Land parcellation reduced spatial redundancy resulting in grassland degradation

Abstract

Degradation is the most immediate threat to biodiversity and sustainable livelihoods of global pastoral grasslands. However, we are still uncertain why grassland degradation is enhanced by the land policy of grassland parcellation. We hypothesized that spatial redundancy of the grazing system, that is, parcellation accelerates land fragmentation and reduces spatial redundancy, leads to land degradation. This study aimed to clarify the relationship between households, livestock and grassland management under the grassland contract policy, and to determine the causes and consequences of grassland fragmentation and degradation. We interviewed households who received land under the grassland contract policy in typical pastoral areas of the Qinghai–Tibetan Plateau, and then, we employed binary logistic regressions, multinomial logistic regressions and fuzzy‐set qualitative comparative analyses to examine the relationship between spatial redundancy and grassland–livestock balance. A very significant (p < 0.001) negative asymptotic exponential trend emerged between grassland area and stocking rate, but with regional differences. Grassland spatial redundancy and stocking rate displayed a significant (r² = 0.96; p < 0.05) negative asymptotic exponential trend, that is, households with a lower spatial redundancy had a higher stocking rate. Based on our results, spatial redundancy can be used as one of the key indices to evaluate landscape stability of grazing systems. In any terrestrial ecosystem, the size of spatial redundancy reflects the carrying capacity of the land and determines the delicate relationship between land, animals and population, that is, in equilibrium or non‐equilibrium. Therefore, taking the Qinghai–Tibetan Plateau as an example, we face the spread of grassland degradation caused by the gradual fragmentation of resources under the current grassland contract policy and parcellation in global pastoral areas. Determining the spatial redundancy size of grassland is an important prerequisite for understanding grassland degradation and formulating grassland management strategies. These novel findings have potential beneficial information for resource utilization, especially land management planning in pastoral areas worldwide.

Read the free Plain Language Summary for this article on the Journal blog.

INTRODUCTION

The tragedy of the commons (Hardin, 1968) is a major challenge for many societies, and the privatization of the commons is often considered to be the solution. However, this has not been the case in many pastoral regions, as, for example, the grassland contract policy in China was based on the privatization of most pastoral areas, and grassland degradation remains widespread (Cao, Holden, et al., 2018a; Conte & Tilt, 2014; Su et al., 2021; Yang et al., 2020). Similar results following land privatization have occurred in other regions of the world such as the fragmentation of the landscape in Iran (Barati et al., 2021), and the degradation and fragmentation of rangelands in southern Ethiopia (Boru et al., 2015) and Kenya (Lesorogol, 2010), all of which are closely related to the changes and development of land management policies. Consequently, Ostrom's (1990) view that privatization of grassland may not be a good solution is supported, but how can this paradox be explained? Why does this land reform lead to degradation of grassland in pastoral areas such as the Qinghai–Tibetan Plateau (QTP)? We believe that it may be related to the weakening of the resiliency of the grassland after parcellation.

Spatial redundancy of renewable natural resources allows more recovery time, which, according to the ecological redundancy theory, is the key to sustainable utilization of resources (Marcus & Colding, 2014; Streit et al., 2019). It leads to high resilience (Gothe et al., 2014; Marcus & Colding, 2014), which depends on the balance between grazing and restoration of the grassland (Streit et al., 2019). Therefore, revealing why grasslands are undergoing rapid degradation following grassland parcellation from the perspective of the spatial redundancy theory (Marcus & Colding, 2014) could provide new insights in rangeland degradation of pastoral areas worldwide, and identify the social–ecological mechanisms that drive grassland degradation.

We argue that a reduction in redundancy weakens the grassland–livestock balance, which, ultimately, could result in the collapse of the grassland ecosystem and the animal husbandry economy. With the increase in world population and the need for more food, the trade-off between productivity and environmental damage is pushing the global ecosystem to the brink of collapse (Funabashi, 2018). Overuse and increasing productivity of the land promote this trade-off in an unfavourable direction, that is, at the cost of consuming natural capital (such as land resources), leading to global ecological overshoot and damaging resource security (such as land degradation; Wackernagel et al., 2021). Therefore, the transformation towards sustainability of the social–ecological system is particularly important in the face of the pressure brought by the ever-increasing population, food demand and the continuous depletion of natural resources (Fischer et al., 2021).

