Aboveground Biomass in China’s Managed Grasslands and Their Responses to Environmental and Management Variations

Abstract

Aboveground biomass (AGB) in managed grasslands can vary across a suite of environmental and management conditions; however, there lacks a quantitative assessment at the national scale of China. Although the potential effects of individual drivers (e.g., species, nutrient fertilization, and water management) have been examined in China’s managed grasslands, no attempts have been made to comprehensively assess the effects of multiple variables on AGB. Using a meta-data analysis approach, we created a database composed of AGB and associated attributes of managed grasslands in China. The database was used to assess the responses of AGB to anthropogenic factors, in addition to a suite of natural variables including climate, soil, and topography. The average AGB in managed grasslands of China is approximately 630 g m⁻² of dry matter, ranging from 55 to 2172 g m⁻² (95% confidence interval). Medicago sativa is the most widely planted species in China’s managed grasslands, followed by Elymus dahuricus and Bromus japonicus. The national average AGB of these three species was around 692, 530, and 856 g m⁻², respectively. For each species, AGB shows a large discrepancy across different places. In general, grassland AGB depends substantially on species, environments, and management practices. The dependence can be well described by a linear mixed-effects regression in which a series of biotic and abiotic factors are used as predictors. We highlight that establishing managed grassland can potentially contribute to not only AGB enhancement, but also grassland restoration on degraded natural grasslands.

1. Introduction

Covering nearly 40% of the Earth’s surface, grassland is the world’s largest terrestrial ecosystem [1], providing food and ecosystem services and contributing substantively to regulating the global carbon cycle and climate change [2]. Due to either anthropogenic activities (e.g., cultivation and overgrazing) or climate change (e.g., drought), grasslands have been degrading across the world, leading to substantial decreases in the production of forages and livestock [2,3]. To combat terrestrial ecosystem degradation, the United Nations (UN) General Assembly declared 2021–2030 the “UN Decade on Ecosystem Restoration”, which outlined the urgent need for global restoration of degraded lands, including grasslands.

China has the world’s second largest area of grasslands, a majority of which has been degrading to some extent due to both climate change and/or anthropogenic activities since more than half a century ago [4,5,6]. Grassland degradation is characterized mainly by decreases in plant species diversity, plant cover, and productivity. Synchronously occurring with degradation, grassland biodiversity and productivity have also been widely declining [7,8]. China’s grassland aboveground biomass (AGB) was generally 30–50% lower than that of approximately 60 to 70 years ago [9]. Decreases in grassland AGB pose massive threats to food security and ecosystem health. To address this issue, the establishment of managed grasslands by planting grass species with high quality and quantity forage production and/or supplying additional nutrients and water has widely been recommended [10,11]. However, to date, managed grasslands make up only approximately 3% of China’s total grassland area, which is substantially lower than those in developed regions such as Europe and Australia. Overall, to meet the growing needs of forage and livestock production, the establishment of managed grasslands is becoming an increasing trend [11].

Over the past two decades, a number of managed grassland experiments aimed at assessing the responses of plant productivity to environmental and anthropogenic factors have been conducted across China (Appendix B). Specifically, in this study, managed grasslands are defined as the grasslands managed by introducing species with high-quality and high-quantity forage production combined with a series of recommended management practices, such as nutrient fertilization and irrigation [11]. Most existing field-scale managed grassland experiments, however, focused on the effect of a single factor, such as plant species [12,13] and nutrient fertilization [14,15], on AGB. It is well known that grassland productivity is coregulated by both natural and anthropogenic attributes, such as irrigation and climatic, edaphic, and topographic conditions [16,17,18,19]. However, the regulating effects of these drivers have seldom been assessed [9], leaving a gap in knowledge on the dynamics of grassland AGB under changing environments and/or management. Moreover, potential forage production through establishing managed grasslands remains unclear at the national scale, which prevents policy makers and local shepherds from projecting the expected benefits from establishing managed grasslands.

