Projected Drought Risk to Vegetation Productivity Across the Mongolian Plateau Under CMIP6 Scenarios

Abstract

In the context of global climate change, a comprehensive understanding of the spatiotemporal impacts of drought on vegetation productivity is essential for assessing terrestrial ecosystem stability. Utilizing outputs from six global climate models (GCMs) from the Coupled Model Intercomparison Project Phase 6 (CMIP6), this study systematically assessed historical and projected drought probability, the drought vulnerability of Net Primary Productivity (NPP), and overall drought risk across the Mongolian Plateau under three Shared Socioeconomic Pathway scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5). Results revealed that the Standardized Precipitation Evapotranspiration Index (SPEI) exhibited a declining trend, whereas NPP showed an overall increasing trend. These changes were most pronounced under the SSP5-8.5 scenario, with the SPEI decreasing at a rate of −0.39/10a and NPP increasing at 25.8/10a. Drought severity exhibited strong spatial heterogeneity, intensifying from northeast to southwest, whereas NPP demonstrated an inverse spatial pattern. The spatial distribution of high-drought-risk zones varied markedly across scenarios: the southwestern region was most affected under SSP1-2.6, the northwestern region under SSP2-4.5, and the southeastern region under SSP5-8.5. Based on 12-month SPEI values and NPP derived from the Carnegie–Ames–Stanford Approach (CASA) model, SSP2-4.5 presented the highest overall drought risk, despite lower emissions. The annual mean NPP drought vulnerability ranked as follows: SSP2-4.5 (0.60 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}\mathrm{r}}^{-1}$) > SSP1-2.6 (−1.03 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}\mathrm{r}}^{-1}$) > SSP5-8.5 (−1.24 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}\mathrm{r}}^{-1}$). Projections indicated a substantial increase in drought occurrence probability during the period 2061–2100, particularly under SSP2-4.5 and SSP5-8.5. Under higher emissions, the spatial extent of areas with negative drought vulnerability values was expected to expand 68%. Wind speed was the dominant factor influencing drought risk under SSP1-2.6 and SSP2-4.5, whereas precipitation became the primary driver (45.34%) under SSP5-8.5. These findings offer critical insights for early drought warning systems and for strengthening ecosystem resilience across the Mongolian Plateau.

1. Introduction

Drought is a climatic phenomenon characterized by prolonged periods when the humidity is lower than normal, resulting in water scarcity [1]. It is widely recognized as one of the most destructive and economically consequential natural hazards among all natural disasters [2]. Global warming and the increase in carbon emissions have led to a continuous rise in droughts, and the economic losses caused by droughts amount to 59% [3]. Droughts are typically classified into four types: meteorological drought, agricultural drought, hydrological drought, and socio-economic drought [4]. Among these, meteorological drought is considered the precursor to the others and has garnered significant attention in recent research [5]. To advance drought monitoring and assessment, several drought indices have been developed, including the Palmer Drought Severity Index (PDSI), the Standardized Precipitation Index (SPI), the Reconnaissance Drought Index (RDI), and the Standardized Precipitation Evapotranspiration Index (SPEI) [6]. SPEI, in particular, is well-suited for drought analysis under climate change conditions due to its ability to account for both drought duration and cumulative severity [7]. Vegetation responses to drought are often time-scale dependent: grasslands and croplands respond rapidly, whereas forests exhibit delayed responses, making SPEI particularly suitable for ecosystem-scale drought assessments [8]. Under global warming, droughts are projected to become more frequent, severe, and spatially widespread [9,10]. The combined pressures of climate change and anthropogenic activities have substantially amplified the impacts of drought on ecosystems and socio-environmental systems [11,12]. Studies have reported increasing drought intensity and frequency in regions such as the Southern Sahara, Central Asia, Western Australia, and Mexico [13]. In China, drought severity has intensified across 78.4% of the country since 2000, with a notable spatial divergence characterized by a “wet-getting-wetter, dry-getting-drier” pattern [14]. Drought negatively affects ecosystems by reducing vegetation growth, lowering canopy cover, and diminishing productivity [15,16], with some impacts persisting even after drought conditions subside [17,18]. Projections from global climate models indicate that future warming will exacerbate drought frequency and intensity [19,20], with profound impacts on terrestrial ecosystems and human societies.

Net Primary Productivity (NPP), defined as the difference between the total carbon gained by vegetation through photosynthesis and the energy expended through respiration, serves as a critical indicator of ecosystem function [21]. Various models have been developed to estimate NPP, each with varying degrees of applicability across different regions [22]. For instance, the process-based Boreal Ecosystem Productivity Simulator (BEPS) has proven effective in capturing gross primary productivity (GPP) changes across 41 flux stations globally, including their vulnerability to drought. The Carnegie–Ames–Stanford Approach (CASA) model, which is based on light utilization efficiency, can effectively characterize vegetation productivity in the Mongolian Plateau [23]. Although the response of NPP to climate change and its long-term trends have been widely studied [24], relatively few investigations have focused on how NPP responds specifically to drought stress. Recent increases in drought frequency have severely disrupted vegetation growth and ecosystem productivity. At the global scale, NPP in arid and semi-arid regions has shown a positive correlation with SPEI, particularly within the 20–50° latitude band, where vegetation exhibits pronounced drought responses [25]. As climate change intensifies, the frequency and intensity of drought events are projected to increase further, potentially posing greater threats to terrestrial ecosystem NPP [26,27]. Regional studies have further confirmed this trend. For example, in Inner Mongolia, droughts of varying intensities lead to significantly different levels of grassland NPP loss, which increases exponentially with drought severity [28]. Notably, under SSP2-4.5, simulations indicate that terrestrial NPP across China may continue to rise, implying a relatively lower risk to ecosystem productivity under moderate emissions [29]. Therefore, comprehensively assessing NPP drought risk has become a critical step in mitigating the impacts of drought on ecosystems.

