved evergreens or by broad-leaved deciduous trees (Waring and Schlesinger, 1984) and whether steppes are dominated by grasses or by shrubs (Haxeltine, et al, 1996).

Vegetation and climate, in turn, combine with substrate features to determine soil characteristics. While climate directly determines rates of substrate breakdown, organic matter decomposition and soil erosion, climate indirectly determines other soil features through its control of vegetation. For example, it permits the dominance of conifers that acidify soils, or of steppe grasses which contribute to high concentrations of soil organic material.

Feedbacks of vegetation and soils to climate are lesser but important features of climate-biosphere interactions. The differences in albedo between one soil or vegetation and another can have significant effects on the radiation balance (Robinson, et al., 1983) and hence on climate, while the evapotranspiration regime of different vegetation types can help determine atmospheric moisture concentrations, cloud properties (Dickinson, et al., 1991), and the regional contribution of soil moisture to precipitation (Lettau, et al, 1979; Shukla, et al, 1990). On smaller scales of a few 10s of km², soil, topography and land use history combine to determine productivity and distribution of vegetation and the land use options available at any specific location.

Vegetation productivity refers to rates and amounts of carbon fixed from the atmosphere. Temperature and soil moisture are known to exert fundamental control of photosynthesis (carbon fixation) and respiration (carbon release), both of which respond differently to those climate variables (Schimel, et al, 1995). Also, the rate of carbon accumulation increases at higher CO₂ concentrations, at least in short-term greenhouse experiments, although there is yet no direct evidence that it does so in natural vegetation. Thus, the rate at which climate-controlling CO₂ accumulates in the atmosphere will be directly related to changing temperature and precipitation patterns and possibly, to changing atmospheric CO₂ concentrations.

Consideration of vegetation productivity leads directly to the examination of biogeochemical cycles. These include the cycling of the elements which are essential to life and which are strongly influenced by living organisms, between the biosphere, atmosphere, hydrosphere and lithosphere. The carbon (discussed above), nitrogen, phosphorus and sulphur cycles are most important because they underlay all primary production and trace gas emission models.

Nitrogen is the principal limitation for the operation of the carbon cycle in most terrestrial ecosystems, including agroecosystems. N₂O and NO are climatically important trace gases. Phosphorus is the principal constraint for the operation of the carbon cycle in many tropical terrestrial ecosystems and most aquatic ecosystems. Phosphorus also controls the input of nitrogen to ecosystems via biological nitrogen fixation. The sulphur cycle operates in tandem with nitrogen and carbon. Sulphur aerosols are very important to the radiative balance of the earth and are the main component of acid rain.

The impact of the four cycles on the climate occurs via the emissions of radiatively-active trace gases as a result of biogeochemical cycling (CO₂, CH₄, N₂O; ozone precursors such as NO and CO; and aerosols) and via the land surface characteristics such as biomass and leaf area which are constrained by biogeochemical considerations. Since biogeochemical cycling is strongly influenced by climate, this constitutes one of the major avenues for both impacts and feedbacks.

It is obviously impossible to separate the vegetation structure and processes of productivity from the atmospheric, soil, and hydrological processes produced by changes in land cover and land use. Shifts in temperature and precipitation will affect energy, carbon and water balance, which in turn will cause changes in soil properties and in distribution and density of vegetation. Climate will have a major influence on land use change, and both of these directly affect the vegetation patterns. In turn, shifts in vegetation patterns feed back directly to hydrological cycles, radiative properties of the atmosphere, and the climate system.

Vegetation

Most of the impacts of climate change on the terrestrial biosphere are expected to manifest themselves through changes in vigour of species populations and the resulting changes in the composition, structure and distribution of vegetation; and through changes in biogeochemical, energy, and water fluxes between ecosystems and the atmosphere. Climate change will affect the terrestrial biosphere through changes in the regional energy balance (Dickinson, 1983) and associated changes in the regional water balance (Eagleson, 1986). Temperature, humidity, wind, precipitation and runoff affect the water balance by influencing soil moisture, evaporation and transpiration. Hence, these are critical observations for understanding the effect of climate change on terrestrial ecosystems and on subsequent feedbacks to climate. Seasonal shifts in the water balance may occur because of changes in precipitation patterns (Lettenmaier and Sheer, 1991) or volume (Dolph, et al, 1992). For example, a significant decrease in runoff and soil moisture will affect soil properties such as carbon content, nitrogen and phosphorus.

