Contribution of Climate Change and Grazing on Carbon Dynamics in Central Asian Pasturelands

Abstract

Reducing the uncertainties in carbon balance assessment is essential for better pastureland management in arid areas. Climate forcing data are some of the major uncertainty sources. In this study, a modified Biome-BGC grazing model was driven by an ensemble of reanalysis data of the Climate Forecast System Reanalysis data (CFSR), the European Centre for Medium-Range Weather Forecasts Interim Reanalysis (ERA-Interim), and the Modern-Era Retrospective Analysis for Research and Applications (MERRA), to study the effect of climate change and grazing on the net ecosystem exchange (NEE) of the pasturelands in Central Asia. Afterwards, we evaluated the performance of corresponding climate datasets over four major pastureland types, and quantified the modeling uncertainty induced by climate forcing data. Our results suggest that (1) a significant positive trend in temperature and a negative trend in precipitation were obtained from the three climate datasets. The average precipitation is apparently higher in the CFSR and MERRA data, showing the highest temperature value among the data sets; (2) pasturelands in Central Asia released 2.10 ± 1.60 Pg C in the past 36 years. The highest values were obtained with the CFSR (−1.53 Pg C) and the lowest with the MERRA (−2.35 Pg C) data set; (3) without grazing effects, pasturelands in Central Asia assimilated 0.13 ± 0.06 Pg C from 1981–2014. Grazing activities dominated carbon release (100%), whereas climate changes dominated carbon assimilation (offset 6.22% of all the carbon release). This study offered possible implications to the policy makers and local herdsmen of sustainable management of pastureland and the adaptation of climate change in Central Asia.

1. Introduction

Occupying about 47% of the global land area, the dryland ecosystem is an essential component to the global carbon cycle [1]. However, most of the arid and semiarid regions are facing serious ecosystem degradation that was caused by climate change, as well as human interferences [2]. Carbon dynamics in the dryland ecosystem are sensitive to climatic variability and human activity, yet considerably less is known about the extent of the influence. Pasturelands in Central Asia were degraded as a consequence of a rapid expansion in livestock numbers and economic reforms initiated in the 1980s [3].

Biogeochemical models constrained by high-quality, gridded atmospheric variables are essential for simulating the response of carbon and water dynamics to environmental changes at a regional scale [4,5]. Reducing simulation uncertainties is becoming increasingly significant to further explore the impacts of climate change on terrestrial carbon cycles [6]. The simulation quality depends on the forcing data accuracy and model structures and parameters [7]. Comparing the uncertainties resulting from model schemes and parameterizations, uncertainties caused by the meteorology-driven data need to be further investigated, especially in arid and semi-arid regions, which can deepen our understanding of climate and ecosystem interactions [8,9] and is helpful in predicting the regional response to environmental change for the purpose of sustainable land use planning.

A number of atmospheric reanalysis datasets with high tempo-spatial resolutions and accuracy were well developed in the past decades, which enable their applications in forcing biogeochemical models. Driven by an ensemble of reanalysis data, ecological models can greatly reduce the simulation uncertainties.

Therefore, in this study, we chose the major pastureland types in Central Asia as the study area, and evaluated the modelling interannual Net Ecosystem Exchange (NEE) variability driven by three climate datasets. We aim to highlight the importance of an accurate climate dataset for terrestrial carbon cycle simulations, and especially for the assessment studies of climate and environmental change involving Central Asia. Our study wishes to offer possible references of grazing management and climate change adaptation to the policy makers as well as local herdsmen of Central Asia.