According to the Fourth Assessment Report (AR4) of the Intergovernmental Panel on Climate Change (IPCC), natural disasters arise from the interaction between hazards and vulnerabilities. In this context, the term ‘risk’ refers specifically to risks associated with the ecosystem carbon cycle. Drought risk refers to the potential adverse impacts and losses caused by drought events on natural ecosystems and socio-economic systems. Thus, drought risk is defined as the product of the probability of drought occurrence and ecosystem vulnerability [30]. Research indicates that most areas of the arid regions in Asia will face more severe forms of drought in the future. Although the frequency of droughts may decrease, their duration and intensity are projected to increase substantially [31,32]. From 1902 to 2018, average drought frequencies in China and Mongolia were 57.4% and 54.6%, respectively, making them the most drought-prone areas in East Asia. The ecosystem drought risk assessment framework developed by Li et al. (2020) [33] indicates that vulnerability is the primary driver of drought risk in Northeast China, while adaptability plays a key role in modulating system sensitivity. However, quantitative projections of vegetation drought risk across emission scenarios remain limited for the plateau. The probability of drought risk in the Mongolian Plateau is projected to increase significantly from 2080 to 2100, especially in areas prone to drought [34]. NPP, as an indicator of ecosystem function, can reflect variations in drought risk. Using historical data, Ren et al. (2023) [35] assessed drought probability, NPP drought vulnerability, and drought risk across the Mongolian Plateau. They found that areas prone to drought were concentrated in the central and northern regions, while the areas with the highest drought vulnerability were located in the northeastern, central, and southeastern regions. However, existing studies are predominantly retrospective, focusing on spatiotemporal patterns of past drought risks. Research examining future drought risks to vegetation under different climate scenarios remains scarce. This highlights an urgent need for forward-looking assessments of drought-vegetation dynamics under climate change scenarios.

As a transitional zone between arid and semi-arid climate, the Mongolian Plateau supports diverse vegetation types and plays a critical role in the global carbon cycle. However, over the past four decades, it has experienced pronounced warming and drying trends, severely threatening ecosystem stability [36]. Since the early 21st century, drought events on the Mongolian Plateau have intensified [37,38], with their spatial extent and severity being widely validated. Consequently, accurately projecting future drought dynamics and associated vegetation responses, as well as systematically assessing the evolving patterns of drought risk, is a critical research priority [39]. The primary objectives of this study were the following: (1) analyze trends in NPP and drought-related indicators (including drought probability, vulnerability, and risk) across the Mongolian Plateau under multiple emission scenarios from 2021 to 2100; (2) quantify NPP drought vulnerability and drought probability to assess overall drought risk; and (3) elucidate the impact mechanisms of drought stress on NPP in the Mongolian Plateau. This study addresses critical knowledge gaps regarding NPP drought risk under different emission scenarios on the Mongolian Plateau and enhances scientific understanding of regional vegetation-drought dynamics. The findings of this study are expected to provide theoretical support for improving disaster risk prevention and management strategies on the Mongolian Plateau.

2. Materials and Methods

2.1. Study Area

The Mongolian Plateau, located in the interior of Eurasia (37°22′–53°20′ N, 87°43′–126°04′ E), primarily encompasses Mongolia and the Inner Mongolia Autonomous Region of China. The region exhibits a west-high, east-low topographic gradient, with an average elevation of 1580 m (Figure 1a). It experiences a typical temperate continental climate characterized by extreme temporal and spatial variability in precipitation. Spatially, humidity increases progressively from the southwest to the northeast. Annual precipitation is approximately 50 mm in the inland areas of the southern and western Gobi Desert, while it exceeds 400 mm in the northeastern forest region (Figure 1c). The eastern part of the Mongolian Plateau is dominated by temperate grasslands with high vegetation coverage; the central area serves as a transitional zone between grasslands and deserts, marked by sparse vegetation; and the western region consists predominantly of deserts with extremely low vegetation cover (Figure 1b). Soil types are distributed in a zonal pattern, transitioning from sandy aridisols in the western deserts to chernozems in the eastern grasslands, influencing regional differences in drought vulnerability. Animal husbandry is the dominant economic activity on the Mongolian Plateau. In Inner Mongolia, animal husbandry has shifted from a nomadic to a more settled form, with the intensity of human activity being greater than that in Mongolia. Compared to Mongolia, Inner Mongolia has a larger population and greater area under cultivation. These unique geographical, climatic, and socio-economic characteristics make the Mongolian Plateau an ideal region for investigating drought mechanisms and vegetation responses.

