Impact of China's Grain for Green Project on the landscape of vulnerable arid and semi-arid agricultural regions: a case study in northern Shaanxi Province

Summary

• 1
  China's Grain for Green Project is a rapid landscape-scale shift in ground cover and land use with significant implications for biodiversity. From 1998 to 2005, we carried out field studies to examine the landscape-level impacts of this project, and to provide a practical example of the successes and failures of a large-scale attempt to restore a vulnerable environment.
• 2
  In a northern part of China's Shaanxi Province, our results indicated that the total vegetation cover in areas covered by this project increased from 29·7% in 1998 to 42·2% in 2005. However, we also found evidence that large-scale afforestation in this vulnerable arid and semi-arid region could increase the severity of water shortages, decrease vegetation cover in afforestation plots, and adversely affect the number of species present. The exclusion of livestock from overgrazed areas and the elimination of cultivation in marginal areas had the biggest effects on the restoration of vegetation cover, whereas tree planting had a strong negative effect in vulnerable areas.
• 3
  Synthesis and applications. In practical terms, the destruction of natural vegetation cover during afforestation should be avoided, as this makes the soil surface more vulnerable to erosion and reduces species diversity. Managers should reduce the intensity of farming and grazing on fragile land rather than relying on afforestation as the primary tool for ecological restoration in arid and semi-arid areas. Afforestation remains a valuable tool but should be limited to the planting of native or other species that will not exacerbate soil water shortages such as stable communities of natural desert steppe, maximum water-use efficiency dwarf shrubs, and possibly even lichen species in more severely degraded environments.

Introduction

Sustainability has become a central concept in environmental planning and policy over the last 20 years but remains poorly understood at the landscape level (Peterseila et al. 2004). Scientists, land managers, and communities face the challenge of meeting human needs while protecting the environment from damaging development at local to global levels (Cash et al. 2003). Managers face an even bigger challenge when they must scale up ecological restoration efforts from individual sites to the landscape level (Lamb, Erskine & Parrotta 2005). This difficulty is exacerbated by a lack of sufficiently detailed knowledge of ecosystems and of human impacts upon them, an absence of sustainability guidelines for specific ecosystems, and a lack of ‘best practices’ for adopting available guidelines (Western 2001). To permit sustainable agricultural development, policymakers require a profound understanding of agricultural systems at the level of individual farms and of how these systems scale up to the regional level (Bontkes & Keulen 2003).

Changes in land use and land cover, and the dynamics of land and water use, are central issues in the study of global environmental change, and must be clearly understood before sustainability will become possible (Fischer & Sun 2001). Land-use changes are most apparent in terms of the changes in vegetation and other cover types, and these changes reflect both how the land is being used and the goals of its users (Verburg, Chen &Veldkamp 2000). In the present study, we studied land-use changes in China's Loess Plateau, in northern Shaanxi province, that have arisen from China's Grain for Green Project (GGP), a conservation set-aside programme designed to restore fragile ecosystems damaged by unsustainable farming and grazing. Our goal was to review the successes and failures of a large-scale attempt to restore a degraded and vulnerable environment to a more stable condition.

China is the most populous country in the world, has one of the largest territories, and has a booming economy. However, forest cover now accounts for only 16·5% of the nation's area. A half-century policy of forest exploitation and monoculture planting in China has led to large decreases in species diversity as a result of the disappearance of natural forest, and large increases in insect and disease problems in monoculture plantations (Liu et al. 2003). During the 1990s, the area of eroded land increased by more than 10 000 km² annually, with the result that 38% of China's total land area is now considered badly eroded. At least 200 plant species have become extinct in China since the 1950s, and more than 61% of wildlife species have suffered severe losses of their habitats (Li 2004).

Chinese environmental problems are among the most severe of any major country and show signs of worsening (Liu & Diamond 2005). The list of problems includes air pollution, losses of biodiversity and cropland, depleted fisheries, desertification, disappearing wetlands, grassland degradation, and an increasing frequency and scale of human-induced natural disasters (Liu et al. 2003). There is also evidence of problems associated with invasive species, overgrazing, interrupted river flow, salinization, soil erosion, accumulation of waste materials, increased severity of floods and dust storms, degradation of forests and landscapes, and water pollution and shortages. These issues are not accounted for in national economic statistics, but are nonetheless causing serious economic losses (both actual and potential), as well as social conflicts and increased health care costs (Liu & Diamond 2005). Many Chinese, including the nation's leaders, are aware of these problems and have tried to tackle them.