Property rights is considered a tactic to solve the problems of land degradation (Li et al., 2007). However, it has been demonstrated that privatization often causes more land degradation than the traditional collective property system (Cao, Xu, et al., 2018b; Sneath, 1998), regardless of whether land ownership is maintained by the government or the individual (Ostrom et al., 1999; Tseng et al., 2021). Some governments are modifying the land use rights system by increasing funds to protect and restore the ecosystem (Ouyang et al., 2016; Su et al., 2021). For example, ‘Separating Three Property Rights’ was the latest tenure reform policy inaugurated in 2017 in China, in which grassland ownership belonged to the state or was collective, and households had the right to parcellate and operate the land (Su et al., 2021). However, grassland degradation was not alleviated under this grassland contract policy (Cao, Holden, et al., 2018a). In the Borana Zone in southern Ethiopia, the government initiated irrigation-based farming in privatized land, but this fragmented the grazing rangelands (Boru et al., 2015).

The parcellation and contracting of common pastureland to individuals has led to fencing of the property and fragmentation of resources (Jürgenson, 2016). This has caused ecological and social problems (Herse & Boyle, 2020), among which the imbalance of household resources has a direct negative impact on livestock production and the livelihood of the households (Tan et al., 2018), as livestock is the main measure of wealth of households. These practices are restricting the pasture availability to livestock, reducing the actual utilization of space (Hirt et al., 2021) and threatening the local biodiversity (Banks-Leite et al., 2020). Resource fragmentation causes the grassland to be divided into many plots, which reduces the area, carrying capacity, vegetation productivity and household's potential income; increases the production cost; and aggravates the degradation of natural resources (Kosmas et al., 2016). Spatial redundancy could provide a novel perspective in understanding the grassland contract policy dilemma in pastoral areas. For this purpose, we put forward the spatial redundancy hypothesis of the grazing system combined with the ecosystem redundancy theory, that is, the regeneration of grassland enables high productivity, which is the equivalent of more grassland area. The spatial redundancy resource could alleviate the pressure of time and promote the function of grassland regeneration. Based on this reasoning, we hypothesized that the acquisition of individual land ownership has accelerated fragmentation and reduced grassland spatial redundancy, resulting in grassland overgrazing and degradation. To test this hypothesis, we interviewed household members of pastoral areas, being guided by the following questions: (1) What are the relationships between grassland area, stocking rate and grassland degradation? (2) What are the drivers for the difference between the grassland–livestock relationships among households? (3) Does grassland parcellation by the grassland contract policy affect spatial redundancy? and (4) Has grassland parcellation restricted grassland services such as providing feed for livestock? These findings could provide valuable information for policy makers on grassland ecological restoration and sustainable development in global pastoral areas.

MATERIALS AND METHODS

Study area

The pastoral areas of the QTP have undergone substantial changes in land management policies, grazing systems, ecological environment and socio-economic development in the past decades (Yang et al., 2020). We selected four typical pastoral counties, namely Zoige, Maqu, Maqin and Naqu, on the QTP for the study (Figure 1). Zoige, Maqu and Maqin are located in the north-eastern part of the QTP (32°56′–35°16′ N, 98°48′–103°39′ E) and cover an area of 34,210 km² at average elevations of 3471, 3700 and 4200 m a.s.l., respectively. The counties are characterized by a typical plateau continental climate of sub-frigid and sub-humid conditions. The average annual air temperature is 1.1, 1.7 and −0.6°C with an annual precipitation of 650, 600 and 513 mm, respectively. Naqu County is located in the hinterland of the QTP (30°31′–31°55′ N, 91°12′–93°05′ E) at an average elevation of 4500 m a.s.l., and covers 16,200 km². It has a highland mountain climate with an average annual air temperature of −2.1°C and an annual precipitation of 400 mm. The livelihood of these four counties relies on animal husbandry, and households raise mainly yaks, Tibetan sheep and horses.

[IMAGE OMITTED. SEE PDF]

Conceptual framework

Figure 2 presents the analytical process of this study in the form of a conceptual framework. Taking the continuous degradation of grasslands on the QTP after the implementation of the grassland contract policy as an example, we verified the issues with Hardin's ‘tragedy of the commons’ and its privatization solution. Based on the theory of ecosystem redundancy, we propose a novel hypothesis of spatial redundancy in grazing systems to explain the process of grassland degradation. To this end, we selected herders as our research subjects and set up variables such as basic household information, grassland and livestock management and cognitive behaviours (Table S1) to analyse the causes of grassland degradation. The conclusions are supported primarily by logistic regression and Fuzzy-set qualitative comparative analysis.

