Global and Planetary Change 48 (2005) 287–302 www.elsevier.com/locate/gloplacha

Changes of major terrestrial ecosystems in since 1960

Tian Xiang YueT, Ze Meng Fan, Ji Yuan Liu

Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Jia No. 11, Datun, Anwai, 100101 Beijing, China

Abstract

Daily temperature and data since 1960 are selected from 735 weather stations that are scattered over China. After comparatively analyzing relative interpolation methods, gradient-plus-inverse distance squared (GIDS) is selected to create temperature surfaces and Kriging interpolation method is selected to create precipitation surfaces. Digital elevation model of China is combined into Holdridge (HLZ) model on the basis of simulating relationships between temperature and elevation in different regions of China. HLZ model is operated on the created temperature and precipitation surfaces in ARC/ INFO environment. Spatial pattern of major terrestrial ecosystems in China and its change in the four decades of 1960s, 1970s, 1980s and 1990s are analyzed in terms of results from operating HLZ model. The results show that HLZ spatial pattern in China has had a great change since 1960. For instance, nival area and subtropical thorn woodland had a rapid decrease on an average and they might disappear in 159 years and 96 years, respectively, if their areas would decrease at present rate. Alpine dry and cool temperate scrub continuously increased in the four decades and the decadal increase rates are, respectively, 13.1% and 3.4%. HLZ patch connectivity has a continuous increase trend and HLZ diversity has a continuous decrease trend on the average. Warm temperate thorn , subtropical wet and cool temperate wet forest shifted 1781.45 km, 1208.14 km and 977.43 km in the four decades, respectively. These HLZ types are more sensitive to than other ones. These changes reflect the great effects of climate change on terrestrial ecosystems in China. D 2005 Elsevier B.V. All rights reserved.

Keywords: Holdridge Life Zone; digital elevation model; interpolation method; spatial pattern; mean-center shift; geographical information system

1. Introduction world’s environments since early 1880s (Schouw, 1823; Griesebach, 1872; Merriam, 1892; Clements, A continuing endeavor had been made in 1916; Koeppen and Geiger, 1930; Thornthwaite, different earth surface sciences to classify the 1931). To improve the drawback of the categories of the prior systems that are too coarse and T Corresponding author. Tel.: +86 10 64889633; fax: +86 10 inapplicable to the whole world, Holdridge (1947) 64889630. devised a classification. The Holdridge Life Zone E-mail address: [email protected] (T.X. Yue). (HLZ) classification relates the distribution of major

0921-8181/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.gloplacha.2005.03.001 288 T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302 terrestrial ecosystems (termed life zones) to the seem more important than direct physiological ones at bioclimatic variables. It is a scheme that uses the the life zone level. The vegetation distribution is three bioclimatic variables derived from standard essentially controlled by climate at large scale (Wood- meteorological data to formulate the relation of ward, 1987). Application of HLZ model was validated climate patterns and broad-scale vegetation distri- by comparing the modeled HLZ distribution for the bution. It has been widely accepted in projecting current climate with the real life zone distribution map impacts of climate change on vegetation distribu- in China (Chen et al., 2003, 2005). It is indicated that tions (Chen et al., 2003, 2005; Chinea and Helmer, HLZ model could be applied to simulate changes of 2003; Kerr et al., 2003; Yang et al., 2002; Yue et major terrestrial ecosystems in China. In this paper, al., 2001; Xu and Yan, 2001; Peng, 2000; Kirilenko spatial pattern of major terrestrial ecosystems in China et al., 2000; Powell et al., 2000; Dixon et al., 1999; and its changes in the four decades of 1960s, 1970s, Metternicht and Zinck, 1998; Belotelov et al., 1996; 1980s and 1990s are analyzed on the basis of digital Smith et al., 1992; Post et al., 1982). elevation model of China by operating HLZ model on Although no consideration of physiological meteorological data from 735 weather stations that are changes is a drawback of HLZ model, climatic effects scattered over China (Fig. 1).

N

WE

S

3000 300 600 km

Fig. 1. Spatial distribution of the weather observation stations in China. T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302 289

2. Methods

2.1. The HLZ model

The HLZ model divides the world into over 100 life zones in terms of mean annual biotemperature in degrees centigrade (MAB), average total annual precipitation in millimeters (TAP), and potential evapotranspiration ratio (PER) logarithmically. Biotemperature is defined as the mean of unit-period temperatures with substitution of zero for all unit-period values below 0 8C and above 30 8C(Holdridge et al., 1971). Evapotranspiration is the total amount of water that is returned directly to the atmosphere in the form of vapor through the combined processes of evaporation and transpiration. Potential evapotranspiration is the amount of water that would be transpired under constantly optimal conditions of soil moisture and plant cover. The potential evapotranspiration ratio is the ratio of mean annual potential evapotranspiration to average total annual precipitation, which provides an index of biological humidity conditions. In other words, MAB, TAP and PER at site (x,y) and in the t-th year have the following formulation:

1 X365 MABðÞ¼x; y; t TEMðÞj; x; y; t ð1Þ 365 j¼1

X365 TAPðÞ¼x; y; t PjðÞ; x; y; t ð2Þ j¼1

58:93MABðÞx; y; t PERðÞ¼x; y; t ð3Þ TAPðÞx; y; t where TEM( j,x,y,t) is the value summing the hourly temperature above 0 8C and below 30 8C on the j-th day and dividing by 24; P( j,x,y,t) is the mean of precipitation on the j-th day. Suppose

; ; ; ; MxðÞ¼y t log2MABðÞx y t ð4Þ

; ; ; ; TxðÞ¼y t log2TAPðÞx y t ð5Þ

; ; ; ; PxðÞ¼y t log2PERðÞx y t ð6Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 2 diðÞ¼x; y; t ðMxðÞ; y; t Mi0Þ þðTxðÞ; y; t Ti0Þ þðPxðÞ; y; t Pi0Þ ð7Þ where Mi0, Ti0 and Pi0 are standards of MAB logarithm, TAP logarithm and PER logarithm at the central point of the i-th life zone in the hexagonal system of HLZs. When dk ðÞ¼x; y; t mini fdiðÞgx; y; t , the site (x,y) is classified into the k-th life zone.

