Spatiotemporal Analysis of Drought Variation from 2001 to 2023 in the China–Mongolia–Russia Transboundary Heilongjiang River Basin Based on ITVDI

Abstract

Drought impacts agricultural production and regional sustainable development. Accordingly, timely and accurate drought monitoring is essential for ensuring food security in rain-fed agricultural regions. Alternating drought and flood events frequently occur in the Heilongjiang River Basin, the largest grain-producing area in Far East Asia. However, spatiotemporal variability in drought is not well understood, in part owing to the limitations of the traditional Temperature Vegetation Dryness Index (TVDI). In this study, an Improved Temperature Vegetation Dryness Index (ITVDI) was developed by incorporating Digital Elevation Model data to correct land surface temperatures and introducing a constraint line method to replace the traditional linear regression for fitting dry–wet boundaries. Based on MODIS (Moderate-resolution Imaging Spectroradiometer) normalized vegetation index and land surface temperature products, the Heilongjiang River Basin, a cross-border basin between China, Mongolia, and Russia, exhibited pronounced spatiotemporal variability in drought conditions of the growing season from 2001 to 2023. Drought severity demonstrated clear geographical zonation, with a higher intensity in the western region and lower intensity in the eastern region. The Mongolian Plateau and grasslands were identified as drought hotspots. The Far East Asia forest belt was relatively humid, with an overall lower drought risk. The central region exhibited variation in drought characteristics. From the perspective of cross-national differences, the drought severity distribution in Northeast China and Inner Mongolia exhibits marked spatial heterogeneity. In Mongolia, regional drought levels exhibited a notable trend toward homogenization, with a higher proportion of extreme drought than in other areas. The overall drought risk in the Russian part of the basin was relatively low. A trend analysis indicated a general pattern of drought alleviation in western regions and intensification in eastern areas. Most regions showed relatively stable patterns, with few areas exhibiting significant changes, mainly surrounding cities such as Qiqihar, Daqing, Harbin, Changchun, and Amur Oblast. Regions with aggravation accounted for 52.29% of the total study area, while regions showing slight alleviation account for 35.58%. This study provides a scientific basis and data infrastructure for drought monitoring in transboundary watersheds and for ensuring agricultural production security.

1. Introduction

Under global climate change, frequent extreme climatic events (e.g., droughts and floods) have exerted severe impacts, resulting in crop failures, ecosystem degradation, and socioeconomic challenges [1,2]. Drought stands as one of the most severe disasters impacting both ecosystems and socioeconomic development [3,4]. Long-term drought events have a direct impact on crop growth, food security, and regional sustainable development [5]. Accurately assessing regional agricultural drought intensity and its spatiotemporal variation is crucial. The Heilongjiang River Basin, a vital grain production region in Northeast Asia, holds a critical position in the global food supply system. As the region predominantly relies on rainfed agriculture, intensified extreme weather events in recent years have triggered frequent drought–flood alternations, posing severe threats to food security. Long-term drought monitoring in the Heilongjiang River Basin using multisource data and scientific research methodologies provides essential support for implementing drought prevention and disaster mitigation strategies as well as for future risk assessment and forecasting.