2. Materials and Methods

2.1. Study Area

The regions of Central Asia, including Kazakhstan, Turkmenistan, Uzbekistan, Tajikistan, Kyrgyzstan, and Xinjiang province in China, extend from 34.3° to 55.4° N latitude and from 46.5° to 96.4° E longitude, encompassing a land area of approximately 5.7 million km². The overall population of Central Asia was approximately 101 million by 2020, which equaled to about 18 people per km², according to the statistics provide by the World Bank (https://data.worldbank.org/indicator/SP.POP.TOTL; accessed 12 February 2022) and China Statistical Yearbook 2021 (http://www.stats.gov.cn/tjsj/ndsj/2021/indexch.htm; accessed 12 February 2022). The climate is arid and semiarid with the annual precipitation less than 300 mm in more than 70% of the region. Coexisting with native and domestic grazers for thousands of years, Central Asian rangelands are the largest contiguous grazed land on this planet. However, they have been heavily degraded [10,11] and subjected to climate changes over the past decades [12,13]. According to the distribution elevation and climate conditions, we divided the pasturelands of Central Asia into three typical types: forest meadow (FM, >1650 m a.s.l., arid index > 0.7), temperate grasslands (TG, 650–1650 m a.s.l., arid index ≈ 0.3), and dry grasslands (DG, <650 m a.s.l., arid index ≈ 0.1) [14] (Figure 1).

2.2. Model Description

In this study, we used the Biome-BGC grazing model which was developed in our previous studies to analyze the NEE dynamics of pasturelands of Central Asia under grazing effects and multi climate dataset. This model was developed to study the hydrological processes and carbon and nitrogen dynamics for grassland ecosystems by incorporating the grazing process into the original Biome-BGC model [15,16]. In the new model, gross primary productivity (GPP) followed the Farquhar photosynthesis function [17], considering both shaded and sunlit foliage. Respiration processes, including autotrophic growth (g_r) and maintenance (m_r) respiration, and soil heterotrophic (h_r) respiration, were mainly regulated by temperature and water availability. Grazing loss of plant biomass carbon (D_r, g C ha⁻¹ d⁻¹) is calculated according to grazing efficiency (G_e, ha/d/sheep unit) and grazing intensity (GI, head/ha).

$$D_r=G_e\times GI\times \left(C_{leaf}-{\left(C_{leaf}\right)}_r\right)\left(0<D_r<D_x\times G_i\right)$$

(1)

where $C_{leaf}$ is the leaf biomass, carbon (g C m⁻²), ${\left(C_{leaf}\right)}_r$ is the residual leaf carbon, which is unavailable to be grazed (g C m⁻²), and $D_x$ is the livestock’s maximum ingestion rate (g C d⁻¹ head⁻¹) [18]. In the model, we further divided $D_r$ into five components: the respiration of livestock, methane release from livestock, carbon in meat production, carbon transfer from feces, and urine into soil.

As a result, NEE is calculated as the difference between GPP and ecosystem total respiration and grazing consumption.

$$NEE=GPP-m_r-g_r-h_r-D_r$$

(2)

where $m_r$ is maintenance respiration, $g_r$ is autotrophic growth respiration, $h_r$. is soil heterotrophic respiration, and $D_r$ is grazing loss of plant biomass carbon.

In our previous studies, the model has been systematically parameterized and calibrated according to field observations [14,15,16,19]. Specifically, the simulated net primary productivity (NPP), both under grazing and protected conditions, were validated against experiment data, and the modeling NEE, as well as water cycle, were validated using eddy covariance flux measurements. In this study, a negative NEE value indicates a carbon loss from the ecosystem.

2.3. Meteorological Data

In this study, we chose three popular and well-performed, long-term gridded reanalysis datasets to force the Biome-BGC grazing model (

Table 1. Forcing data of the Biome-BGC grazing model.

Data Sets	Spatial Resolution/Time Steps
Climate data
--Climate Forecast System Reanalysis Data (CFSR)	0.31 × 0.31/daily (1981–2014)
--European Centre for Medium-Range Weather Forecasts Interim Reanalysis (Era-Interim)	0.75 × 0.75/daily (1981–2014)
--Modern-Era Retrospective Analysis for Research and Applications (MERRA)	0.5 × 0.5/daily (1981–2014)
Soil property --texture, bulk density, soil depth	1 km × 1 km
Atmospheric CO2	annual
Topographical data --elevation, slope, and aspect	30 m × 30 m
Grazing intensity --Kazakhstan & Uzbekistan --The rest of Central Asia
	State Country	/1981–2014

). Reanalysis data used data assimilation methods by merging meteorological observations and climate model forecasts to improve the accuracy and reliability of climate description. (1) Provided by The European Centre for Medium-Range Weather Forecasts (ECMWF), the European Centre for Medium-Range Weather Forecasts Interim Reanalysis (ERA-Interim) is a global atmospheric reanalysis product with a spatial resolution of about 80 km, which updates about one month later from the real time [20]. (2) Modern-Era Retrospective Analysis for Research and Applications (MERRA) is produced by using the Goddard Earth Observing System Data Assimilation System Version 5 (GEOS-5), which assimilates satellite and ground observations [21]. (3) The Climate Forecast System Reanalysis Data (CFSR) is a new-released reanalysis dataset with a relatively high tempo-spatial resolution, and has been widely used in relating studies [22].