Changes in the hydrological cycle also influence moisture fluxes between vegetation and the atmosphere. These fluxes strongly depend on the area of photosynthetic surface available to accept radiation, to transpire water, and so on. The usual surface measure is the leaf area index (LAI) which is defined as one half of the total projected green leaf per unit ground area. LAI is hypothesized to describe a fundamental property of the plant canopy in its interaction with the atmosphere, especially as it concerns the momentum, radiation, energy, and gas exchange of the atmosphere (Monteith and Unsworth, 1990). Typically, the LAI will vary along moisture gradients, being high enough to ensure a maximum amount of photosynthetic surface and productivity, but low enough to prevent damage to leaves and plants during periods of moderate drought conditions (Woodward, 1987).

Since the structure of vegetation (species composition, density and three-dimensional configuration) is controlled by a number of environmental factors such as seasonal water availability, temperature, length of the growing season and nutrient availability, climate change will affect the structure as well as the distribution of the system components. For example, as water availability decreases, LAI should decline, thus decreasing productivity but conserving water. Conversely, the parallel increase in concentrations of CO₂ may actually stabilize or increase LAI by increasing water use efficiency of leaves (Melillo, et al., 1996). At the stand level, radiometer analyses, or allometric equations linking tree diameters to leaf area, permit estimates of LAI. In estimating variables such as LAI and net primary production (NPP) over large areas, remote sensing should play an increasingly important role. Spectral vegetation indices such as the Normalized Difference Vegetation Index (NDVI) derived from ratios of near infrared and red spectral reflectances or radiances have correlated with estimated LAI especially well (Sellers, 1989) and future availability of data from active microwave sensors (radars) should be useful in determining water-related properties of vegetation.

Recent model projections of global vegetation distributions controlled by climate change suggest that up to half of the natural vegetation of the earth could undergo change from a climate characteristic of one biome type to that of another (Kirschbaum, et al., 1996). Vegetation responses to changes of this magnitude will obviously affect biodiversity through habitat modification, and it will affect ecosystem function through increasing mortality and decreasing productivity. Hence, variables describing habitat deterioration and those measuring mortality could be particularly useful. Because especially slowly growing forests are unlikely to keep pace with rapidly shifting climates, the monitoring programme should take cognizance of processes by which species migrate (seed transport and establishment followed by growth, maturity, seed production and seed transport again) and of the environmental constraints on both migration and subsequent forest development. Seedling variations should be monitored because seedling establishment is usually the limiting process in both forest development and in immigration.

The terrestrial role in the global carbon cycle defines a critical set of measurements. The terrestrial biosphere stores three times as much carbon as the atmosphere. On an annual basis, the flux of carbon between the atmosphere and land surfaces is equal to about 20% of the atmospheric content. Thus, the CO₂ flux to the atmosphere is an important observation. The standing stock of ecosystem carbon at any given time is highly variable from place to place but can be broadly estimated from carefully placed measurements of the biomass (both above and below ground) and the necromass. In addition, wildfire releases measurable quantities of carbon from the land surface (Goldammer, 1990). Hence, carbon standing stocks, necromass and areas burned are important observations for understanding the carbon cycle.

To understand, model or monitor carbon or any of the other three biogeochemical cycles (N, P, and S), information needs to be obtained on the size of the pools, the magnitude of the fluxes, and the factors which control the fluxes. In practice, all cycles can be simplified to some degree by considering only the major fluxes and pools. Generally, only one flux will be 'rate-limiting' at a time, and its controlling factors need to be known in detail. For example, carbon assimilation by plants is a rate-limiting step in the carbon cycle, while nitrogen mineralization is limiting in the nitrogen cycle.

The fluxes of CO₂ are largely controlled by photosynthesis and respiration (autotrophic and heterotrophic). Temperature and moisture are critical elements that affect photosynthesis and respiration which in turn are important in determining the NPP. CO₂, concentrations may also be found to control NPP. NPP along with ecosystem respiration determine the net ecosystem production (NEP). NEP characterizes the status and ability of an ecosystem to store carbon and consequently is critical in understanding the carbon cycle in terrestrial ecosystems. To detect changes in the storage of carbon, observations on NPP and particularly NEP are required. Existing terrestrial ecosystem sites routinely make these measurements, but space-based observations are required to obtain NPP/NEP estimates over large areas.

Changes in terrestrial vegetation composition. distribution, and structure affect terrestrial carbon storage and thus provide a feedback to climate. Equilibrium analyses using different vegetation models and climate scenarios generally show that under doubled CO₂ conditions a terrestrial biosphere without agricultural land use will store more carbon than today, and will act as a negative feedback to climate change (Smith, et al., 1992); with agricultural land uses, the terrestrial biosphere cannot store as much carbon as today and acts as a positive feedback on climate change (Solomon, et al., 1993; Dixon, et al., 1994). Analyses of the more realistic situation which includes transient response of vegetation to climate change suggest that vegetation redistribution could also lead to a transitional release of CO₂ to the atmosphere for a period of about 200 to 400 years (King and Neilson, 1992; Smith and Shugart, 1993; Solomon, 1996).

