Spatiotemporal Variations of Chinese Terrestrial Ecosystems in Response to Land Use and Future Climate Change

Abstract

Terrestrial ecosystems in China are threatened by land use and future climate change. Understanding the effects of these changes on vegetation and the climate-vegetation interactions is critical for vegetation preservation and mitigation. However, land-use impacts on vegetation are neglected in terrestrial ecosystems exploration, and a deep understanding of land-use impacts on vegetation dynamics is lacking. Additionally, few studies have examined the contribution of vegetation succession to changes in vegetation dynamics. To fill the above gaps in the field, the spatiotemporal distribution of terrestrial ecosystems under the current land use and climate baseline (1970–2000) was examined in this study using the Comprehensive Sequential Classification System (CSCS) model. Moreover, the spatiotemporal variations of ecosystems and their succession under future climate scenarios (the 2030s–2080s) were quantitatively projected and compared. The results demonstrated that under the current situation, vegetation without human disturbance was mainly distributed in high elevation regions and less than 10% of the national area. For future vegetation dynamics, more than 58% of tundra and alpine steppe would shrink. Semidesert would respond to climate change with an expansion of 39.49 × 10⁴ km², including the succession of the steppe to semidesert. Although some advancement of the temperate forest at the expense of substantial dieback of tundra and alpine steppe is expected to occur, this century would witness a considerable shrinkage of them, especially in RCP8.5, at approximately 55.06 × 10⁴ km². Overall, a warmer and wetter climate would be conducive to the occurrence and development of the CSCS ecosystems. These results offer new insights on the potential ecosystem response to land use and climate change over the Chinese domain, and on creating targeted policies for effective adaptation to these changes and implementation of ecosystem protection measures.

1. Introduction

Climate change and terrestrial ecosystem interactions at the local scale affect the nature, ecology, magnitude, distribution, and formation of biomes [1]. Climate variables, specifically temperature and precipitation, have been recognized as a direct determinant of large-scale ecosystem development and ranges [2,3,4,5,6,7]. Meanwhile, biogeophysical and biogeochemical processes such as lower surface albedo and increased absorbed solar radiation can produce feedback of changes in vegetation distribution and growth on the climate, exacerbating the original warming [3,6,8,9]. However, the mechanism of the response of vegetation to climate change is complex and not necessarily deterministic, which similar climates could support different vegetation ecosystems [10,11,12,13]. Steep gradients in topography and soil with similar climates, all of which influence the heterogeneity of biome patterns [14]. The evolutionary history and multi-stability of various ecosystems further contribute to this complexity [15]. Therefore, knowledge of climate change-vegetation pattern interactions is particularly important in highly heterogeneous environments such as China. The pronounced steep gradients in climate and topography of China have influenced the formation and development of biome patterns, which contribute to diversity and richness in vegetation types, which provide different ecosystem services and functions [13,16]. A forest ecosystem characterized by dense tree cover and competition for light primarily is conducive to the dynamics and structural complexity of the vegetation [17]. The grass biomass of savannas provides a range of ecosystem goods and services to people and high densities of large grazers and supplies fuel for frequent fires [17]. Therefore, understanding the terrestrial ecosystem richness and climate change-vegetation interactions of China are urgent needs for ecologists. However, a general assessment of the current patterns of the rich terrestrial ecosystems and a projection of their dynamics in response to climate change is still not sufficient.

The great diversity of terrestrial ecosystems across the China domain is affected by the synergistic effects of both climate change and anthropogenic causes, according to observational evidence [18,19]. Among climate factors, many simulated results have indicated that both rapidly rising temperatures and the precipitation events of the past decades have significantly influenced the occurrence, distribution, growth, and production potential of vegetation [2]. In the future, the mean annual temperature is projected to rise by 2–5 °C across the globe at the end of this century [2], while the mean annual precipitation would continue to increase over this century by 3–8%. Wang et al., for example, speculated that the Guangxi and Yunnan regions of southern China may face increased rainfall shortages in the future [20]. Furthermore, by 2100, China is expected to lose half of its original forests and half of its ecosystems [21]. These effects restrained the preadaptation of terrestrial ecosystems to new habitats and caused a decrease in resilience [13,22,23,24], especially for some climate-sensitive and climate-vulnerable vegetation. Among other synergistic factors, the increasing pressure of current land use also has severe consequences for the provision of ecosystem services and vegetation dynamics [25]. The practice of slash and burn, especially cutting and burning forests, also known as shifting cultivation, degrades ecosystems, causes more carbon emissions in a specific region and inhibits time to recover, which leads to land transformation [26]. In addition, overgrazing has had significant implications on land transformation in recent years, which has posed a great threat to the terrestrial ecology in China [13,27]. Therefore, although future greenhouse gas (GHG) emissions have uncertainties, exploration of the resulting climate change and current human land use in China, how they significantly affect vegetation dynamics, and where they are distributed, are essential for understanding climate-vegetation interaction and climate adaptation. However, land-use impacts on vegetation are often neglected in the exploration of terrestrial ecosystems.

Potential natural vegetation (PNV) is a direct proxy for the predicted growth stage of mature vegetation in the absence of human intervention [28,29], which is regulated only by water factors and heat and can reflect the actual relationship between climate and vegetation. Some definitions of terrestrial ecosystems are vegetation ecosystems based on the concept of PNV. As a result, precise PNV estimates will boost confidence in projections for the potential effects of future climate change and existing land use impacts on ecosystems under various climate and management scenarios [30,31]. Several prognostic models [32] have emerged and developed in recent decades to assist with these studies for estimating PNV response to climate change across various spatial scales. PNV models, such as bio-geographical models (e.g., Holdridge Life Zone (HLZ) [33]) and equilibrium vegetation models (e.g., BIOME4) [34], provide a wealth of information on the understanding of terrestrial vegetation in relation to past and future climatic variation at global and regional scales [33,35,36,37]. The HLZ model, in particular, divides the system into 38 ecological units known as Life Zones, which are defined by bio-temperature (BT), average total annual precipitation (P), and potential evapotranspiration ratio (PER). However, it was not exempt from criticism that this classification scheme does not consider the seasonality in climatic parameters and the difference in the quantitative standard between horizontal and vertical zones [38]. The BIOME4 model simulates PNV distribution by relating PNV to the geographic distribution of climate parameters, assuming both vegetation and climate are at equilibrium [34]. So, model-simulated PNV varies significantly depending on the model structure as well as the quality and characteristics of the input (or baseline) climatology data used to run the models [39,40]. The development of dynamic global vegetation models (DGVMs) now integrates vegetation dynamics and biogeochemical processes, providing a unique opportunity to investigate the transient responses of terrestrial ecosystems to rapid climate change [35,40]. Significant uncertainties, however, remain in current estimates of ecological processes at regional to global scales, coupled with the complexity of input data required, which makes it challenging to be widely applied. The Comprehensive Sequential Classification System (CSCS) model synthesizes information on bioclimatic variables, edaphic conditions, and vegetation characteristic variables in relation to water and heat conditions in a specific environment [41,42]. On a global and regional scale, the model has become a critical tool for creating strategies to mitigate the consequences of climate change on ecosystems. It has been successfully utilized to model biomes since its continuous optimization and evolution in recent decades.