2.2. Data Sources

Meteorological observations for this study were sourced from the ERA5-Land reanalysis dataset provided by the European Centre for Medium-Range Weather Forecasts (ECMWF), covering the period 1981–2014. The dataset included variables such as total precipitation (P), 2 m air temperature (t2m), 10 m u-component (u10) and v-component (v10) of wind, surface net solar radiation (ssr), and 2 m dewpoint temperature (d2m) (https://cds.climate.copernicus.eu/datasets/reanalysis-era5-land?tab=overview (accessed on 3 December 2023)). ERA5-Land is the land component reanalysis of the ERA5 climate reanalysis, featuring higher spatial and temporal resolution than the ERA5 dataset. ERA5-Land is produced via a single uncoupled land surface simulation and is not integrated with the atmospheric module of the ECMWF Integrated Forecasting System (IFS) or the IFS wave model. Consequently, it is updated more rapidly and does not require data assimilation, thereby supporting high-precision and scientifically robust climate analyses [40,41]. These meteorological data were used to calculate the Standardized Precipitation Evapotranspiration Index (SPEI) and served as foundational inputs for assessing the applicability of CMIP6 models and evaluating drought risk. The original data at 0.1° resolution were resampled to 0.25° using bilinear interpolation, and the optimally performing climate models for the Mongolian Plateau region were selected based on the 0.25° dataset [42].

Vegetation productivity was modeled using the Carnegie–Ames–Stanford Approach (CASA), which is primarily driven by the normalized difference vegetation index (NDVI), monthly mean temperature, monthly solar radiation, and monthly precipitation. NDVI data ware obtained from the GIMMS NDVI3g version 1q global-scale dataset generated by several advanced high-resolution radiometers (AVHRR) of the National Oceanic and Atmospheric Administration (NOAA) of the United States. This dataset has a 15-day temporal resolution covering the period from July 1981 to December 2015 and a spatial resolution of 0.0833°. The 15-day NDVI composites were processed using the maximum value composite(MVC) method to generate a monthly NDVI time series [43]. Meteorological inputs (monthly mean temperature, precipitation, and solar radiation) for the period 1982–2014 were obtained from the China Integrated Meteorological Information Service System and the Mongolia Meteorological Agency. These were interpolated using the Kriging method to match the resolution of CMIP6 NPP data. The Kriging interpolation method offers superior performance and high accuracy in regions with flat terrain and evenly distributed meteorological stations [44]. All data were preprocessed and aggregated to an annual scale for subsequent calculations of drought vulnerability and risk.

Climate model projections were obtained from the NASA Earth Exchange Global Daily Downscaled Projections based on CMIP6 (NEX-GDDP-CMIP6, https://www.nccs.nasa.gov/services/data-collections/land-based-products/nex-gddp-cmip6 (accessed on 20 October 2024)). This dataset provides high-resolution, bias-corrected climate projections, making it well-suited for fine-scale climate change assessments. Based on data completeness, coverage of key meteorological variables (monthly temperature, precipitation, minimum and maximum temperatures, relative humidity, wind speed, and solar radiation), and model performance (requiring R > 0.8 for both precipitation and PET simulations), seven global climate models (GCMs) were selected (

Table 1. Summary of seven climate models from NEX-GDDP-CMIP6 used in this study.

	Model Name	Country	Grid (Lon × Lat)
1	ACCESS-CM2	Australia	0.25° × 0.25°
2	ACCESS-ESM1-5	Australia	0.25° × 0.25°
3	EC-Earth3	Europe	0.25° × 0.25°
4	EC-Earth3-Veg-LR	Europe	0.25° × 0.25°
5	MPI-ESM1-2-HR	Germany	0.25° × 0.25°
6	MPI-ESM1-2-LR	Germany	0.25° × 0.25°
7	MRI-ESM2-0	Japan	0.25° × 0.25°

). This NEX-GDDP-CMIP6 dataset does not include NPP projections [45], so corresponding NPP outputs from the same seven CMIP6 GCMs were used (

Table 2. Summary of seven climate models from CMIP6 used in this study.

	Model Name	Country	Lattice Points
1	ACCESS-ESM1-5	Australia	192 × 145
2	CanESM5	Canada	128 × 64
3	CMCC-ESM2	Italy	288 × 192
4	EC-Earth3-Veg	Europe	512 × 256
5	EC-Earth3-Veg-LR	Europe	320 × 160
6	INM-CM4-8	Russia	180 × 120
7	MPI-ESM1-2-LR	Germany	192 × 96

). Historical data (1981–2014) and future projections (2015–2100) were analyzed under three emission scenarios: SSP1-2.6, SSP2-4.5, and SSP5-8.5. The CMIP6 NPP data were resampled via bilinear interpolation to match the 0.25° × 0.25° spatial resolution of the NEX-GDDP-CMIP6 data. A multi-model ensemble (MME) was generated using an equal-weight averaging method to reduce prediction noise and enhance the reliability of results [46].

2.3. Methodology

2.3.1. Standardized Precipitation Evapotranspiration Index (SPEI)

The SPEI is a multi-scalar drought index based on the imbalance between precipitation and potential evapotranspiration (PET) [47]. In this study, SPEI was calculated using the Penman–Monteith equation to analyze drought conditions over the Mongolian Plateau [48]. The calculating equation is as follows:

(1)

Calculation of potential evapotranspiration:

$$PET={\displaystyle \frac{0.408∆\left(R_n - G\right) + \gamma \frac{900}{T + 273}{U}_2(e_s - e_a)}{∆ + \gamma (1 + 0.34U_2)}}$$

(1)

where $R_n$ is net radiation, G is the soil heat flux, $\gamma$ is the psychrometric constant, $∆$ is the slope of the saturation vapor pressure curve, T is air temperature, $U_2$ is wind speed at a height of 2 m, $e_s$ is the mean saturation vapor pressure, and $e_a$ is the actual vapor pressure, respectively. The wind speed in 10 m was converted into wind speed in 2 m based on the Equation (2).

$$U_2=U_{10}{\displaystyle \frac{4.87}{ln(67.7\times 10-5.42)}}$$

(2)

$U_2$ and $U_{10}$ are the wind speed at 2 and 10 m, respectively.

