Introduction

The Tibetan Plateau, with an average altitude of about 5000 m, is one of the most important regions in the world. It affects terrestrial ecosystems in China because of its unique location, high elevation (topography), and special climate systems (Zhang 1983). The biomes on the Tibetan Plateau are special (Zhang 1983). The reason is that the plateau biomes are not completely tundra and ice, as simulated by some global models, eg, TEM (Melillo et al 1993; Xiao et al 1998), BIOME1 (Prentice et al 1992), and BIOME3 (Haxeltine and Prentice 1996). In fact, there is special high-cold vegetation at three altitudinal levels, ie, subnival and nival (ca >5200 m above sea level [asl]); alpine meadow and scrubland, alpine steppe and shrubland, alpine desert in the central plateau, plateau temperate steppe, and plateau temperate desert in the periphery (ca 4000–5200 m asl); and tropical and subtropical mountain forests in the southern Himalayan mountains and subalpine conifer forests in the southeastern part of the plateau (ca <4000 m asl) (Editorial Committee for Vegetation of China 1980; Zhang et al 1996). The vertical and horizontal vegetation distribution on the Tibetan Plateau was simulated with climatic warming (Zhang et al 1996). Horizontal vegetation distribution is very sensitive and vulnerable to global change, mainly because the Tibetan Plateau is located in marginal land areas where the growth and distribution of plants depend heavily on local climate conditions (Zhang et al 1996). It can therefore be used as an indicator for monitoring global climate change (Zhang et al 1996).

The biomes of the world, including the Tibetan Plateau, were simulated by different global models, eg, the biogeography model (BIOME1: Prentice et al 1992), the biogeochemistry model (TEM: Melillo et al 1993; McGuire et al 1997), a combination of these two types (BIOME3: Haxeltine and Prentice 1996), and the integrated biosphere model of land surface processes, terrestrial carbon balance, and vegetation dynamics (IBIS: Foley et al 1996). Chinese biomes were recently also predicted by BIOME3 (Ni et al 1999) and TEM (Xiao et al 1998). In these simulations, the biome types of the Tibetan Plateau were mainly considered as tundra and ice/polar desert owing to prediction at global and regional scales. Nevertheless, special high-cold vegetation, as described above, does exist. Therefore, the simulation of biomes and their response to global change on the Tibetan Plateau should be improved at a local scale to avoid confusion and mistakes due to very simple biome classification.

Zhang et al (1996) studied the response of natural vegetation on the Tibetan Plateau to global climate change using several bioclimatic methods, such as the Holdridge life zone classification system, the montane vegetation belt system, the permafrost model, and the improved net primary production (NPP) model. However, Zhang et al (1996) used statistical models. The simulation of biomes was not numerically compared with the potential natural vegetation distribution. Modeled NPP was not compared with observed NPP. The scenarios used in the paper were for increments of 4°C temperature and 10% precipitation, but did not involve output of general circulation models (GCMs). The CO₂ ecophysiological effect was not taken into account.

Zheng (1996) classified the physicogeographical regions system of the Tibetan Plateau based on thermal conditions (days of above 10°C daily mean temperature and mean temperature of the warmest month), moisture regimes (annual aridity index, ie, evapotranspiration/precipitation and annual precipitation), and variation in landform (topography). This is a new physicogeographical regions system, but the ecophysiological constraints of plants, as shown in the papers of Prentice et al (1992) and Haxeltine and Prentice (1996), were not yet considered.

In this study, a simulation of biomes on the Tibetan Plateau was made using BIOME3China, an improved version of the equilibrium biosphere–atmosphere model BIOME3 (Haxeltine and Prentice 1996). The responses of biomes to changed climate and elevated CO₂ under a GCM scenario at the Hadley Centre (Johns et al 1997; Mitchell et al 1995) are also modeled. At the same time, BIOME3China simulates the annual NPP of each biome at the present climate and scenario. Permafrost at the present climate and its response to climate warming are predicted, with the simulation of vegetation and NPP on the Tibetan Plateau at a local scale.