If GHG emissions, climate change, and related disturbances continue at or above current rates, Fischlin et al. [39] have predicted that the resilience and adaptability of many terrestrial ecosystems and biodiversity will face an unprecedented challenge by the mid-21st century and beyond from a combination of changes in climate and associated disturbances. The negative effects of future climate change on natural ecosystems are becoming more widely recognized, as is the significant contribution of monitoring and attributing spatiotemporal variations of terrestrial ecosystems in China in response to future climate change to quantitative evaluation of climate change-vegetation growth interactions. Fortunately, Tapiador et al. have reported projections of high-resolution GCMs for future climates under the assumptions of three Representative Concentration Pathways (RCP2.6, RCP4.5, RCP6.0, and RCP8.5), which can be used as a primary source for climate change and global warming studies [43], and which have been incorporated into the Intergovernmental Panel on Climate Change IPCC Fifth Assessment Report (AR5-CMIP5) [44]. By the end of the century, the four RCPs represent radiative forcing levels of 8.5, 6, 4.5, and 2.6 W/m², respectively [44,45]. RCPs provide a unique set of data for extensive climate modeling, global and regional analysis, and impact assessment due to their comprehensiveness in terms of sources addressed as well as geographic scale features [46].

Previous efforts to study PNV over China using this dataset, based on the CSCS model, have primarily focused on the implications of climate change on terrestrial ecosystems under a single future climate scenario or for a certain period [42] or a specific period [47]; few have tried to compare the PNV variations in China under different climate scenarios. Furthermore, a deep understanding of land use impacts on vegetation dynamics is lacking, and their succession and evolution are being ignored in the estimation of PNV across spatial and temporal scales. In addition, detailed comparisons between the CSCS model and the HLZ model on a national scale have only been implemented minimally. This situation creates a significant incentive to investigate the synergistic effects of climate change and land-use factors, as well as future intercomparison research, especially between different RCPs. To fill the gaps in this subject, it is necessary to evaluate potential changes in vegetation caused by current land use and future climate change under various RCPs.

Using the CSCS model, this study intends to quantify terrestrial vegetation over the China domain under the current baseline scenario (1970–2000) and three future climatic scenarios (the 2030s–2080s). Under the baseline scenario, PNV distribution and the effects of land use on their growth were assessed, as well as PNV variations under other climate scenarios were compared. First, a detailed comparison with the HLZ model and a vegetation map to validate the performance of the CSCS model were used. The vegetation distribution for the China domain is then simulated using a series of climatic data (average temperature, precipitation) layers under the current climate baseline (1970–2000). To examine human factor impacts on vegetation, DEM data from 1980 to 2000 were used to represent regional and temporal variability of pixels of unchanging land cover-type categories. In addition, PNV changes and compared the changes between future climatic scenarios and the existing climate baseline (1970–2000) were forecasted. The effects of succession and evolution on vegetation growth and dieback were investigated. This study provides a scientific foundation for a fair assessment and forecast of the effects of present land use on CSCS ecosystems across the China domain, as well as how they will adapt to climate change in the next decades.

2. Materials and Methods

2.1. Study Area

China is located in East Asia, on the western shore of the Pacific Ocean, with a land area of roughly 9.63 × 10⁶ km² (Figure 1a,b). China has a diverse climate due to its extraordinary topographical variations and complexity. From northwest to southeast, extra-arid zones, arid zones, semiarid zones, subhumid zones, humid zones, and perhumid zones exist. Different climate zones dominate different regions in China, and the diversity in climate environments has bred rich and diverse vegetation types. Frigid zones, cold temperate zones, cool temperate zones, warm temperate zones, warm zones, subtropical zones, and tropical zones dominate different regions in China [48,49]. Changes in climate indices, such as heat and humidity levels, that China may experience in the future would have a significant impact on vegetation growth in the context of global warming. Under various climate scenarios, the relationships between vegetation and climate change are inconsistent. More extreme scenarios would result in wider ecosystem responses [50], which provides potential value for our subsequent study of vegetation change in China under different climate scenarios.

2.2. Data and Processing

2.2.1. The Baseline Scenario Data

A comparison climatic scenario is often used as a reference for the scenarios and climate change forecasts is the 1970–2000 time period. The climatic dataset is from WorldClim 2.1, which is available from World Climate Data (

Table 1. Datasets used in this study.

Datasets	Time (year)	Temporal Resolution	Spatial Resolution	URL
The baseline scenario data	1970–2000	Monthly	30 arc-s (1 km2)	https://www.worldclim.org/ accessed on 1 June 2020
The future scenario data (RCP2.6, RCP4.5, and RCP8.5)	2030s (2020–2049) 2050s (2040–2069) 2070s (2060–2089) 2080s (2070–2099)	Monthly	30 arc-s (1 km2)	http://ccafs-climate.org/ accessed on 10 August 2020
The land use data (LU)	1980, 1990, 1995, 2000	Monthly	1 km2	http://www.resdc.cn/ accessed on 20 July 2020
SRTM-DEM data	-	-	30 arc-s (1 km2)	https://www.worldclim.org/ accessed on 25 June 2020
A vegetation map in China at a scale of 1:1,000,000			1 km2	http://www.resdc.cn/ accessed on 11 May 2020

). This gridded dataset is provided at four spatial resolutions ranging from 30 s (~1 km²) to 10 min (~340 km²). At a 30 arc-second resolution (~1 km²), these data were utilized to construct finer estimates for month-by-month temperature and precipitation variables. By selecting the best-performing model for each location and variable, the forecast accuracy for temperature variables improves by 5–15% (0.07–0.17 °C) [51]. The data is used to drive the CSCS model, which is used to simulate PNV under the baseline scenario. The modeling findings are utilized to verify the CSCS model’s capacity to assess PNV on a national scale. In addition, using the simulated PNV, the current PNV zones in China are created under the baseline scenario.

2.2.2. The Future Scenario Data

To strengthen the study’s robustness, three RCPs were used to offset the bias caused by a single scenario as follows: one very low forcing level mitigation scenario (RCP2.6), one medium stabilization scenario (RCP4.5), and one extremely high baseline emission scenario (RCP8.5) [52]. Each future climatic scenario was separated into the following four time periods: 2020–2049 (2030s), 2040–2069 (2050s), 2060–2089 (2070s), and 2070–2099 (2080s) (Table 1). One GCM from those in the bias-corrected CMIP5 Coupled Model Intercomparison Project (CMIP5) [53] was used to run a series of climatic classifications, as follows: Meteorological Research Institute Coupled General Circulation Model version 3 (MRI–CGCM3) from the IPCC Fifth Assessment Report (IPCC–AR5) at spatial resolutions of 30 s (~1 km²). The CMIP5 dataset has been used for global changes in precipitation and in terms of seasonal changes [54]. CMIP5 projections have also been crucial for knowledge of atmospheric dynamics, present and future biases and uncertainties regarding greenhouse gas forcing, and attribution to human activities of the disruptions in precipitation cycles [55]. In addition, the MRI–CGCM3 is expected to perform well in reproducing seasonal temperature fluctuations and precipitation distribution differences between north and south [56,57,58]. All projected climate data, including monthly average temperature and monthly average precipitation, were the output data with 30 s resolution (~1 km²), which were incorporated into the vegetation model as inputs for the projected mean annual precipitation (MAP) and cumulative annual temperature above 0 °C (∑θ). All the gridded datasets are publicly available in the World Data Center for Climate due to the preponderance of decreased climate model bias by 50–70% [59].

2.2.3. The Land Use Data

The Resource and Environment Data Cloud Platform of China (RESDC) provided land use (LU) data with a 1 km² resolution [60,61]. The grid cells of the Chinese land use (CLU) data displayed information on the vegetation ecosystem type in the area. The change in vegetation type of the pixels is expected to occur over time and through the potential impacts of humans, which could affect vegetation succession [62]. To avoid uncertainty in the results caused by particular or single-year data, the study, therefore, selected those pixels of unchanged vegetation ecosystem types from LU data during 1980, 1990, 1995, and 2000 (Table 1) because multiperiodic ensembles capture comprehensive information of both geographic and temporal variability from all participating years. These pixels were extracted to estimate the existing PNV of China more robustly under the baseline scenario. The CLU data in 1980 are reconstructed based on the Landsat-MSS remote sensing image, while the remote sensing interpretation of the data for 1990, 1995, and 2000 are performed by employing Landsat-TM/ETM [63,64]. Farmland, forest, grassland, water area (which includes water, snow, and ice) [65], residential land, and unused land are among the six groups of LU data based on land resources and their use attributes [66]. These land-use groups are combined into three categories in this study (vegetation covering the area, no vegetation area, and cultivated vegetation area) to examine how existing terrestrial vegetation ecosystems are responding to climate change. The LU dataset, the most precise product of remote sensing monitoring data of land use in China, is notably useful for concerns such as national land resource surveys, hydrology, and ecological study, with an accuracy of nearly 95% [67].