(2)

The climatic water balance for month j is calculated as the difference between precipitation and potential evapotranspiration.

$$D_j=P_j-{PET}_j$$

(3)

where $P_j$ and ${PET}_j$ are precipitation and potential evapotranspiration in month j (mm).

(3)

The cumulative difference between precipitation and PET over different time scales was calculated, denoted as $X_{i,j}^k$. The variable $X_{i,j}^k$ represents the cumulative water deficit over a time scale of k months, calculated as the sum of the deficits from the current month j and the preceding k − 1 month in year i. The resulting time series was then fitted to a log-logistic probability distribution to obtain the standardized SPEI values.

$$SPEI=\mathrm{w}-{\displaystyle \frac{{\mathrm{C}}_0+C_{1}w+C_2{w}^2}{1+d_{1}w+d_2{w}^2+d_3{w}^3}}$$

(4)

When p ≤ 0.5, $w=\sqrt{-2lnP}$, and when p > 0.5, $w=\sqrt{-2\mathrm{l}\mathrm{n}(1-P)}$, where p is the cumulative probability of $X_{i,j}^k$ at a given time scale, ${\mathrm{C}}_0$ = 2.515517, $C_1$ = 0.802853, $C_2$ = 0.010328, $d_1$ = 1.432788, $d_2$ = 0.189269, and $d_3$ = 0.001308. The SPEI on a 12-month time scale was used to characterize drought in the present study. As the SPEI represents standardized anomalies relative to a long-term climatology, no additional detrending was applied, unlike NPP, which required detrending to isolate climate-driven interannual variability. The drought severity was classified according to the SPEI, as shown in

Table 3. The SPEI drought index categories.

Grade	Type	SPEI Value
0	Normal	more than −0.5
1	Mild drought	(−1.00, −0.5]
2	Moderate drought	(−1.50, −1.00]
3	Severe drought	(−2.00, −1.50]
4	Extreme drought	less than −2.00

Note. SPEI (Standardized Precipitation Evapotranspiration Index).

.

2.3.2. Carnegie–Ames–Stanford Approach (CASA) Model

The CASA model is a land surface process model that simulates the interaction between the land surface and the atmosphere, based on the maximization of light energy utilization. It fully takes into account the conditions related to the ecosystem and the environment, as well as the impact of vegetation on the estimation of net primary productivity. The NPP is calculated by considering the absorption of photosynthetically active radiation (APAR) by plants and the actual light energy utilization rate (ε) [49]:

$$\mathrm{N}\mathrm{P}\mathrm{P}(\mathrm{x},\mathrm{t})=\mathrm{A}PAR\left(\mathrm{x},t\right)\times \epsilon (x,t)$$

(5)

where $\mathrm{A}PAR\left(\mathrm{x},t\right)$, $\epsilon (x,t)$, and $\mathrm{N}\mathrm{P}\mathrm{P}(\mathrm{x},\mathrm{t})$ are the photosynthetically absorbed active radiation, the efficiency of actual light energy, and the net primary productivity of vegetation, respectively, where t is the month and x is the geographical location.

$$APAR\left(x,t\right)=0.5SOL\left(x,t\right)\times FPAR(x,t)$$

(6)

$$\epsilon \left(x,t\right)=T_{\epsilon 1}\left(x,t\right)\times T_{\epsilon 2}\left(x,t\right)\times W_{\epsilon }\left(x,t\right)\times {\epsilon }_{max}$$

(7)

where 0.5 denotes the ratio of photosynthetically active radiation available to vegetation, $SOL\left(x,t\right)$ is the total solar radiation, FPAR is the fraction of photosynthetically active radiation absorbed by the vegetation canopy, $T_{\epsilon 1}\left(x,t\right)$ and $T_{\epsilon 2}\left(x,t\right)$ represent the effects of low-temperature and high-temperature stress, respectively, on the light use efficiency of vegetation at location x. $W_{\epsilon }\left(x,t\right)$ represents the water stress scalar that modifies the light use efficiency for a specific vegetation type, while ${\epsilon }_{max}$ denotes the maximum light use efficiency for that vegetation type. Due to the relatively high uncertainties of MODIS NPP in arid and semi-arid areas and the scarcity of ground observations, the CASA model was selected. Its integration of remote sensing and meteorological inputs enables a more accurate and spatially continuous characterization of NPP dynamics.

2.3.3. Assessment of Drought Risk of NPP

Based on Van Oijen et al. [30] and He et al. [22], the drought risk of NPP is calculated using the following equation:

$${NPP}_r=P_d\times {NPP}_v$$

(8)

where ${NPP}_r$ is the drought risk of NPP, $P_d$ is the drought probability, and ${NPP}_v$ represents the drought vulnerability of NPP. The value of $P_d$ is determined based on a given threshold (e.g., −0.5) of the SPEI. Based on the climatic characteristics of the previous study area and the historical drought statistics, the drought occurrence probability was calculated using SPEI < −0.5 [50,51,52]. ${NPP}_v$ was computed as average de-trended NPP in all years without drought (SPEI-12 > −0.5) minus average de-trended NPP in all drought years. Large ${NPP}_r$ represents the higher drought vulnerability of NPP, and vice versa.