In addition to carbon, methane, and nitrous oxides also clearly depend on terrestrial ecosystem processes and will respond to changes in vegetation types. For example, the production of methane from wetland areas depends on the duration of inundation and temperature, both of which are likely to change with a changing climate. Currently, we lack sufficient data to accurately predict the net effect of climate on the production of methane on a global basis. Hence, at least the areas covered by freshwater wetlands (streams, lakes, bogs, marshes, swamps) should be monitored through remote sensing.

Soils

Soils are a key component of terrestrial ecosystems. They are important in global biogeochemical cycles, accounting for 2/3 of the terrestrial carbon stocks. Their physical properties are critical to climate because they can control albedo and act as a porous medium for water storage and movement. Soils are the site of mineralization and provide the majority of nutrients that control plant growth. The interactions between these properties and climate determine the rates of decomposition, carbon storage, moisture availability, trace gas production, and nutrient supply. Thus, observations on soil carbon, nitrogen, phosphorus, bulk density, surface state, and particle size distribution are needed.

Changes in soil moisture, nutrients and evapotranspiration will have large impacts on biome types, since the distribution and abundance of these resources are controlled to a large extent by the volume and seasonality of available soil moisture (Woodward, 1987). Where changes in the regional water balance are significant, major shifts in vegetation patterns and condition are a likely result (Kirschbaum, et al., 1996). Modelled effects of global climate change on the water balance produced substantial loss in the amount of the forest cover in the continental U.S. (Neilson, 1995), assuming that humidity and wind are the primary mediators of evapotranspiration of soil moisture. Data documenting rooting depth are important in this regard because soil rooting depth determines the volume within which moisture, carbon and nutrients are stored to support plant growth. Adequate spatial and temporal resolution of soil moisture observations in the rooting zone are particularly hard to obtain because of the transient nature of soil moisture.

Soil scientists have developed two principal types of information which will be of value to GTOS and GCOS. These are soil maps of regional and global coverage, and chemical and physical measurements on soils. When combined, they provide important spatial information concerning soil characteristics. The shortcomings in both data sets limit their present usefulness. Currently, the best global soil map is the UNESCO-FAO Soil Map of the World, released in digital form at a scale of 1:5 million. While this represents a comprehensive and valuable resource, at this scale most mapping units consist of a mixture of soil types, thus soil conditions cannot be characterized with sufficient accuracy.

Land cover and land use

Changes in land cover and land use, including agricultural use, can both drive changes in the climate system and result from climatic change. The physical and biological characteristics of the soil and vegetation surface influence the absorption of energy at the surface, which in turn influences the relative amounts of latent and sensible heat exchange, local hydrology and transfer of momentum from the atmosphere (Dickenson, 1983). Human activities that modify the Earth's vegetation cover, for example by changing agricultural practices, deforestation and biomass burning, represent a significant influence on the global climate. Land use also has a significant effect on the carbon fluxes that influence atmospheric CO₂, concentration. For example, tropical deforestation in the 1980s was estimated to contribute between 2.0 and 2.8 Gt carbon per year (Houghton, et al., 1990), or approximately 25-35% of the total biosphere emissions of carbon.

Land cover information is needed by existing general circulation models (grid scale 250 km). For those climate models which will be implemented in the next five years the identification of incipient change requires a stratified approach, with some observations at much higher spatial scales (20-500 m) over limited areas. Local models of terrestrial biosphere dynamics and interactions with climate also depend on land surface processes at the same fine resolutions, but globally-comprehensive biosphere models require land cover data at minimally 1 km covering all terrestrial areas.

Multi-temporal satellite observations can be used to define land cover and land condition. The most consistent data are from NOAA Advanced Very High Resolution Radiometer (AVHRR) satellites. Global data sets at a resolution of 8 km and a temporal resolution of 10 days are being assembled and will be used to prepare the required parameters on a continuing basis. An experimental global data set at 1 km resolution has been collected for 1992-1996, although these data will not portray all the heterogeneity of terrestrial ecosystems which occurs at finer spatial resolutions. High resolution data from sensors such as the Landsat Thematic Mapper (30 m) and the Satellite Probatoire d'Observation de la Terre (SPOT) High Resolution Visible (HRV) (10-20 m) are available in principle, but their high cost has limited their use to date. Nevertheless, this cost is much less than that of conventional monitoring methods, especially for large and remote areas. While the heterogeneity of a single ecological unit would not be described, high resolution data collected on a systematic global basis would substantially capture the heterogeneous nature of terrestrial ecosystems. Algorithms could be developed that would allow statistical interpretation of the heterogeneity of the 1 km AVHRR data sets (GCOS-15).