In this study, we conducted a data synthesis of 101 publications to collate AGB observations in China’s managed grasslands. By employing a suite of environmental and management covariates, we further clarified the regulating effects of these predictors on managed grassland AGB in China.

2. Materials and Methods

2.1. Data Compilation

Using the keywords of aboveground biomass, productivity, production, China, and managed grasslands (or managed pasture), peer-reviewed publications reporting AGB measurements in managed grasslands in China were collected by searching Web of Science (WoS; since the publication year of 1990) to construct a synthesis dataset during October–December 2021. The online datasets used for the literature search in WoS included the WoS Core Collection and Chinese Science Citation Database (articles written in Chinese). We ultimately obtained 101 studies (19 in English and 82 in Chinese) by screening the return publications using the following criteria: (1) managed grasslands were involved; (2) grass species were specified; (3) amount of AGB was reported; (4) experimental locations with precise latitude and longitude coordinates were specified; (5) observation year was directly reported; (6) if fertilization was adopted, type and amount of fertilizers [e.g., nitrogen (N), phosphorus (P) and potassium (K)] were reported; and (7) water management (i.e., rainfed or irrigation) was reported or could be obtained through personal communications. We finally obtained 864 individual AGB measurements widely distributed in China’s main grassland areas (Figure 1). In all the managed grassland experiments included in our study (Appendix B), zero-grazing was adopted. In general, the standing tissue of grasses was cut and harvested manually for measuring AGB during its peak amount for a year (e.g., July or August). All other agronomy regarding fertilization and water management for each experiment was summarized in a dataset, which is publicly obtainable via this link: https://figshare.com/articles/dataset/AGB_of_managed_grasslands_in_China/19641654 (accessed on 1 July 2022).

We obtained a set of global and/or national layers of environmental covariates, including edaphic, climatic, and topographic variables (

Table A1. Environmental covariates.

Covariates	Code	Description	Unit
Edaphic variables	CFRAG	Coarse fragments (>2 mm)	%
	BULK	Bulk density	g cm−3
	ORGC	Organic carbon	g kg−1
	SDTO	Sand content	%
	CLPC	Clay content	%
	STPC	Silt content	%
	TAWC	Available water capacity	cm m−1
	TOTN	Total nitrogen	g kg−1
	CNrt	C:N ratio	-
	PHAQ	pH measured in H2O	-
Bioclimatic variables	T1	Annual mean temperature	°C
	T2	Mean diurnal range	°C
	T3	Isothermality (T2/T7×100)	%
	T4	Temperature seasonality (standard deviation×100)	°C
	T5	Max temperature of warmest month	°C
	T6	Min temperature of coldest month	°C
	T7	Temperature annual range (T5–T6)	°C
	T8	Mean temperature of wettest quarter	°C
	T9	Mean temperature of direst quarter	°C
	T10	Mean temperature of warmest quarter	°C
	T11	Mean temperature of coldest quarter	°C
	P1	Annual precipitation	mm
	P2	Precipitation of wettest month	mm
	P3	Precipitation of driest month	mm
	P4	Precipitation seasonality (coefficient of variation)	%
	P5	Precipitation of wettest quarter	mm
	P6	Precipitation of driest quarter	mm
	P7	Precipitation of warmest quarter	mm
	P8	Precipitation of coldest quarter	mm
Topographic variables	Elevation	Elevation	m
	Roughness	Roughness	-
	TRI	Terrain Ruggedness Index	-
	TPI	Topographic Position Index	-
	VRM	Vector Ruggedness Measure	-
	Aspectcosine	Aspect Cosine	-
	Aspectsine	Aspect Sine	-
	Slope	Slope	-
	Eastness	Index from −1 to 1 of how east or west a site faces	-
	Northness	Index from −1 to 1 of how north a site faces	-
	PCURV	Profile curvature	°
	TCURV	Tangential curvature	°
	dx	First order partial derivative (E-W slope)	-
	dy	First order partial derivative (N-S slope)	-
	dxx	Second order partial derivative (E-W slope)	-
	dyy	Second order partial derivative (N-S slope)	-