Drought monitoring and early warning are prominent research topics. Based on data acquisition methods and monitoring techniques, this research can be primarily classified into two categories. First are studies based on traditional ground monitoring stations. Common traditional drought monitoring indices include the Palmer Drought Severity Index (PDSI) [6,7,8,9], Self-Calibrated Palmer Drought Severity Index (scPDSI) [10,11,12], Standardized Precipitation Index (SPI) [13,14], Standardized Precipitation Evapotranspiration Index (SPEI) [15,16], and so on. These methods are advantageous owing to their extensive data sources and straightforward computations, enabling the assessment of drought conditions across various temporal scales; however, the PDSI has limitations in characterizing drought features at multiple scales [17]; the SPI is developed solely based on precipitation data and fails to comprehensively consider key environmental variables such as evapotranspiration [18]; although the SPEI incorporates a water balance model, it does not fully incorporate elements like soil moisture [19]. The second category involves the use of satellite remote sensing data for drought monitoring. Advances in remote sensing technology have enabled rapid monitoring, wide coverage, and high spatial resolution; therefore, this has gradually become the primary technical approach for drought monitoring supported by ecological or environmental factors. The Normalized Difference Vegetation Index (NDVI) [20] is sensitive to drought stress in vegetation; however, its performance is inadequate in areas with unevenly distributed or sparse vegetation. It exhibits saturation in dense vegetation zones and is susceptible to noise interference [21]. The Enhanced Vegetation Index (EVI) effectively mitigates the saturation phenomenon in high-density vegetation areas and reduces the effects of atmospheric aerosols and other variables [22,23], while partially addressing limitations of NDVI in vegetation monitoring applications. However, drought monitoring based only on vegetation indices may introduce substantial uncertainties and result in limited generalizability. Composite drought indices constructed from multiple individual remote sensing indices increase the accuracy of remote sensing drought monitoring substantially [24,25,26]. The Temperature Vegetation Dryness Index (TVDI) combines the Normalized Difference Vegetation Index (NDVI) and land surface temperature (LST), capturing the physical interaction dynamics between soil moisture distribution patterns and vegetation growth status; it has been widely applied in remote-sensing-based drought monitoring [20,27]. Numerous studies have demonstrated that the TVDI exhibits distinct advantages over other drought indices [28,29], including a superior ability to monitor the soil moisture content.

However, the TVDI monitoring model has some limitations. First, the traditional TVDI model does not effectively account for the influence of elevation on land surface temperature. In regions with complex climates and topography, altitude and latitude influence LST, which inevitably introduces errors into TVDI inversion [30,31]. For example, in the Heilongjiang River Basin, terrain undulations are significant, and the latitudinal span is substantial. These topographic and latitudinal differences can cause deviations in land surface temperature data, thereby affecting the accuracy of drought monitoring. Second, traditional linear fitting methods do not accurately capture the nonlinear relationship between LST and NDVI in complex situations, such as areas with bare soil or low vegetation cover and high surface evaporation. This limits their ability to comprehensively and accurately reflect the actual drought conditions. This study utilized MODIS NDVI and land surface temperature (LST) products spanning June to August (i.e., the crop growing season) from 2001 to 2023 supported by DEM data correction, and a constraint line method to improve the drought monitoring index, resulting in the development of the Improved Temperature Vegetation Dryness Index (ITVDI). This newly developed ITVDI was used to systematically explore the spatiotemporal evolution of summer droughts in the Heilongjiang River Basin.

2. Data and Methods

2.1. Study Area

The China–Mongolia–Russia transboundary Heilongjiang River (known as the Amur River in Russia) Basin (Figure 1) refers to the watershed area of the Heilongjiang and its tributaries, spanning China, Mongolia, and Russia. It is located between 107°31′E–141°14′E and 41°42′N–55°56′N. The Heilongjiang River Basin features diverse geomorphological types, including the Greater Khingan Range, Lesser Khingan Range, and Changbai Mountains, as well as the Mongolian Plateau, Songnen Plain, Sanjiang Plain, and lakes such as Lake Xingkai and Lake Baikal [32].

2.2. Data Sources

The NDVI and LST data for 2001–2023 were sourced from NASA’s MODIS Normalized Difference Vegetation Index product (MOD13A1) and land surface temperature product (MOD11A2), with temporal resolutions of 16 days and 1 day, respectively, and a spatial resolution of 1 km for both products. To perform pixel-wise correction on LST products and ensure consistent spatial resolution, DEM data were sourced from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model dataset, with a spatial resolution of 30 m, resampled to a resolution of 1 km. Precipitation data were sourced from the National Centers for Environmental Information (NCEI) under the National Oceanic and Atmospheric Administration (NOAA) of the United States.

2.3. Methods

DEM data were used to correct land surface temperature, and a constraint line method was introduced to improve the drought monitoring index, resulting in the development of the ITVDI. Based on MODIS NDVI and LST products, this study systematically investigated the spatiotemporal patterns of drought in the China–Mongolia–Russia transboundary Heilongjiang River Basin. The technical route included data preprocessing, constraint line fitting, constructing and visualizing the ITVDI, and analyses (Figure 2).

