CHAPTER FIVE

The Ecology of Grassland Enclosures and Changing Patterns of Livestock Grazing: A Vegetation Analysis

Introduction (back to top)

~~[SHOULD MOVE SOME OF THIS CHAPTER’S INTRO TO THE CHAPTER’S OR THE THESIS’ DISCUSSION…~~]

Large tracts of Qinghai’s alpine grasslands currently are being enclosed (fenced) or undergoing other dramatic changes that affect how they are utilized by pastoralists (Luosan 1996). When ecologically important variables such as the total area of available grassland or the number of livestock are held constant, it is clear that fences affect, first and foremost, a change in the spatio-temporal pattern of livestock grazing. Even though fences can be used to limit livestock numbers in some areas for limited periods of time, they do not of themselves affect changes in overall grazing pressure. Clearly, with a constant total livestock population, if fewer animals are present inside fenced areas – as is generally the case, which is why fencing often has (wrongly) been equated directly with “grassland protection” – then livestock simply are more numerous outside such fenced areas, to the detriment of the unfenced lands. Unless the total number of livestock is decreased, fencing alone will do little to ameliorate the overall condition of the grasslands. As Holzner and Kriechbaum (1999) observed in Tibet, there are even real “dangers of fencing: as the pastures inside [are] protected, the areas outside (the pastures “where you can do what you want”) are overgrazed automatically…. A preferable alternative would be an open, flexible system to treat all the pasture land in a proper way!” Thus the enclosure of grassland is an important variable that can, and indeed should, be examined independently of other variables such as total area of grassland or livestock numbers. This chapter focuses specifically on the ecological (botanical) impacts of fencing and of related changes in livestock grazing patterns.

The “fencing treatment” in this study represents not only the simple presence of fences, but more importantly the presence of a whole new grazing pattern: mainly winter-grazed grassland inside the fenced areas, and a more continuously grazed grassland (with both winter and summer grazing) outside the fenced areas. If continuously grazed grassland shows any sign of degradation, as ecological theory would suggest (Coughenour 1991, Humphrey and Sneath 1996, Miller 1996) and experience has shown (Li et al. 1993, Williams 1996), then a significant redirection of pastoral development funds away from fencing schemes and other aspects of land use intensification and sedentarization should be seriously considered. Yet enormous funds continue to be spent on enormous fencing schemes, both by local government bureaus in Qinghai and by large international development agencies (e.g., Japan International Cooperation Agency). To avoid further unnecessary (and possibly detrimental) spending habits, several important questions should be asked. In particular, How does fencing affect species richness, plant cover, and species composition of the grassland? Or, in more general terms, How does fencing impact grassland quality? This chapter examines these questions in one specific area near the southeastern shore of Qinghai Lake, an area that is representative of a large portion of Qinghai’s alpine region (Bian 1987, Su 1993). In principle, the move to enclose grasslands in China is closely linked with the quasi-privatization of land (through the Household Responsibility System). This effort is meant to lead toward more uniform grazing pressure on the land, with each individual family using resources more proportionate with their own needs and their livestock’s needs. In this scenario, grassland fencing, land privatization, building permanent houses, and even the shift toward a market-driven economy are all part of one common drive to render the Tibetan Plateau grasslands more “productive” within the larger context of China’s national economy. Raising fences is also so tightly linked with building permanent houses and other aspects of the sedentarization of pastoralists that its full impact may neither be entirely predictable (Skånes 1997) nor in the desired direction of change in ecologically fragile grassland habitats (Li et al. 1993, Williams 1996). The strong linkages between fencing, increasingly sedentary lifestyles, and resultant changes in livestock grazing patterns cannot be stressed sufficiently. A recent newspaper article gives a glimpse into the depth and the degree of these interactions in present-day Qinghai:

“Provincial government officials … hope that herdsmen can gradually give up their traditional nomadic herding and turn to modern production methods. … The province will concentrate on infrastructure construction in the grassland … planting new grass, setting up fences…, building sheds to prevent animals from suffering illnesses from the cold or freezing to death, and constructing homes to help bring herdsmen into permanent settlements. … A total of 423 million yuan ($51 million) was invested in grassland infrastructure development from 1992 to 1996. Half the funds were raised by herdsmen themselves and through bank loans. … So far, 10.4 per cent of herding families in the province have benefited from these projects and 66,800 herding families, or 67 per cent of the total number of families in the province, have settled down” (Xie 1997).

Holzner and Kriechbaum (1999) clearly maintain that “if nomads move into houses from tents, … the result will be the overgrazing of the surrounding areas, as can be seen in [many places in Tibet].”

In the present study, the “fencing treatment” thus represents not only the simple presence of fences, but more importantly the presence of a whole new grazing pattern: mainly winter-grazed grassland inside the fenced areas, and a more continuously grazed grassland (with both winter and summer grazing) outside the fenced areas. If the continuously grazed grassland shows any sign of degradation, as basic ecological theory would suggest (Coughenour 1991, Humphrey and Sneath 1996, Miller 1996), then a significant redirection of pastoral development funds away from fencing schemes and other aspects of land use intensification and sedentarization should be seriously considered. Development investments should instead be directed toward promoting more flexible and opportunistic management alternatives, both traditional and modern (Galaty and Johnson 1990, Briske and Heitschmidt 1991, Sheehy 1993, 1996, Miller and Jackson 1994, Tainton et al. 1996 cited in Holzner and Kriechbaum 1999, Wester 1997).