In summary, vegetation structure, biogeochemistry, soil condition and the energy and water balance are tightly coupled. Detecting and assessing the response of any one of these to climate change requires observations of the others. To understand the effects climate change is likely to have on vegetation and water resources, it is essential that the analysis integrate both hydrologic and biogeochemical processes. Anthropogenic influences on land cover and land use must also be considered in projecting future climate-biosphere interactions. In this regard, the implementation of this Plan should be done in cooperation with the IGBP Land Use and Cover Change (LUCC) project which is linked to the intersection of natural and human/societal research on global change.

2.3 Modelling Approaches

There are three principal types of models relevant to simulating climate-biosphere interactions at the watershed to global scales: physiognomic vegetation models (biogeographic models), vegetation succession models (gap and migration models), and biogeochemical models (NPP models). Simulating physiognomy (the species composition and 3-dimensional arrangement within plant communities) is useful in assessing the potential limits to physical feedbacks between vegetation and climate. Vegetation physiognomy affects surface albedo, roughness and evapotranspiration. Vegetation succession is an expression of temporal processes which control lagged (transient) vegetation responses to a rapid climate change. These responses must be understood a priori for proper use of monitoring data: decades and even centuries may be required before many critical vegetation responses to current changes can become evident. Biogeochemical processes, such as annual carbon cycling and primary productivity, affect trace gas concentrations in the atmosphere. Furthermore, components of all three model types are interrelated and affect each other. Biogeochemical processes are critical for determining the distribution of vegetation, just as species available and physiognomy affects biogeochemical processes. Consequently, in order to simulate climate-biosphere interactions, these three model components need to be linked together in a biologically and hydrologically consistent manner (Melillo, et al., 1996).

Biogeographic modelling approaches

Geographic redistribution of vegetation to be expected at some time following cessation of directional environmental change (Prentice and Solomon, 1990), can be projected by several available models. Commonly, an empirical classification is developed that relates derived climatic variables with biome distributions, which is then used to draw new biome distributions from a different climate distribution. The fundamental flaw of this approach is that dynamic equilibrium of vegetation distribution cannot be expected very soon, not even for centuries, following an environmental equilibrium which itself is not likely in the foreseeable future.

The Holdridge Life Zone Classification (Holdridge, 1967) was applied in a globally-comprehensive way to estimate vegetation geography (Emanuel, et al., 1985a and 1985b) and forest standing stock (Sedjo and Solomon, 1989). However, the Holdridge system and its modified versions (e.g., Prentice and Fung, 1990; Smith and Shugart, 1993) share the problem that variables with which it classifies "life zones" (annual warmth, annual precipitation) do not control the current distribution of vegetation. To solve this problem, a biogeography model was developed for assessing geographic responses of vegetation zones to expected climate changes (BIOME; Prentice, et al., 1992). Unlike other static models, it utilizes physiological thresholds of temperature minima, warmth and water stress maxima in order to define a set of plant functional types (PFTs), each characterized by the minimal set of known climate constraints and each free to recombine under different climates with PFTs which are not now associates. BIOME was used to project future biogeographic patterns and associated carbon stocks under past, present and future doubled-CO₂ climate (Prentice, et al., 1993b), and in the absence and presence of agriculture (Solomon, et al., 1993).

More recently, global integrated assessment models (GIAM) have been developed which combine biogeography changes with simulations of land use and cover, the latter as a function of changing human population size, resource demand, atmospheric chemistry and climate. The goal is to calculate the resulting atmospheric concentrations of greenhouse gases. The most comprehensive GIAM, IMAGE 2 (Alcamo, 1994; Leemans, et al, 1997) was applied in several chapters of the Intergovernmental Panel on Climate Change (IPCC) Second Assessment Report to calculate future land use effects on forest growing stocks (Solomon, et al., 1996), maximum potential carbon storage in forests grown to take up atmospheric carbon (Brown, et al, 1996), and to back-calculate carbon emission chronologies necessary to reach atmospheric carbon concentration goals at specified times in the future (Alcamo, 1994).