2.2. Mean center model

The model of the mean center was introduced in the USA census of population in 1870 (Shaw and Wheeler, 1985). Such spatial statistics has been used in geography since 1950s (Hart, 1954; Warntz and Neft, 1960; Ebdon, 1978; Shaw and Wheeler, 1985; Yue et al., 2003). The major applications were those concerned with central tendency, especially the weight mean center of spatial distribution. The model is formulated as (Yue et al., 2003), 290 T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302

XIj s ðÞt X ðÞt x ðÞ¼t ij ij ð8Þ j S t i¼1 jðÞ

XIj s ðÞt Y ðÞt y ðÞ¼t ij ij ð9Þ j S t i¼1 jðÞ where t is the variable of time; Ij(t) is patch number of HLZ type j; sij(t) is area of the i-th patch of HLZ type j; Sj(t) is total area of HLZ type j;(Xij(t), Yij(t)) is longitude and latitude coordinate of the geometric center of the i-th patch of HLZ type j;(xj(t), yj(t)) is the mean center of HLZ type j.

2.3. Ecological diversity index

On the basis of analyzing all diversity indexes and conducting theoretical demonstration, the ecological diversity index is expressed as (Yue et al., 1998, 2001, 2003, in press-b),

!2 mXðÞe 1 ln ðÞpiðÞt 2 i 1 dtðÞ¼ ¼ ð10Þ lnðÞe where t is the variable of time; pi(t) is probability of the i-th ecotope; m(e) is total number of the 1 A investigation ecotopes; e ¼ e þ s , A is area of studied region in hectares or area of the sampling quadrat, s is spatial resolution of land cover data or the smallest crown diameter of the sampled individuals, and e equals 2.71828.

2.4. Patch connectivity index

Landscape connectivity is distinguished into patch connectivity, line connectivity, vertex connectivity and network connectivity (Yue et al., 2003). The patch connectivity index is formulated as (Yue et al., 2003, 2004),

XmtðÞ nXiðÞt COðÞ¼t pijðÞt SijðÞt ð11Þ i¼1 j¼1

AijðÞt where t is the variable of time; pijðÞ¼tffiffi A , Aij(t) the area of the j-th patch in the i-th HLZ type and A the total p pffiffiffi 8 3AijðÞt area under investigation; SijðÞ¼t 2 ,Prij(t) is the perimeter of the j-th patch in the i-th HLZ type and 8 3 the ðPrijðÞt Þ ratio of the square of perimeter to the area of a hexagon; 0VC(t)V1.1 and when all patches have the shape of hexagon (6-gon), C(t)=1.0.

3. Simulation process

3.1. Interpolation

Our analyses on data from the 735 weather observation stations show that there is no statistically significant correlation between precipitation and elevation. Therefore, Kriging interpolation method is selected T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302 291 to create 1 km1 km precipitation surfaces. The Kriging is a generalized linear regression that is used to formulate an optimal estimator in a minimum mean square error sense (Li et al., 2004). It was introduced to avoid systematic errors in interpolation by the South African mining engineer, D.G. Krige (Kleijnen and van Beers, 2005). The Kriging is referred to as the best linear unbiased estimate because some study-cases showed that the expected value for the estimate error equals zero and whose variance is a minimum. The Kriging model has a higher precision than Inverse Distance Weighted (Yue et al., in press-a). Daily temperature and daily precipitation data from 1960 to 2002 are selected from the 735 weather observation stations. Temperature surfaces are created on the basis of improving the interpolation method, gradient-plus-inverse distance squared (GIDS). The GIDS (Nalder and Wein, 1998) was developed on the basis of comparatively analyzing interpolation methods such as inverse distance weight (Lee and Angelier, 1994) and the Kriging (Olea, 1999). The applications of GIDS in interpolating every-10-days and annual mean temperature in China showed that GIDS has a much less error than the inverse distance weight and the Kriging (Lin et al., 2002; Pan et al., 2004). GIDS is further improved on the basis of iterative simulation and formulated as, Â À À Á À Á À ÁÃ !, XNjk XNjk Ti þ a Ejk EiÞþCX Xjk Xi þ CY Yjk Yi þ CE Ejk Ei =2 1 T jk ¼ 2 2 ð12Þ i¼1 dijk i¼1 dijk where a is decreasing rate of temperature with elevation increase; CX, CY and CE are linear regression coefficients between temperature and longitude, latitude and elevation over; Xi, Yi, Ei and Ti are, respectively, longitude, latitude, elevation and temperature at the weather observation station i; Xjk, Yjk, Ejk and Tjk are, respectively, longitude, latitude, elevation and temperature at point ( j,k) to be interpolated; Njk is number of weather observation stations that are involved in the interpolation of the point ( j,k); dijk is distance from the observation station i to the interpolation point ( j,k). Although the decrease rate of temperature with elevation increase, a=0.0065 8C/m, has been adopted globally, it is incorrect to be applied in China. Our simulation results show that decrease rate of temperature with elevation increase has a great difference in different regions because of the considerable topographical variety over China. In terms of data from the 735 weather observation stations, the decrease rate of temperature (Table 1)is0.0028 8C/m in Changbai , 0.0040 8C/m in Da-Xiao Hinggan mountains, 0.0029 8C/m in Qinghai_Xizang Plateau, 0.0054 8C/m in , 0.0036 8C/m in Loess Plateau, 0.0060 8C/m in , 0.0037 8C/m in , 0.0040 8C/m in mountains, 0.0050 8C/m in Taihang and Lvliang mountains, 0.0038 8C/m in Tianshan mountains, 0.0055 8C/m in , 0.0044 8C/m in , 0.0064 8C/m in Himalaya mountains, 0.0046 8C/m in Yanshan mountains, and 0.0045 8C/m in Yunnan- Plateau. On an average, the decrease rate of temperature in China is a=0.0046 8C/m and the multivariate regression model of relationship between temperature T, longitude X, latitude Y and elevation E is simulated as,