The main argument generally used in favor of fencing is that, although livestock are now owned and managed by individual households (and sometimes by schools, villages, or other corporate entities), it remains extremely difficult for individuals to ensure that the grassland allotted to them is never used by outsiders unless it is permanently guarded – which is virtually impossible – or, alternately, physically enclosed or fenced. Most social mechanisms that traditionally have guarded against cheating were lost during the intense upheavals associated with “liberation” in the 1950s, the Great Leap Forward (1958-62), the Cultural Revolution (1966-1976), and the commune era in general (Goldstein and Beall 1990, Geoffrey 1993, Becker 1996, Smith 1996, Wu N. 1997, Wozencraft 2000).

The second main argument commonly used to promote fencing – i.e., that grassland fencing equals grassland protection – is based on the erroneous assumption that the mere presence of fences is sufficient to protect resources. A simple analogy is found in the national park systems of the world, which have led to the creation of small islands of less degraded land within a much larger sea of land where “anything goes” and where resources are generally depleted or degraded in unsustainable ways (McNeely and Thorsell 1991, Whelan 1991, Noss and Cooperrider 1994, Stevens 1997, Mowforth and Munt 1998, Holzner and Kriechbaum 1999). A more accurate view of fencing would therefore be to consider it as only one of many possible tools that can be used to manage the grassland. Presently, fencing schemes in China are aimed as much toward an intensification of land use as they are toward the long-term protection of grasslands. As already stated, fencing ultimately only leads to a redistribution of livestock on the land, not to a reduction in their total numbers or overall grazing pressure. Thus, in and of itself, fencing affects only the temporal and spatial pattern of livestock grazing in an area.

In promoting a uniform and spatially fixed presence of herders on the Tibetan Plateau’s alpine grasslands, local governments and international development agencies alike are also assuming (wrongly) that these ecosystems have fixed carrying capacities. However, not only is it very difficult to determine with certainty any carrying capacity, even in more stable ecological environments, but such “capacities” may not even exist in most of the plateau’s extremely variable (unpredictable) environments. Successional theory and equilibrium-based management strategies likely do not apply in many of the world’s arid and semi-arid, climatically variable environments (Coughenour 1991). Specifically, Qinghai’s alpine grasslands are more likely to operate as state-and-transition systems (Westoby et al. 1989a, 1989b) and/or to have a “dynamic carrying capacity” (Cincotta et al. 1992, Miller 1995, Miller and Craig 1997). The variability of these grasslands’ environment also requires that as much flexibility be maintained as possible in any grazing system, traditional or modern, in order to allow herders to respond rapidly to the normal fluctuations in seasonal and annual primary productivity as well as to periodic catastrophic events. Overall, flexibility is crucial. However, flexibility is limited by fences as well as by individual legal titles (contracts) to the land.

Unless the total number of livestock is decreased, fencing alone will likely do very little to ameliorate the overall condition of the grasslands. Yet enormous funds are now being spent on vast fencing schemes, both by the local government and by several international agencies (e.g., Japan International Cooperation Agency). To avoid unnecessary (and possibly detrimental) spending habits, several important questions should be asked. In particular, How does fencing affect species richness, plant cover, and the species composition of the grassland? Or in general, How does fencing impact grassland quality? This chapter examines these questions in one specific area near the southeastern shore of Qinghai Lake, an area that is typical or representative of a large portion of Qinghai’s alpine region (Bian 1987, Su 1993).

Methods and Results (back to top)

General methods (back to top)

In total, 194 plots (each plot 0.25 m²) in alpine meadow vegetation were examined in the summer of 1997. The study area was located at the base of the Nanshan Mountains near the southeastern shore of Qinghai Lake (in Daotanghe Township, Gonghe County, Hainan Tibetan Autonomous Prefecture). Sixty-eight plots were studied between 6 - 12 June (in early summer) and 126 plots were examined around one and a half months later between 29 July and 12 August (in late summer). A further 38 plots were located in nearby but different habitat types for comparative purposes. In each plot, as many plant species as possible were identified in the field. However, when a species could not be identified, several specimens were collected and later identified by a research assistant, Simo Tolvanen, in Helsinki, Finland. Two identification guides were used, the five-part illustrated Flora of China, Iconographia Cormophytorum Sinicorum, and Hao (1938). When possible, plant identifications were checked by comparing with specimens available at the Helsinki Botanical Museum. On the field, the percent cover was estimated for each plant species in each sampling plot. For analytical purposes, if a plant was noted as present but covered less than one percent of a plot’s area, it was given an assumed value of 0.5 percent cover.

Plots were located on 37 different transects at 10 - 20 m intervals. One-third of the transects were paired inside and outside fenced areas in order to remove as many confounding factors as possible, thus providing a more solid foundation by which to compare the relative effects of seasonal (winter) grazing versus year-round (summer and winter) grazing by livestock. Further, to more adequately compare between plots and species, only the plots observed in late summer are included in most analyses (n = 118; this excludes eight plots that were heavily trampled along a travel route). The relative impacts of fencing and grazing patterns on biodiversity (species richness), vegetation cover, and species composition thus are studied in one area of alpine meadow near Qinghai Lake in the northeastern part of the Tibetan plateau.