2.2.4. The SRTM-DEM Data

The SRTM-DEM data with 30 s (~1 km²) resolution are derived from WorldClim 2.1 provided by World Climate Data. The data is mainly used to extract the typical existing PNV areas (Table 1).

2.2.5. A Vegetation Map at a Scale of 1:1,000,000

In order to select a better vegetation model, a vegetation map of China at a scale of 1:1,000,000 with a 1 km resolution (Table 1) is used to validate the CSCS model and HLZ model. The vegetation map with 46 original vegetation types can reflect the actual vegetation distribution status in more detail. The data has been widely used as critical scientific information and a foundation in the domains of climate change, biodiversity, and vegetation protection and monitoring.

2.3. Methods

Figure 2 depicts the research methodologies used in this study. The following are the key processes: (1) validate the CSCS model through CSCS vs. HLZ, CSCS vs. a vegetation map, and HLZ vs. a vegetation map; (2) extract pixels of unchanged vegetation ecosystem types during 1980, 1990, 1995, and 2000 to estimate the existing PNV under the baseline climate scenario; (3) carry out the PNV dynamics analysis from 1970 through the 2080s and compare PNV variations under present and future climatic scenarios.

2.3.1. The CSCS Model

The CSCS model, a complete, sequential categorization system of climate–soil–vegetation, is developed by grouping or clustering units with comparable moisture and thermal properties [41,42,68]. There are the following three levels in the system: class, subclass, and type [41,68,69]. This synthetic classification scheme attempts to account for bioclimatic conditions, edaphic conditions, and vegetation characteristics variables, and the bioclimatic conditions are supposed to be the most stable [42]. Therefore, the class level is the basic unit, and the classification of vegetation ecosystems mainly determined by bioclimatic conditions. The subclass vegetation differs in edaphic conditions, including landscape and soil, and vegetation characteristics classify the type level. Subclasses are integrated into classes based on the indexes of temperature and moisture, which capture the natural occurrence of vegetation ecosystems.

At the class level, two primary drivers of the CSCS model are annual cumulative temperature above 0 °C (${\displaystyle \sum \theta }$) (Growing Degree-Days on 0 °C base, GDD0) and humidity index (K), which are expressed by the following:

$$K=MAP/(0.1\times {\displaystyle \sum \theta })=MAP/(0.1\times GDD0)$$

(1)

where $MAP$ is the mean yearly precipitation (mm); ${\displaystyle \sum \theta }$ is the cumulative annual temperature above 0 °C; 0.1 is an empirical parameter. A humidity index (K) is the ratio of $MAP$ to ${\displaystyle \sum \theta }$ Values of ${\displaystyle \sum \theta }$ and K are divided into seven heat zones and six humidity levels, respectively (the index chart is available in Figure 3). The forty-two CSCS ecosystem can be theoretically produced. For explicit reflection and further comparison of the spatiotemporal distribution of PNV in China, 42 classes were regrouped into 10 super-classes based on the zonal characteristics (Figure 3) [41].

In the baseline scenario of the CSCS model, the numbers of PNV are separated into 36 classes and 10 super-classes over the China domain for the independent input variables employed as a grid format to build a PNV map (Figure 4). However, the baseline scenario did not include subtropical-extra arid subtropical desert (VIA), tropical-extrarid tropical desert (VIIA), warm-arid warm subtropical semidesert (VB), subtropical arid subtropical desert brush (VIB), tropical arid tropical desert brush (VIIB), and warm-semiarid subtropical grasses-fruticous steppe (VC).

2.3.2. The HLZ Model

The Holdridge Life Zone (HLZ) classification, an empirically based system, characterizes vegetation types in equilibrium with their annual climate [43]. HLZ does this by dividing the system into 38 ecological units known as Life Zones. A hexagon is drawn along the axes of biotemperature (BT), average total yearly precipitation (P), and potential evapotranspiration ratio (PER) to define each Life Zone (Figure 5) [70]. The specific calculation formula of the ecological unit is as follows:

$$BT(x,y,t)=1/12\times {\displaystyle {\sum }_{j=1}^{12}T(j,x,y,t)}$$

(2)

$$PET(x,y,t)=58.96\times BT(x,y,t)$$

(3)

$$PER(x,y,t)=PET(x,y,t)/P(x,y,t)=58.96\times BT(x,y,t)/P(x,y,t)$$

(4)

$$D_k(x,y,t)=\sqrt{{[BT(x,y,t)-B{T}_k]}^2+{[P(x,y,t)-P_k]}^2+{[PER(x,y,t)-PE{R}_k]}^2}$$

(5)

where $BT\left(x,y,t\right)$ is the mean biological temperature (°C) the xth row and the yth column grid in the period t in the spatial raster data, with values below 0 °C substituted with 0 °C and above 30 °C substituted with 30 °C; $BT\left(x,y,t\right)$, $PET\left(x,y,t\right)$, $P\left(x,y,t\right)$, and $PER\left(x,y,t\right)$, respectively, represent average annual biotemperature (°C), potential evapotranspiration (mm), mean total yearly precipitation (mm), and potential evapotranspiration ratio of the xth row and the yth column grid in the period t in the spatial raster data. $B{T}_k$, $P_k$, and $PE{R}_k$ are the central climate indexes of the kth hexagon in the Life Zones; $D_k\left(x,y,t\right)$ is the distance from the hexagon center in the kth Life Zone to xth row and the yth column grid.

2.3.3. Vegetation Classification Schemes

To evaluate the models’ simulated distribution of vegetation under the baseline climate scenario, the following two climate classifications and actual vegetation maps are used here: the CSCS system (42 classes), the HLZ classification (38 classes), and vegetation map (46 classes). These types are described in

Table A1. The PNV classes for the CSCS model and the HLZ model.