3. Results

3.1. Model Evaluation

To evaluate model performance over the Mongolian Plateau, Taylor diagrams were employed to compare global climate model simulations against historical observational data. Taylor diagrams assess the agreement between simulations and observations using four key metrics: the spatial correlation coefficient, the centered root mean square error (RMSE), and the standard deviations of both simulated and observed fields [53]. Figure 2a presents the precipitation simulation results across multiple climate models. Most models exhibit correlation coefficients between 0.86 and 0.91, indicating a good ability to reproduce the spatiotemporal distribution of precipitation. Among the seven models, ACCESS-ESM1-5 aligned most closely with observational data, and demonstrated the best simulation performance. Figure 2b illustrates the simulation performance for potential evapotranspiration (PET), which showed greater accuracy than precipitation and the SPEI, with correlation coefficients of approximately 0.93 across all models. Figure 2c displays the simulation of the annual SPEI, where correlation coefficients range from 0.3 to 0.6 and RMSE values lie between 1.0 and 1.5. The multi-model ensemble (MME) mean outperforms individual models, with correlation coefficients of 0.92 for precipitation, 0.93 for PET, and 0.69 for SPEI-12, highlighting the advantages of ensemble approaches in climate simulations. Accordingly, this study adopted the MME for subsequent analyses of future drought conditions over the Mongolian Plateau under different emission scenarios.

In this study, the CASA model was used to estimate NPP over the Mongolian Plateau during 1982–2014, utilizing these results as observational data to evaluate the vegetation productivity simulation capabilities of seven global climate models. Previous studies have shown that the CASA model can effectively reflect the actual NPP of vegetation on the Mongolian Plateau [54,55]. During model evaluation, a pixel-by-pixel extraction method at consistent spatial resolution was employed, with validation conducted through linear regression analysis. The results demonstrate a significant correlation between CMIP6-simulated NPP and CASA-modeled NPP, with a coefficient of determination (R²) reaching 0.6106 (p < 0.01). Additionally, the mean relative error (MRE) of 0.24 indicates 76% simulation accuracy for CMIP6 models in this region. Meanwhile, the RMSE between CMIP6-simulated NPP and CASA-NPP was 110.98 g C m⁻², with a mean absolute error (MAE) of 85.76 g C m⁻². The results demonstrate that CMIP6-based NPP simulations can effectively capture the spatial distribution characteristics of NPP across the Mongolian Plateau (Figure 3).

3.2. Spatiotemporal Dynamics of SPEI and NPP over the Mongolia Plateau

Figure 4 depicts the spatiotemporal evolution of annual drought conditions during the historical period and under three future emission scenarios. The results indicate a progressive intensification of drought across the Mongolian Plateau in the future. Under the SSP1-2.6 scenario, drought trends gradually ease, whereas the SPEI curves under SSP2-4.5 and SSP5-8.5 scenarios show substantial declines. The most severe drought conditions are observed under SSP5-8.5, with a maximum decline rate of −0.39/10a (Figure 4a).

Spatially, the historical period and the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios exhibit consistent drought trend patterns, with the southwestern region identified as the core area of drought intensification. During the historical period, drought severity gradually increased from northeast to southwest, with the most severe drought occurring in the southwestern region, where the maximum trend rate reached −0.87/a. Under the SSP1-2.6 scenario, drought conditions showed an overall amelioration, with notable wetting observed, particularly in northern regions (Figure 4c). In contrast, under the SSP2-4.5 scenario, drought exhibited an intensifying trend, with trend rates ranging from −0.23 to −0.39/a (Figure 4d). The SSP5-8.5 scenario displayed the most severe drought conditions, with trend rates concentrated between −0.41 and −0.39/a. Notably, drought intensity was observed to be higher in Mongolia than in Inner Mongolia, with drought intensity increasing from northeast to southwest.

Figure 5 illustrates the spatiotemporal variation in NPP across the Mongolian Plateau under the three emission scenarios. As shown in Figure 5a, NPP increases markedly across all scenarios, with the fastest growth under SSP5-8.5 at 25.8/10a. During the historical period, only a small section of the southwestern region exhibited slight declines in NPP (−0.1 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}}^{-1}$), while the remaining regions showed increasing trends, particularly the northeast, where productivity gains reached 2.8 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}}^{-1}$ (Figure 5b). Under SSP1-2.6, the entire region exhibited increasing NPP trends ranging from 0 to 1 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}}^{-1}$ (Figure 5c). With increasing emissions, NPP growth rates intensified, especially in Inner Mongolia (1.15 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}}^{-1}$) compared to Mongolia (0.92 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}}^{-1}$) (Figure 5d). Under SSP5-8.5, the most pronounced NPP increases were observed, with a peak of 4.7$\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}}^{-1}$ (Figure 5e). Overall, the spatial distribution of NPP exhibited significant heterogeneity, characterized by a southwest-to-northeast increasing gradient.

3.3. Future Changes in Drought Risk on the Mongolian Plateau

Drought probability ($P_d$) was determined using SPEI values below −0.5. Drought vulnerability (${NPP}_v$) was calculated as the difference between NPP in non-drought and drought years. The product of $P_d$ and ${NPP}_v$ was used to assess drought risk to NPP. Figure 6 shows the drought probability, NPP vulnerability, and potential drought risk to NPP over the Mongolian Plateau under the three emission scenarios.

Drought probability exhibits substantial spatial heterogeneity under different emission scenarios (Figure 6a–c). As shown in

Table 4. Migration distances and directions of the centers of gravity of drought occurrence probability, drought vulnerability of NPP, and drought risk on the Mongolian Plateau.