Most statistical analyses were done with the Data Analysis Tool in Microsoft Excel (version 7.0), though factor analysis (principal components analysis) was done in SPSS for Windows (version 6.0).

Species richness (back to top)

A simple (though only partial) indicator of biodiversity is species richness, measured as the average number of plant species observed per plot. On average, 11.3 species were observed in the seasonally grazed (winter, fenced) plots, with a 95 percent confidence interval, CI, equal to 0.6 (sample size, n = 67 plots), while only 9.8 species were observed in the continuously grazed (summer and winter, unfenced) plots (CI = 0.7; n = 51). Because variances differ significantly (F-value = 1.7704, p = 0.0150), a t-test assuming unequal variances was done to determine if these average species richness values are statistically different. This test shows that the difference in average number of plant species observed inside and outside the fenced areas is statistically very significant (t Stat = -3.4939, p = 0.0007).

In total, over 40 plant species (in 12 families) were observed in late summer plots, including 3-4 sedges, 8-11 grasses, 4-5 legumes, and at least 25 other forbs (Table 7).

Table 7. Plant species observed in alpine meadow vegetation, arranged in approximate order of abundance in the study area (species identification by Simo Tolvanen; classification based on Hao 1938, Mitchell and Rook 1979, Missouri Botanical Gardens 2000)

FAMILY Species Notes

CYPERACEAE - Sedge Family

Kobresia humilis Sedges are the most abundant graminoids in

Kobresia sp. The Tibetan plateau rangelands, especially

Carex sp. Kobresia species.

an unidentified sedge

POACEAE (GRAMINAE) - Grass Family

Poa tibetica Most alpine grasslanad formations in

Poa bulbosa Qinghai are comprised of Kobresia spp. and

Poa sp. one or more grasses. Stipa purpurea, for

Stipa purpurea example, is a dominant whose center of

Stipa capillata importance is on the Tibetan plateau

Stipa sp. (Change 1981).

Festuca ovina

Festuca sp.

Elymus nutans

Achnatherum splendens

an unidentified grass

LEGUMINOSAE (FABACEAE) - Bean Family

Medicago ruthenica Several Astragalus spp. Are endemic or

Astragalus polycladus native to China.

Astragalus alpinus

Astragalus sp.

Oxytropis falcata

COMPOSITAE (ASTERACEAE) - Daisy (Sunflower) Family

Artemisia campestris China native

Leontopodium nanum China native

Leontopodium himalayanum

Taraxacum sp.

Anaphalis lactea

Crepis flexuosa

Heteropappus altaicus

Heteropappus hispidus

an unidentified compositae

THYMELAEACEAE - Mezereum Family

Stellera chamaejasme See Figure 25.

LABIATAE (LAMIACEAE) - Mint Family

Dracocephalum heterophyllum

ROSACEAE - Rose Family

Potentilla multicaulis Potentilla multicaulis is endemic to China,

Potentilla nivea and Potentilla nivea is native to China.

Potentilla cuneata (P. ambigua)

Potentilla bifurca

Potentilla chinensis

Coluria longifolia

PLANTAGINACEAE - Plantain Family

Plantago asiatica

UMBELLIFERAE (APIACEAE) - Carrot Family

Bupleurum scorzonerifolium

CAMPANULACEAE - Bellflower Family

Adenophora gmelini

GENTIANACEAE - Gentian Family

Gentiana dahurica

Gentiana szechenyii

Gentiana sp.

POLYGONACEAE - Buckwheat Family

Polygonum viviparum

Polygonum sp.

Unidentified family

unidentified forb (no. 1)

unidentified forb (no. 2)

Species-area curves (back to top)

A fuller view of species richness can be seen in species-area curves, which display the cumulative number of plant species observed in function of the total area sampled. Species-area curves were obtained by randomly (re)sampling the same plots examined earlier for 15 iterations, each time taking note of the cumulative number of plant species observed in function of the total area sampled. Twenty plots (5 m²) were examined in each set of iterations. Multiple iterations were needed in order to calculate 95 percent confidence intervals for the species-area curves. The species-area curves inside and outside fenced pastures are shown in Figure 33. “W-mean” is the average number of species observed in winter (fenced) pastures, “S-mean” is the average number of plants observed in summer (unfenced) pastures, and “Log” and “Ln” represent the logarithmic function. (It should again be noted that the unfenced areas denoted here as “summer” pastures traditionally have been winter pastures, but with the recent agricultural reforms – including the introduction of fences – traditional winter areas that have not yet been fenced have increasingly been grazed in the summer as well). The algebraic formulae for the two species-area curves also are given in Figure 33, where y is the cumulative number of species observed, and x is the total area (in m²). Species richness in fenced and unfenced pastures tends to plateau at around 38 and 35 species. Table 8 also shows that differences in species richness between the two pasture types becomes statistically significant when around 3.75 m² (15 plots) of alpine meadow is examined.