Since 1999, China's government has pursued one of the most ambitious conservation set-aside programmes in the developing world – the GGP – to prevent soil erosion (Uchida, Xu & Rozelle 2005). Priority areas include upstream regions of major river systems, and especially the Yellow, Yangtze, and Songhuajiang river basins, which have sustained massive ecological and environmental degradation during the past 50 years (Zhang et al. 2000). The government plans to spend US$40 billion on the project to convert 147 million ha of farmland into forest and grassland and 173 million ha of wasteland (grassland) into forest in 25 provinces in western China from 1999 to 2010 (Tao, Xu &Xu 2004). In most cases, the project focuses on cultivated land on steep slopes (≥ 25°), since these locations are most likely to experience severe erosion and other adverse impacts resulting from cultivation. By the end of 2003, 72 million ha of farmland (33 million ha in 2003 alone) in all pilot areas had been transformed into forest or grassland. This amounts to 49% of the 2010 target. In addition, 79·3 million ha of suitable grassland had been planted with trees, accounting for 46% of the 2010 target (Tao, Xu &Xu 2004; Uchida, Xu & Rozelle 2005).

Both its scale and the magnitude of the investment make the GGP the largest ecological restoration programme in the world (Zhang et al. 2000; Liu & Diamond 2005; Uchida, Xu & Rozelle 2005). As part of this project, the government of China appears to be making aggressive changes in forestry-related policies, which formerly emphasized economic returns (Zhang et al. 2000). Accordingly, the focus of the new policies is on how to grow more forests and how to shift from natural vegetation to human-made forests (i.e. afforestation) as a fast way to promote restoration of the landscape. In the present study, our goals were to evaluate potential links between the GGP environmental policy and the environmental sustainability of the project. Most importantly, the research was intended to provide a case study that reports the results of a major environmental remediation policy in terms of environmental factors.

Materials and methods

study area

Our study focused on five randomly selected counties (Jingbian, Ansai, Baota, Yanchang, Luochuan) out of the 25 counties in northern Shaanxi Province (Supporting Information, Fig. S1), that are covered by the GGP. Severe soil erosion has occurred in these areas historically, at an average rate of approximately 15 000 tons km⁻² year⁻¹. The annual average temperature is 9·4 °C, and annual average precipitation is 547·4 mm, with an average of 413·6 mm falling during the rainy season from July to September based on meteorological records from 1950 to 2000. The annual frost-free season is 147 days. From 1998 to 2005, total annual precipitation averaged 461·7 mm, ranging from 366 mm in Jingbian County to 609·4 mm in Luochuan County (Supporting Information, Table S1). The potential evapotranspiration for the study area averages 793·7 mm yr⁻¹, and is thus well above precipitation in most years. The topography is very rough, with many steep slopes and deep ravines. Loess is the main soil type at our study sites. It covers more than 75% of the area, averages 50 to 200 m in depth, and is found primarily on hill tops, level areas, and the upper slopes of valleys. Its composition is > 20% sand (1 to 0·05 mm) and < 30% clay (0·010 to 0·001 mm), with 3·8 g kg⁻¹ mean organic matter content and a mean porosity of 52·1% (Yan’an City Soil and Water Conservation Team, 1985).

observations and analysis

In each of the five counties, we selected five representative villages that had afforestation plots and abandoned land (land on which grazing and agriculture had been prohibited under the GGP). Because of the different land-use types in the study area, we established 3 to 8 randomly selected 0·5 ha plots at each village: 1 to 6 tree-species plots, a farmland plot, and an abandoned land plot. Our total sample size was thus 25 villages, and n = 25 villages × 3 to 8 plots per village. In each plot that underwent afforestation under the GGP, we assessed every tree species from 1998 to 2005 each year during the summer. Five species of non-native trees were planted in single-species plantations (Robinia pseudoacacia, Prunus armeniaca, Hippophae rhamnoides, Platycladus orientalis, and Caragana korshinskii) and in mixed-species plantations (R. pseudoacacia with C. korshinskii). The five species and the mixed-species plantations accounted for 21·1%, 19·6%, 15·5%, 4·1%, 31·2%, and 8·5% of each tree species, respectively, in the plantations that we studied in the five counties (Supporting Information, Table S2). Tree survival rates were monitored by means of random onsite sampling (i.e. at each afforestation plot we monitored a randomly determined sample of 100 trees per species).