	Emission Scenarios	Distance of the Center of Gravity Shift from 2021 to 2100 (Km)	Direction of Migration (Azimuth)
Drought	SSP1-2.6	213.69	Southeast (175°)
probability	SSP2-4.5	1514.13	Southwest (245°)
	SSP5-8.5	1130.3	Southeast (140°)
Drought	SSP1-2.6	836.19	Southeast (117°)
vulnerability	SSP2-4.5	255.87	Southeast (166°)
	SSP5-8.5	653.39	Southeast (122°)
Drought	SSP1-2.6	1023	Southeast (131°)
risk	SSP2-4.5	426.22	Southwest (210°)
	SSP5-8.5	1677.20	Southeast (134°)

, under the low emission scenario (SSP1-2.6), drought probability distribution exhibits a gradient, with lower values in the northeast and higher values in the southwest; the spatial centroid of drought occurrence probability is located in Central Mongolia. Under the medium emission scenario (SSP2-4.5), drought probability increases significantly in the northwestern region, with the drought centroid migrating toward the west-southwest (245°), reaching a maximum distance of 1514.13 km. Under the high-emission scenario (SSP5-8.5), drought probability shows relatively uniform spatial distribution (0.33–0.37), and the drought centroid shifts southward from its initial northernmost position toward the central area, exhibiting an overall southeastward migration of 642.06 km. Notably, as emissions intensify, the mean drought probability rises from 0.32 (SSP1-2.6) to 0.35 (SSP5-8.5), with the greatest centroid migration distance observed under SSP2-4.5.

Figure 6d–f illustrate drought vulnerability and its centroid migration. Under SSP1-2.6, high vulnerability (5–8 g C m⁻² yr⁻¹) is concentrated in northern, central, and southern regions of the Mongolian Plateau, indicating potentially substantial annual NPP declines in these areas. In contrast, northeastern regions exhibit lower NPP drought vulnerability (2–10 g C m⁻² yr⁻¹), suggesting relatively minor drought impacts on NPP. The center of gravity shifts towards the central region over time. Overall, the center of gravity shifted southeastward (117°), with a displacement of 836.19 km (Figure 6d and Table 4). Under SSP2-4.5, 96% of the region displays drought vulnerability between −2 and 8 g C m⁻² yr⁻¹, with the centroid located in Eastern Mongolia. In contrast, under SSP5-8.5, drought vulnerability declines across 71% of the region, falling between −2 and 4 g C m⁻² yr⁻¹, and the centroid shifts southeastward (122°) by 653.39 km (Figure 6f and Table 4). In summary, the annual mean NPP drought vulnerability across the Mongolian Plateau ranks as follows: SSP2-4.5 (0.60 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}\mathrm{r}}^{-1}$) > SSP1-2.6 (−1.03 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}\mathrm{r}}^{-1}$) > SSP5-8.5 (−1.24 $\mathrm{g}\mathrm{C}{\mathrm{m}}^{-2}{\mathrm{y}\mathrm{r}}^{-1}$), with the centroid under all scenarios migrating southeastward.

Drought risk patterns (Figure 6g–i) show that under SSP1-2.6, the highest risk values (2.5–3.5) are concentrated in the central-southern region, with the lowest risks observed in the southeast (−5). The centroid is located centrally and migrates southeastward (Figure 6g). Under SSP2-4.5, 64% of the area exhibits risk values between 0 and 1.5, with the highest risk (2.5) in the southeast. From 2021 to 2100, the centroid shifts 462.22 km from Northeastern Mongolia to the central region (Figure 6h and Table 4). Under SSP5-8.5, the highest risk is observed in the central region and the lowest risk in the north. The centroid migrates from the northernmost part of Mongolia to the eastern part of Inner Mongolia during 2021–2060 (Figure 6i and Table 4). The mean annual risk values have the following order: SSP2-4.5 (0.20) > SSP1-2.6 (−0.33) > SSP5-8.5 (−0.43). Notably, SSP2-4.5 has the highest proportion of areas with positive changes (88.9%), indicating that NPP drought vulnerability contributes most significantly to regional drought risk under this scenario and explaining the higher risk levels compared to SSP5-8.5.

Figure 7 presents variation patterns of drought risk indicators across climate zones. As shown in Figure 7a, median values of drought occurrence probability increase across all zones with higher radiative forcing. Under SSP5-8.5, the semi-arid zone shows the highest probability. In humid zones, median drought probability exceeds that of semi-humid zones under SSP1-2.6 and SSP2-4.5, yet exhibits optimal stability under SSP5-8.5. Figure 7b illustrates drought vulnerability across climate zones. Under SSP2-4.5, all zones except arid regions reach peak vulnerability, while semi-arid zones show the greatest stability. Vulnerability is higher under SSP1-2.6 and SSP2-4.5 compared to SSP5-8.5 in the semi-arid to humid zones, suggesting that NPP is more susceptible to drought impacts under low- and medium-emission scenarios. Drought risk trends (Figure 7c) closely mirror vulnerability patterns. Under SSP2-4.5, risk increases from arid to humid zones, whereas under SSP5-8.5, risk demonstrates a decreasing gradient. In summary, while arid zones show the highest drought probability, humid and semi-humid zones exhibit the highest sensitivity and drought risk under SSP2-4.5.