), as potential predictors of managed grassland AGB. These variables have widely been used to assess the possible regulation effects of environment on changes in grassland aboveground biomass [20] and terrestrial carbon cycles [21]. Here, 10 soil physical and chemical properties (Table A1) were obtained from the ISRIC-WISE soil profile database [22] with a spatial resolution of 1 km². We also obtained 16 topographic attributes with the same resolution as the WISE database from Amatulli, et al. [23]. In addition, we determined 19 bioclimatic attributes (T1–T11 and P1–P8; Table A1) quantifying biologically meaningful variables using monthly maximum and minimum temperature and precipitation [24] at each location in the observation year of the experiment. Specifically, for the observation year, we first extracted the monthly maximum and minimum temperature and precipitation from Peng, et al. [25] using the location information of each measurement. Then, the 19 bioclimatic variables (Table A1) were calculated using the biovars function in the R package dismo. Global gridded rasters of these 19 bioclimatic attributes (representing the period of 1980–2000) with a spatial resolution of 1 km² were derived from WorldClim [24]. More details of these global spatial covariate layers are described in Table A1. The collated AGB measurements and their associated predictors are documented and publicly available from: https://figshare.com/articles/dataset/AGB_of_managed_grasslands_in_China/19641654 (accessed on 1 July 2022). The spatial distribution of grasslands was obtained from the National Land Cover DataSets (NLCD) of China developed from Landsat TM digital images [26].

2.2. Drivers of Managed Grassland AGB

We first produced boxplots characterizing the mean, median, and interquartile range of AGB among different groups of environmental and management attributes. Then, we used linear mixed-effects regression (LMER) to examine the relationship of a suite of predicting variables with AGB. We assumed that AGB is potentially associated with soil properties, climatic factors, topographic conditions (Table A1), plant species, and management practices. In fitting the LMER, plant species was treated as a random effect, i.e., the coefficients of other predictor variables were modified by species, and all numerical variables were standardized to unit variance; therefore, the absolute magnitude of the coefficients for the predictor variables reflected their relative importance [27]. A principal component analysis (PCA) was applied to eliminate potential correlations in the 10 edaphic variables, 19 climatic variables, and 16 topographic variables (Table A1). The most important principal components (PCs) with variances greater than 1 were retained for the regression [28]. PCA and LMER were performed using procomp in the R package stats and lmer in the R package arm, respectively, in R 4.0.3 [29].

Furthermore, we performed a machine learning-based regression (i.e., a random forest model) to assess the drivers of AGB. Before fitting the regression, the variance inflation factor (VIF) was calculated and used to minimize the multicollinearity of environmental covariates. Specifically, the variables with a VIF value larger than 10 were excluded from further regressions. Treating the remaining covariates together with species, fertilization, and irrigation regimes as independent variables, and AGB as a dependent variable, we then fitted the machine learning-based model, which inherently quantifies the variable importance of predictors using importance scores for each predictor in the regression.

3. Results

3.1. AGB in China’s Managed Grasslands

In total, 16 dominant plant species were identified in these studies (Figure A1). In general, the top four species with the highest frequencies identified among the 107 studies, including Medicago sativa, Elymus dahuricus, and Bromus japonicus, were generally more widespread than the remaining species (i.e., Poa pratensis, Leymus chinensis, Elymus sibiricus, Agropyron cristatum, Lolium perenne, Onobrychis viciifolia, Dactylis glomerata, Festuca ovina, Trifolium repens, Astragalus adsurgens, and Phleum pratense; Figure A1 and Figure A2).