2.3.1. DEM-Based Correction of Surface Temperatures

Due to the impacts of air temperature and atmospheric turbulence, land surface temperature exhibits significant variation across different elevations [33,34]. The topographic variation and latitudinal differences within the study area can affect the MODIS land surface temperature data, leading to errors in the calculation of the TVDI. Therefore, it is necessary to obtain the corrected land surface temperature (T_H) as follows [35]:

$$T_H=Ts+a\times H+b\times L+c$$

(1)

where T_s is the uncorrected land surface temperature, H is elevation, L is latitude, a is the elevation correction coefficient, and b and c are the latitude correction coefficients. The model quantifies the contribution of “latitude-dominated solar radiation heat differences” to LST by introducing the latitude factor L, enabling the corrected land surface temperature to focus more on thermal anomalies induced by local water stress.

2.3.2. Constraint Line Method for Fitting Boundaries

The constraint line method [36] is an effective approach for accurately characterizing the boundary patterns of data and is used to extract the boundaries of scatter point clouds. The Segmented Quantile Approach [37] is a constraint line extraction method integrating logical slicing and quantile regression. Through the workflow of interval-based statistics, high quantile point extraction, and multi-model fitting optimization, it can better fit the nonlinear relationships in the boundaries of scatter plots. Compared with traditional linear fitting, its core advantage lies in balancing statistical rigor and ecological explanatory power. This study used the segmented quantile approach to extract the dry–wet boundary. The limiting variable NDVI was uniformly divided into 100 parts according to the range of values; to eliminate the effect of outliers, the points on the 99.95% or 0.05% quantiles were selected as the boundary points for dry and wet edge fitting, the wet and dry boundary points were then fitted to obtain the constraint line.

$$T_{\min }^{\prime }=g_1(NDVI)$$

(2)

$$T_{\max }^{\prime }=g_2(NDVI)$$

(3)

where g₁ is the wet-side fitting equation and g₂ is the dry-side fitting equation.

2.3.3. Improved Temperature Vegetation Drought Index (ITVDI)

The TVDI is a soil moisture monitoring method based on the vegetation index–land surface temperature feature space [38]. It can overcome the limitations of using vegetation indices alone for drought monitoring, including latency, or the drawbacks of using land surface temperature alone, including susceptibility to soil background temperature effects. The formula is as follows:

$$TVDI={\displaystyle \frac{T-T_{\min }}{T_{\max }-T_{\min }}}$$

(4)

$$T_{\min }=a\times NDVI+b$$

(5)

$$T_{\max }=c\times NDVI+d$$

(6)

where T is the pixel value of the land surface temperature product, and T_min and T_max are the minimum and maximum land surface temperatures corresponding to the NDVI in the generalized feature space. Coefficients a and e represent the slope and intercept of the wet edge, while b and d denote those of the dry edge in the NDVI-LST feature space, respectively.

In the NDVI-LST feature space, dry and wet edges are defined as straight lines related to NDVI. However, when bare soil or vegetation cover is low and land surface evapotranspiration is high, the traditional linear fitting cannot capture the nonlinear relationship between LST and NDVI [39]. Therefore, this study introduced the constraint line method to represent the nonlinear dry and wet edges (Figure 3).

$$ITVDI={\displaystyle \frac{T-T_{\min }^{\prime }}{T_{\max }^{\prime }-T_{\min }^{\prime }}}$$

(7)

where ITVDI is an improved TVDI that employs a constraint line, with values ranging from 0 to 1. Values approaching 1 indicate more severe drought, whereas values near 0 indicate wetter conditions. The ITVDI is used as a classification indicator to divide drought into five levels [40] (

Table 1. ITVDI drought classification criteria.