3.4. Relative Changes in Future Drought Risk on the Mongolian Plateau

This study divided the period 2021–2100 into two distinct phases (2021–2060 and 2061–2100) to facilitate a comparative analysis of the spatiotemporal dynamics of drought probability, vulnerability, and risk evolution across the Mongolian Plateau. Regarding drought probability (Figure 8a–c), the SSP1–2.6 scenario exhibited clear regional shifts, with approximately 69% of the study area displaying an increasing trend relative to the earlier period. This increase was particularly pronounced in the southwestern desert regions. In contrast, the SSP2–4.5 and SSP5–8.5 scenarios revealed an emission-intensity-dependent amplification, with the SSP5–8.5 scenario showing the most substantial escalation in drought probability during the latter phase (2061–2100) relative to the early phase (2021–2060). Drought vulnerability exhibited a more intricate spatial distribution (Figure 8d–f). Under SSP1–2.6, approximately 23% of the Mongolian Plateau displayed negative vulnerability values, indicating that the impact of drought on net primary productivity (NPP) was greater during 2061–2100 than during 2021–2060. Under SSP2–4.5, highly vulnerable regions were concentrated in Eastern Mongolia and Central Inner Mongolia. In the SSP5–8.5 scenario, NPP vulnerability to drought appeared mitigated, with negative values spanning 42% of the area, primarily in the northeastern and northwestern regions. The drought risk analysis (Figure 8g–i) revealed a distinct contraction and spatial concentration of areas with positive risk values as emission intensity increased. Under SSP1-2.6, areas with elevated drought risk covered 74% of the study region and were broadly distributed. By contrast, in SSP5-8.5, these areas contracted to 32%, forming a spatially concentrated core in the central region. Collectively, these findings suggest that the potential risk of drought to vegetation productivity across the Mongolian Plateau may decline during 2061–2100 under higher emission scenarios.

3.5. Analysis of Dominant Factors Influencing Drought Vulnerability

Drought vulnerability constitutes a critical component of overall drought risk [33,56]. To elucidate the independent effects of various climatic variables on drought vulnerability, partial correlation analysis was employed to identify the dominant climatic drivers [57]. Figure 9 illustrates the spatial distribution of these drivers under the various SSP scenarios. In the SSP1–2.6 scenario, wind speed was the dominant controlling factor in the eastern region, accounting for 44% of the area, while precipitation dominated in 22.67% of the central region. Both increased wind speed and decreased precipitation have been shown to reduce NPP [58]. Under SSP2–4.5, drought vulnerability in the eastern region was jointly influenced by wind speed, temperature, and relative humidity, with the magnitude of their impact ranked as follows: wind speed > precipitation > relative humidity > temperature. In the SSP5–8.5 scenario, precipitation emerged as the dominant factor, influencing 45.34% of the study area, while the spatial influence of temperature increased significantly, covering 16.37%, a notable rise compared to the other scenarios. The overall ranking of climatic drivers in this scenario was as follows: precipitation > wind speed > temperature > relative humidity. In summary, the relationship between drought vulnerability and climatic factors on the Mongolian Plateau exhibits pronounced regional heterogeneity. Although certain climatic factors dominate in some regions, drought vulnerability is ultimately driven by the synergistic effects of multiple factors, highlighting the importance of considering compound influence mechanisms.

4. Discussion

4.1. Future Trends of Drought over the Mongolia Plateau

Drought conditions and vegetation dynamics across the Mongolian Plateau demonstrate pronounced spatial heterogeneity. Consistent with previous studies (e.g., Tong et al. (2018) [52] and Mei et al. (2024) [59]), our results indicate a persistent aridification trend during both the historical period and under all three future emission scenarios. The aridification trend exhibits distinct spatial heterogeneity, with significantly greater intensity in northern regions compared to southern areas. Additionally, drought severity is more pronounced in Mongolia compared to Inner Mongolia. These patterns are corroborated by Li et al. (2020) [51], who reported greater interannual variability in the SPEI under RCP8.5 than RCP4.5, supporting the findings of the present study that higher emission scenarios are associated with increasingly severe drought conditions. The intensification of drought under future scenarios is driven by complex and multifaceted mechanisms. Climate warming enhances evapotranspiration—a key component of the hydrological cycle—thereby accelerating aridification. Specifically, elevated temperatures, intensified desertification, and increased surface net radiation collectively raise potential evapotranspiration and reduce soil moisture availability, ultimately exacerbating drought severity [60]. With respect to vegetation productivity, NPP across the Mongolian Plateau shows a consistent upward trend during both the historical and future periods. Historical observations indicate that the increasing trend in NPP is particularly pronounced during the growing season and spring [61]. Among the future scenarios, the SSP5-8.5 pathway exhibits the most rapid increase in NPP, aligning with findings from global scale-studies. For example, Gang et al. (2017) [62] projected a continuous rise in global terrestrial NPP throughout the 21st century. Zhang et al. (2022) [19] reported a similar trend for the NPP of terrestrial ecosystems in China during the 21st century, implying a relatively low risk to future ecosystem productivity. Yin et al. (2022) [54] also observed sustained increases in NPP across most of the Mongolian Plateau. To accurately assess drought vulnerability, the NPP time series was subjected to linear detrending to eliminate the influence of long-term trends in the analysis, thereby isolating the effects of interannual climate variability on vegetation productivity.