Figure 33. Species-area curves observed in fenced pastures (seasonally grazed, winter only) and unfenced pastures (continuously grazed, summer and winter) in alpine meadow vegetation near Qinghai Lake

Table 8. Cumulative number of plant species observed in sampled plots

Area Plots Fenced Pastures Unfenced Pastures Level of Significance

(m²) (#) (and 95 Percent CI) (and 95 Percent CI) (*: p < .05; ns: not sign.)

0.25 1 11.93 (1.09) 9.33 (1.26) ns

0.50 2 17.47 (1.40) 15.93 (1.15) ns

0.75 3 21.20 (1.17) 19.21 (1.11) ns

1.00 4 24.20 (1.40) 21.50 (1.03) ns

2.00 8 30.07 (1.23) 27.36 (1.44) ns

3.00 12 33.27 (1.42) 30.71 (1.04) ns

3.75 15 35.27 (1.00) 32.36 (0.68) *

4.00 16 35.53 (0.97) 32.86 (0.65) *

5.00 20 37.33 (0.71) 34.00 (0.72) *

The above measures of species richness represent a-diversity, or the diversity of species within a single habitat type. When other habitat types are also examined, many new plant species are observed. Inter-habitat diversity is called b-diversity. Several other types of vegetation are present in the vicinity of the present study area, including tussock grassland near the lakeshore, wetland habitat, Salix shrubland on the slopes of the Nanshan Mountains, and a different alpine meadow type near the summit of the mountains. According to Su (1993), the specific types of rangeland near the study area include “temperate forage improved grassland” as well as Iris lactea lowland meadow, Achnatherum splendens temperate steppe, Stipa purpurea high-cold steppe, and Kobresia pygmaea and Kobresia humilis alpine meadow.

~~These measures of species richness represent~~ a~~-diversity, or the diversity of species within a single habitat type. When other habitat types are also examined, many new plant species are observed. Inter-habitat diversity is called~~ b~~-diversity. Several other types of vegetation are present in the vicinity of the present study area, including tussock grassland near the lakeshore, wetland habitat,~~ ~~Salix~~ shrubland on the slopes of the Nanshan Mountains, and a different alpine meadow type near the summit of the mountains. According to Su (1993), the specific types of rangeland near the study area include “temperate forage improved grassland” as well as ~~Iris lactea~~ ~~lowland meadow,~~ ~~Achnatherum splendens~~ ~~temperate steppe,~~ ~~Stipa purpurea~~ ~~high-cold steppe, and~~ ~~Kobresia pygmaea~~ ~~and~~ ~~Kobresia humilis~~ ~~alpine meadow.~~

Vegetation cover (back to top)

Vegetation cover also varies between fenced and unfenced pastures. On average, the seasonally grazed (winter, fenced) plots had 55.0 percent cover, while the continuously grazed (summer, unfenced) plots had only 40.5 percent cover. Because variances do not differ significantly (F-value = 1.0606, p = 0.4175), a t-test assuming equal variances was done to determine if the differences in percent cover are statistically significant. This test shows that the difference in average percent cover of plots located inside and outside the fenced areas is indeed statistically very significant (t Stat = 4.5126, p = 0.0000). However, it is still not known which individual species, or groups of species, contribute the most to the observed differences in plant cover between the different land management schemes.

To study the contributions of individual species or groups of species to total vegetation cover, a simplified dataset was used that includes only species with an average percent cover over one percent and that were present in at least one-quarter of the study plots. Despite the reduced number of species (22 species; Table 9), this list still accounts for over 90 percent of the total vegetation cover in both types of pasture. However, since the distribution of most species’ percent cover in the sampled plots is not normal (i.e., most species have skewed distributions, with a very low percent cover in most plots, and much higher cover in only a few plots), it is not possible to calculate whether the observed differences between a species’ cover in each of the two treatments is statistically significant (i.e., between the “inside” and “outside” fencing treatments, or between the seasonally grazed versus more continuously grazed pastures).

Table 9. Most common species in alpine meadow vegetation, with average percent covers and the proportion of plots with species present inside and outside the fenced areas