[IMAGE OMITTED. SEE PDF]

Field survey

A total of 51 pure pastoral villages, including 17 in Zoige, 13 in Maqu, 12 in Maqin and 9 in Naqu, were selected randomly for semi-structured interviews from June to August, 2020. All interviewed households participated in the grassland contract policy programme, which was confirmed by their government certificates (Figure S1). The interviewee was usually the head of the household, but occasionally, was an older household member, and the interviews were carried out in the Tibetan language. A total of 374 households were interviewed, of which 74 were discarded due to incomplete information, and, therefore, 300 interviews, including 80 from Zoige, 63 from Maqu, 84 from Maqin and 73 from Naqu, were analysed. Total milk collected by the households on the day of the interview was weighed, and the daily milk production per yak was calculated. The descriptive statistics of the households surveyed are presented in Table S2. The average age of the household head was 44 years, while 78.3% had no formal education. The household size and number of labourers averaged 5.7 and 2.6 persons, respectively, and the household annual income averaged 67,760 yuan.

Evaluation of grassland-livestock balance

Nine households in each county were selected randomly from July to August, 2020 for pasture measurements. Three random quadrats (0.5 × 0.5 m size) in each grassland were selected, all vegetation was harvested at ground level and fresh weight was determined. We used the grazing pressure index (GPI) and the grassland–livestock balance index (GLBI) to calculate the carrying capacity of each household's grassland (Xu et al., 2012; Zhang et al., 2014). Details are presented in the Supporting Information.

Evaluation of grassland spatial redundancy

The regenerative ability of grassland determines the availability and spatial redundancy of resources within grazing units, ensuring sustained support for grazing activities. An ideal grazing unit should have enough space to meet livestock demands, relieve time pressure from space to promote regeneration of pasture and maintain sustainable use of grassland. However, when the grassland area is small, the spatial redundancy of the grazing unit is reduced, and the stocking rate can exceed the carrying capacity. We proposed a grassland spatial redundant factor for each household, based on the definition of the structural strength redundant factor (Frangopol & Curley, 1987). The method of calculation is presented in the Supporting Information.

Logistic regression model construction

We selected 19 independent variables (Table S2) for the 300 interviews, and the sample size met the requirements to generate a logistic regression (Yang et al., 2020). The variance inflation factor (VIF) was used to test for multi-collinearity among variables, and a value of less than 10 was accepted (James et al., 2013; Table S3).

We used the GPI to evaluate whether the stocking rate of the grassland, as the dependent variable of the binary logistic regression (Supporting Information), was at carrying capacity. Independent variables were screened through stepwise regressions (forward, LR method), and the final model retained only significant variables (p < 0.05). The odds ratio (OR) was used as a measure of relative risk. The accuracy of the logistic regression was determined using the receiver operating characteristic curve analysis (ROC; Fawcett, 2006), in which the accuracy of the prediction model increases with an increase in the area under the curve (Lin et al., 2011). The degree of grassland–livestock balance was used as the dependent variable for multinomial logistic regression analysis. In this study, the ordinal regression model was chosen initially; however, due to the violation of the parallel line assumption, multinomial logistic regression was used for the analysis (Jamil et al., 2023).

According to GLBI, the dependent variable was divided initially into five levels of stocking rate in relation to carrying capacity, namely, understocked, at carrying capacity, overstocked, seriously overstocked and extremely overstocked (Table S5). However, the classification accuracy ratio of the seriously overstocked model was 0%, which may have been caused by too small a sample size (36), and, consequently, the model was meaningless. These samples were then included in the overstocked group, and, therefore, there were four groups (understocked, at carrying capacity, overstocked and seriously overstocked). At carrying capacity was used as a reference variable, and the remaining three variables as dependent variables for the multinomial logistic regression model. Three models were generated to identify the drivers that determine the degree of grassland–livestock balance.