Using a steel tape, we measured the crowns of 20 randomly selected trees of each species in each plot each year during the middle of the growing season (between the last 10 days of June and the end of August) to determine crown area, which we used to represent the vegetation coverage of the site. We measured the maximum and minimum crown radii and modelled the crown as an ellipse, with these radii representing the semi-major and semi-minor axes, and calculated the mean canopy area for each species using geometric mean values to account for extreme values. Tree cover (the proportion of the total site area accounted for by a vertical projection of the elliptical crowns of the trees) was calculated by multiplying the mean crown area of a species in a given year by the number of surviving trees in that year, then dividing this total by the total area planted with that species. Canopies that overlapped were combined and treated as a single canopy to avoid double-counting of crown area; that is, we calculated the total crown area as if the two trees were a single tree, then divided the resulting crown area by 2 to produce a mean value per stem.

In each non-forest plot, we performed line-intersect sampling using two 10-m transects at right angles to each other to survey non-tree vegetation; in forest plots, we used three randomly located 4-m² circular quadrats. We measured the vegetation cover and the presence of lichen species in the abandoned land and afforestation plots by means of line-intersect sampling in each plot using a 10-m transect perpendicular to the edges of the plots. We identified every species of plant in these transects every year at the same time (between the last 10 days of June and the end of August). Total vegetation cover (combined cover of tree and herbaceous vegetation, i.e. grasses, forbs, herbs) for a given area (e.g. a county or village) was calculated by multiplying the mean cover value for a given type of plot by the proportion of the total area occupied by that type of plot.

To describe vegetation species diversity in the study plots, samples of all plant species were collected annually from each plot in August; samples were brought to China Agriculture University if their identity needed to be confirmed. Each year during the growing season (May to October), we sampled volumetric soil moisture content in each plot at three randomly chosen locations. Using a 10-mm-diameter auger, samples were obtained on the 10th, 20th, and 30th day of each month at 20-cm intervals to a depth of 600 cm below the surface (0 to 20, 20 to 40, ... , and 580 to 600 cm). Each sample was placed in a sealed steel box to retain moisture until it could be weighed to determine its fresh weight; the samples were then oven-dried (24 h at 105 °C) to calculate the water content. These values were averaged to provide a mean value for the growing season.

spss software (SPSS Inc., Chicago, IL, USA) was used for comparisons of field data, with F tests used to compare results for different categories and post hoc tests used to distinguish differences among means. Significance is accepted at P < 0·05.

Results

vegetation cover

Our field measurements in the five counties of northern Shaanxi Province indicate that total vegetation cover for the study area (combined for planted forest and herbaceous vegetation, i.e. grasses, forbs, herbs) increased from 29·7% in 1998 to 42·2% in 2005 as a result of the GGP (P < 0·01). Yanchang County showed the greatest rate of vegetation restoration, whereas Jingbian County showed the slowest (Table 1). However, the average vegetation cover in the afforestation plots was 30·2% in 2005, which was 32·3% lower than the corresponding cover in the abandoned plots (44·6%; P < 0·05). In the afforestation plots, R. pseudoacacia and P. armeniaca had greater levels of vegetation cover (38·8% and 34·1%, respectively) than the other afforestation species. Luochuan County showed the greatest vegetation cover (46·3%) in the afforestation plots, whereas Jingbian County showed the least (20·5%) (Supporting Information, Table S3).

Table 1.  Total vegetation cover in the plots located in five counties from 1998 to 2005

Year	Vegetation cover per county (%)
	Jingbian	Ansai	Baota	Yanchang	Luochuan	Average
Values in a column followed by different letters differ significantly among the years (P < 0·05). The total vegetation cover represents the average value for the afforestation plots and the abandoned plots. A linear regression of the vegetation cover against the precipitation showed a strong and positive correlation (R= 0·96, P < 0·01).
1998	19·5a	22·1a	28·5a	21·5a	56·9a	29·7a
1999	19·6a	22·7a	28·4a	22·9b	57·2a	30·1ab
2000	21·5b	24·0ab	29·7ab	24·5bc	58·2ab	31·6b
2001	22·0b	25·5b	31·5b	26·1cd	59·8b	32·9b
2002	23·7bc	27·7b	34·8c	28·9cd	62·0b	35·4c
2003	25·9cd	31·1c	37·1cd	32·7d	64·9c	38·3c
2004	26·4cd	33·0c	39·4d	35·9d	66·5c	40·3cd
2005	27·9d	35·3d	41·0d	38·6e	67·9a	42·2d
Precipitation (mm)	366	435·6	437·4	460·1	609·4	461·7