The three datasets have been evaluated against the site meteorological observations in Central Asia [23,24,25], and were proved to be effective in representing the climate changes of this region. All the datasets were merged into daily time-step for the demand of the Biome-BGC grazing model.

2.4. Other Forcing Data

Soil textures (i.e., sand/silt/clay ratio, bulk density, and soil depth) were obtained from the Harmonized World Soil Database (HWSD) (FAO/IIASA/ISRIC/ISS-CAS/JRC, 2012). Atmospheric CO₂ concentrations were derived from observations taken at Mauna Loa (https://gml.noaa.gov/webdata/ccgg/trends/co2/co2_annmean_mlo.txt; accessed on 25 April 2017). Topographical information, including elevation, slope, and aspects, was derived from the ASTER with the resolution of 30 m × 30 m (the Advanced Spaceborne Thermal Emission and Reflection Radiometer) Global Digital Elevation Model Version 2 dataset (ASTER GDEM, v2) (http://gdem.caersdac.jspacesystems.or.jp/; accessed on 17 May 2017). Grazing intensity data were obtained from the FAO’s Animal Production and Health Division [26], and the detailed data processing can be found in Han et al., 2016.

3. Results

3.1. Comparison of the Climate Data Sets

During the study period, the three datasets agreed that the grassland in Central Asia experienced a significant warming trend (0.03 ± 0.01 °C year⁻¹, R² = 0.12) and a drying trend (2.18 ± 2.04 mm year⁻¹, R² = 0.21) (Figure 2). The three datasets present a significant consistency in the mean air temperature annual variability with each other (

Table 2. The linear correlation coefficients of annual total precipitation (Prec, the lower triangle matrix) and annual mean air temperature (Temp, the upper triangle matrix) between datasets. *: p < 0.05; **: p < 0.01.

	Temp	CFSR	ERA-Interim	MERRA
Prec
CFSR			0.94 **	0.93 **
ERA-Interim		0.21		0.98 **
MERRA		0.43 *	0.84 **

), but the long-term mean (1981–2014) value shows a disparity with the MERRA, being approximately 1.3 °C warmer than the other two datasets. Precipitation cohmparison shows that the CFSR differed from the other two datasets, both in annual values and temporal pattern (Table 2 and Figure 2). Its long-term average is 118 mm and 167 mm higher than the ERA-Interim and MERRA, respectively, and the decreasing rate was most dramatic (−4.06 mm year⁻¹), at over four times faster than that in the ERA-Interim and MERRA.

Figure 3 showed overall agreements in the spatial pattern of long-term climate average conditions. However, cross datasets comparison showed that the precipitation in temperate grassland was most abundant in the CFSR, while least in the MERRA. The temperature and precipitation in dry grassland showed little variation among the three datasets. The three datasets have notable differences in representing the temperature and precipitation over forest meadow, which is mainly located on the Tianshan Mountains.

3.2. Spatio-Temporal Variations in NEE

Simulated annual carbon fluxes over the pasturelands from 1981 to 2014, from three versions of reanalysis data, are shown in Figure 4. The model simulated an annual total terrestrial NEE of −0.06 ± 0.04 Pg C year⁻¹, which means that pasturelands in Central Asia released 2.10 ± 1.60 Pg C in the past 36 years. We compared the average state (1981–2014) of the terrestrial carbon cycle among climate datasets (Figure 5), and the NEE differed among the simulations. The highest values were obtained with the CFSR (−15.42 g C m⁻² year⁻¹) and the lowest with the MERRA (−23.69 g C m⁻² year⁻¹) dataset.