Fuzzy-set qualitative comparative analysis

Qualitative comparative analysis (QCA) is a set theory configuration analysis method based on Boolean algebra. By examining the sufficient and necessary subset relations between antecedent conditions and results, it can determine how complex social problems caused by multiple concurrent causations occur (Rihoux & Ragin, 2008). Compared with traditional quantitative research methods, QCA provides a new perspective on complex causal relationships such as concurrent causations, equivalence and asymmetry (Pappas & Woodside, 2021). Currently, QCA methods can be used as an independent alternative to existing standard regression analysis methods, or as a complementary component to hybrid methods combined with standard regression analysis (Greckhamer et al., 2013). For example, regression analysis can only provide the importance ranking of different independent variables according to the influence and contribution degree of each independent variable to the dependent variable (El-Habil, 2012). However, QCA provides the core condition configuration of the results from a holistic perspective, indicating that there are various paths to the results, each one composed of different conditions, and the condition combination of the results differ from that of the results (Ragin & Fiss, 2008). In addition, complex social phenomena may be more easily understood in terms of set relations, and, therefore, the set theory approach is suitable (Schneider & Wagemann, 2013). Why the grassland is still degraded under the grassland contract policy on the QTP is a common social problem that is caused by many factors and is also a typical configuration problem. Therefore, this study attempts to examine how grassland degradation is caused by employing the combination of regression analysis and QCA, which is also the first application of QCA in the sustainable management of pastoral areas.

QCA includes mainly crisp set QCA (csQCA), fuzzy-set QCA (fsQCA) and multi-value QCA (mvQCA). csQCA can deal only with binary variables, fsQCA can characterize continuous variables as any number between 0 and 1, that is, membership score, while mvQCA is an extension of csQCA, but it is not fully utilized compared with the first two types (Thiem & Dusa, 2013). Most of the variables in this study were continuous variables, so fsQCA was selected for subsequent analyses. The number of conditions is the key factor to consider in building the fsQCA model. If there are too many conditions, the number of configurations will easily exceed the number of cases, and the problem of limited diversity of cases will appear (Ragin & Fiss, 2008). It has been suggested that a small sample size limits the model to seven conditions, but even with a large sample size, too many conditions may complicate the interpretation of the findings (Greckhamer et al., 2013). Usually, the appropriate condition selection is based on theoretical or empirical knowledge, such as clarifying the causal relationship between the condition and the result through mainstream statistical methods, and further demonstrating the joint impact of each condition on the result from the perspective of configuration (Greckhamer et al., 2018). In this study, we used binary logistic regressions and multinomial logistic regressions to analyse the main determinants of the grassland–livestock relationship, and finally obtained eight conditions, including grassland area, number of livestock, days of feed supplementation, feed costs, family members, livestock deaths, whether livestock damage the grassland and whether to increase the number of livestock. Since fsQCA deals with continuous variables, the two binary variables of whether livestock damage the grassland and whether to increase the number of livestock were eliminated, and the remaining six variables were selected as antecedent conditions to analyse the core configuration of grassland space redundancy. Furthermore, considering the stability of fsQCA analysis results, a boxplot detected outliers, while the cases inside the box targeted cases for analysis (Schwertman et al., 2004). In total, 111 cases inside the box were obtained.

The analysis used fsQCA3.0 software, and the following procedures: Step 1: Calibration of causal and outcome conditions. Three membership scores were used to calibrate all conditions data between 0 and 1, cut-off points of fully-in and fully-out were set at 0.95 and 0.05, and the maximum ambiguity was set to 0.50 (Ragin & Fiss, 2008). Step 2: Assessment whether necessary conditions were required for the result to occur. A consistency threshold of the condition greater than 0.9 with a non-trivial coverage was considered necessary (Schneider, 2018). The consistency of all conditions was less than 0.9, and the negation of the results was also less than 0.9 (Table S4). Step 3: Constructing a truth table. The minimum case frequency was set to 1, and the consistency threshold to 0.8 (Douglas et al., 2020). Through standard analyses, three kinds of solutions were produced—complex, parsimonious and intermediate. Intermediate solutions were usually the first choice for fsQCA (Schneider & Wagemann, 2013). Step 4: Test for robustness. The robustness of the findings was assessed by changing the cross-over point of the causal conditions (Fiss, 2011). FsQCA was applied for the first time for theoretical research on pasture-livestock trade-offs and results are presented following the conditions described by Fiss (2011).