T ¼ 43:312 0:106X 0:469Y 0:00361E ð13Þ where correlation coefficient is 0.9817. Therefore, the improved GIDS can be specifically formulated as,

ÂÃÀÁ ÀÁ ÀÁ ÀÁ XNjk 2Ti 0:0046 Ejk Ei þ 0:106 Xjk Xi þ 0:469 Yjk Yi þ 0:0036 Ejk Ei T jk ¼ ! ð14Þ XNjk i¼1 2 2 2dijk dijk i¼1 292 T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302

Table 1 The decrease rate of temperature with elevation increase in different regions of China Regions Correlation coefficients Decreasing rate (8C/m) 0.962828 0.0028 Da-Xiao Hinggan mountains 0.84284 0.0040 Qinghai_Xizang Plateau 0.953225 0.0029 Hengduan mountains 0.953959 0.0054 Loess plateau 0.870578 0.0036 Nanling mountains 0.909555 0.0060 Qilian mountains 0.993717 0.0037 Qinling mountains 0.764436 0.0040 Taihang and Lvliang mountains 0.978394 0.0050 Tianshan mountains 0.850741 0.0038 Wuling mountains 0.945071 0.0055 Wuyi mountains 0.905499 0.0044 Himalaya mountains 0.852985 0.0064 Yanshan mountains 0.984928 0.0046 Yunnan-Guizhou Plateau 0.854687 0.0045

In terms of formulation (12), all primary temperature surfaces of 1 km1 km grid data are created by iterative interpolations, in which search radiuses for each temperature surface are, respectively, defined as 150 km, 200 km, 250 km and 500 km. Then, the primary temperature surfaces are corrected in terms of 1 km1 km digital elevation model of China (Fig. 2) and relationships between temperature and elevation (Fig. 3).

3.2. HLZ model operation

The calculation process includes the following 10 steps: (1) modularizing file names of all raw temperature and precipitation data to unify file-name formats of all base data by ACDSee, (2) converting daily data into annual data and transforming the data format into Txt file by means of VC++, (3) transforming the annual data into Point Coverage by AML program of ARC/INFO, (4) operating the improved GIDS model by ArcObjects VBA Control and creating annual mean biotemperature surfaces of the period from the years 1960 to 2002 on the basis of digital elevation model and Point Coverage of annual mean biotemperature; (5) creating annual mean precipitation surfaces by Kriging interpolation and transforming the Point Coverage into 1 km Grid data by means of Spatial Analyst module of ARC/INFO; (6) converting the 1 km Grid data into ASCII data by ArcToolbox module and AML program of ARC/INFO; (7) calculating the potential evapotranspiration in ASCII data by modularized programming of VC++ and then transforming the results into Grid data; (8) conducting Grid computation of biotemperature, precipitation and evapotranspiration; (9) operating HLZ model on the ASCII data by VC++; (10) transforming the ASCII data from calculation of HLZ model into 1 km Grid data and then converting the 1 km Grid data into Polygon Coverage (Table 2).

4. Results tundra and alpine rain tundra are distributed in the eastern Qinghai-Xizang Plateau and Tianshan moun- 4.1. Spatial distribution of HLZs in China tains. Boreal dry scrub is mainly distributed in Tianshan mountains and the south of Qinghai- The results from spatially analyzing HLZs during Xizang Plateau and boreal is mainly dis- the period from 1960 to 2002 on an average (Fig. 4) tributed in the north of Qinghai-Xizang Plateau and show that nival area, alpine dry tundra and alpine the edge of southern Tarim basin. The tropical moist tundra are mainly distributed in the middle desert is mainly distributed in the central areas of and western Qinghai-Xizang Plateau; alpine wet Tarim basin and Turpan basin; warm temperate

294 T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302

Changbai mountains Da-Xiao Hinggan mountains Hengduan mountains 10.00 25.00 15.00 8.00 20.00 T= -0.0054E + 24.542 R2 = 0.91 10.00 T= -0.0028E + 9.3481 6.00 15.00 R2 = 0.927 4.00 T= -0.004E + 8.3638 10.00 5.00 2.00 R2 = 0.7104 5.00 Temperature Temperature Temperature 0.00 0.00 0.00 0 1000 2000 3000 0 200 400 600 800 0 1000 2000 3000 4000 5000 Elevation Elevation Elevation

Loess plateau Nanling mountains Qilian mountains 12.00 25.00 12.00 10.00 10.00 20.00 T= -0.006E + 18.996 R2 = 0.8273 8.00 8.00 T= -0.0036E + 14.111 15.00 6.00 T= -0.0037E + 14.787 2 10.00 4.00 2 6.00 R = 0.7579 R = 0.9875