Species Name Percent Cover of Species Proportion of Plots w/ Species

Inside Outside Inside Outside

Kobresia humilis 11.2% 11.4% 65.7% 60.8%

Medicago ruthenica 6.3% 4.7% 88.1% 84.3%

Poa tibetica 4.5% 0.6% 71.6% 21.6%

Stipa purpurea 3.3% 0.4% 37.3% 11.8%

Festuca ovina 3.2% 1.9% 79.1% 56.9%

Elymus nutans 3.0% 0.9% 53.7% 25.5%

Astragalus polygladus 2.5% 1.8% 68.7% 64.7%

Carex sp. 2.5% 0.4% 46.3% 17.6%

Artemisia campestris 2.4% 0.6% 58.2% 37.3%

Stellera chamaejasme 2.0% 0.9% 59.7% 41.2%

Potentilla multicaulis 1.6% 2.1% 61.2% 76.5%

Leontopodium nanum 1.6% 0.4% 17.9% 13.7%

Astragalus alpinus 1.3% 0.6% 20.9% 13.7%

Dracocephalum heterophyllum 1.3% 0.8% 37.3% 52.9%

Taraxacum sp. 0.9% 0.8% 44.8% 51.0%

Poa bulbosa 0.8% 0.6% 25.4% 19.6%

Kobresia sp. 0.8% 2.7% 13.4% 23.5%

Potentilla nivea 0.6% 0.4% 31.3% 31.4%

Potentilla cuneata 0.4% 1.5% 31.3% 58.8%

Potentilla bifurca 0.3% 1.8% 22.4% 43.1%

Plantago asiatica 0.2% 0.9% 14.9% 33.3%

Achnatherum splendens 0.0% 1.2% 0.0% 21.6%

Based on the above highlights, each factor has been defined according to main plant types (Table 11). Factor 1, for example, has an important sedge component, along with a legume and another forb, while Factor 2 includes an important grass component along with a legume and sedge. Transformation of the data with principal components analysis (PCA) provides statistically independent factors with normal distributions (as well as a reduction in the number of factors to be analyzed). If we consider only the factors that explain over 5 percent of the total variability in percent cover, the 22-species matrix is reduced even further to only 8 independent factors (Table 10). To increase the interpretability of each factor, all loadings were maximized using Varimax rotation on the entire matrix. Simple factor definitions are given in Table 11, and the entire factor matrix is shown in Table 12.

Table 10. Main principal component analysis (PCA) factors

Variable Eigenvalue Percent of Variation Cumulative Percent

Factor 1 2.3386 10.6 10.6

Factor 2 2.0470 9.3 19.9

Factor 3 1.9922 9.1 29.0

Factor 4 1.5334 7.0 36.0

Factor 5 1.4969 6.8 42.8

Factor 6 1.3366 6.1 48.8

Factor 7 1.2973 5.9 54.7

Factor 8 1.1384 5.2 59.9

Table 11. PCA factor definitions (based on the rotated factor matrix; see Table 12), with the statistical significance of the differences of each PCA factor between fenced and unfenced pastures

Factor Factor Name Main Plant Species in PCA Factor Differences

No. Name Between

Treatments

Fct 1 SEDGES-LEGUMES Carex sp., Kobresia sp., Astragalus polycladus p = 0.0239 (*)

Fct 2 GRASSES-LEGUMES Poa tibetica, Poa bulbosa., Astragalus polycladus p = 0.0232 (*)

Fct 3 GRASSES-FORBS Stipa purpurea, Taraxacum sp. p = 0.1181 (ns)

Fct 4 FORBS-GRASSES 1 Potentilla spp., Achnatherums splendens p = 0.0027 (*)

Fct 5 FORBS-GRASSES 2 Potentialla cuneata, Plantago asiatica, E. nutans p = 0.5517 (ns)

Fct 6 LEGUMES 1 Astragalus alpinus p = 0.0000 (*)

Fct 7 FORBS Stellera chamaejasme, Leontopodium nanum p = 0.1087 (ns)

Fct 8 LEGUMES 2 Medicago ruthenica p = 0.0096 (*)

Table 12. Principal components analysis (PCA) factor matrix after Varimax rotation (see Table 9 for full species names)

Species Fct. 1 Fct. 2 Fct. 3 Fct. 4 Fct. 5 Fct. 6 Fct. 7 Fct. 8

Name

K. humilis .27122 -.48912 -.20599 -.10948 -.19046 .00157 -.18632 -.13473

M. ruthenica .02738 -.01045 . 19644 -.12347 -.05313 -.11777 -.09907 .76830

P. tibetica -.02974 -.77025 . 00630 -.17191 -.16887 .12625 -.08360 .07300

S. purpurea -.02885 -05886 . 74286 .09286 -.12597 .13345 .16043 .05580

F. ovina -.29255 .15125 -.25281 -.03071 -.46190 -.01110 .35895 .18520

E. nutans .04761 .01424 . 28567 .01166 .51031 .09423 .17405 .37959

A. polycladus .43698 .49723 . 11078 -.26151 .07970 -.49291 -.13228 .02404

Carex sp. .69345 .05409 -.09986 -.16011 -.08069 -.13153 .09197 .20024

A. campestris .53988 .1999 -.29342 .18668 -.11701 .14104 .04452 .16038

S. chamaejasme .04523 -.01753 -.02279 -.12146 -.06546 .07452 .75672 -.02341

P. multicaulis .05046 -.32211 -.00888 -.34931 -.10709 .04607 -.25049 -.40398

L. nanum -.07347 -.01794 . 27347 -.12746 .01889 -.06319 .49751 -.46801

A. alpinus .11019 -.01998 -.08707 .00950 .11333 .75506 .20357 -.03535

D. heterophyllum -.09262 .16645 . 30408 -.06076 -.01301 .68917 -.34003 -.10296

Taraxacum sp -.04794 .13357 . 73618 -.09016 .07349 -.06639 -.13899 .11347

P. bulbosa .41984 .68692 -.08376 .03870 .04970 -.08444 -.02493 -.15515

Kobresia sp. -.68488 .08437 -.13028 -.12090 -.06903 -.10717 .07832 .19838

P. nivea -.11156 .12944 -.16946 .61758 -.26624 -.27741 -.09101 -.21665

P. cuneata .04550 -.04145 -.19199 .29123 .68350 .03409 .07018 .01261

P. bifurca .10879 -.07333 .07036 .73046 .02636 .33095 -.06026 .02637

P. asiatica -.24453 .08864 -.04041 -.12352 .67180 .03347 -.18815 -.08582

A. splendens .02329 -.12349 .01767 .55506 .22901 -.07528 -.11098 .02709

How should each PCA factor be defined? To begin, each species can be examined to see if it is “loaded” mainly on only one or a few factors – a box has been drawn around the main loading(s) (see Table 12). Then, in a second step, each factor can be examined to see which of its component ~~species have~~species have the highest loadings – these loadings have been shaded in Table 12 [boxes and shading not available in web-format].