The data synthesis suggested that, by averaging across the 101 sites in China (Figure 1), AGB in managed grasslands was estimated to be 630 g m⁻² of dry matter (ranging from 55 to 2172 g m⁻², lower and upper limit of 95% confidence interval). Among the four most widely distributed species (Figure A1), Bromus japonicus had the highest average AGB (856 g m⁻²), followed by Medicago sativa (692 g m⁻²) and Elymus dahuricus (530 g m⁻², Figure 2). A large variability existed in the observed AGB (Figure 2). For example, the uncertainty (expressed as CV, i.e., standard deviation divided by mean) of observed AGB for Medicago sativa was 96% (data not shown).

Regardless of other factors, such as species and water management, AGB under fertilization (e.g., nitrogen, phosphorus, and potassium) was on average 50% higher than that under zero fertilization (Figure 3a). Similarly, the adoption of irrigation enhanced the average AGB by approximately 100% compared with that under rainfed conditions (Figure 3b). When pooling all data together, we found that AGB was generally higher in regions with higher mean temperatures during the plant growing seasons (i.e., April–October; Figure A3a), while the correlation between AGB and P_G (accumulated precipitation during the growing season) was much weaker (Figure A3b). By excluding the impacts of fertilization and irrigation, we found that the AGB of Medicago sativa, for example, was generally higher in regions with a warmer and wetter growing season (Figure 3c,d).

3.2. Comprehensive Assessment of Drivers on Managed Grassland AGB

We further introduced three sets of environmental factors, i.e., soil, climate, and topography. Principal component analysis (PCA) suggested that the first two soil principal components (PCs), four climate PCs, and four topography PCs could explain 71% (Figure 4a), 94% (Figure 4c), and 77% (Figure 4e) of the variances in the 10 soil attributes, 19 climate variables, and 16 topography properties (Table A1), respectively. For the first two PCs of soil, the most important contributing variables were ORGC (organic carbon) and CLPC (clay content) (Figure 4b and Figure A4a). For the first four PCs of climate, the most important contributing attributes were P1 (annual precipitation), T10 (mean temperature of warmest quarter), T3 (isothermality), and P4 (precipitation seasonality) (Figure 4d and Figure A4b). For the first four PCs of topography, the most important contributing variables were TRI (topographic position index), Northness, PCURV (profile curvature), and Aspectsine (aspect sine) (Figure 4f and Figure A4c).

Using these PCs as predictors, together with species, which were treated as random effects, and nutrient fertilization attributes as co-predictors, a linear mixed-effects regression (LMER) was fitted to the observed AGB (Figure 5). On average, the fitted LMER explained 68% of the variances in AGB (Figure 5). In general, AGB was significantly and positively correlated with nitrogen fertilization (Figure 5). AGB was also positively and significantly correlated with the first two PCs of climate variables (Figure 5). Thus, annual precipitation (P1, the most contributing variable of PC1 of climate, Figure A4b) and mean temperature of the warmest quarter (T10, the most contributing variable of PC2 of climate, Figure A4b) were generally positively correlated with AGB. The first PC of soil was generally negatively correlated with AGB (Figure 5). Variations in AGB were also regulated by topographic factors, and AGB was significantly correlated with the first three PCs of topography (Figure 5).

The fitted machine learning-based model (i.e., random forest) indicated that 57% of the variances in AGB can be explained by the species, management of fertilization and irrigation, and the environmental attributes selected in Section 2.2 (Figure 6a). As indicated by the random forest model, three climatic variables (i.e., T8, P2, and T4) are the top three most important regulators of AGB (Figure 6b).