Temperature 5.00

Temperature 2.00 Temperature 4.00 0.00 0.00 600 800 1000 1200 1400 1600 0 500 1000 1500 0 1000 3000 4000 5000 Elevation Elevation Elevation

Qinghai-Xizang Plateau Qinling mountains Taihang and Lvliang mountains 10.00 20.00 T= -0.0029E + 14.442 15.00 8.00 R2 = 0.9086 15.00 T= -0.005E + 14.53 6.00 10.00 R2 = 0.9573 10.00 T= -0.004E + 15.681 4.00 R2 = 0.5844 5.00 2.00 5.00 Temperature Temperature 0.00 0.00 Temperature 0.00 1500 2500 3500 4500 5000 0 200 400 600 800 1000 1200 0 500 1000 1500 2000 2500 Elevation Elevation Elevation

Tianshan mountains Wuling mountains 20.00 Wuyi mountains 20.00 T= -0.0038E + 13.858 15.00 30.00 15.00 T= -0.0044E + 18.848 R2 = 0.7238 20.00 R2 = 0.8199 10.00 10.00 T= -0.0055E + 18.112 10.00 5.00 5.00 R2 = 0.8932 Temperature Temperature Temperature 0.00 0.00 0.00 0 500 1000 1500 2000 0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 Elevation Elevation Elevation Yanshan mountains 12.00 Himalaya mountains Yunnan-Guizhou Plateau 10.00 10.00 25.00 T= -0.0045E+ 19.979 8.00 20.00 R2 = 0.7305 6.00 T=-0.0046E + 12.258 5.00 T= -0.0064E + 30.978 15.00 4.00 R2 = 0.9701 R2 = 0.7276 10.00 Temperature Temperature 2.00 0.00 Temperature 5.00 0.00 3000 3250 3500 3750 4000 4250 4500 0 500 1000 1500 2000 2500 100 600 1100 1600 Elevation Elevation Elevation

Fig. 3. Relationships between temperature and elevation in different regions of China.

4.2. HLZ area change the decadal decrease rate is 10.4%. Nival zone and subtropical thorn woodland might disappear in 159 Analyses on HLZs of China in 1960s, 1970s, years and 96 years, respectively, if their areas would 1980s and 1990s (Fig. 5) shows that nival area decrease at present rate. Alpine dry tundra and cool continuously decreased in the four decades and the temperate scrub continuously increased in the four decadal decrease rate is 6.3%. Subtropical thorn decades and the decadal increase rates are, respec- woodland had a rapid decrease on an average and tively, 13.1% and 3.4%. T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302 295

Table 2 The classification standards of MAB, TAP and PER at the central point of the i-th life zone in the hexagonal system of HLZs Code HLZ Standard of Standard of Standard MAB (8C) TAP (mm) of PER 1 Nival 0.2670 88.3880 0.1770 1 Nival 0.5300 88.3880 0.3540 1 Nival 0.5300 177.7770 0.1770 1 Nival 1.0610 88.3880 0.7070 1 Nival 1.0610 177.7770 0.3540 1 Nival 1.0610 353.5520 0.1770 2 Alpine dry tundra 2.1210 88.3880 1.4140 3 Alpine moist tundra 2.1210 177.7770 0.7070 4 Alpine wet tundra 2.1210 353.5520 0.3540 5 Alpine rain tundra 2.1210 707.1070 0.1770 6 Boreal desert 4.2430 88.3880 2.8280 7 Boreal dry scrub 4.2430 177.7770 1.4140 8 Boreal moist forest 4.2430 353.5520 0.7070 9 Boreal wet forest 4.2430 707.1770 0.3540 10 Boreal rain forest 4.2430 1414.2130 0.1770 11 Cool temperate desert 8.4850 88.3880 5.6750 12 Cool temperate scrub 8.4850 177.7770 2.8280 13 Cool temperate steppe 8.4850 353.5520 1.4140 14 Cool temperate moist forest 8.4850 707.1070 0.7070 15 Cool temperate wet forest 8.4850 1414.2130 0.3540 16 Cool temperate rain forest 8.4850 2828.4270 0.1770 17 Warm temperate desert 14.2700 88.3880 11.3140 18 Warm temperate desert scrub 14.2700 177.7770 5.6750 19 Warm temperate thorn steppe 14.2700 353.5520 2.8280 20 Warm temperate dry forest 14.2700 707.1070 1.4140 21 Warm temperate moist forest 14.2700 1414.2130 0.7070 22 Warm temperate wet forest 14.2700 2828.4270 0.3540 23 Warm temperate rain forest 14.2700 5656.8540 0.1770 24 Subtropical desert 20.1810 88.3880 11.3140 25 Subtropical desert scrub 20.1810 177.7770 5.6750 26 Subtropical thorn woodland 20.1810 353.5520 2.8280 27 Subtropical dry forest 20.1810 707.1070 1.4140 28 Subtropical moist forest 20.1810 1414.2130 0.7070 29 Subtropical wet forest 20.1810 2828.4270 0.3540 30 Subtropical rain forest 20.1810 5656.8540 0.1770 31 Tropical desert 33.9410 88.3880 22.6270 32 Tropical desert scrub 33.9410 177.7770 11.3140 33 Tropical thorn woodland 33.9410 353.5520 5.6750 34 Tropical very dry forest 33.9410 707.1070 2.8280 35 Tropical dry forest 33.9410 1414.2130 1.4140 36 Tropical moist forest 33.9410 2828.4270 0.7070 37 Tropical wet forest 33.9410 5656.8540 0.3540 38 Tropical rain forest 33.9410 11,313.7100 0.1770