To increase further our confidence in the analysis of whether or not fencing (or the grazing pattern that fencing represents) has a significant impact on each PCA factor, a subset of the 118 plots examined to date (i.e., the 67 plots inside and 51 plots outside fenced areas) were paired in the sampling design in order to remove or to minimize as much as possible the effect of as many confounding variables as possible (e.g., altitude, short-term patterns of livestock grazing, proximity to water, proximity to tents). These paired plots (n = 76, or 38 pairs) provide repeated measures of percent cover (transformed through principal component analysis into factor values) in similar habitat types and at the same distance on both sides of the fences. The difference of cover between inside fenced areas and outside fenced areas was examined separately for each factor with a paired t-test. The significance levels of these differences are indicated in Table 11. Most strikingly, four of the five PCA factors that exhibit significant differences between inside and outside fenced areas have relatively important “leguminous” components (factors 1, 2, 6, and 8). As already indicated, Factor 1 also has an important sedge component, and Factor 2 a large grass component, while Factor 4 is comprised mainly of other forbs. It is evident, then, that fencing – and the changes in land use pattern that fencing tends to represent – has a measurable and statistically significant effect on a variety of plant species and plant types. Surprisingly, however, no significant difference was observed for one factor comprised mainly of two weed species (Factor 7). It is generally recognized, for example, that Stellera chamaejasme is a poisonous plant inedible by livestock, and thus a species that is likely to increase in the biota because of selective livestock grazing. This species can therefore serve as an indicator of overgrazing (Environment Science and Technology 1998b). Qualitative visual observations indicate that Stellera chamaejasme was more abundant outside of fenced areas, in the year-round grazed pastures, but the overall sample size in this study was too small to statistically detect this trend.

Another way to compare groups of species is to categorize them according to main types, that is, grasses, sedges, legumes, and other forbs. An examination of the paired dataset (n = 76) according to these four main types reveals in particular that grasses and all forbs combined (i.e., including legumes) have a significantly lower percent cover outside of fenced areas, but that the percent cover of sedges remains unchanged. Furthermore, regarding the individual contributions made by each plant type to the total vegetation cover, grasses tend to comprise a smaller proportion of the total cover outside of fenced areas, while sedges tend to comprise a larger proportion. Legumes and other forbs, on the other hand, contribute similarly to overall vegetation cover whether inside or outside the fenced areas (Table 13). When all plots are examined together (i.e., more than just the paired plots; n = 118), it also is noted that grasses and leguminous plants have significantly lower percent covers outside fenced areas. Finally, while grasses’ contribution to total cover is significantly less outside fenced areas, the contribution of forbs tends to be higher (Table 14). Thus, on the whole, the treatment of continuous grazing (outside fences) does not appear to affect sedges very much, but it apparently leads to a decrease in the absolute percent cover of grasses and legumes, as well as to a decrease in the proportion of grasses in the overall vegetation assemblage. Based on Table 14, this treatment also appears to lead to a slight increase in the proportion of non-leguminous forbs in the overall vegetation assemblage.

Table 13. Average percent covers and contributions to total vegetation cover of four plant types (paired plots only, n = 78), with p-value of paired t-test for differences between fencing treatments

Plant Type Percent Cover of Each Plant Type Proportions of Total Vegetation Cover

Inside Outside (p-value) Inside Outside (p-value)

Grasses 12 % 5 % (0.0013) 22 % 17 % (0.1450)

Sedges 16 % 16 % (0.8034) 30 % 37 % (0.1187)

Legumes 10 % 7 % (0.0127) 21 % 20 % (0.6737)

Other Forbs 12 % 8 % (0.0033) 27 % 26 % (0.8963)

Table 14. Average percent covers and contributions to total vegetation cover of four plant types (all plots, n = 118), with p-value of t-test (assuming unequal variances) for differences between fencing treatments

Plant Type Percent Cover of Each Plant Type Proportions of Total Vegetation Cover

Inside Outside (p-value) Inside Outside (p-value)

Grasses 15 % 6 % (0.0000) 28 % 17 % (0.0002)

Sedges 14 % 14 % (0.9966) 26 % 32 % (0.1919)

Legumes 10 % 7 % (0.0040) 21 % 21 % (0.9692)

Other Forbs 11 % 10 % (0.4072) 25 % 30 % (0.0956)