Methods

Permafrost

Figure 2a shows a geographical distribution of frozen ground on 109 grid cells, digitized on the basis of data observed on the Tibetan Plateau. An air frost index (AFI) according to Nelson and Outcalt (1987) is often used to quantify the geographical distribution of permafrost, and is expressed as

DDF and DDT are the freezing and thawing indices (°C day) for a study area, respectively. The two indices can be calculated from DDF = –T_wL_w and DDT = T_sL_s,where

and

where T_max and T_min are, respectively, the air temperatures of the warmest and coldest months. Therefore, T and A approximately define the annual mean air temperature and the amplitude of annual temperature cycle, respectively. B is the frost angle defined as a point along the time axis at which the annual temperature curve crosses the line 0°C.

No permafrost can be observed in areas with an AFI less than 0.4, while permafrost is continuously distributed in areas where the AFI is above 0.6. In areas with an AFI between 0.4 and 0.6, the distribution of permafrost is intermittent rather than continuous.

Results

Discussion

The Tibetan Plateau is one of the Earth's unique physicogeographical regions. Generally, the spatial differentiation of physicogeographical regions on the plateau is determined mainly by topographic configuration and atmospheric circulation, warm and humid in the southeast and cold and arid in the northwest (Zheng 1996). In these districts, temperature and precipitation decrease gradually with the increment in latitude, indicating that the climatic resources controlling plant growth diminish from the southeastern regions to the northwestern regions. The reduction in climatic resources toward the northwest is the most important reason for the simplification of species complexity in vegetation and of regional differentiation in vertical vegetation (Zhang et al 1996). The vegetation type in this district changes gradually from marine humid montane (tropical seasonal and rain forest, warm-temperate broad-leaved evergreen forest, temperate deciduous forest, and conifer forest) in the southeastern region to continental semiarid montane (temperate shrubland/meadow, steppe, alpine meadow/shrubland, and steppe) in the middle region to continental arid montane (temperate desert, alpine desert, and ice/polar desert) in the northwestern region (Figure 1a). BIOME3China simulated the biome distribution with generally good agreement in central and northwestern Tibet (ΔV = 0.36 for nonforests) based only on climatic constraints (GDD0 and precipitation) and with fair agreement in the southeast (ΔV = 0.48 for forests) based on both climatic conditions and biogeochemical factors (Table 2). This vegetation pattern is similar to that modeled by Zhang et al (1996) using the Holdridge life zone classification system. The problem is that the actual vegetation distribution in Tibet is mosaic (Figure 1a), but the biome predicted by BIOME3China was mostly continuous (Figure 1b). The predicted alpine steppe biome moved further north (Figure 1b) than its actual distribution (Figure 1a). Four forest types and alpine meadow/shrubland modeled by BIOME3China were concentrated only in the southeastern region (Figure 1b), but their actual distribution could expand to the southern and central Plateau (Figure 1a). The reason is that the actual vegetation was derived from the digitized vegetation map of China. The actual distribution of vegetation on the Tibetan Plateau was determined not only by climate but also by soil, topography, and human disturbance, which is usually extreme in terms of spatial heterogeneity. On the other hand, potential vegetation driven by the interpolated continuous climatic data and by constraints in a large range would be mostly continuous when climate conditions are similar around one place.