Classification System		Class Name and Code
CSCS	IA	Frigid-extrarid frigid desert, alpine desert	ID	Frigid subhumid moist tundra, alpine meadow steppe
	IIA	Cold temperate-extrarid montane desert	IID	Cold temperate subhumid montane meadow steppe
	IIIA	Cool temperate-extra arid temperate zonal desert	IIID	Cool temperate-subhumid meadow steppe
	IVA	Warm temperate-extra arid warm temperate zonal desert	IVD	Warm temperate sub humid forest steppe
	VA	Warm-extra arid subtropical desert	VD	Warm-subhumid deciduous broad leaved forest
	VIA	Subtropical-extra arid subtropical desert	VID	Subtropical-subhumid sclerophyllous forest
	VIIA	Tropical-extrarid tropical desert	VIID	Tropical-subhumid tropical xerophytic forest
	IB	Frigid-arid frigid zonal semidesert, alpine semidesert	IE	Frigid-humid tundra, alpine meadow
	IIB	Cold temperate-arid montane semidesert	IIE	Cold temperate-humid montane meadow
	IIIB	Cool temperate-arid temperate zonal semidesert	IIIE	Cool temperate-humid forest steppe, deciduous broad leaved forest
	IVB	Warm temperate-arid warm temperate zonal semidesert	IVE	Warm temperate-humid deciduous broad leaved forest
	VB	Warm-arid warm subtropical semidesert	VE	Warm-humid evergreen- broad leaved forest
	IB	Frigid-arid frigid zonal semidesert, alpine semidesert	VIE	Subtropical-humid evergreen broad leaved forest
	IIB	Cold temperate-arid montane semidesert	VIIE	Tropical-humid seasonal rain forest
	IC	Frigid-semiarid dry tundra, alpine steppe	IF	Frigid-perhumid rain tundra, alpine meadow
	IIC	Cold temperate semiarid montane steppe	IIF	Cold temperate perhumid taiga forest
	IIIC	Cool temperate-semi arid temperate typical steppe	IIIF	Cold temperate perhumid mixed coniferous broad leaved forest
	IVC	Warm temperate-semiarid warm temperate typical steppe	IVF	Warm temperate perhumid deciduous broad-leaved forest
	VC	Warm-semiarid subtropical grasses-fruticous steppe	VF	Warm-humid deciduous-evergreen broad-leaved forest
	VIC	Subtropical-semiarid subtropical brush steppe	VIF	Subtropical perhumid evergreen broad-leaved forest
	VIIC	Tropical-semiarid savanna	VIIF	Tropical-humid rain forest
HLZ	1	Polar desert	20	Warm temperate dry forest
	2	Subpolar dry tundra	21	Warm temperate moist forest
	3	Subpolar moist tundra	22	Warm temperate wet forest
	4	Subpolar wet tundra	23	Warm temperate rain forest
	5	Subpolar rain tundra	24	Subtropical desert
	6	Boreal desert	25	Subtropical desert scrub
	7	Boreal dry scrub	26	Subtropical thorn woodland
	8	Boreal moist fores	27	Subtropical dry forest
	9	Boreal wet forest	28	Subtropical moist forest
	10	Boreal rain forest	29	Subtropical wet forest
	11	Cool temperate desert	30	Subtropical rain forest
	12	Cool temperate desert scrub	31	Tropical desert
	13	Cool temperate steppe	32	Tropical desert scrub
	14	Cool temperate moist forest	33	Tropical thorn woodland
	15	Cool temperate wet forest	34	Tropical very dry forest
	16	Cool temperate rain forest	35	Tropical dry forest
	17	Warm temperate desert	36	Tropical moist forest
	17	Warm temperate desert scrub	37	Tropical wet forest
	19	Warm temperate thorn scrub	38	Tropical rain forest
1:1,000,000 vegetation distribution map	1	Cold temperate and temperate mountain coniferous forest	23	Subalpine hard-leaved evergreen broad-leaved thickets
	2	Temperate coniferous forest	24	Evergreen coniferous thickets in subalpine mountains
	3	Subtropical coniferous forest	25	Temperate dwarf semiarbour desert
	4	Tropical coniferous forests	26	temperate shrub deserts
	5	Subtropical and tropical montane coniferous forests	27	Temperate steppe shrub desert
	6	Temperate coniferous and deciduous broad-leaved mixed forest	28	Temperate semi-shrub, dwarf semi-shrub desert
	7	Mixed forest of coniferous, evergreen broad-leaved and deciduous broad-leaved in subtropical mountain areas	29	Temperate succulent halophyte dwarf shrub desert
	8	Temperate deciduous broad-leaved forest (8), temperate deciduous leaflet forests	30	Temperate desert of annual herbs
	9	Temperate deciduous leaflet forests	31	Alpine cushion-like dwarf semi-shrub desert
	10	Subtropical deciduous broad-leaved forest	32	Temperate grass and miscellaneous grass meadow grassland
	11	Subtropical evergreen and deciduous broad-leaved mixed forest	33	Typical grassland of temperate tufted grasses
	12	Subtropical evergreen broad-leaved forest	34	Temperate fascicled dwarf grass and dwarf semi-shrub desert grassland
	13	Subtropical monsoon evergreen broad-leaved forest	35	Alpine grasses and moss grasslands
	14	Subtropical hardleaf evergreen broad-leaved forest and copse	36	Temperate grass
	15	Monsoon rain forest	37	Subtropical and tropical grass
	16	Tropical rain forests	38	Temperate grasses and miscellaneous grasses meadow
	17	Subtropical and tropical bamboo forests and clusters	39	Temperate grasses, liverworts and miscellaneous grasses bog meadow
	18	Deciduous thickets in temperate zone	40	Temperate grasses and miscellaneous grasses salt meadow
	19	Subtropical, tropical evergreen broad-leaved, deciduous broad-leaved scrub (often containing rare trees)	41	Alpine hyssop and miscellaneous grass meadow
	20	Tropical coral limestone fleshy evergreen broad-leaved thickets and copses	46	Alpine tundra
	21	Subtropical and tropical xerophytic evergreen succulent prickly thickets	47	Alpine cushioned vegetation
	22	Deciduous broad-leaf thickets in subalpine mountains	48	Alpine sparse vegetation

. As a novelty compared to the previous work, the HLZ classification method was implemented, and for comparison of CSCS model output with that of the HLZ model, the original classes of these two systems and vegetation map were aggregated into the same five broad categories (

Table 2. Aggregation schemes of climatic classes for CSCS, HLZ, and Vegetation map. Class codes refer to Table A1 in the Appendix A.

Code	Broad Vegetation Types	Class Codes (a) of CSCS (42 Classes)	Class Codes (b) of HLZ (38 Classes)	Vegetation Code (c) in the Vegetation Map
1	Tundra	IA, IB, IC, ID, IE, IF	1, 2, 3, 4, 5	46, 47, 48
2	Desert	IIA, IIIA, IVA, IIB, IIIB, IVB, VB, VA, VIA, VIIA	6, 7, 11, 12, 17, 18, 19, 24, 25, 31, 32	25, 26, 27, 28, 29, 30, 31
3	Boreal and temperate forest	IVD, IIIE, IVE, IIF, IIIF, IVF	8, 9, 10, 14, 15, 16	1, 2, 6, 7, 8, 9, 18, 22, 23, 24
4	Subtropical and tropical forest	VD, VID, VE, VIE, VF, VIF, VIID, VIIE, VIIF	20, 21, 22, 23, 26, 27, 28, 29, 30, 33, 34, 35, 36, 37, 38	3, 4, 5, 7, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20, 21, 22
5	Grassland	IIC, IIIC, IVC, VC, IID, IIID, IIE, VIB, VIIB, VIC, VIIC	13	32, 33, 34, 35, 36, 37, 38, 39, 40, 41

).

2.3.4. Combination of Land Use/Cover (CLU) to PNV

To distinguish the non-vegetation area and the cultivated vegetation area from the natural vegetation area in the PNV map, the classification system of the CLU dataset was combined into three new classes (

Table 3. Combination of LU remote sensing monitoring data classification system.

The First Level	The Second Level (CLU Number)	The Third Level (CLU Number)
Vegetation coverage area	Forest (2)	Forest land (21), shrub forest (22), sparse forest land (23), other forest land (24)
	Grassland (3)	High-coverage grassland (31), medium-coverage grassland (32), low-coverage grassland (33)
	Unused land (6)	Gobi (62), marshland (64), bare land (65), bare rock texture (66), others (67)
	Other areas with low coverage vegetation	Coverage Tidal flats (45), Beaches (46)
No vegetation coverage area	Water area (4)	Canals (41), lakes (42), reservoirs and ponds (43), permanent glaciers and snow (44)
	Urban and rural, industrial and mining, residential land (5)	Urban land (51), rural residential area (52), other construction land (53), ocean (99)
	Unused land (6)	Sandy land (61), saline land (63)
Cultivated vegetation area	Farmland (1)	Paddy field (11), dry land (12)

).