4. Discussion

Managed grasslands in China are mainly distributed in the northern and western regions, from Inner Mongolia to the Qinghai–Tibetan Plateau (Figure 1). These regions are generally characterized by a relatively cold and/or dry climate due to high altitudes and/or more northern latitudes [30,31]. As such, both temperature and precipitation were found to be positively correlated with grassland AGB (Figure 3c,d). As expected, fertilization significantly increased AGB compared with zero fertilization (Figure 3a), which is consistent with existing findings that addition of nutrients (e.g., nitrogen, phosphorus, and potassium) can enhance the nutritional quality of plant tissues, thereby promoting grassland AGB [32,33,34]. This is because in the world’s most terrestrial ecosystem [35,36], including China’s grasslands [37,38], plant productivity is widely acknowledged to be nutrient-limited. Apart from nutrients, water availability is another strong constraint on plant productivity in global terrestrial ecosystems, particularly in grasslands [39,40]. This can help to explain the general promoting effect of water irrigation on AGB (Figure 3b). Moreover, in a system with sufficient water supply, the impact of precipitation on AGB can be eliminated [41], which underpins our results that, when irrigation was involved, precipitation seemed to have limited influence on AGB (Figure A3b). In addition, our results demonstrated the coregulating effects of edaphic and topographic attributes on AGB (Figure 5 and Figure 6b), which are generally comparable with the findings in the literature [42,43,44].

Regardless of plant species, AGB in managed grasslands averaged approximately 630 g m⁻² of dry matter, ranging from 6 to 7 times China’s national natural grassland AGB [45,46,47,48]. Although grasslands account for approximately 40% of China’s land [4,49], only around 3% of the grassland area is under managed conditions (i.e., managed grasslands), which is substantially lower than those in developed regions such as Europe, Australia and New Zealand [11]. In addition, improving management practices and species varieties can help to enhance AGB. It has been reported that Medicago sativa is one of the world’s most popular species due to its high forage productivity, high nutritional quality, and wide adaptability to different climatic and edaphic conditions [50]. Our results indicate that, on average, the AGB of Medicago sativa can reach approximately 700 g m⁻² (Figure 2), i.e., ~7 Mg ha⁻¹, which is only half the production of that in developed countries, such as the USA [51]. This could be attributed to the fact that most managed grassland experiments in China (Appendix B) were conducted on existing degraded lands with relatively poorer soil nutrient condition.

Despite the positive effect on AGB, management of grasslands at large scales should be undertaken with caution due to the possible negative consequences on other ecosystem functionalities. For example, growing evidence has suggested that nutrient enrichment due to fertilization can lead to widespread decreases in grassland biodiversity [52], which is deemed another key characteristic of grassland degradation [53]. Moreover, conversion from natural lands to managed grasslands can possibly result in significant losses of ecosystem carbon stock. In natural grasslands, root biomass, rather than AGB, constitutes the majority of the total plant biomass carbon stocks [45,47]. Furthermore, the largest carbon reservoir of grassland is soil, containing more than 90% of the carbon in the whole grassland system [54]. A number of studies have indicated that conversion from natural grasslands to managed grasslands would not only reduce root biomass [55,56,57] but also substantially decrease the soil carbon pool [58,59], thereby causing a net warming effect on climate. In contrast, in regions that have already suffered from degradation, the establishment of managed grasslands can significantly enhance both root biomass productivity and soil carbon content [60,61]. On this basis, we suggest that priority of managed grassland establishment should be given to grasslands that have been severely degraded to benefit not only forage production but also ecosystem restoration. In addition, since managed grassland is in general established in areas with soil degradation possibly caused by water limitation, we highlight a need for introducing the plant species with high drought tolerance and/or deep rooting capacity in the future.

5. Conclusions

We comprehensively and quantitatively assessed the aboveground biomass (AGB) of dominant plant species in China’s managed grasslands. We found that the establishment of managed grasslands via practices such as introducing advanced species, fertilization, and irrigation can potentially increase aboveground biomass. The magnitude of enhanced AGB through establishing managed grasslands is associated with a series of biotic and abiotic factors, such as species and climatic, edaphic, and topographic attributes. We highlight the need for a more extensive establishment of managed grasslands, particularly in areas suffering degradation of natural grasslands, thereby favoring not only livestock production, but also grassland restoration. Apart from improving management practices, introducing plant species with drought tolerance and/or deep rooting capacity may also contribute to productivity enhancement and grassland restoration.