4.2. Assessment of Drought Risk on the Mongolia Plateau

Previous studies have confirmed that drought risk varies substantially under different emission scenarios. Li et al. (2020) [51] noted that drought frequency is markedly higher under the RCP8.5 scenario than under RCP4.5, reinforcing projections of increased drought severity in the Mongolian Plateau in the future. This conclusion is highly consistent with the findings of the present study, which indicate that drought risk is greater under the medium emission SSP2-4.5 scenario than under the high emission SSP5-8.5 scenario. This may be due to the inability of the SSP2-4.5 scenario to reap the carbon dioxide benefits brought about by high emissions, thereby intensifying water stress in humidity-sensitive regions. Cao et al. (2022) [13] similarly noted that lower NPP losses under RCP8.5 might be attributed to the fact that the high future CO₂ concentration counteracted the negative anomaly of NPP. This observation aligns with the drought risk assessment in the present study. Furthermore, SSP2-4.5 may enhance vapor pressure deficit without sufficient CO₂ compensation, intensifying physiological stress in humidity-sensitive ecosystems.

NPP drought vulnerability is a complex phenomenon influenced by multiple factors beyond drought severity alone. Spatial heterogeneity in vegetation types and climatic zones also exerts substantial influence. For example, under SSP2-4.5, higher drought vulnerability and risk are observed in humid and semi-humid zones. These patterns likely reflect the synergistic effects of changes in temperature, precipitation, and solar radiation, all of which intensify the temperature sensitivity of ecosystems [63]. This finding is supported by Sun et al. (2022) [64], who demonstrated that variations in precipitation, temperature, growing season length, and soil temperature collectively regulate NPP dynamics. Drought vulnerability plays a decisive role in shaping the spatial configuration of drought risk [65,66]. To further clarify its climatic drivers, correlation analyses between drought vulnerability and climatic factors were conducted. Results show that, in the central Mongolian Plateau, drought vulnerability is positively correlated with precipitation (59%), while in the east, it is negatively correlated with wind speed (Figure 10a–d). This is consistent with findings from Li et al. (2023) [67], which showed that grasslands in Inner Mongolia are more sensitive to precipitation fluctuations than their counterparts in Mongolia. Under SSP2-4.5, drought vulnerability is positively correlated with relative humidity in the eastern and northwestern regions (22%), and negatively correlated with other climatic variables (Figure 10e–h). Under SSP5-8.5, vulnerability shows a positive correlation with both precipitation and wind speed, with wind speed exerting a particularly strong influence in the northern and western parts of the Mongolian Plateau. Temperature exhibits only a weak correlation with drought vulnerability in this scenario (Figure 10i–l). These results clearly indicate that the influence of a single climatic factor on drought vulnerability varies significantly across emission scenarios and geographic regions, highlighting the need for region-specific strategies in climate change impact assessments.

4.3. Limitations and Perspectives

This study has certain limitations. The use of the SPEI to characterize meteorological drought may not fully capture the characteristics of ecological drought, potentially affecting the accuracy of NPP response assessments. Future research should integrate additional indicators, such as soil moisture and plant physiological indicators, to develop more comprehensive ecological drought indices. Moreover, the drought threshold used in this analysis directly influences risk assessment outcomes and warrants further sensitivity analysis. The linear detrending of the NPP time series might oversimplify the actual response process of vegetation productivity to climate change. In addition, the projections based on CMIP6 models inherently contain certain uncertainties. From a methodological perspective, this study examined only the qualitative associations between drought vulnerability and climatic variables; future research should employ quantitative methods such as structural equation modeling to assess the relative importance of climatic drivers across different vegetation types and seasons. Integration of process-based models and machine learning approaches may further elucidate nonlinear interactions between climate and vegetation. Finally, the fusion of multi-source datasets should be strengthened to enhance the spatiotemporal precision of ecological drought risk projections. These advancements would contribute to the development of more accurate early warning systems for ecological droughts.

5. Conclusions

Based on CMIP6 global climate models, this study systematically assessed the spatiotemporal evolution of net primary productivity and drought dynamics across the Mongolian Plateau. A comprehensive analysis of regional variations in drought probability, vulnerability, and overall risk was conducted. The main conclusions are as follows:

(1) Across all three scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5), SPEI shows a persistent decline, indicating increasing drought severity, while NPP displays an overall increasing trend. A spatial gradient in drought severity is observed, intensifying from northeast to southwest, with Mongolia experiencing more severe drought conditions than Inner Mongolia.

(2) Drought risk exhibits marked spatial differences across emission scenarios. Under SSP1-2.6, the highest drought probability is found in the southwestern region. Under SSP2-4.5, the northwestern region becomes the primary high-risk area. Under SSP5-8.5, drought risk is most prominent in the southeastern part of the Mongolian Plateau. Notably, both drought vulnerability and overall risk are most severe under SSP2-4.5. Temporally, drought probability under SSP2-4.5 and SSP5-8.5 increases significantly during 2061–2100 relative to 2021–2060. However, overall drought risk declines with increasing emission levels.

(3) Drought probability increases with stronger radiative forcing, with the highest occurrence observed in arid regions. Under SSP2-4.5, drought vulnerability increases along a gradient from arid to humid zones, with the spatial pattern of drought risk closely mirroring that of vulnerability. Humid and semi-humid regions exhibit the highest sensitivity to drought events.

(4) The primary climatic factors driving drought vulnerability vary across emission scenarios. Under SSP1-2.6, wind speed dominates in the eastern region, while precipitation dominates in the central region. Under SSP2-4.5, drought vulnerability in the east is jointly influenced by wind speed, temperature, and relative humidity. Under SSP5-8.5, precipitation becomes the dominant controlling factor throughout the Mongolian Plateau.