ITVDI Value Range	Soil/Vegetation Conditions	Degree of Drought
(0, 0.46)	Land surface wetting	Non-drought
[0.46, 0.57)	Normal soil moisture or dry air near land surface	Mild drought
[0.57, 0.76)	Soil surface drying, leaves drying and yellowing due to lack of water	Moderate drought
[0.76, 0.86)	Soil shows a dry layer, dry yellow leaves	Severe drought
[0.86, 1)	Drying out and death of surface plants	Extreme drought

).

2.3.4. Correlation Analysis

Using precipitation data from NOAA meteorological stations, the Pearson correlation coefficient method was used. A correlation analysis between drought intensity derived from both methods (TVDI and ITVDI) was used to validate the applicability of the improved approach in the study area.

2.3.5. Sen–MK Trend Analysis

The Theil–Sen median slope estimation reduces the effects of missing data, data outliers, and data distribution patterns in the analysis process, enabling more accurate trend estimation. The Mann–Kendall (M-K) trend test is a non-parametric method that does not require the data to follow a specific distribution and is not affected by missing values or outliers [41]. The two methods are often used in combination to assess trends in data characteristics and to estimate the significance of the trends. They are widely used in long-term drought time-series analyses [42,43,44].

$$\beta =Median\left({\displaystyle \frac{ITVD{I}_j-ITVD{I}_i}{j-i}}\right),\hspace{1em}1\le i<j\le n$$

(8)

where ITVDI_i and ITVDI_j are the ITVDI of a given pixel at moments i and j, respectively. Median indicates that the median is taken. β > 0 indicates an increasing trend in the ITVDI, β < 0 indicates a decreasing trend in the ITVDI, and β = 0 indicates a stabilizing trend in the ITVDI.

$$\mathrm{S}=\sum _{i=1}^{n-1}\sum _{j=i+1}^n\mathrm{sgn}\left(ITVD{I}_j-ITVD{I}_i\right)$$

(9)

where S is the defined test statistic, n denotes the length of the time series, and sgn is the sign function, which is denoted as follows:

$$\mathrm{sgn}(\theta )=\left\{\begin{array}{c}1(\theta >0)\\ 0(\theta =0)\\ -1(\theta <0)\end{array}\right.,\theta =ITVD{I}_j-ITVD{I}_i$$

(10)

When n ≥ 8, the test statistic S is approximately normally distributed, and S is standardized:

$$\mathrm{Z}=\left\{\begin{array}{c}{\displaystyle \frac{\mathrm{S}-1}{\sqrt{\mathrm{Var}(\mathrm{S})}}}(\mathrm{S}>0)\\ 0(\mathrm{S}=0)\\ {\displaystyle \frac{\mathrm{S}+1}{\sqrt{\mathrm{Var}(\mathrm{S})}}}(\mathrm{S}<0)\end{array}\right.$$

(11)

$$\mathrm{Var}(\mathrm{S})={\displaystyle \frac{n(n-1)(2n-5)}{18}}$$

(12)

where Z is the statistic of the standardized M-K trend test, obeying a standard normal distribution. |Z| > μ_1−α/2 indicates a statistically significant trend in the time series at the α significance level; μ_1−α/2 is the standard normal distribution, and α is the significance test level.

3. Results

3.1. Spatial and Temporal Distribution Pattern of Drought

The ITVDI of the Heilongjiang River Basin from 2001 to 2023 was divided according to the drought class classification criteria (Table 1) to evaluate the spatial distribution of the drought classes (Figure 4). The drought levels in the Heilongjiang River Basin tended to be high in the west and low in the east. Extreme drought conditions were primarily concentrated in the western region of the Heilongjiang River Basin (including parts of Mongolia). The eastern part of the Heilongjiang River Basin (Jewish Autonomous Oblast, Khabarovsk, Primorsky Krai, etc.) was wetter, with drought levels mainly characterized as non-drought or low drought risk. The central part of the Heilongjiang River Basin (Heilongjiang Province, Jilin Province, Amur Oblast, etc.) was characterized by varying degrees of drought.