Areas of other HLZs had been ups and downs in warm temperate desert, warm temperate desert four decades. The results from comparing their areas scrub, warm temperate dry forest, warm temperate in 1990s with the ones in 1960s show that the life wet forest, subtropical dry forest subtropical moist zones, of which areas increased, include alpine forest, subtropical wet forest, and tropical moist moist tundra, alpine wet tundra, boreal desert, boreal forest. The ones, of which areas decreased, include wet forest, boreal rain forest, cool temperate steppe, boreal dry scrub, boreal moist forest, cool temperate 296 T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302

Fig. 4. The spatial distribution of Holdridge life zones on an average during the period from 1960 to 2002 in China. desert, cool temperate moist forest, cool temperate HLZ patch connectivity has a continuous increase wet forest, warm temperate thorn steppe, warm trend that can be regressively formulated as, temperate moist forest, tropical desert and tropical dry forest. The areas of warm temperate moist forest : : and subtropical wet forest had the biggest increase COðÞ¼t 0 0075t þ 0 122 ð16Þ and increased by 139% and 200%, respectively, in the four decades on an average. The areas of tropical where DI(t) is HLZ diversity; CO(t) is HLZ patch desert and tropical dry forest had the biggest connectivity; t is time variable and t=1, 2, 3 and decrease and decreased by 14% and 10.7%, respec- 4, respectively, correspond to 1960s, 1970s, 1980s tively (Table 3). and 1990s. Correlation coefficients of regression Eqs. (15) and (16) are, respectively, 0.94 and 4.3. Changes of HLZ diversity and patch connectivity 0.99.

The calculation results by operating the diver- 4.4. Shift trend of the mean center sity index and the patch connectivity index on the data corresponding to Fig. 5 show that HLZ The analysis on shift trend of the mean center diversity has a continuous decrease trend on the (Table 5 and Fig. 6) shows that HLZ types, which average (Table 4) that can be regressively formu- have a larger shift range, include warm temperate lated as, thorn steppe, subtropical wet forest, cool temperate wet forest, boreal moist forest, boreal wet forest, DIðÞ¼t 0:0005t þ 0:1422 ð15Þ alpine moist tundra, warm temperate dry forest, T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302 297

Fig. 5. HLZs in 1960s, 1970s, 1980s, and 1990s.

cool temperate steppe, cool temperate moist forest, thorn woodland that shifted 50.98 km in the four warm temperate moist forest, nival area and tropical decades. desert. They shifted 1781.45 km, 1208.14 km, In general, mean center of warm temperate thorn 977.43 km, 963.05 km, 826.41 km, 793.59 km, steppe shifted towards northwest in the four decades 750.11 km, 642.02 km, 613.54 km, 574.69 km, because the warm temperate thorn steppe is being in 518.46 km and 517.35 km in the four decades, succession to warm temperate dry forest with the respectively. These HLZ types are more sensitive to increase of temperature and precipitation in the climate change than other ones such as subtropical southeast of Himalaya mountains and in the mean- 298 T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302

Table 3 Area change of HLZs in the four decades in China (units: million hectare) Holdridge life zone type 1960s 1970s 1980s 1990s Code Name Area Proportion Area Proportion Area Proportion Area Proportion (%) (%) (%) (%) 1 Nival 121.05 12.75 97.78 10.30 95.27 10.04 90.53 9.54 2 Alpine dry tundra 6.40 0.67 6.90 0.73 8.93 0.94 9.75 1.03 3 Alpine moist tundra 9.76 1.03 51.09 5.38 11.90 1.25 10.50 1.11 4 Alpine wet tundra 30.18 3.18 32.54 3.43 31.41 3.31 36.60 3.86 5 Alpine rain tundra 42.85 4.51 21.04 2.22 58.31 6.14 51.82 5.46 6 Boreal desert 8.74 0.92 9.24 0.97 8.83 0.93 9.22 0.97 7 Boreal dry scrub 11.14 1.17 11.41 1.20 12.09 1.27 10.52 1.11 8 Boreal moist forest 48.55 5.11 54.50 5.74 42.04 4.43 42.54 4.48 9 Boreal wet forest 18.31 1.93 16.47 1.74 21.87 2.30 21.76 2.29 10 Boreal rain forest 0.39 0.04 0.39 0.04 0.51 0.05 0.48 0.05 11 Cool temperate desert 39.46 4.16 43.54 4.59 34.37 3.62 34.20 3.60 12 Cool temperate scrub 47.92 5.05 48.40 5.10 53.19 5.60 54.49 5.74 13 Cool temperate steppe 82.78 8.72 99.86 10.52 90.93 9.58 97.08 10.23 14 Cool temperate moist 103.30 10.88 87.39 9.21 97.72 10.29 93.86 9.89 forest 15 Cool temperate wet 5.40 0.57 4.46 0.47 6.23 0.66 4.54 0.48 forest 17 Warm temperate desert 65.98 6.95 75.85 7.99 82.23 8.66 75.43 7.95 18 Warm temperate desert 6.78 0.71 4.97 0.52 7.49 0.79 10.34 1.09 scrub 19 Warm temperate thorn 1.14 0.12 0.65 0.07 1.27 0.13 0.84 0.09 steppe 20 Warm temperate dry 65.12 6.86 71.07 7.49 68.44 7.21 79.09 8.33 forest 21 Warm temperate moist 127.78 13.46 122.70 12.92 129.09 13.60 111.76 11.77 forest 22 Warm temperate wet 0.57 0.06 1.82 0.19 1.23 0.13 3.55 0.37 forest 26 Subtropical thorn 0.77 0.08 0.60 0.06 0.61 0.06 0.45 0.05 woodland 27 Subtropical dry forest 5.45 0.57 6.93 0.73 5.64 0.59 6.01 0.63 28 Subtropical moist forest 75.50 7.95 70.72 7.45 70.72 7.45 82.50 8.69 29 Subtropical wet forest 0.09 0.01 0.61 0.06 0.24 0.03 0.81 0.09 31 Tropical desert 23.69 2.50 8.24 0.87 8.51 0.90 10.44 1.10 35 Tropical dry forest 0.21 0.02 0.10 0.01 0.21 0.02 0.12 0.01 36 Tropical moist forest 0.082 0.009 0.027 0.003 0.100 0.011