In response to climate change and grazing effects, regional NEE estimated by the three reanalysis datasets increased 0.43 g C m⁻² year⁻¹ (R² = 0.30) in the CFSR, 0.40 g C m⁻² year⁻¹ (R² = 0.20) in ERA-Interim, and 0.54 g C m⁻² year⁻¹ (R² = 0.35) for the MERRA, respectively. Precipitation is more significantly related with NEE interannual variability than temperature at the regional scale (Figure 4).

Figure 5 and Figure 6 show the spatial distribution of annual mean values of the NEE in DG, TG, and FM. The three datasets all agreed that DG, TG, and FM were C sources during the study period, as well as the strongest carbon loss in TG (−35.47 ± 7.75 g C m⁻² year⁻¹). The simulation results under the MERRA showed the largest carbon release in all the three grassland types, while the CFSR drove the weakest carbon source. In addition, the NEE from the three versions of the reanalysis data differed significantly. In DG and TG regions, CFSR data performed differently with the other two, by more than 35% on average. In contrast, there was a smaller difference in FM from the CFSR and ERA-Interim (

Table 3. Averaged NEE during 1981–2014 under simulations simulated by three sets of reanalysis data. DG, Dry Grassland; TG, Temperate Grassland; FM, Forest Meadow.

PFT	NEE (g C m−2 Year−1)
	CFSR	ERA	MERRA	Average
DG	−9.28	−13.32	−14.21	−12.27
TG	−26.63	−38.64	−41.13	−35.47
FM	−10.34	−10.72	−15.73	−12.26
CA	−15.42	−20.89	−23.69	−20.00

).

3.3. Contribution of Climate Change and Grazing Effects

Projections without grazing effects result in higher NEE than projections with grazing effects. Compared with simulations with grazing effects, the NEE is projected to increase to 1.32 ± 0.61 g C m⁻² year⁻¹, which means pasturelands in Central Asia assimilated 0.13 ± 0.06 Pg C from 1981–2014. All grassland types have changed to carbon sinks except for MERRA simulation, for which the NEE is projected to be negative. Using CFSR data, the NEE is projected to increase from −15.42 ± 9.73 g C m⁻² year⁻¹ under the grazing scenario, to 2.29 ± 1.18 g C m⁻² year⁻¹ without grazing. Using ERA data, NEE is projected to substantially increase by 22.72 ± 16.18 g C m⁻² year⁻¹ after eliminating the grazing impact.

We assessed the relative contribution of climate change and grazing on the NEE for different pastureland types (Figure 7). For the average situation in Central Asia, 100% of carbon release was caused by grazing activities, 6.22% of which was offset by climate change (Figure 7d). Effects of grazing were more profound on TG than on FM and DG (Figure 7a–c). In the TG region, grazing released 35.47 ± 7.76 g C m⁻² year⁻¹ and climate change only offset 4.34%. This situation improves in the DG region, where grazing released 12.27 ± 2.63 g C m⁻² year⁻¹ and climate change increased 12.50% of carbon fixation. As a whole, grazing was the dominant factor for carbon release in Central Asia.

4. Discussion

4.1. Uncertainty in Carbon Dynamics in Central Asia

The Biome-BGC model has been used to assess the responses of carbon dynamics to climate change in Central Asia [14] and China [19]. However, only one climate dataset was used in these studies. The current paper presents the application of multiple input datasets to analyze the carbon dynamics of Central Asian pasturelands, giving uncertainty in modeling.

Modeling can be an effective tool to address experiments’ limitations and to depict the response mechanism of plants to environments; however, researchers demonstrate the need to systematically investigate the uncertainty resulted from modeling. One of the main uncertainties came from forcing datasets themselves, such as from meteorology [6,27,28]. Using multiple data sources can give the magnitude and range of carbon flux, which can ultimately help to reduce the errors in future modeling.

The model simulated that pasturelands in Central Asia released 2.10 ± 1.60 Pg C, and the NEE range from three input climate datasets was −1.53~−2.35 Pg C in the past 34 years. Our results also show that the annual NEE of Central Asia generally had an increasing trend from 1981–2014, which is in line with some previous studies [13,14,29,30]. Our results indicate that pasturelands in Central Asia play an important role in the global carbon cycle. However, a weak carbon source was observed in this region, which can be attributed to two reasons. Firstly, high grazing intensity had been sustained especially in Kazakhstan [31,32,33], although the situation was relieved since the disintegration of the Soviet Union in 1991–1992. Secondly, a significant positive trend in temperature (0.03 ± 0.01 °C year⁻¹) and a negative trend in precipitation (2.18 ± 2.04 mm year⁻¹) may lower the gross primary productivity and enhance the respiration [34,35,36,37].