When cultivation of farmland and grazing were prohibited, the number of lichen species increased steadily in the study area, leading to obvious lichen growth on the soil surface (Supporting Information, Fig. S2). The afforestation plots showed significantly lower development of lichens than did the abandoned plots in all counties (Supporting Information, Table S4). The average proportion of the surface covered by lichen species in the afforestation plots reached an average of only 38·4% by 2005, which was 39·2% lower than the corresponding values in the abandoned plots (63·2%). Luochuan County showed the greatest development of lichen species in both plot types, and Jingbian County the least.

contribution to increased vegetation cover

Tree survival rates in the afforestation plots averaged 55·7% in the first year after planting and 49% in the seventh year (Supporting Information, Table S5). Robinia pseudoacacia showed the highest survival (68·2%) in year 7 (2005), whereas the lowest survival was observed for P. armeniaca (11·4%), P. orientalis, (12·7%), and for trees in the mixed-species plantation (15·3%). The canopy area in the afforestation plots averaged 1·2 m² in year 7 (Supporting Information, Table S5). The mixed-species plantation and R. pseudoacacia showed the greatest mean canopy areas (2·1 and 2·0 m², respectively).

There was widespread destruction of natural vegetation during planting. This included the removal of natural herbaceous vegetation (i.e. grasses, forbs, herbs), to promote tree growth and vegetation that was destroyed during the construction of trenches designed to channel precipitation towards the tree. These activities led to a 30·5% decrease in the overall vegetation cover in the afforestation plots by year 7 (2005) (Supporting Information, Table S5). Cover of Platycladus orientalis showed the greatest decrease (49·2%) and the mixed-species plantation the smallest (22·2%), but even here, cover had not recovered by year 7. Compared with the natural vegetation in the abandoned plots, it was clear that the overall vegetation cover was negatively affected by afforestation, with a net decrease in cover (after accounting for the increase in cover due to growth of the trees) equal to 6·1% (Fig. 1). Most of the increase in vegetation cover that occurred outside these plots was thus attributed to the prohibition of grazing (99% increase) and decreases in the area of cultivated land (10·4% increase), but increased road construction and mining both had a negative impact on vegetation cover (3·3% decrease). Table S6 (Supporting Information) provides a detailed summary of the factors that contributed to increased or decreased vegetation cover.

Figure 1. Net contribution of land use to changes in vegetation cover in the northern part of Shaanxi Province. All bars significantly different from one another at P < 0·0001.

plant species

Our results suggest that the number of plant species in plots where cultivation was abandoned increased to between 21 and 31 species, compared with a range of 9 to 14 species in the afforestation plots (Supporting Information, Table S7). The main community types included Artemisia gmelinii, Artemisia giraldii, and Artemisia japonica communities. Several shrubs and shrubby species, including Lespedeza dahurica, Periploca sepium, Ziziphus jujuba var. spinosa, and Sophora viciifolia, also appeared after the abandonment of cultivation. Luochuan County showed the highest plant species diversity in the afforestation plots, and Jingbian County showed the lowest. With a total of 15 plant species, the C. korshinskii plot showed the highest species diversity of the afforestation plots.

soil moisture

Soil moisture in the afforestation plots decreased nearly continuously compared with the abandoned plots, in which the average soil moisture content in the soil to a depth of 6 m fluctuated in response to changing precipitation but showed no long-term decreasing trend, (Supporting Information, Fig. S3). Soil moisture content in the soil to a depth of 6 m in the afforestation plots increased during the first 2 years after afforestation, but decreased steadily from the third year onwards. The soil water content increased briefly during a year with heavy rainfall (2003; Supporting Information, Fig. S3 and Table S1), but the increase was only transient.