Statistical analyses

Non-parametric testing and logistic regression all used SPSS22 (SPSS Inc., Chicago, IL, USA), and Forestplot in R 4.1.2 software displayed the logistic regression results. The asymptotic exponential function described the relationship between grassland area and GLBI. Origin Lab 2018 (OriginLab Corporation, Northampton, MA, USA) was used for asymptotic exponential fitting and box plots. p < 0.05 was accepted as the level of significance.

RESULTS

Relation between grassland and livestock

There were differences among counties in livestock stocking rate in relation to the carrying capacity of the rangeland (Figure 3a, Table S5). Among the 300 households, 67 (22%) were understocked, 48 (16%) were at carrying capacity, 60 (20%) were overstocked, 36 (12%) were seriously overstocked and 89 (30%) were extremely overstocked. No leased grassland was found in Naqu County (Figure 3b), which had the most (p < 0.01) days of feed supplementation (Figure 3c). Households in Zoige County had the highest average feed costs and annual income (Figure 3d,e).

[IMAGE OMITTED. SEE PDF]

An asymptotic exponential model was then fitted to the relationship between grassland area and GLBI in the four counties (Figure 4). With an increase in grassland area, the GLBI decreased exponentially and significantly (r² = 0.377), with the highest correlation in Naqu County (r² = 0.661).

[IMAGE OMITTED. SEE PDF]

Key drivers of overstocking grassland

A binary logistic regression model identified the key drivers that determined whether the household's grassland was overstocked (Figure 5a). The area under the ROC curve was 0.950 (Figure S2), that is, greater than 0.9, indicating that the binary logistic model could distinguish the binary variables. Grassland area, livestock number, days of feed supplementation, feed costs, grazing damage of pasture and increasing livestock number were significant (p < 0.05) drivers for determining whether the grassland was overstocked. Decreasing grassland area resulted in greater overstocking (OR = 0.996), feed costs had no impact (OR ≈ 1), while increases in livestock number (OR = 1.021), days of feed supplementation (OR = 1.193), grazing damage of pasture (OR = 8.21) and increasing livestock number (means expanding the scale of breeding; OR = 4.524) all lead to overstocking.

[IMAGE OMITTED. SEE PDF]

Key drivers of the relationship between grassland and livestock

The larger the grassland area (OR = 0.997), the lower the stocking rate. For understocked grassland (Figure 5b), understocking increased with larger grassland area (OR = 1.001) and greater number of household members (OR = 1.352). In addition, lesser livestock (OR = 0.987) and household policy not to increase livestock number (i.e. households believed they cannot continue to expand scale of breeding) (OR = 0.106), lowered understocking. For overstocked grassland (Figure 5c), overstocking increased with an increase in livestock number (OR = 1.011) and mortality number (OR = 1.029), and when households believed that grazing did not harm the grassland (OR = 7.071). Seriously overstocked grassland was similar to overstocked grassland (Figure 5d).

The relationship between carrying capacity and spatial redundancy of grassland

We proposed a grassland spatial redundancy factor (R) to measure the spatial utilization of grassland resources of the household. R > 1 indicated that the grassland owned by the household met the spatial resource needs of the grazing livestock when there was a moderate increase in grazing pressure; R = 1 meant that the grassland just met the space resource needs of the grazing livestock; and R < 1 meant that the grassland area did not meet the space resource needs of the grazing livestock. The relationship between stocking rate and spatial redundancy of grassland was analysed by curve fitting with the GLBI (Figure 6). There was a very significant (r² = 0.96, p < 0.05) negative exponential asymptotic relationship, that is, the higher the GLBI, the lower the R, indicating that the lower the spatial redundancy, the greater the overstocking.

[IMAGE OMITTED. SEE PDF]

Sufficiency analyses for grassland spatial redundancy

Results of the spatial redundancy of grassland by the configuration of the six conditions are presented in Table 1. There were four configurations for the occurrence of grassland spatial redundancy, and, whether a single solution or overall solution was employed, the consistency was higher than the acceptable minimum standard of 0.75 (Pappas & Woodside, 2021).

TABLE 1 Configurations sufficient for spatial redundancy of grassland by the fuzzy-set qualitative comparative analysis (fsQCA) model.