while cool temperate steppe is being in succession to Table 4 Changes of HLZ diversity and patch connectivity in the four warm temperate thorn steppe in the northwest of decades on the average Tianshan mountains. Mean center of nival areas Periods 1960s 1970s 1980s 1990s Decadal shifted towards northwest, which tallies with the increase phenomena that snow line has continuously shrunk rate (%) back and area of nival areas has continuously Holdridge life zone 0.1418 0.1408 0.1408 0.1401 0.0030 decreased in recent decades. Mean center of warm diversity temperate desert scrub shifted towards southeast, Patch connectivity 0.1303 0.1365 0.1424 0.1532 0.0439 which tallies with the phenomenon that grassland in T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302 299

Table 5 Shift trend of each HLZ type in the four decades in China (units: kilometer) Holdridge life zone type From 1960s to 1970s From 1970s to 1980s From 1980s to 1990s Code Name Distance Direction Distance Direction Distance Direction 1 Nival 412.68 Toward north 88.16 Toward northwest 17.62 Toward northeast 2 Alpine dry tundra 108.41 Toward southwest 56.15 Toward southwest 39.56 Toward southwest 3 Alpine moist tundra 379.55 Toward southeast 352.62 Toward northwest 61.42 Toward southeast 4 Alpine wet tundra 169.47 Toward southeast 220.66 Toward northwest 11.09 Toward north 5 Alpine rain tundra 188.63 Toward northeast 296.42 Toward southwest 12.03 Toward northeast 6 Boreal desert 12.31 Toward southwest 83.86 Toward southwest 142.67 Toward southwest 7 Boreal dry scrub 53.40 Toward southwest 65.89 Toward southwest 113.53 Toward southwest 8 Boreal moist forest 270.49 Toward northeast 329.20 Toward southwest 363.36 Toward southwest 9 Boreal wet forest 254.14 Toward southwest 518.35 Toward northeast 53.92 Toward west 10 Boreal rain forest 160.81 Toward southwest 58.57 Toward northeast 197.16 Toward southwest 11 Cool temperate desert 51.35 Toward northwest 82.98 Toward southeast 17.76 Toward northwest 12 Cool temperate scrub 30.44 Toward northeast 48.33 Toward southwest 40.74 Toward southwest 13 Cool temperate steppe 101.09 Toward northeast 377.27 Toward southwest 163.68 Toward southeast 14 Cool temperate moist 16.21 Toward southeast 352.93 Toward northeast 244.40 Toward southeast forest 15 Cool temperate wet 479.94 Toward southwest 189.61 Toward northeast 307.88 Toward southwest forest 17 Warm temperate desert 203.33 Toward northwest 217.28 Toward southeast 55.51 Toward west 18 Warm temperate desert 94.59 Toward south 87.75 Toward northwest 209.96 Toward southeast scrub 19 Warm temperate thorn 119.36 Toward northwest 778.94 Toward northeast 883.15 Toward northwest steppe 20 Warm temperate dry 353.72 Toward southwest 238.16 Toward northeast 158.23 Toward southwest forest 21 Warm temperate moist 23.49 Toward northeast 214.68 Toward northeast 336.52 Toward southwest forest 22 Warm temperate wet 44.08 Toward northwest 72.35 Toward northeast 77.17 Toward southwest forest 26 Subtropical thorn 22.88 Toward southwest 2.75 Toward southwest 25.35 Toward southwest woodland 27 Subtropical dry forest 95.30 Toward north 206.73 Toward southwest 191.51 Toward northwest 28 Subtropical moist forest 120.32 Toward south 21.64 Toward south 140.22 Toward north 29 Subtropical wet forest 469.80 Toward southeast 95.83 Toward southwest 642.51 Toward northeast 31 Tropical desert 251.28 Toward northeast 127.47 Toward northeast 138.60 Toward southwest 35 Tropical dry forest 112.91 Toward northwest 55.17 Toward southeast 51.62 Toward northwest 36 Tropical moist forest 60.93 Toward northeast 243.09 Toward northeast

has continuously degraded in recent (10.66%), cool temperate moist forest (10.07%), decades. cool temperate steppe (9.76%), warm temperate desert (7.89%), subtropical moist forest (7.89%), warm temperate dry forest (7.47%), cool temperate 5. Discussion and conclusions scrub (5.37%), boreal moist forest (4.94%) and alpine rain tundra (4.58%). Their area accounts for Twenty-eight HLZ types can be found in China 81.56% of the whole of China (Table 6). Although during the period from 1960 to 2002 (Fig. 4), these 10 HLZ types are dominant ones in China, among which tropical moist forest has appeared other 18 HLZ types play an important role in HLZ since 1970s. The 10 dominant HLZ types include diversity and most of them, such as warm temperate warm temperate moist forest (12.94%), nival area thorn steppe, subtropical wet forest, cool temperate 300 T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302