2.3.5. Kappa Statistic

The Kappa statistic is used to compare and judge the accuracy and consistency of the results and is widely used to evaluate the simulated vegetation types and their spatial distributions [39]. In this study, based on GIS technology, 750 sample points were randomly selected (Figure 6). The Kappa statistic was used to evaluate the similarity and consistency between the vegetation distribution map simulated by the CSCS model, the vegetation distribution map simulated by the HLZ, and the 1:1,000,000 vegetation distribution map of China. Three potential vegetation types of distribution error matrices were generated, respectively. Then, the statistical value of the Kappa statistic of the sample vegetation type is as follows:

$$K=(N{\displaystyle {\sum }_i^n{P}_{ii}}-{\displaystyle {\sum }_i^n(P_{i+}{P}_{+i})})/(N^2-{\displaystyle {\sum }_i^n(P_{i+}{P}_{+i})})$$

(6)

where $N$ denotes the total number of samples; $n$ is the number of rows and columns in the error matrix; $P_{ii}$ is the individual entry for the row and column on the main diagonal of the constructed error matrix; $P_{i+}$ is the row total for each category $i$; $P_{+i}$ is the column total for each category $i$.

In general, the Kappa statistic (k) ranges from 0.0 to 1.0, with 0.0 representing totally different patterns and 1.0 indicating complete agreement. The generic evaluation thresholds of the statistic are shown in

Table 4. The evaluation thresholds of Kappa static.

Kappa Statistic	Agreement
0.0	Totally different patterns
0.0–0.2	No-to-poor agreement
0.2–0.4	Poor-to-fair agreement
0.4–0.55	Fair-to-good agreement
0.55–0.70	Good-to-very good agreement
0.70–1.0	Very good-to-perfect agreement
1.0	Complete agreement.

.

2.3.6. PNV under Current Baseline Scenario

It is important to note, in the real environment, that the PNV of China was altered to a great extent by the influence of human activities; with that in mind, a map of the current PNV map based on the CSCS model was produced. Our goal in choosing a strategy for gathering credible data for the present PNV map of China was to create a system that supplied numerous samples of each land cover class from many years in order to reflect spatial and temporal heterogeneity [71]. To create a contemporary PNV map of China, pixels of the unchanging land cover of the first level from reclassification LU data for the years 1980, 1990, 1995, and 2000 as a data source were used. Specifically, spatial overlay analysis between the unchanged pixels of cultivated vegetation or no vegetation in reclassification LU data and PNV map was carried out in the ArcGIS, obtaining a distribution map including vegetation area, no vegetation area, and PNV area.

2.3.7. Typical Areas of PNV under Current Baseline Scenario

Although the scope of human activities is everywhere, there are still some areas where the PNV is consistent with the PNV under current baseline scenario, such as some coasts and river banks in the high latitude areas of the Arctic and Antarctic and high altitude areas [72]. In order to eliminate human interference factors to determine the typical areas current PNV in China, the altitude element and then were selected as the typical distribution areas of existing PNV the regions of DEM > DEMmean + DEMstd from each PNV type [73]. This work aims to provide a scientific basis for field investigation and selection of biological characteristics of different PNV types [74].

3. Results

3.1. Performance of the CSCS Model

To evaluate the performance of the CSCS model, a comparison with classification results (Figure 7a) from 1970 to 2000 using the HLZ model (Figure 7b) generates 36 and 30 PNV classes, respectively, with a kappa index of 0.75; the CSCS PNV and the vegetation distribution map at a scale of 1:1,000,000 generates good-to-very good agreement (K value of 0.6), while the HLZ PNV and the map of actual vegetation generate fair-to-good agreement (K value of 0.4). The previous relevant study [75] also calculated the K value (0.74) for CSCS PNV vs. HLZ PNV. The good-to-very good agreement for CSCS PNV vs. the vegetation distribution map at a scale of 1:1,000,000 (overall K value of 0.6) is similar to that of the HLZ PNV vs. the map of actual vegetation (overall Kappa statistic of 0.4). It is notable that the CSCS model has a higher K index than the HLZ model. In other relevant studies, Zheng et al. [76] and Yates et al. [70] also calculated similar results. Therefore, we think the CSCS model performs better in simulating the PNV in China. Figure 8 shows the aggregated graphs, which provide a visual account of the commonalities (Figure 8a) and differences (Figure 8b,c) between different terrestrial ecosystems derived from the CSCS model and the HLZ model. The results show a consensus of 743.96 × 10⁴ km² between the outputs of these two models directly over the maps, covering nearly 77% of the total national area. However, there are significant mismatches between these two models in, for instance, the cases of the North China Plain, Northeast plain, Greater Hinggan Mountains, Inner Mongolia Plateau, Altai Mountains, Zhungeer Basin, Kunlun Mountains, Qilian Mountains, Tibetan plateau, Hengduan Mountains, Yunnan-Guizhou Plateau, Arbitrary Mountains, and Taiwan Mountains. This is more apparent in the relatively high-altitude regions, with inconsistent areas of more than 216.41 × 10⁴ km², amounting to 37% of the total land area of China, where a slight difference in the annual mean of temperature and precipitation variables may decide the classes to which they belong.

3.2. Spatial Distribution of PNV in the Baseline Scenario

In the application of such CLU dataset, for extraction of those pixels of unchanged land cover-type categories during 1980, 1990, 1995, and 2000 to capture more both geographic and temporal variability, the map is driven by CLU data is composed of the following three zones: PNV areas, cultivated vegetation areas, and no vegetation areas (Figure 9). Spatial patterns of the current PNV areas, 36 PNV classes from 10 super-classes performed by the CSCS model, are highly variable over the China domain in the baseline scenario (Figure 10). In general, the 36 PNV types from 10 super-classes were estimated to occur in the northwest, west, and south of China, with 664.24 × 10⁴ km², covering nearly 69% of the total national area. Specifically, frigid perhumid rain tundra and alpine meadow (IF) are the most distributed vegetation, followed by cool temperate-sub humid meadow steppe (IIIE) and warm temperate-extrarid warm temperate zonal desert (IVA) (

Table A2. The areas, percentage, and DEM of each PNV type in China in the baseline scenario.