4.2. Contribution of Different Factors to NEE

An accurate assessment of carbon dynamics-controlling factors was vital for preventing grassland degradation [38]. Many studies have been conducted to assess the relative roles of climate change and human activities at regional and global scales [6,38,39,40]. In this research, 100% of the carbon release in Central Asia was caused by grazing, and climate change offset 6.22% on average. Some researchers have already found that grazing was the main cause for desertification [11,33]; however, in this study, the range and extent of the influence were given.

The role of climate change and grazing in Central Asian carbon dynamics is varied among different eco-regions. Effects of grazing were more evident on TG, which can be attributed to its relatively high temperature and low precipitation (Figure 3), and the resulting ecological vulnerability.

4.3. Socio-Economic Impacts and Future Implications

Central Asia has great carbon sequestration potential, but it is still among the most poorly known area in global carbon cycle [1]. Central Asian countries experienced one of the most widespread land use changes in the 20th century due to the collapse of the former Soviet Union in early 1990s. Large tracts of arable land were abandoned to grassland or left unutilized, which led to remarkable carbon sinks of 70 g C m^–2 year^–1 in Ukraine, Belarus, and European Russia [40], 240 g C m^–2 year^–1 in Kazakhstan [41], and 95 g C m^–2 year^–1 in Russia [42]. Nevertheless, the impacts of land use change are still highly uncertain, because due to the shortage of refined ground observational data, most modeling studies had to rely on sensitivity analysis rather than solid validation [43], and long-term, spatially-explicit land use data that could reflect the large scale land dynamics were also highly limited [44]. Recently, using integrated frameworks were proven to be effective choices to solve complex interactions between ecological and economic processes, especially for those lacking parameter values [45,46,47]. These studies generally combined field measurements, mathematical cooperative models, and geographic information system models. This study attempted to use climate forcing data ensembles to minimize the simulation uncertainties; however, the uncertainties in other input data, as well as in model structure and parameters, deserve no less attention.

Central Asia is projected to suffer from above-average warming and drought occurrences in coming decades [48]. Meanwhile, the demands for socio-economic developments, such as urban expansion, mining exploitation, and transport infrastructure, are rapidly growing, which, together with future climate change, will inevitably affect the ecosystem functioning, land productivities, and production choices. These issues deserve high priority, because Central Asia is vulnerable to climatic and anthropogenic disturbances [1], and is already facing poverty and environment and ecosystem degradation. Unfortunately, in the absence of grazing and other specific land use information under future socio-economic scenarios, our present work could not thoroughly solve these issues.

5. Conclusions

This work quantified the temporal and spatial effects and the relative impacts of recent climate change and grazing activities on the NEE for each country or region (Xinjiang), by using the Biome-BGC grazing model forced by three climate datasets (CFSR, ERA-Interim, and MERRA). During 1981–2014, Central Asia experienced a significant warming trend (0.03 ± 0.01 °C year⁻¹) and a drying trend (−2.18 ± 2.04 mm year⁻¹), and the average annual precipitation in the CFSR was 45.14% and 80.47%, higher than that in the ERA-Interim and the MERRA, respectively. The simulated annual total NEE is −0.06 ± 0.04 Pg C year⁻¹, which equals to a total carbon source of −2.10 ± 1.60 Pg C in the pasturelands of Central Asia in 1981–2014. The NEE varied among the three simulations; the highest NEE was obtained with the CFSR (−15.42 g C m⁻² year⁻¹) and the lowest with the MERRA (−23.69 g C m⁻² year⁻¹). In addition, we found that climate changes dominated carbon assimilation, while grazing resulted in 100% of carbon release, and its relative contribution differed among grassland types, which indicates that effective and regional-specific grazing controls should be implemented. For the long run, this study could serve as a possible reference for decision makers, social organizations, and individuals, in light of regional sustainable development and the international CO₂ trading. A more systemic, interdisciplinary approach to evaluate the carbon budget of Central Asia is promising for future studies.