Our results indicate more severe aridity in the soil to a depth of 6 m in the afforestation plots than in the abandoned plots 7 years after the abandonment of agriculture (Table 2); the soil moisture content averaged 7·2% during the 2005 growing season, which is 37·4% lower than the mean value of 11·5% in the abandoned plots (anova, P < 0·05). The soil moisture content in the afforestation plots averaged 6·1% and 6·0% at depths of 0 to 1·0 m and 1·0 to 2·0 m, respectively, corresponding to values 63·2% and 42·8% lower than those in the abandoned plots (P < 0·05). The differences between the mean soil moisture contents at a given depth in the afforestation plots as a function of vegetation cover were generally not significant when the vegetation cover was less than 20%.

Table 2.  Soil moisture content (%) to a depth of 6 m 7 years after afforestation and the abandonment of cultivation in northern Shaanxi Province (Values were measured during the growing season of 2005)

Soil layer (m)	Abandoned plots	Soil moisture content (%)
		Afforestation plot (based on % vegetation cover)
		Average	<10	10·1–20	20·1–30	30·1–40	40·1–50	>50
Values in a row followed by different letters differ significantly among the vegetation cover percentages (P < 0·05). A linear regression of the vegetation cover against the total soil moisture to a depth of 6 m showed a strong and negative correlation in the afforestation plots (R=–0·92, P < 0·01).
0–1·0	16·60a	6·14c	8·49b	7·86b	5·78bc	5·30c	5·04c	3·50d
1·0–2·0	10·48a	5·99bc	11·28a	8·66ab	6·60b	4·12c	3·59c	3·35c
2·0–3·0	10·27a	7·49b	11·84a	9·77a	7·31a	5·85b	5·75ab	4·34b
3·0–4·0	11·42a	7·54b	12·39a	9·17a	7·22b	5·31bc	6·27b	4·65c
4·0–5·0	10·05a	8·13a	13·07b	9·58a	8·42a	6·40c	6·28c	4·77d
5·0–6·0	10·04a	8·10a	13·28b	9·20a	8·33a	6·42ac	6·27c	4·80d
Average	11·48a	7·23b	11·72a	9·04a	7·28b	5·57bc	5·53bc	4·23c

Discussion

Land-use and land-cover changes can lead to significant environmental changes at both local and global scales. These changes have important consequences for ecosystems through their impacts on soil and water quality, biodiversity, and global climatic systems (Chen et al. 2001). Planting trees in vulnerable arid and semi-arid agricultural regions such as those of the study area in northern Shaanxi Province during China's GGP produced a net decrease in total vegetation cover (Fig. 1) in the number of plant species (Supporting Information, Table S7) and in soil moisture (Table 2). Part of the problem may be that when trenches were dug during planting (Supporting Information, Fig. S4a), vegetation in the path of the trenches was destroyed (Cao et al. 2007); in addition, herbaceous vegetation (i.e. grasses, forbs, herbs) was manually removed under the trees to promote tree growth by reducing competition for moisture (Supporting Information, Fig. S4b). Table S5 (Supporting Information) summarizes the areas of vegetation that were destroyed during the afforestation process. The destruction of vegetation to promote tree growth and the low seedling survival (49% mean survival rate in year 7, with rates ranging from 11·4 to 68·2%; Supporting Information, Table S5), combined with an inappropriate emphasis on trees at the expense of other vegetation cover, explains the overall negative impact of afforestation on total vegetation cover during the project (a 6·1% decrease) in afforestation areas. Our results suggest that the policies of prohibiting cultivation and grazing in steep terrain were significantly more effective than the afforestation policy, and thus offer a valuable strategy for environmental restoration in similar remote rural regions, both in China and around the world.

Drought is a major constraint to growth of common vegetation types such as forests worldwide, and revegetation of arid regions such as those in China is primarily water-limited (Jackson et al. 2002). Since 1978, the overall survival of trees planted during afforestation projects was only 15% at the Three Norths Shelter Forest System Project across arid and semi-arid northern China (Su 2004). The mean survival rate for trees planted during the GGP afforestation projects in our study area was only 55·7% in the first year, and decreased to 49% by the seventh year, with some values much lower (Supporting Information, Table S5). Luochuan County had the most rainfall and the highest vegetation cover, whereas Jingbian County had the least rainfall and the lowest vegetation cover (Table 1).