ID	Class_Code	Class_Name	PNV						Specific PNV
			Area (104 km2)	Percentage (%)	DEM Value (m)				Area (104 km2)	Percentage (%)	DEM Value (m)
					Min	Max	Mean	Std			Min	Max	Mean	Std
1	IA	Frigid-extrarid frigid desert, alpine desert	0.23	0.02	3618	4797	4336.38	288.83	0.03	<0.01	4607	4797	4669.84	32.44
2	IB	Frigid-arid frigid zonal semidesert, alpine semidesert	4.27	0.45	3279	5259	4583.30	333.07	0.48	0.05	4670	5376	4979.95	53.84
3	IC	Frigid-semiarid dry tundra, alpine steppe	8.10	0.86	3124	5403	4701.97	470.04	0.34	0.04	4942	5476	5213.00	40.84
4	ID	Frigid-subhumid moist tundra, alpine meadow steppe	9.31	0.99	2095	5514	4764.47	475.26	0.75	0.08	4999	5630	5292.18	47.88
5	IE	Frigid-humid tundra, alpine meadow	19.48	2.06	1672	5705	4748.41	595.42	1.47	0.16	4934	5958	5417.45	63.62
6	IF	Frigid perhumid rain tundra, alpine meadow	149.85	15.88	1422	6356	4684.98	610.82	19.00	2.00	4812	6554	5522.18	180.07
7	IIA	Cold temperate-extrarid montane desert	4.07	0.43	2297	4536	3156.01	396.39	0.62	0.07	3082	4611	3891.48	291.26
8	IIB	Cold temperate-arid montane semidesert	5.53	0.59	1767	4694	3448.84	510.28	0.89	0.09	3748	4737	4436.43	127.98
9	IIC	Cold temperate-semiarid montane steppe	3.82	0.40	1272	4822	3235.99	819.09	0.66	0.07	4057	4832	4486.18	105.73
10	IID	Cold temperate-humid montane meadow	3.55	0.38	728	4823	2887.80	1031.05	0.58	0.06	3820	4882	4576.41	165.52
11	IIE	Cool temperate-humid forest steppe, deciduous broad-leaved forest	12.03	1.27	313	4847	2571.67	1285.21	2.65	0.28	3558	4973	4434.08	252.78
12	IIF	Cold temperate perhumid taiga forest	53.97	5.72	223	4930	2444.93	1540.63	10.94	1.15	3272	5123	4317.24	237.31
13	IIIA	Cool temperate-extrarid temperate zonal desert	21.23	2.25	803	3448	1799.55	589.06	4.71	0.50	2029	3808	2748.94	178.07
14	IIIB	Cool temperate-arid temperate zonal semidesert	21.81	2.31	414	3178	1410.25	529.23	3.34	0.35	1786	3344	2343.10	332.86
15	IIIC	Cool temperate-semiarid temperate typical steppe	18.54	1.96	414	3135	1278.56	423.59	2.73	0.29	1394	3160	2036.38	319.00
16	IIID	Cool temperate-subhumid meadow steppe	17.92	1.90	125	2762	1111.02	454.47	2.46	0.26	1396	2762	1822.29	222.14
17	IIIE	Cool temperate-sub humid meadow steppe	35.14	3.72	24	4063	970.00	665.15	3.89	0.41	1243	4258	2187.05	623.15
18	IIIF	Cold temperate perhumid taiga forest	35.17	3.73	9	4032	1235.98	1102.80	8.13	0.86	1824	4107	3007.66	327.55
19	IVA	Warm temperate-extrarid warm temperate zonal desert	37.05	3.93	170	2335	1055.85	331.97	4.46	0.47	1226	2457	1665.05	218.65
20	IVB	Warm temperate-arid warm temperate zonal semidesert	6.50	0.69	193	2041	777.36	388.47	1.57	0.17	1129	2071	1365.98	144.91
21	IVC	Warm temperate-semiarid warm temperate typical steppe	0.40	0.04	10	1679	1118.62	401.22	0.02	<0.01	1512	1752	1580.68	45.51
22	IVD	Warm temperate-subhumid forest steppe	2.70	0.29	−3	1742	430.43	376.71	0.54	0.06	611	1742	1006.19	159.04
23	IVE	Warm temperate-humid deciduous broad-leaved forest	19.00	2.01	−111	3100	1007.21	801.14	4.02	0.42	1410	3444	2322.52	235.71
24	IVF	Warm temperate perhumid deciduous broad-leaved forest	21.11	2.24	−1	3188	1551.15	655.11	4.34	0.46	1513	3558	2522.13	224.16
25	VA	Warm-extrarid subtropical desert	0.60	0.06	−153	411	99.12	137.32.	0.11	0.01	221	411	295.86	38.68
26	VB	Warm-arid warm subtropical semidesert	-	-	-	-	-	-	-	-	-	-	-	-
27	VC	Warm-semiarid subtropical grasses-fruticous steppe	-	-	-	-	-	-	-	-	-	-	-	-
28	VD	Warm-sub humid deciduous broad-leaved forest	0.08	0.01	52	1813	156.24	140.97	<0.01	<0.01	305	2039	633.42	608.85
29	VE	Warm-humid evergreen-deciduous broad-leaved forest	12.63	1.34	−3	2454	975.76	812.94	3.81	0.40	1332	2655	1994.38	131.18
30	VF	Warm-perhumid deciduous-evergreen broad-leaved forest	29.58	3.13	−2	2465	711.80	490.20	4.55	0.48	677	2559	1656.42	269.86
31	VIA	Sub tropical-extrarid subtropical desert	-	-	-	-	-	-	-	-	-	-	-	-
32	VIB	Subtropical arid subtropical desert brush	-	-	-	-	-	-	-	-	-	-	-	-
33	VIC	Subtropical-semiarid subtropical brush steppe	0.04	<0.01	919	1371	1184.18	104.50	<0.01	<0.01	1252	1559	1322.04	42.17
34	VID	Subtropical-subhumid sclerophyllous forest	0.61	0.06	658	1912	1445.06	227.77	0.11	0.01	1307	2041	1744.19	64.32
35	VIE	Sub tropical-humid evergreen broad-leaved forest	12.01	1.27	0	2000	840.21	529.02	2.45	0.26	1000	2144	1586.09	145.13
36	VIF	Subtropical perhumid evergreen broad-leaved forest	37.00	3.92	0	2045	475.45	389.26	5.61	0.59	459	2209	1245.05	254.06
37	VIIA	Tropical extrarid tropical desert	-	-	-	-	-	-	-	-	-	-	-	-
38	VIIB	Tropical arid tropical desert brush	-	-	-	-	-	-	-	-	-	-	-	-
39	VIIC	Tropical-semiarid savanna	<0.01	<0.01	926	1106	1035.88	46.82	<0.01	<0.01	1085	1106	1095.33	8.58
40	VIID	Tropical-subhumid tropical xerophytic forest	0.01	<0.01	1	34	14.03	8.17	<0.01	<0.01	23	34	28.36	3.75
41	VIIE	Tropical-humid seasonal rain forest	2.63	0.28	−2	994	212.62	181.28	0.42	0.04	158	1005	552.91	117.86
42	VIIF	Tropical perhumid rain forest	3.23	0.34	−5	918	157.55	188.08	0.42	0.04	199	1008	577.44	154.71

). Among them, IF is estimated to dominate in the cold and wet areas of the Qinghai-Tibet Plateau, the Tianshan Mountains, and the Altai Mountains; the IIIE shows prominent distribution characteristics of the south side of Xiaoxing’anling, Daxing’anling, and Nanshan of Qinghai; the occurrence and development regions of IVA are the Tarim Basin, the Inner Mongolia Plateau, and the Zhungeer Basin. In addition, the cultivated vegetation zones, including artificially planted arable land, farmland, and dry land, appear in the North China Plain, the Sichuan Basin, the Middle and Lower Yangtze Plain, the Northeast Plain, the Junggar Basin, the Tarim Basin, and the Yinshan Mountains, which are simulated to cover 103.71 × 10⁴ km², accounting for 11% of the total national area. Moreover, no vegetation zones, at 55.58 × 10⁴ km², accounting for 6% of the total area of China, consist of water areas and residential land, which are scattered in the Junggar Basin, Tianshan Mountains, Kunlun Mountains, the middle and lower reaches of the Yangtze River, the Northeast Plain, and the North China Plain.

As to PNV types, the lowest DEMmean values (14.03 m) are observed in the tropical-subhumid tropical xerophytic forest (VIID) (Figure 10), followed by the warm-extrarid subtropical desert (VA) (99.12 m), the warm–subhumid deciduous broad-leaved forest (VD) (156.24 m), and the tropical-humid seasonal rain forest (VIIF) (157.55 m) in the baseline scenario (Table A2 and Figure 11). The results suggest that above-vegetation from the forest or warm desert grassland with higher thermal grade occurs in tropical/warm areas, for instance, the Turpan Basin. Meanwhile, VIID, with the smallest DEMrange of only 33 m (DEMmax-DEMmin), is also the vegetation with the most concentrated vertical distribution. However, frigid-subhumid moist tundra, alpine meadow steppe (ID) is the vegetation with the highest DEMmean reaching 4764.47 m, followed by frigid-humid tundra, alpine meadow (IE) (4748.41 m), frigid-semiarid dry tundra, alpine steppe (IC) (4701.97 m), and frigid perhumid rain tundra, alpine meadow (IF) (4684.98 m). Moreover, this vegetation from the tundra and alpine steppe with the lowest heat level is projected into frigid and relatively humid regions, such as the Qinghai-Tibet Plateau, the Kunlun Tianshan, and the Altai Mountains. Moreover, the highest DEMrange values are estimated for IF, the most extensive distribution vegetation, reaching 4934 m.