A final very important question is, Are all the above results simply artifacts of more livestock being grazing outside of fenced areas in the late summer when most plots were observed, or are these results truly indicative of longer-term impacts? If short-term effects were most important, then it would be expected that opposite conclusions would be drawn from the observations made earlier in the summer – that is, in early summer when livestock still were inside the fenced pastures (or just recently had moved away), relatively more vegetation cover would be observed in the still largely ungrazed pastures outside of the fenced areas. However, the data do not support this hypothesis since there is no significant difference in vegetation cover between fenced and unfenced areas in early summer, and the direction is similar to that observed in late summer. On average, fenced and unfenced plots had 60.4 and 53.7 percent cover, respectively. Measures of species richness (the average number of species per plot) also are not significantly different between the two treatments in early summer (10.5 and 11.0 species per plot inside and outside fenced areas, respectively). And, finally, the species richness values measured in both seasons (i.e., in early and late summer) are not significantly different (t-test, p = 0.8970). These few simple observations indicate that the observed patterns of vegetation cover, species richness, and species composition are unlikely to be caused solely by recent grazing events (i.e., short-term effects), but they are rather likely to result from the joint effect of both short-term and longer-term land use management practices.

General Discussion (back to top)

The observations and analyses presented in this chapter indicate that grassland degradation is occurring in measurable ways at a very local level as well as at the broader scale that ~~one can~~is more readily observed when traveling through the province’s vast alpine regions. This degradation takes the form of species loss, changes in plant species composition, and a decrease in average vegetation cover. Furthermore, these changes are impacted by fencing, which is tightly linked (correlated) with new land use patterns, particularly with decreased seasonal mobility and increased year-round grazing in some pastures (Sheehy 1993, Williams 1996, Fernández-Giménez 1997).

In principle, the move to fence grasslands is tied up with the quasi-privatization of land in China (through the Household Responsibility System). The endeavor to fence grassland is meant to lead toward a more uniform grazing pressure on the land, with each individual family using resources more proportionate with their own needs and their livestock’s needs. In this scenario, grassland fencing, land privatization, the construction of permanent houses, and even the shift toward a market-driven economy are all a part of a single drive to make the Tibetan plateau grasslands more productive within the larger socio-political context of China’s national economy. Raising fences also is so tightly linked with building houses and other aspects of the sedentarization that its full impact may neither be entirely predictable (Skånes 1997), nor in the desired direction of change in these ecologically fragile grassland habitats (Li et al. 1993, Williams 1996).

The strong linkages between fencing, increasingly sedentary lifestyles, and resultant changes in livestock grazing patterns cannot be stressed sufficiently. A recent newspaper article gives a glimpse into the depth and the degree of these interactions in present-day Qinghai:

However, as Holzner and Kriechbaum (1999) maintain, “if nomads move into houses from tents, … the result will be the overgrazing of the surrounding areas,” as can be seen in many localities on the Tibetan plateau and around the world. And regarding the specific impact of fencing,

“enclosures, as implemented through village level social context, actually compound grazing problems for most residents and the wider ecosystem. Expanding household enclosures intensify hyper-critical stocking ratios on highly vulnerable rangeland, exacerbating wind and soil erosion process across vast territories only to protect small isolated fields dedicated to poorly financed fodder cultivation” (Williams 1996).

The main argument generally used in favor of fencing is that, although livestock are now owned and managed by individual households (and sometimes by schools, villages, or other corporate entities), it remains extremely difficult for individuals to ensure that their grassland is never used by outsiders (i.e., cheaters) unless it is permanently guarded – which is virtually impossible – or, alternately, physically enclosed or fenced. Most of the social mechanisms that traditionally guarded against cheating were lost during the intense upheavals associated of the recent past, including “liberation” in the 1950s, the Great Leap Forward (1958-62), the Cultural Revolution (1966-1976), and the commune era in general (Goldstein and Beall 1990, Geoffrey 1993, Becker 1996, Smith 1996, Wu N. 1997, Wozencraft 2000).

The second main argument commonly used to promote grassland fencing is that it protects the grassland from overgrazing. However, this argument is based on the erroneous assumption that the mere presence of fences is sufficient to protect natural resources. A simple analogy is found in the national parks of the world, which have become small islands of less degraded land within a larger matrix of land where “anything goes” and where resources are degraded or depleted in unsustainable ways (McNeely and Thorsell 1991, Whelan 1991, Noss and Cooperrider 1994, Stevens 1997a, Mowforth and Munt 1998, Holzner and Kriechbaum 1999). A more accurate view of fencing is to consider it as only one of several tools that can be used to manage grassland ecosystems. Presently, fencing schemes in China are used as much to intensify land use as they are to promote the long-term protection of grasslands. As already stated, fencing ultimately leads only to a redistribution of livestock on the land, not to a reduction in their total numbers or in overall grazing pressure. Thus, in and of itself, fencing affects only the temporal and spatial pattern of livestock grazing in an area.