Successional modelling approaches

Other vegetation models have been developed to simulate vegetation transient responses. Probably most important is plant succession, a process whereby rapidly-growing shade-intolerant species are gradually replaced by more slow-growing, shade-tolerant species. During the succession. carbon storage values are thought to follow an s-curve, increasing early and after some 50-500 years, asymptotically approaching a maximum that depends on soil and climate characteristics. Forest gap models (Botkin, et al, 1972: Shugart, 1984) were developed to simulate forest succession after a tree death caused a gap in the forest canopy, a common means by which mid and low latitude forests reproduce themselves in the absence of more catastrophic forms of mortality (e.g., wind throw, wildfires, pest epidemics, etc.). These models simulate annual changes in establishment, growth and mortality of mixed species on small forest plots based on the relative abilities of tree species to compete for sunlight, water, nutrients and space. Several different forest gap models have been developed for different purposes and applied to simulate the response of forest stands to climate change (Solomon, 1986; Prentice, et al., 1993a; Bugmann and Solomon, 1996). Unfortunately, we lack the natural history information from less-studied regions required to parameterize the models (e.g., maximum known height, diameter, and age; shade tolerance, cold tolerance, drought tolerance, etc.). Hence, they are not yet ready to apply on a global scale. These parameters may be gleaned eventually from a combination of tier 4 observations with isolated data already available.

The other major transient process, species migration, is only now becoming amenable to global-scale modelling. Measurable tree migration responses will probably not become locally visible for 100 to 200 years (Solomon and Bartlein, 1992; Solomon, 1996), but their potential is great thereafter for inducing regional extinctions and reducing carbon carrying capacity. Current research (e.g., Collingham, et al, 1997; King, et al., 1997) demonstrates that development of reliable predictive models will depend on uniform global knowledge of species seed production, weights and transport modes. Predicting migration of species that depend on animal transport will also require mapped distribution of birds of the Corvidae genus (crows and jays), land use and land cover mapped at 0.2-1 km resolution, and characterization of natural histories of the species. Species depending on wind transport will also require estimates of vertical and horizontal wind structure, especially the frequency and intensity of highest wind speeds (as a proxy for up and down drafts).

Biogeochemical modelling approaches

Two conceptual approaches have been used to develop biogeochemical models: correlating net primary productivity directly with climate, and modelling ecosystem processes based on ecophysiological principles. The correlation approach is straightforward in that regression is used to relate observed net primary productivity to climate variables (Lieth and Whittaker, 1975). This approach can be used to simulate future NPP, but they provide results of unknown reliability because the relationship between the few climate variables used and NPP could change under future climates due to the complexity of climate-vegetation interactions.

More complex biogeochemical models are based on physiological principles. They assume that leaf-level processes can be scaled directly to the level of the biosphere (hence the name "big leaf models). They typically treat the carbon and nitrogen cycles (Rastetter, et al., 1991) and may incorporate the hydrologic cycle (BIOME-BGC, Running and Hunt, 1995). Approaches to scaling up have included interpolations of outputs between sites (Kittel and Coughnor, 1987); running a gridded model version based on interpolated climate data (Raich, et al., 1991); and initializing a gridded model with satellite remote-sensing data and driving it with interpolated climate surfaces (Running et al., 1989). Although these models can be used to simulate changes in NPP and biogeochemical processes under changed climates, they cannot simulate vegetation redistribution. Therefore, they cannot be used alone to either simulate biophysical feedbacks (surface roughness and albedo) to the climate system, or the carbon dynamics associated with vegetation redistribution.

A new approach combines the biogeographic models (Section 2.3) with big-leaf physiological models (VEMAP, 1995). These calculate photosynthesis, respiration, and nutrient uptake as a function of leaf area permitted by available moisture, sunlight, warmth and nutrient availability, and scaled to values characteristic of each biome (e.g., BIOME 2, Haxeltine, et al, 1996; DOLY, Woodward, et al., 1996; MAPPS, Neilson, 1995). The transient processes (Section 2.3) are not yet simulated in the integrated biogeography/geochemistry models, but rather instant migration and ecosystem development is assumed to match any climate change. Terrestrial NPP and NEP measurements will be critical to validating and modifying the models to increase their accuracy.

In summary, to understand the impacts of climate on terrestrial ecosystems, and the feedbacks from the terrestrial systems to climate, the variables listed in Table 2.1 are critical. Current models use vegetation structure, climate data, land cover and land use to infer several critical parameters such as surface roughness, albedo and evapotranspiration. The models now require more complete and consistent information than is readily available. Much of the information now in use is anecdotal and inconsistent. Techniques to provide consistent information in terms appropriate for the models are available, and the required data are being acquired, but not necessarily saved or fully processed.

Table 2.1 - Key variables for biosphere observations.