3.2. Characterization of the Drought Distribution at the Regional Scale

To intuitively reflect the spatiotemporal distribution characteristics of drought in the Heilongjiang River Basin, according to the provincial administrative units, the area share of drought levels in 16 provincial administrative regions of the Heilongjiang River Basin for 2001–2023 was obtained (Figure 5). There was significant spatiotemporal variability in drought within the basin. Extreme drought exhibited regional differentiation. In Dornod and Govisümber provinces of Mongolia, the proportion of area affected by extreme drought remained consistently above 95% over an extended period (e.g., both reached 100% in 2009), suggesting that the drought state was persistent. Extreme drought in China’s Heilongjiang province declined from its peak in 2013 (9234 km²) to 620 km² in 2023, reflecting drought mitigation dynamics. Furthermore, the drought severity showed obvious geographic variation. The Mongolian Plateau and adjacent grassland regions (e.g., Inner Mongolia and Zabaykalsky Krai) have been identified as drought hotspots, where the annual average extreme drought area exceeds 20,000 km². These regions exhibit ecosystem vulnerability and high precipitation variability, primarily driven by the synergistic effects of low-precipitation and high-variability climate patterns and degradation of grassland water-retention functions. Atmospheric circulation anomalies and North Atlantic sea surface temperature forcing induce unstable water supply, while the grassland soil system further compromises water retention capacity. The forest belt of the Russian Far East (e.g., Khabarovsk and Primorsky Krai) is predominantly characterized by non-drought class (comprising over 90%), reflecting its strong drought resistance derived from the regulatory roles of a humid climate and vegetation cover. This resistance stems from the interaction between a stable humid climate and forest regulatory functions. The coniferous–broadleaved mixed forests form an efficient “sponge system” through canopy interception, litter water retention, and deep soil regulation. They also promote local water circulation via transpiration, thereby enhancing drought-resilient capacity. Regarding drought class structures, extreme-drought-dominated areas (Dornogovi and Sühbaatar provinces of Mongolia) contrasted with the multi-class mixed areas (Heilongjiang and Jilin provinces of China). The latter was dominated by mild–moderate drought (e.g., moderate drought accounted for 51.0% in Jilin Province, China in 2023), with significant interannual variability (extreme drought accounted for 9.3% in Jilin Province, China in 2013). An analysis of anomalous years shows that the area of extreme drought in Inner Mongolia of China increased abruptly to 63,743 km² in 2017.

A comparative analysis across countries showed that the drought patterns in China, Mongolia, and Russia are significantly differentiated: The distribution of drought severity in Northeast China and Inner Mongolia showed significant spatial heterogeneity. Extreme drought events frequently occurred in Inner Mongolia, where the area of extreme drought reached 63,743 km² in 2017 (accounting for 20.8% of the regional area), posing a serious threat to the stability of grassland ecosystems. Heilongjiang Province was dominated by mild–moderate drought (the area of mild drought was 188,350 km², accounting for 42.3% in 2023). Certain years saw a sudden increase in extreme drought (e.g., the area of extreme drought was 9234 km² in 2013), posing a direct threat to yields in major agricultural-producing areas. Jilin Province had an annual proportion of mild–moderate drought area exceeding 70% (e.g., moderate drought covered 80,649 km², accounting for 62.3% in 2021); persistent water stress needs to be prevented to avoid crop yield reduction. Mongolia exhibited a significant trend toward homogenization in drought severity levels across regions, with extreme drought proportions substantially higher than those of other regions. In Dornogovi Province, the average annual proportion of extreme drought area exceeded 95% from 2001 to 2023, with values for non-drought areas consistently at zero. In Sühbaatar Province, the area of extreme drought reached 4042 km² in 2011 (accounting for 67.1% of the regional area). In Töv Province, the area of extreme drought increased from 1547 km² in 2013 to 4640 km² in 2020, triggering a decline in livestock inventory, reflecting the sustained erosion of rangeland productivity due to increasing drought variability. The Russian Far East had an overall low drought risk, though local moderate-severe droughts posed potential threats to the stability of agricultural exports. In Khabarovsk, the proportion of non-drought areas annually exceeded 90% (317,846 km², accounting for 99.4% in 2015), and the area of moderate drought increased suddenly to 6312 km² in 2017. In Primorsky Krai, the area of mild drought was 33,094 km² (33.0%) in 2020. In Amur Oblast, the area of severe drought reached 7576 km² in 2012.