Configuration	Solutiona
1	2	3	4
Grassland area	●	●	●	●
Number of livestock	$\otimes$	•		$\otimes$
Livestock deaths	•	•	•	•
Days of feed supplementation		$\otimes$	•	•
Feed costs	$\otimes$	●	●	•
Household members	•		$\otimes$
Consistency	0.934	0.937	0.936	0.956
Raw coverage	0.353	0.271	0.338	0.345
Unique coverage	0.143	0.028	0.009	0.027
Overall solution coverage	0.564
Overall solution consistency	0.930

Grassland area was the core condition shared by the four solutions, indicating that grassland area was most important for grassland spatial redundancy. Grassland area and number of livestock (absence) were core conditions shared by solutions 1 and 4. When combined with the marginal conditions of livestock deaths, feed costs (absence) and household members, a high grassland spatial redundancy resulted for solution 1, and when combined with the marginal conditions of livestock deaths, days of feed supplementation and feed costs, the same results were obtained for solution 4. Grassland area and feed costs were core conditions shared by solutions 2 and 3. When combined with the core condition of feed costs and the marginal conditions of feed supplementation (absence), number of livestock and number of livestock deaths, a high grassland spatial redundancy resulted for solution 2. When not considering the number of livestock, the core condition of household members (absence), combined with the marginal conditions of livestock deaths and days of feed supplementation, solution 3 was also obtained.

DISCUSSION

Balance between grassland and livestock

The trade-off between grassland health and livestock production usually depends on the management by the household, and the size and quality of the grassland (Cao et al., 2019). In the grassland–livestock relationship in the four counties, the larger the grassland area, the lower the degree of overstocking and the more likely of a grassland–livestock balance. In contrast, the greater the stocking rate of the grassland, the greater the degradation of the grassland. This answers our first question on the relation among grassland area, stocking rate and grassland degradation.

Livestock is the basic currency in pastoral areas, and households with a small grassland area increase the number of livestock to increase income. However, this practice ignores the input of production cost and the carrying capacity of the land, causing ecological damage to the grassland (Barati et al., 2021). The current transition of grassland rights in the pastoral areas from public to private is one of the main reasons. The privatization of land, and practices such as fencing, leads to fragmentation of the grassland resources. It has profound impacts on the sustainability of grassland and livestock production (Hobbs et al., 2008). On the QTP, the spatiotemporal heterogeneity of food resources necessitates that grazing livestock, such as yaks, migrate between areas to obtain energy, nutrients and water. There is evidence that yaks walk 0.11 km per hectare each day while foraging (Liu et al., 2019). However, when a grassland is divided into smaller plots by fencing, it limits livestock access to abundant resources, which affects their health and normal production capacity. This also poses threats to the survival of wildlife, as fencing hinders wildlife movement (Sun et al., 2020), which in turn threatens the maintenance of biodiversity. This is similar to the fragmentation, degradation and loss of grazing rangelands in Kyrgyzstan and Tajikistan (Kerven et al., 2012), Kenya (Lesorogol, 2010; Nkedianye et al., 2020), Ethiopia (Boru et al., 2015) and Mongolia (Sneath, 2003), driven by factors such as land ownership and policies, fencing and government priorities.

The current grassland contract policy and parcellation of grassland resources has caused inequality in resource utilization, increased grassland fragmentation and adversely affected grassland health, especially for small-scale households (Cao, Holden, et al., 2018a). For example, the implementation of the Grassland Ecological Protection Award Policy (GEPAP) has not been successful, and overgrazing continues (Yin et al., 2019). This policy provides subsidies to households based mainly on grassland area and household size. Households with small grassland areas and small families receive lower subsidies than households with large grassland areas and large families. As a result, small-scale households are reluctant to comply with the policy, and attempt to improve their livelihoods by increasing the number of livestock. Similar findings have been reported in studies of other pastoralist groups (Hou et al., 2021; Zhen et al., 2014). It is believed that the effectiveness of agricultural policies for households depends largely on their willingness to accept government policies (Conte & Tilt, 2014). Based on this study, small-scale households were reluctant to accept the policies. In addition, most households who received compensation did not know the purpose of the payment, which may be related, at least in part, to their low level of education (Wang et al., 2016). Therefore, while implementing these policies, the government should organize more training sessions to help pastoralists understand the purpose of the policies, especially regarding grazing management. Additionally, the subsidy policies should favour small-scale households.