Although cover differed among the five species used for afforestation and mixed planting, all afforestation plots had significantly lower vegetation cover than the abandoned plots (Supporting Information, Table S3) and the soil moisture worsened steadily in the afforestation plots (Supporting Information, Fig. S3; Table 2). An alternative approach would be to use fast-growing but short-lived tree species (equivalent to pioneer species) to create an initial canopy cover under which other vegetation can become established (Lamb, Erskine, &Parrotta 2005). However, there will be high demands for water as a consequence of the fast growth of pioneer species which may exacerbate soil water shortages, at least in the short term (Xu 2006). The poor afforestation performance observed in the present study may relate to an inappropriate choice of tree species (particularly the use of non-native species), given the study area's environmental constraints (particularly the low water availability). Because trees generally have lower water-use efficiency than other forms of vegetation, afforestation is a more suitable choice in areas where precipitation is suitable, but in the vulnerable arid and semi-arid agricultural regions of the present study, it would take considerable research to identify suitable species. Areas in which grazing and cultivation had been prohibited occupied the largest area in the GGP plots; hence, they contributed the highest proportion of the vegetation cover (Fig. 1). Our results show that although the annual precipitation did not differ significantly during the field investigations (except for Jingbian County in 1999 and 2000, and Yanchang County in 2003), the ecosystems were able to recover quickly with little or no human intervention when unsustainable previous land uses were prohibited before they could severely degrade the soils or deplete the local species reservoir (Mitchell & Ricardo 2004).

Biodiversity plays a vital role in the functioning of ecosystems in changing environments (Norberg et al. 2001). Biodiversity protection has thus become a fundamental policy objective both internationally and locally in the discussion of global environmental change (Yliskyla-Peuralahti 2003). Experience has clearly shown that careful choice of which areas to protect against unsustainable use can promote biodiversity (Lamb, Erskine & Parrotta 2005). The low species diversity we observed suggests that the structure and composition of the initial communities that form during the recovery process were not stable and will progress to an ecosystem with a different species assemblage. This is not surprising because previous research in the study region (Yan’an City Soil and Water Conservation Team 1985) revealed that the succession from bare ground to a stable climax community can take 20 to 40 years, and that several pioneer and mid-successional stages will occur before this succession is complete. Additional monitoring will be required for at least this time period to determine whether succession in the study area follows the same sequence as in natural ecosystems and how long it takes for stable communities to arise.

Tree planting is a popular approach to restoring degraded sites. However, plantation failures can occur when inappropriate species are selected or early stand management is inadequate or inappropriate (Lamb, Erskine & Parrotta 2005). Natural ecosystems comprise many individuals of multiple species that interact with each other and the abiotic environment to produce complex structures and dynamics (McNeill 2004). In any stable ecosystem, a dynamic equilibrium develops in which the biotic components of the ecosystem evolve over time to produce a stable community that can sustainably use the available resources, and subsequent maintenance of that stability requires maintenance of the ecosystem's biodiversity (Hooper & Vitousek 1997; Loreau et al. 2001; Dirzo & Loreau 2005; Brooks et al. 2006). The introduction of alien species into an ecosystem, as in afforestation with non-native species, alters biodiversity (Groombridge & Jenkins 2003), and if this alteration disrupts the balance of the original ecosystem, stability is lost and the ecosystem may degrade or transition to another, potentially undesirable, state. In the present study, afforestation decreased soil moisture in the grassland areas of northern Shaanxi Province (Table 2), and decreased the number of plant species at the afforestation site by an average of 52% by the seventh year after planting (Supporting Information, Table S7). Regressions of the vegetation cover against the soil moisture content showed a strong and negative correlation in the afforestation plots (Table 2), possibly because the increased vegetation cover resulted primarily from growth of the trees, which increased overall evapotranspiration (Cao 2008). Therefore, the government's overemphasis on afforestation using non-native species appears likely to increase the risk of ecological degradation in this region.