3.3. Spatial Distribution of PNV in Typical Areas in the Baseline Scenario

An additional and valuable application of DEM data is the simulation of the spatial distribution of special PNV areas in the baseline scenario by selecting as the typical areas of PNV the pixels of DEM > DEMmean + DEMstd from each PNV type (Figure 12 and Figure 13). As to total areas of PNV, the typical areas of PNV are projected to be only 96.10 × 10⁴ km², accounting for 10% of current PNV regions or 11% of national land areas. These results are consistent with those of Feng et al. [67]. From the spatial distribution perspective, these vegetations occur in the Tianshan Mountains, the Kunlun Mountains, the Altai Mountains, the Himalaya Mountains, and the Hengduan Mountains. There are apparent environmental characteristics of high altitude, complex topography, and sparse population in these distributed regions, resulting in the absence of human intervention. This result is encouraging because it provides the scientific basis for field investigation and selection of biological characteristics of different PNV types. However, according to our results, population expansion and broad-scale agricultural activities are two dominant factors that affect the occurrence and development of terrestrial vegetation.

Although the relationship between the DEMmean range of PNV in typical areas and their thermal level is observed to be overall similar to that of existing PNV, both the DEMrange and DEMstd of each PNV in typical areas decrease significantly relative to existing PNV, especially for frigid–humid tundra and alpine meadow (Table A2 and Figure 13).

3.4. Changes of PNV Super-Classes under Future Climate Scenarios

Figure 14 and Figure 15 and

Table 5. The areas of each PNV type in China in the future climate scenario.

ID Super_Classes Name		Area (104 km2)
		T0	RCP2.6				RCP4.5				RCP8.5
			T1	T2	T3	T4	T1	T2	T3	T4	T1	T2	T3	T4
1	Tundra and alpine steppe	191.67	171.53	170.61	168.12	165.05	168.19	153.27	148.49	145.51	161.37	135.26	106.16	88.84
2	Cold desert	127.03	122.50	122.36	121.61	118.89	119.78	111.51	88.05	80.02	119.84	76.45	62.08	52.91
3	Semidesert	57.49	89.95	92.10	92.88	94.95	94.21	93.54	93.98	94.41	94.26	101.96	103.54	101.58
4	Steppe	33.55	62.60	77.38	69.53	18.70	77.77	67.86	67.51	68.81	80.72	91.24	79.56	76.02
5	Temperate humid grassland	55.16	60.48	52.19	54.95	46.76	54.86	57.30	58.97	56.96	55.33	51.72	48.92	51.02
6	Temperate forest	303.23	240.83	223.27	234.37	287.73	228.17	245.99	247.05	248.17	229.95	228.50	247.15	258.68
7	Subtropical forest	178.71	183.32	186.70	182.20	193.05	183.31	186.26	186.57	187.31	181.32	184.22	182.11	180.75
8	Tropical forest	12.50	25.61	28.72	30.11	31.30	28.45	31.80	35.20	36.88	28.69	36.75	48.30	58.80
9	Warm desert	0.97	1.54	4.10	4.13	3.99	2.51	10.34	32.43	39.79	5.24	47.49	62.85	69.69
10	Savanna	0.07	2.71	3.65	3.21	0.68	3.82	3.26	2.91	3.30	4.40	7.58	20.51	22.91

indicate the temporal changes in regional distribution, areal extents, and fractions of PNV super-classes under future climatic scenarios. Forests, the most widely dispersed vegetation, are projected to account for around 48% of the whole nation’s land area. Temperate forests are the most widely dispersed vegetation, accounting for 247.93 × 10⁴ km², or about 47% of the total country area. Although there are growing tendencies in the area of temperate forest from the 2030s to the 2080s under the three scenarios, the vegetation is expected to decline the most compared to the current climate baseline condition among the three forest types, according to the findings (Figure 14). With 55.06 and 15.5 × 10⁴ km², respectively, the RCP4.5 and RCP2.6 scenarios would have the greatest and smallest losses in the temperate forest areas. Tropical forests, on the other hand, are expected to spread consistently over the next century, increasing most of the three forest types. The RCP8.5 scenario has the largest area growth of 46.30 × 10⁴ km², whereas the RCP2.6 scenario has the smallest increase of 18.80 × 10⁴ km². In the RCP2.6, RCP4.5, and RCP8.5, subtropical forests would have fluctuating increases (rise at first, reduce later, and increase eventually), steady increases, and fluctuating declines (increase and then decrease). However, in the three scenarios, the vegetation’s areal extents show small increasing tendencies relative to the current baseline, with the RCP2.6 scenario showing the most apparent changes of 14.34 × 10⁴ km².

Grasslands are the second most widely dispersed vegetation, accounting for about 29% of the total national area. Although tundra and alpine steppe are expected to be the most widely dispersed grasslands, they will suffer and face significant dieback as a result of the warmer and wetter environment over this century. In the RCP2.6, RCP4.5, and RCP8.5 scenarios, tundra and alpine steppe areas would decrease by 26.62, 46.16, and 102.83 × 10⁴ km², respectively. Similarly, the steppe area would also decrease by 14.85 × 10⁴ km² under the RCP2.6, while the shrinkage of 8.4 × 10⁴ km² and 4.14 × 10⁴ km² in temperate humid grassland would occur in the RCP2.6 and RCP8.5, respectively. By contrast, savanna would benefit from future climate change, and the distributed extents are expected to expand continuously across all three scenarios. The most dramatic change in savanna would occur in the RCP8.5 by 22.84 × 10⁴ km², and the least expansion is predicted to happen in RCP2.6 by 0.61 × 10⁴ km².

Deserts, which are the least widely dispersed flora, are projected to cover just 26% of the total nation’s area. Under all three scenarios, the cold desert with the broadest distribution would consistently shrink the most over the next century among the three desert types. At the end of this century, the RCP8.5 is anticipated to lose the most land in cold desert areas, by 74.12 × 10⁴ km², while the RCP2.6 will lose the least land, by 8.14 × 10⁴ km². By contrast, although a fluctuating trend (increase at first and decrease later) in the warm desert is observed from the 2030s to the 2080s under the RCP2.6 scenario, a significant expansion would occur relative to the baseline scenario situation across all the three scenarios. The area of warm desert would increase by 3.02, 38.82, and 68.72 × 10⁴ km² in the RCP2.6, RCP4.5, and RCP8.5, respectively. Similarly, semidesert would present fluctuating trends from the 2030s to 2080s in the RCP4.5 and RCP8.5, which would decrease at first and increase later in the RCP4.5 and expand and then shrink in the RCP8.5, respectively. However, there is an apparent increasing variation in the areal extent of semidesert compared with the baseline climate scenario state under the three climate scenarios, by 37.46, 36.92, and 44.09 × 10⁴ km², respectively.

3.5. Succession among PNV Super-Classes over Time

The significant successions among different PNV super-class types during this century are projected across all three RCP scenarios (Figure 16). Additionally, a Sankey diagram is used to visualize the succession patterns (Figure 17).

From the future climate scenarios perspective, future climate change significantly impacts vegetation under the RCP8.5 scenario, where PNV super-classes exhibit the most apparent succession, at approximately 534.85 × 10⁴ km², followed by RCP2.6 (333.80 × 10⁴ km²) and RCP4.5 (357.28 × 10⁴ km²). As a result, in the future context of climate change, more harsh climates produced by climate change will result in larger ecological responses.