Fig. 6. The shift trend of the mean center of each HLZ in the four decades in China.

wet forest, are much more sensitive to climate province, which implies that both temperature and change and human activities than the 10 dominant precipitation have an increase trend in Hainan HLZ types. province and its northeast neighboring area in The analysis on spatial distribution of HLZs recent decades. Area of nival zone has continuously shows that warm temperate dry forest is distributed decreased since 1960 and the mean center of nival in the area between the 600 mm isohyet and 1200 zone has shifted towards northwest, which tallies mm isohyet, which is transitional belt from arid and with climate warming and precipitation increasing semiarid area to semi-humid and humid area. Warm in southwest Tibet and . Cool temperate moist forest is distributed in the south of temperate steppe, which accounts for 9.76% of warm temperate dry forest, between 1200 mm the whole terrestrial area of China, shifted towards isohyet and 1400 mm isohyet. Subtropical dry forest, southwest about 500 km. of which some is scattered in the middle and upper Although HLZ types increased from 27 to 28 reaches of river and subtropical moist forest because tropical moist forest appeared in Hainan are mostly distributed in the area with precipitation province in 1970s, on the average, HLZ diversity has from 1400 mm to 1800 mm. a decrease trend in the four decades and its decadal Tropical moist forest has appeared in China decrease rate is 0.003% (Table 4). Meanwhile, HLZ since 1970s and its mean center has shifted patch connectivity has continuously increased in the northeast from Hainan province to Guangzhou four decades and the decadal increase rate is T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302 301

Table 6 References The area and proportion of each HLZ type on an average from 1960 to 2002 (units: million hectares) Belotelov, N.V., Bogatyrev, B.G., Kirilenko, A.P., Venevsky, S.V., Code Name Average Proportion 1996. Modelling of time-dependent shifts under global area (%) climate changes. Ecological Modelling, 87, 29–40. 1 Nival 101.16 10.66 Chen, X.W., Zhang, X.S., Li, B.L., 2003. The possible response of 2 Alpine dry tundra 8.00 0.84 life zones in China under global climate change. Global and 3 Alpine moist tundra 20.81 2.19 Planetary Change, 38, 327–337. 4 Alpine wet tundra 32.68 3.44 Chen, X.W., Zhang, X.S., Li, B.L., 2005. Influence of Tibetan 5 Alpine rain tundra 43.51 4.58 Plateau on vegetation distributions in East Asia: a modeling 6 Boreal desert 9.01 0.95 perspective. Ecological Modelling, 181, 79–86. 7 Boreal dry scrub 11.29 1.19 Chinea, J.D., Helmer, E.H., 2003. Diversity and composition of 8 Boreal moist forest 46.91 4.94 tropical secondary recovering from large-scale clearing: 9 Boreal wet forest 19.61 2.07 results from the 1990 inventory in Puerto Rico. Forest Ecology 10 Boreal rain forest 0.44 0.047 and Management, 180, 227–240. 11 Cool temperate desert 37.89 3.99 Clements, E., 1916. Climax formations of North America. Plant 12 Cool temperate scrub 51.00 5.37 Succession: an Analysis of the Development of Vegetation, 13 Cool temperate steppe 92.66 9.76 Publication, vol. 242. Carnegie Institute of Washington. 14 Cool temperate moist forest 95.57 10.07 Dixon, R.K., Smith, J.B., Brown, S., Merara, O., Mata, L.J., 15 Cool temperate wet forest 5.16 0.54 Buksha, I., 1999. Simulations of forest system response and 17 Warm temperate desert 74.87 7.887 feedbacks to global change: experiences and results from the 18 Warm temperate desert scrub 7.39 0.78 U.S. Country studies Program. Ecological Modelling, 122, 19 Warm temperate thorn steppe 0.97 0.10 289–305. 20 Warm temperate dry forest 70.93 7.47 Ebdon, D., 1978. Statistics in Geography. Basil Blackwell, 21 Warm temperate moist forest 122.83 12.94 Oxford. 22 Warm temperate wet forest 1.80 0.19 Griesebach, A.H.R., 1872. Die Vegetation der Erde nach ihrer 26 Subtropical thorn woodland 0.61 0.06 klimatischen Anordnung. W. Engelmann, Leipzig. 27 Subtropical dry forest 6.01 0.63 Hart, J.F., 1954. Central tendency in areal distributions. Economic 28 Subtropical moist forest 74.86 7.886 Geography, 30, 48–59. 29 Subtropical wet forest 0.437 0.046 Holdridge, L.R., 1947. Determination of world plant formations 31 Tropical desert 12.72 1.34 from simple climate data. Science, 105 (2727), 367–368. 35 Tropical dry forest 0.16 0.02 Holdridge, L.R., Grenke, W.C., Hatheway, W.H., Liang, T., Tosi, 36 Tropical moist forest 0.05 0.01 J.A., 1971. Forest Environments in Tropical Life Zones. Pergamon Press, Oxford. Kerr, S., Liu, S.G., Pfaff, A.S.P., Hughes, R.F., 2003. Carbon dynamics and land use choices: building a regional-scale 0.0439% on the average. Because the patch con- multidisciplinary model. Journal of Environmental Manage- nectivity is defined as movement efficiency of ment, 69, 25–37. migrants in patches of a region under consideration Kirilenko, A.P., Belotelov, N.V., Bogatyrev, B.G., 2000. Global (Yue et al., 2003, 2004), the continuous increase of model of vegetation migration: incorporation of climatic patch connectivity signifies the strengthening ability variability. Ecological Modelling, 132, 125–133. Kleijnen, J.P.C., van Beers, W.C.M., 2005. Robustness of Kriging of human utilizing ecosystem services. In short, the when interpolating in random simulation with heterogeneous simulated results on the spatial pattern changes variances: some experiments. European Journal of Operational reflect the great effects of climate change on Research, 165 (3), 826–834. terrestrial ecosystems in China. Koeppen, W., Geiger, R., 1930. Handbuch der Climatologie. Teil I D, Borntraeger, Berlin. Lee, J.C., Angelier, J., 1994. Paleostress trajectory maps based on the results of local determinations: the blissageQ program. Acknowledgment Computers and Geosciences, 20 (2), 161–191. Li, H., Wang, Q.X., Lam, K.Y., 2004. Development of a novel This work is supported by National Basic Research meshless Local Kriging (LoKriging) method for structural Priorities Program (2002CB412506) of Ministry of dynamic analysis. Computer Methods in Applied Mechanics and Engineering, 193, 2599–2619. Science and Technology of the People’s Republic of Lin, Z.H., Mo, X.G., Li, H.X., Li, H.B., 2002. Comparison of three China, and by Projects of National Natural Science spatial interpolation methods for climate variables in China. Foundation of China (40371094). Acta Geographica Sinica, 57 (1), 47–56 (in Chinese). 302 T.X. Yue et al. / Global and Planetary Change 48 (2005) 287–302