In promoting a uniform and spatially fixed presence of herders on the Tibetan plateau’s alpine grasslands, local governments and international development agencies alike also (wrongly) assume that these ecosystems have fixed carrying capacities. However, not only is it very difficult to determine with certainty any carrying capacity, even in more stable ecological environments, but such “capacities” may not even exist in most of the plateau’s extremely variable (unpredictable) environments. Successional theory and equilibrium-based land management strategies likely do not apply in arid and climatically variable environments (Coughenour 1991). Specifically, Qinghai’s alpine grasslands are more likely to operate as a state-and-transition system (Westoby et al. 1989a, 1989b) or to have a “dynamic carrying capacity” (Cincotta et al. 1992, Miller 1995, Miller and Craig 1997). The variability of these grasslands’ environment also requires that as much flexibility as possible be maintained in any grazing system, whether traditional or modern, in order to allow herders to respond rapidly to the normal fluctuations in seasonal and annual primary productivity, as well as to periodic catastrophic events. Overall, flexibility is crucial. However, flexibility is limited by fences as well as by individual legal titles (contracts, leases) to the land.

While livestock grazing was traditionally very seasonal, with distinct summer and winter pastures, now only the summer pastures (usually situated high in the mountains) tend to be grazed in one season only. With all the “improvements” recently brought to the winter areas, including permanent dwellings and livestock shelters, these areas – the traditional winter pastures – are now increasingly being used to some degree in every season of the year. According to Cincotta et al. (1992), almost all the “technological improvements that have entered this production system after household responsibility was assumed [have focused] around activities on the privately-managed winter grazing area.” This is certainly the case in the Qinghai Lake area, and in most other parts of the province as well. However, it is only fenced winter pastures that are truly managed as private land – if unfenced, most pastures are treated simply as “free access” land, with all of the associated pitfalls (Monbiot 1998, Ho 1998, 1999, Holzner and Kriechbaum 1999). Furthermore, despite the Chinese government’s effort to “eradicate poverty,” there are still large differences in wealth even among pastoralists. It is such differences that originally led to some pastures in the vicinity of the study area being fenced in 1985, and to other pastures not being fenced. (The “snapshot” presented in this chapter was thus taken approximately 12 years after fences were built). Today, encouraged by the government, wealthier pastoralists (or the poorest, aided by government loans and subsidies) are continuing to make the transition toward a more intensive grazing management system, with new infrastructures, grassland “improvements” (artificial grasslands), and new management strategies (Su 1993, Wei 1993, Western Resources and Environment Research Center 1994, Ma et al. 1995, Drandui 1996, Dorje 1997). As noted earlier in this study, however, any improvement observed inside fenced areas is usually paralleled by some degradation outside the fenced areas.

Fencing all the grassland also is not the answer, since this would only render the entire grazing system dependent on external government subsidies, as well as lock the pastoral population into a sedentary grazing system that cannot respond adequately to high environmental variability. The deteriorating conditions of grasslands in the vicinity of settlements in Mongolia’s grasslands give a vivid example of what could happen if more mobility is lost in pastoral systems – and just as Mongolia is re-considering the value of traditional grazing patterns, so to should the government and international development agencies capitalize more on time-proven, flexible livestock grazing strategies (Li et al. 1993, Mearns 1993, 1995, Sheehy 1993, 1996, Müller 1995, Germeraad and Enebisch. 1996, Williams 1996, Fernández-Giménez 1997). Indeed, as Holzner and Kriechbaum (1999) explain,

“the causes of overgrazing [are] complex and varied, and are not [only] ecological but cultural, social and economic … [including] change or abandonment of the seasonal grazing pattern (induced by changing political or administrative boundaries, or by stopping the migrations of nomads by law or economic stimuli like building houses).”

According to Su (1993), and based on discussions with local pastoralists, livestock has traditionally been herded at the base of the Qinghai Nanshan Mountains and near Qinghai Lake in the cold season, and herded much higher in the mountains in the warm season. Rarely, if ever, was livestock grazed in the present study area during the warm season. Although a traditional grazing pattern is still practiced in large part today, there is an increasing amount of summer grazing in the traditional winter pastures as well. This trend has been noted in many other parts of the province as well. Thus, at least in the present study, fenced pastures may represent not so much the “new” management as they do more traditional forms of grazing – that is, seasonal grazing. Counter-intuitively, it is the unfenced land that best represents the newer, more intensive grazing management that now is being promoted throughout the province. As such, fences are not so much “improving” the grassland as they are “protecting” it from the new trend of continuous grazing. And, as already noted, it is continuously grazed grasslands that are relatively less diverse, and have a poorer vegetation cover. In short, new forms of intensive grassland management in Qinghai may not be ecologically sustainable (Cincotta et al. 1992, Miller 1995), and ~~both~~ more seasonal mobility and flexibility should be sought in future development initiatives. Development investments should also be directed toward promoting more opportunistic management alternatives, both traditional and modern (Galaty and Johnson 1990, Briske and Heitschmidt 1991, Sheehy 1993, 1996, Miller and Jackson 1994, Tainton et al. 1996 cited in Holzner and Kriechbaum 1999, Wester 1997).

(back to top)

_________________________

Table of Contents

Chapter Five

The Ecology of Grassland Enclosures and Changing Patterns of Livestock Grazing: A Vegetation Analysis

5.1. Introduction

5.2. Methods and Results

5.2.1. General methods

5.2.2. Species richness

5.2.3. Species-area curves

5.2.4. Vegetation cover

5.3. General Discussion

(back to top)