The great expansion in human activity is causing significant changes in natural conditions. Among these changes, the increase in GHGs is expected to have great impact on climate and very important implications for the broad-scale distribution of terrestrial ecosystems (IPCC WGI 1996), especially in the mountains (Haslett 1997). A recent study by a Chinese scientist (Liu et al 1998) shows that the Plateau climate has generally been warming during this century and precipitation has generally been increasing over the last 30 years. According to climate output from the Hadley Centre GCM (Johns et al 1997; Mitchell et al 1995), temperature will increase 2–3.6°C by the end of the next century in the Plateau, and precipitation will increase by 0–300 mm in the central and eastern parts of the Plateau and decrease by 0–550 mm in the southwest (Figure 3a,b). The greatest increases in temperature, by 3.0–3.6°C, are expected to occur in the north and the southwest. An increase of 2.0–3.0°C will occur in the central and eastern Plateau (Figure 3a). The greatest increases are likely to be in the winter. The pattern of precipitation on the Tibetan Plateau might change gradually, from an increase in the north to a decrease in the southwest (Figure 3b). After climate change with CO₂ concentration of 500 ppmv, biomes of the Plateau shifted significantly to the north and northwest (Figure 1c). The ΔV value of each biome between the current climate and the scenario reflected the dissimilarity of the biome under different climatic conditions. The greater the ΔV value, the greater the dissimilarity of biome between present climatic conditions and the scenario (Sykes et al 1999). In addition to a great reduction in the area of temperate desert, alpine steppe, desert, and ice/polar desert and a large increase in the area of cold-temperate conifer forest, temperate shrubland/meadow, and temperate steppe, great change would occur in temperate shrubland/meadow, temperate steppe, temperate desert, alpine meadow/shrubland, and ice/polar desert under changed climatic conditions (Table 2). Meadow, shrubland, and steppe would increase in area because increased temperature and water stress favor C₄ grasses over C₃ woody plants. With the inclusion of physiological CO₂ effects, however, there would be a marked decrease in the area occupied by temperate desert, alpine desert, and ice/polar desert (Table 2). This may be due to C₃ plants becoming more competitive relative to C₄ grasses as well as a general reduction in stomatal conductance, which favors the dominance of woody plants by reducing water stress.

The response of life zones on the Tibetan Plateau to climate change predicted by Zhang et al (1996) showed that natural vegetation would shift northward from each present position due to climate warming. This changing pattern is quite similar to that described in the present paper. Tropical and subtropical forests and montane and alpine steppe will increase in area. Ice and polar (frigid) desert will decrease in area. These predictions are similar to those of Zhang et al (1996). The difference is that conifer forest and alpine meadow would largely decrease, while montane and alpine deserts would significantly increase according to Zhang, whereas the opposite is true in the present study. This may be because Zhang et al (1996) considered only an increase in temperature and precipitation of 4° and 10%, respectively. In the present study, however, both GCM output and CO₂ concentration were taken into account. On the other hand, the difference between biome classification in this paper and the Holdridge life zone in Zhang et al (1996) might result in dissimilarities.

Continuous global climate change is expected to have an impact not only on biome geographic distributions but also on biogeochemical cycles in the terrestrial ecosystem. In addition to simulation of biome distribution, BIOME3China also predicted NPP for each grid cell. Under present climatic conditions, tropical, warm-temperate, and temperate forests would have the highest annual NPP (ca 220–810 gC/m²/y). Temperate meadow/shrubland, steppe, and desert would have an intermediate NPP (250–290 gC/m²/y). The lowest production was found in alpine meadow/shrubland, steppe, and desert, whereas predicted annual NPP for ice/polar desert was zero (Figure 4). Although there was a lack of NPP data for meadow, steppe, and desert on the Plateau, 160 forest NPP data were obtained from the Chinese Forest NPP database (Ni and Zhang unpublished manuscript). A comparison was made between predicted annual NPP and 160 field forest NPP mea surements (Figure 5). There are many problems associated with such comparisons, including the quality of the data and the fact that the model simulated the average NPP over a grid square, while NPP measurements were made at particular sites (Haxeltine and Prentice 1996). Some sites with positively measured NPP had zero predicted NPP on the Tibetan Plateau (Figure 5) owing to very high elevation with very low temperatures. Despite these problems, the resulting comparison showed an agreement between the predicted NPP and measurements with a linear regression (Figure 5).