Drivers for the relationship between grassland and livestock

Binary logistic regressions revealed that grassland area, livestock number, days of feed supplementation, feed costs, grazing damage of pasture and increasing livestock number were the main drivers affecting the stocking rate of the grassland. The smaller the grassland area, the more likely it was overstocked, which was consistent with the results of the previous part of this study. Households with a small grassland area faced the problem of limited flexibility in grazing management, and the grassland was more likely to be overstocked, which led to grassland degradation, as was reported in an earlier study (Zhang et al., 2020). Livestock was the main source of livelihood for the households, and it also affected the utilization of resources and changes in the ecological environment (Herrero et al., 2009). The greater the number of livestock, the more likely of overstocking the grassland, which is related to the carrying capacity of the grassland (Fetzel et al., 2018). It has been shown that carrying capacity per unit area increases with increasing area of land because of increasing adaptive foraging options (Fynn, 2012). However, due to the high variability of the pastoral environment, grassland-carrying capacity can currently be defined only as a long-term average, which also has important implications for the management of grazing systems (Seifollahi-Aghmiuni et al., 2022). Therefore, how to optimize grazing management according to the carrying capacity of grassland is a common worldwide problem faced by social ecosystems in pastoral areas.

Households who believed that livestock damage the grassland but can continue to increase the number of livestock had a high probability of overstocking the grassland. The households' perception and grassland management were not consistent, they believed that livestock damaged the grassland and, yet, increased the stocking rate. This may be related to the mismatch between the compensation received and the land size and number of livestock of the households (Zhen et al., 2014). The households received the livestock reduction policy passively, and it was basically ignored. A similar finding was reported in a study on consumers' cognitive behaviour of environmental protection, in which the consumers expressed willingness to protect the environment but did not participate in the programme (Nguyen et al., 2019). In the present study, the households' actions were inconsistent with the government's forage–livestock balance policy. Many households owned more livestock today than 10 years ago, and insisted that the grassland was not overstocked and could even support more livestock, which was similar to the findings of Hou et al. (2014).

Multinomial logistic regressions indicated that the number of livestock deaths and number of household members affected grassland–livestock relationships. Households with more livestock deaths had more serious grassland overstocking. These households raised a large number of livestock, resulting in a lack of forage in the cold season, which increased mortality. This was particularly true when natural disasters such as heavy snow storms occurred, leading to the death of livestock (Yin et al., 2019). This can be compared to the raising of cattle by the Borana in southern Ethiopia where extreme drought and high stocking rate interacted to often cause livestock mortality of over 60%. Here, stocking rate increased, as livestock is a measure of wealth, and ‘herd accumulation remains the primary and most effective means of risk management for pastoralists’ (Gebru & McPeak, 2004). The key to mitigating this dilemma is to reserve enough fodder for the cold or dry season, which requires pastoralists to understand that keeping some grassland for producing forage in the growing season is not a waste of resources, but to store grass for the dormant season. Second, another benefit to encourage pastoralists to adopt this strategy is that this practice restores grassland and can improve soil carbon storage. Pastoralists can then apply for carbon credits, providing an extra source of income.

In the current study, feed costs were not affected by stocking rate; however, households with many days of feed supplementation tended to overstock. We speculated that this may be related to the feed and management in the regions. For example, In Naqu county, although the number of days of feed supplementation was greater than the other counties (Figure 3c), the households in this county did not lease grassland (Figure 3b), and the investment in feed was the least, so the problem of insufficient feed could not be alleviated, and the overstocking was becoming more serious. This suggests that a rational feed supplementation plan is critical for the health of livestock. The logistic regression models answered our second question: What are the drivers for the difference between the grassland–livestock relationships among households?

Spatial redundancy of grassland

We first proposed the grassland spatial redundancy factor from the perspective of resource space requirements, and examined the causal conditions determining the spatial redundancy of grassland by fsQCA. Grassland area and number of animals were the key factors determining spatial redundancy. In addition, livestock deaths, days of feed supplementation, feed costs and household size were important in different configurations.