As to vegetation types, grasslands are projected to experience the most succession across all the three scenarios, which would evolve into deserts and grasslands at approximately 52.35 × 10⁴ km² and 33.59 × 10⁴ km² during T0 to T4, accounting for 61% and 39% of their losses, respectively. Among all the conversions from four grasslands to three forests, it is expected that the area of tundra and alpine steppe converted into the temperate forest in the future would be the most. The most significant succession from tundra and alpine steppe to temperate forest is projected in the RCP8.5 scenario during T0–T4 with 68.71 × 10⁴ km², at approximately 66% of losses for the tundra and alpine steppe. The succession would mainly occur in the western and southern parts of the Tibetan Plateau, the Qilian Mountains, the Hengduan Mountains, the Himalayas, the Tianshan Mountains, and the Altai Mountains. Concurrently, 28.93 × 10⁴ km² of evolution from the steppe into semidesert, the most dramatic change pattern among all types of evolution from grasslands to deserts, is also projected for RCP8.5 for 97% of grasslands losses. The evolution would happen in the Inner Mongolia Plateau, the Taihang Mountains, Qilian Mountains, and Tianshan Mountains.

Forests, the dominant vegetation type, are observed to undergo the second-largest succession across all three scenarios. During T0–T4, there are 50.96 × 10⁴ km² and 1.00 × 10⁴ km² of forests degraded into grasslands and deserts, accounting for 98% and 2% of their losses, respectively. The sharpest degeneration from forests to grasslands and deserts would occur in RCP8.5 in T0-T2 periods. Based on this result, more than 49.42 × 10⁴ km² and 1.88 × 10⁴ km² of temperate forest located adjacent to the North China Plain and Northeast China Plain are predicted to change to steppe and semidesert, respectively.

There are minor losses in deserts among all the three vegetation types that are expected to convert into grassland of 1.33 × 10⁴ km² and forests of only 0.02 × 10⁴ km² from T0 to T4. In the Northeast Plain, 2.74 × 10⁴ km² of semidesert would be converted into steppe during T2–T4 under RCP8.5, which may be most succession from deserts into grasslands.

4. Discussion

4.1. Discussion of the Methodology

The CSCS model and theory is a statistical model based on bioclimate, which was used in this study to quantitatively simulate the vegetation maps under baseline scenarios as well as predict their dynamics and succession trends in different periods (the 2030s–2080s) under RCPs scenarios. Our results exhibit that the CSCS model can not only successfully assess the spatial distribution of grassland, forest, and desert at the national scale but also performs better in simulating the PNV in China than the HLZ model. As well as known, the CSCS was established by climate–soil–vegetation [65]. The humidity index, determined by mean yearly precipitation (MAP) and cumulative temperature, is the main input in the CSCS. However, the precipitation variable does not consider other supply factors, especially underground water and melt water effects, which may cause the underestimation of the water variable and reduce the accuracy of the humidity index. In addition, this system does not take into account the terrain effects, which are likely to have negative impacts on the accuracy in regions with complicated underlying surfaces. Therefore, the terrestrial vegetation was simulated for a 30-year period on a national scale. To minimize the biases of PNV in China, 42 classes were regrouped into 10 super-classes based on the zonal characteristics. Moreover, land use and altitude factors were taken into account in this study. The CSCS enables a feasible approach to assessing the spatiotemporal distribution of terrestrial vegetation under climate conditions globally and regionally. Thus, the CSCS performs promising applications in the field of past and future change, especially for regions or periods with limited data.

4.2. GHG Adaptation Measures

Incrementing GHG concentrations yields increasingly more profound changes, with much of the world shifting towards different climates from those in the present. However, ecosystem-climate feedbacks and interactions are often highly non-linear and non-additive. Although our study has not focused in detail on the impacts of specific vegetation afforestation and vegetation progression on future climate change, some general concepts are discussed below.

The two main sources of uncertainty of climate impacts on vegetation are projections of future climate and their resulting impacts on vegetation. Since changes in climate and vegetation ecosystems depend on complex interactions between the atmosphere, biosphere, and precipitation cycle, there is a need for effective and comprehensive climate impact assessments. A more integrated representation of vegetation in climate models should therefore allow more robust quantification to assess the impacts of climate change in the past, current, and future on vegetation. Based on the CSCS model, we focus on predicting quantitatively the spatiotemporal dynamics of the vegetation ecosystem in China using RCPs datasets. In particular, the ecological responses (e.g., succession from tundra and alpine steppe to temperate forest) of the western and southern parts of the Tibetan Plateau and the Qilian Mountains would be significant in RCP8.5 from 1970–2000 to the 2050s, since the local vegetation development would be primarily influenced by temperature and precipitation changes due to GHG emissions.

Future climate change impacts on vegetation changes in China will depend on the balance between changes in productivity and rates of decomposition and GHG emissions, both of which depend on climatic, land use, and management factors. Changing land management practices within vegetation land uses could also alter the climate, which in turn affects vegetation. In addition, vegetation productivity is likely to be reduced by the precipitation redistribution process in China. Changes in future water management practices will alter the effectiveness of vegetation mitigation strategies, such as ecological vulnerability and water supply for irrigation. Adaptation in the ecological protection sector could potentially provide additional benefits such as vegetation development for reduced flood risk and increased drought resilience in China.

4.3. Uncertainties

In this study, uncertainties exist in the simulation of PNV. The accuracy of the baseline and future scenario datasets may appear to be less certain. The predicted results of PNV depend on the accuracy of the dataset used to drive the CSCS model. Tapiador et al. have suggested that GCMs from those in the bias-corrected CMIP5 are still too limited to extract definite estimates of the regional and world climates, even for present climate conditions [43]. Although the dataset used from MRI–CGCM3 with a decreased climate model bias by 50–70% is expected to perform well in reproducing seasonal temperature fluctuations and precipitation distribution differences in China, the error source of this dataset has been reported to possibly come from the homogenization and interpolation methods. Some regions where the observation data are relatively few or the observation sites are scattered may result in some errors in the classification accuracy of the CSCS model.

5. Conclusions

Using the CSCS model, terrestrial vegetation was quantitatively estimated over the China domain under the present baseline scenario (1970–2000) and three future climatic scenarios (2030s–2080s). We focused on evaluating the CSCS model’s performance and PNV distribution under the present baseline scenario as well as examining PNV fluctuations as a function of climate change.

Present-day land use and future climate change imply heavy pressure on China’s ecosystems and biodiversity. Although their future trajectories may be uncertain, the CSCS model has become an indispensable tool for developing strategies to mitigate the consequences of global climate change on ecosystems, which has a wide range of potential applications in understanding climate-vegetation interactions under different future climate scenarios.

From the current situation, the regions of typical current PNV occupy less than 10% of the total national area caused by human land use and cropland expansion. Moreover, these ecosystems only dominate in high altitudes, complex topography, and sparse populations, where they are not prone to human intervention under the current baseline scenario. A closer view of these ecosystems raises that the higher altitude may be conducive to the occurrence and development of some vegetation ecosystems to a certain extent, especially for low thermal vegetation.

By the end of the 21st century, a variety of climatic impacts on vegetation across the China domain are expected, with consequences varying greatly across various terrestrial ecosystems. Harsher climate induced by climate change might result in larger ecosystem responses, according to the simulated results based on the CSCS model in the context of future climate change. As a consequence of a warming climate, despite some advancement of temperate forest at the expense of substantial dieback of tundra and alpine steppe from 1970–2000 to the 2080s, all emission scenarios agree that this century would witness a considerable shrinkage of temperate forest.

In general, a warmer and wetter climate would be conducive to the occurrence and development of terrestrial vegetation to a certain extent. However, these changes would restrain some ecosystems, which would suffer irreversible degradation in the future, particularly in the mid-/high elevations. Ultimately, although future greenhouse gas (GHG) emissions have uncertainties, our potential CSCS-based analyses of current and future terrestrial ecosystem processes in China are paramount to enhancing our understanding of human intervention impact on vegetation and suggesting possible future scenarios for terrestrial vegetation ecosystems’ dynamic and diverse biosphere-future climate change interactions under different future climate scenarios. Adaptation in the ecological protection sector could potentially provide additional benefits to vegetation development, such as reduced flood risk and increased drought resilience in China.