If climate change and CO₂ concentration enrichment occur at the end of the next century, production in terrestrial ecosystems will be changed correspondingly. The pattern of predicted annual NPP in the scenario with a CO₂ concentration of 500 ppmv was the same as that for the present climate, but the NPP of all biomes increased significantly (Figure 4). This may be largely because of the positive effects of increased CO₂, precipitation, and temperature in most of regions on the Tibetan Plateau.

The frozen ground is formed and maintained by severe climate in winter. It is very sensitive to temperature change. Therefore, the area of frozen ground, especially the permafrost area, would be expected to decrease considerably in response to global climatic warming. Frozen ground has two environmental impacts. The first is a release of soil carbon to the atmosphere. Warming of frozen soils can thaw continuous permafrost, increase drainage, and hasten decomposition (Oechel et al 1994). The second is a northward movement of vegetation zones and permafrost in step with shifting isotherms (Melillo et al 1996). Simple climate/permafrost correlations applied to doubled CO₂ GCM simulations forecast permafrost well north of current distributions in the world (Anisimov and Nelson 1996). The movement (expansion and/or decline) of permafrost under changing climates, therefore, could be used as an indicator of changes in global climate and would help provide an understanding of how terrestrial ecosystems will respond to changes in the global climate.

The permafrost area on the Tibetan Plateau is about 1,500,000 km², which is about 70% of the frozen land area in China (Zhang et al 1996). Permafrost plays a very important role in determining vegetation distribution and dynamics and in indicating changes and fluctuations in climate. Under present climatic conditions, observed data showed that the continuous permafrost occurred mainly in central and northwestern Tibet, ie, in about 30% of the total permafrost area. The alpine permafrost occurred in northeastern and southwestern peaks. The seasonal frozen ground was found in the north and east, and the instantaneous and isolated permafrost occurred in the south (Figure 2a). The permafrost pattern predicted by AFI is similar to observed data, apart from the alpine permafrost occupying part of the continuous permafrost areas in the central plateau and apart from isolated permafrost taking the place of the seasonal frozen ground in the eastern margin (Figure 2b). As with climatic gradient, there is permafrost in the northwest but not in the southeast. The permafrost pattern with climate gradient determines the vegetation distributions on the Tibetan Plateau. Namely, forests and some shrubland/meadow occur in the nonpermafrost areas; temperate steppe, desert, and some shrubland/meadow occur on the seasonal frozen ground; and alpine steppe, desert, and ice/polar desert occur in the continuous permafrost areas.

Under climate warming conditions at the end of the next century, the boundary between continuous and discontinuous permafrost would shift toward the north of the Plateau by about 1–2° in latitude (Figure 2c). The continuous permafrost would largely disappear because increased temperatures would melt the underground ice, and about 70% of the total area on the Tibetan Plateau (Figure 2c) would be occupied by nonpermafrost area (instantaneous and isolated permafrost). The movement of permafrost would happen with the shift of vegetation zones to the north. Disappearance of permafrost and expansion of nonpermafrost would accelerate desertification in the plateau. This may be mainly because temperature increases would result in an increase in evaporation, although ice below ground would melt and precipitation would increase in most regions.

High mountains as vulnerable environments will also be impacted first and with the greatest intensity by anthropogenically induced climate change as a result of both physiological responses arising from the increased CO₂ available for photosynthesis and the responses of organisms to physical climatic changes (Haslett 1997). The equilibrium pattern of Tibetan biomes and the equilibrium responses of biomes to climate change and CO₂ enrichment were modeled using the improved BIOME3China model, but the dynamics of vegetation and carbon budget on the Plateau still need to be explicitly simulated. This is because there is a lack of information on the responses of phenology, growth, and reproduction in major vascular plant species to climate variations and environment manipulations in the Tibetan Plateau. The vertical changes in vegetation on the Tibetan Plateau are also very important. BIOME3China did not model these because of the lack of accurate climatic data in all high mountains and the shortcomings of BIOME3China itself in modeling vertical vegetation changes. This will clearly require further efforts.