Logistic regressions revealed that large households had a low probability of grassland overstocking, while small households displayed an opposite trend. In this study, this occurred because large households still had enough grassland and labour to avoid grassland overstocking. However, under the land contract policy, as the household grows and the parcellation of land continues, the possibility of future overstocking increases. In the survey based on the GLBI, the proportion of young households with overstocked grassland was greater than with understocked grassland (Figure 3a). Therefore, we reasoned that families with few members were mainly newly married young couples. Such families had few resources, a weak economic foundation, and overstocking was common. Resources are limited, and the continuous growth of the population increases the competition for resources (Wackernagel et al., 2021). For large-scale households, they have sufficient grassland for rotational grazing, with grassland designated for recovery in each season. This is especially useful for alleviating the severe shortage of forage in winter on the QTP (Jing et al., 2022). In this situation, the carrying capacity of the grassland increases over time, that is, the spatial redundancy that we propose increases. In contrast, small-scale households have limited grassland areas and no resources for rotation, making them more prone to overgrazing. This indicates that parcellation of grassland resources reduced spatial redundancy indirectly, which supported our hypothesis of grazing system spatial redundancy. This answered the third question: Does grassland parcellation affect spatial redundancy?

The implementation of the grassland contract policy made the limited original public grassland resources available for personal control (Figure 7b). Although this can avoid the tragedy of the commons, the implementation of the family property inheritance system has further exacerbated the parcellation of the grassland, so that small-scale households possess less resources, reducing spatial redundancy. Most households increased the number of livestock in order to escape poverty, and although the income increased in the short term, the increased demand of feed for livestock resulted in insufficient feed availability in the long term (Li et al., 2015). This answered the fourth question: Has grassland parcellation, due to grassland contract policy and inheritance rights, restricted grassland provisioning service such as feed for livestock?

[IMAGE OMITTED. SEE PDF]

Land fragmentation and degradation are still increasing worldwide, creating environmental problems (Banks-Leite et al., 2020; Kosmas et al., 2016). It was reported that increased competition for land resources will continue to cause social and political instability, and exacerbate food insecurity, poverty, conflict and migration in the future (Decorte et al., 2016). Therefore, how to reform land management, improve the effectiveness of common land governance and minimize or halt land fragmentation are the key in addressing land degradation and improving the well-being of millions of people who are highly dependent on land resources.

The spatial redundancy hypothesis of the grazing system proposed in the current study provides new insights for achieving the above goals. Redundancy constraints within household pastures can increase spatial redundancy and enhance resilience of grassland through coupling across scales or units (Gothe et al., 2014). A novel ‘two settlement model’ of pastoralism was presented that could improve constraints of the grassland area (Shang et al., 2014). This proposal, combined with our proposed grazing system spatial redundancy hypothesis, could form the basis of sustainable development of social–ecological grazing systems in the future (Figure 7) as: (1) ecological protection projects should be based on the spatial redundancy of grasslands, the economic status of households and regional differences; (2) grassland parcellation should be discontinued and small-scale households should be encouraged to practice cooperative management, improve the spatial utilization of resources and promote sustainable landscape transformation; (3) the government should set a threshold for the grassland area per household based on the actual conditions of the region. This threshold should be the critical area required to support a household from livestock and ensure that there is spatial redundancy in grassland; (4) households' motivation to protect grassland ecology should be enhanced by grazing management technology training and the dissemination of ecological protection concepts; and, (5) cooperation with experienced households should be encouraged to strengthen the protection and management of the ecosystem through the combination of theory and practice.

AUTHOR CONTRIBUTIONS

ZhanHuan Shang conceived the ideas and designed methodology; JianXin Jiao, ShanShan Li, WenYin Wang and LinYan Qi collected the data; JianXin Jiao and YanFu Bai analysed the data; JianXin Jiao, Ting Jiao and A. Allan Degen led the writing of the manuscript. All authors contributed critically to the drafts and approved the paper for publication.

ACKNOWLEDGEMENTS

This work was supported by the Second Tibetan Plateau Expedition (2019QZKK0302), by the Science-based Advisory Program of The Alliance of National and International Science Organizations for the Belt and Road Regions (ANSO-SBA-2023-02), the Natural Science Foundation of China (U21A20183; 31961143012), the ‘111’ Programme 2.0 (BP0719040). We would like to thank TianHua Jia, Cairangdongzhi and AngLuo Bao for help in fieldwork.

CONFLICT OF INTEREST STATEMENT

The authors do not have any conflicts of interest regarding the article.

DATA AVAILABILITY STATEMENT

The data are available in figshare digital repository: .

References

Banks‐Leite, C., Ewers, R. M., Folkard‐Tapp, H., & Fraser, A. (2020). Countering the effects of habitat loss, fragmentation, and degradation through habitat restoration. One Earth, 3, 672–676. [DOI: https://dx.doi.org/10.1016/j.oneear.2020.11.016]