When farmland and grassland become woodland in areas with adequate levels of precipitation, the environment often improves because trees can protect the soil and retain moisture (Schume, Jost & Hager 2004). The soil moisture status determines the capacity of the soil to absorb water and thus to buffer runoff. Replenishment of soil water storage depends on the crown architecture of trees and on canopy interception of precipitation, which vary widely between tree species (Liu, Wang &Wang 2004). In arid and semi-arid northern China, soil moisture is generally deficient in planted forests as a result of low annual precipitation, unsuitable choice of tree species, and an overly high planting density (Wang, Liu & Zhou 2003; Zhao and Li 2005; Xu et al. 2006). This has led to large-scale mortality of plantations, accompanied by substantial decreases in species diversity (compared with the hundreds of species found in natural forests of the region) in China's northern Hebei, Shanxi, Shaanxi, and Gansu Provinces, and in the Ningxia, Xinjiang, and Inner Mongolia administrative regions in drought years (Wang, Liu & Zhou 2003). There is a clear negative relationship between precipitation and environmental changes when grassland and farmland are invaded by woody vegetation (Raffaelli 2004; Liu, Wang &Wang 2004) due to the large amounts of soil moisture consumed by fast-growing trees; this moisture cannot be replenished during the rainy season in vulnerable arid and semi-arid agricultural regions.

Soil moisture is an important factor that controls the growth of trees and forage species (Jackson et al. 2002). The decreased soil moisture that is available in the afforestation plots, combined with reduced sunlight under the growing tree canopies, reduce the growth of understorey vegetation, and this has decreased overall vegetation cover in the afforestation plots (Fig. 1), even when trees could not form a closed canopy. Previous research in the study region (Wang, Liu & Zhou 2003; Liu, Wang &Wang 2004) has revealed that, compared with farmland, runoff from afforestation plots decreases by an average of 77·1% (ranging from 57·9 to 96·3%). Although this decreased runoff suggests increased retention of precipitation within forested sites and increased soil moisture in the early years after afforestation, the retained moisture is often used more rapidly by the trees than can be replenished during the rainy season (Table 2). Although soil moisture can be replenished temporarily when rainfall increases in certain years (e.g. 2003; Supporting Information, Fig. S3), precipitation shortages are more common in arid and semi-arid areas. As a result, the decreased runoff from afforestation plots in vulnerable arid and semi-arid areas demonstrates that planting trees not only decreased soil moisture but also decreased the supply of water to rivers because the evapotranspiration resulting from afforestation increased steadily.

The destruction of existing natural vegetation to promote growth of newly planted trees (Supporting Information, Fig. S4a,b) explains the negative impact of afforestation on the occurrence of lichen species (Supporting Information, Table S4) in afforestation areas. When lichens and surface crusts are damaged, infiltration of water into the soil improves, but the high potential evapotranspiration of the study area means that both evapotranspiration and infiltration may increase (Zhang et al. 2006). A linear regression of vegetation cover against soil moisture to a depth of 6 m showed a strong and negative correlation in the afforestation plots. As a result, soil moisture in the upper layers of the soil is lost faster in the afforestation plots (Table 2), potentially damaging the growth and development of herbaceous vegetation. This led to significant decreases in total vegetation cover and in numbers of plant species in these plots, thereby increasing the risk of desertification (Supporting Information, Fig. S4c; Cao 2008).

synthesis and applications

The results of our study, and particularly the results in plots where cultivation and grazing were prohibited, show that landscape-scale restoration in vulnerable arid and semi-arid agricultural regions is difficult but possible. However, restoration cannot be accomplished by large government expenditure alone; expenditure must target the real problems and avoid creating new ones. An appropriate choice of tree species, combined with prohibition of the destruction of natural vegetation during planting and subsequent tending of new trees, may increase the success of afforestation efforts, but this must be confirmed through long-term monitoring. In contrast, the abandonment of farming and the removal of livestock from overgrazed areas had large and positive effects on vegetation cover and landscape restoration. These results show that an arbitrary mixture of policies to attain a single end is not necessarily efficient and can have unintended side effects if the policies are not tailored to the specific characteristics of each site. In terms of revegetation strategies, planners must understand that different environments support different vegetation communities, and thus require different solutions. Because the destruction of the natural vegetation cover during the afforestation process has had serious adverse impacts on the GGP in the study areas, forests are clearly not a suitable vegetation choice in all areas. In particular, afforestation is an inappropriate choice where mean annual precipitation is near or below the potential evapotranspiration.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (NSFC40871136) and the 11th Five Scientific and Technological Sustaining Research Program of China (2006BAD03A02). We thank Yufeng Li, Guangdong Li, Suifang Duan, Bing Wu, Jun Chen, and the offices of the GGP in Yan’an City and Luochuan, Baota, Yanchang, Ansai, and Jingbian counties for their assistance in our investigation. We also thank Prof. Richard Dawson at the China Agriculture University and Geoffrey Hart for their help in writing this paper.