Merriam, C.H., 1892. The geographic distribution of life in North Thornthwaite, C.W., 1931. The climates of North America according America. Proceedings of the Biological Society of Washington, to a new classification. Geographical Review, 21 (4), 633–655. 7, 1–74. Warntz, W., Neft, D., 1960. Contributions to a statistical method- Metternicht, G.I., Zinck, J.A., 1998. Evaluating the information ology for areal distributions. Journal of Regional Science, 2, content of JERS-1 SAR and Landsat TM data for discrimination 47–66. of soil erosion features. ISPRS Journal of Photogrammetry and Woodward, F.I., 1987. Climate and Plant Distribution. Cambridge Remote Sensing, 53, 143–153. University Press, Cambridge, UK. Nalder, I.A., Wein, R.W., 1998. Spatial interpolation of climatic Xu, D.Y., Yan, H., 2001. A study of the impacts of climate change normals: test of a new method in the Canadian boreal forest. on the geographic distribution of Pinus koraiensis in China. Agricultural and Forest Meteorology, 92, 211–225. Environment International, 27, 205–210. Olea, R.A., 1999. Geostatistics for Engineers and Earth Scientists. Yang, X., Wang, M.X., Huang, Y., Wang, Y.S., 2002. A one- Kluwer Academic Publishers, Boston. compartment model to study soil carbon composition rate at Pan, Y.Z., Gong, D.Y., Deng, L.M., Li, J., Gao, J., 2004. Smart equilibrium situation. Ecological Modelling, 151, 63–73. distance searching-based and DEM-informed interpolation of Yue, T.X., Haber, W., Grossmann, W.D., Kasperidus, H.D., 1998. surface air temperature in China. Acta Geographica Sinca, 59 Towards the satisfying model for biological diversity. Ekologia, (3), 366–374. 17 (Suppl. 1), 129–141. Peng, C.H., 2000. From static biogeographical model to dynamic Yue, T.X., Liu, J.Y., Jbrgensen, S.E., Gao, Z.Q., Zhang, S.H., Deng, global vegetation model: a global perspective on modelling X.Z., 2001. Changes of HLZ diversity in all of China over half a vegetation dynamics. Ecological Modelling, 135, 33–54. century. Ecological Modelling, 144, 153–162. Post, W.M., Emanuel, W.R., Zinke, P.J., Stangenberger, A.G., 1982. Yue, T.X., Liu, J.Y., Jbrgensen, S.E., Ye, Q.H., 2003. Landscape Soil carbon pools and world life zones. Nature, 298, 156–159. change detection of the newly created wetland in Yellow River Powell, G.V.N., Barborak, J., Rodriguez, M., 2000. Assessing Delta. Ecological Modelling, 164, 21–31. representativeness of the protected natural areas in Costa Rica Yue, T.X., Xu, B., Liu, J.Y., 2004. A patch connectivity index and its for conserving biodiversity: a preliminary gap analysis. Bio- change on a newly born wetland at the Yellow River Delta. logical Conservation, 93, 35–41. International Journal of Remote Sensing, 25 (21), 4617–4628. Schouw, J.F., 1823. Grundzuege einer allgemeinen Pflanzengeog- Yue, T.X., Du, Z.P., Liu, J.Y., Chen, S.P., Fan, Z.M., Wang, Y.A., raphie. Berlin. Chen, W.H., Qiu, H.D., in press-a. High precision surface Shaw, G., Wheeler, D., 1985. Statistical Techniques in Geographical modelling: progress (I). ISPRS Journal of Photogrammetry and Analysis. John Wiley & Sons, New York. Remote Sensing. Smith, T.M., Shugart, H.H., Bonan, G.B., Smith, J.B., Woodward, Yue, T.X., Liu, J.Y., Chen, S.Q., Li, Z.Q., Ma, S.N., Tian, Y.Z., F.I., 1992. Modeling the potential response of vegetation to Ge.F., in press-b. Considerable effects of diversity indices and global climate change. Advances in Ecological Research, 22, spatial scales on conclusions relating to ecological diversity. 93–113. Ecological Modelling.