3.3. Analysis of Drought Trends in the Heilongjiang River Basin

Based on a time-series per-pixel drought trend analysis, Theil–Sen median slope estimation was performed on summer ITVDI values (2001–2023) for each pixel in the Heilongjiang River Basin to obtain the spatial distribution of trend slope β. The Mann–Kendall test was used to determine the significance of the spatial distribution in ITVDI changes. The combined results of slope estimation and trend testing were classified into five drought change types: significantly aggravated, slightly aggravated, relatively stable, slightly alleviated, and significantly alleviated (Figure 6). The area ratio was calculated for each severity category.

As shown in Figure 6, drought in the Heilongjiang River Basin exhibited a general trend characterized by alleviation in the west and aggravation in the east. In the Heilongjiang River Basin from 2001 to 2023, more than half of the areas showed drought aggravation, with the main type being slightly aggravated. The area of this change type accounted for 52.29% of the total study area, while areas with slightly alleviated drought accounted for 35.58%. Areas with significantly aggravated drought were mainly concentrated in the Chinese part of the Heilongjiang River Basin, surrounding cities, such as Qiqihar, Daqing, Harbin, Changchun, Baicheng, and Suihua. In addition, some areas showed significantly aggravated drought trends in Amur Oblast, Northern Khabarovsk, Southern Primorsky Krai, and Western Zabaykalsky Krai.

4. Discussion

4.1. ITVDI Suitability Evaluation

The development of indices in the field of drought monitoring shows a trend of multisource data integration and deepening of physical mechanisms. Traditional drought indices have contextual limitations in parsing drought information in basins with complex underlying surfaces due to data dependency characteristics, model assumption premises, or differences in regional adaptability. A comparison between the ITVDI and mainstream drought indices is conducted (

Table 2. Analysis of advantages and disadvantages of different drought indices.

Index		Advantages	Limitations	Characteristics
Drought index based on meteorological stations	PDSI	Reflects soil moisture balance, superior for agricultural drought assessment.	Computationally intensive; slow response; poor regional portability.	Sparse spatial distribution, limited coverage, low timeliness.
	scPDSI	Self-calibrating to climatic conditions; automated correction via station data.	Soil parameters are difficult to obtain; lacks real-time capability.
	SPI	Simple computation; applicable across multiple timescales.	Ignore the evaporation effect; lag in drought response.
	SPEI	Improves SPI by incorporating evapotranspiration stress.	Relies on empirical formulas for evapotranspiration; limited applicability in alpine zones.
Drought index based on remote sensing data	VCI	Directly reflects physiological vegetation stress; highly suitable for agriculture.	Sensitive to phenology and crop-type variations.	Rapid monitoring, high spatiotemporal resolution, broad coverage.
	EVI	High sensitivity to vegetation dynamics; widely applicable.	The calculation is complex and relies on multi-band data; data availability and continuity are limited.
	TVDI	Integrates vegetation–temperature data; clear physical basis; scalable for large areas.	Distorted in complex terrain; affected by mixed pixels.
	ITVDI	Combines topographic correction and nonlinear fitting; suited for heterogeneous surfaces.	The statistical properties of the mean value may smooth out extreme fluctuations at short time scales, failing to identify abrupt changes in moisture conditions at decadal or weekly scales.

) to clarify the complementary advantages and methodological limitations of different indices. This study used DEM data to perform topographic correction of surface temperatures and introduced a constraint line method to replace the traditional linear fitting of dry and wet boundaries to construct the ITVDI for accurate drought monitoring. Compared with the traditional TVDI, the ITVDI was able to better reflect the drought conditions in the study area, demonstrating the advantages of the constraint line method in fitting wet and dry boundaries.

Correlation analysis between TVDI/ITVDI and precipitation showed that both indices exhibited significant negative correlations with precipitation (p < 0.05), indicating that lower precipitation corresponded to higher drought intensity. The correlation between the ITVDI and precipitation was significantly higher than that between the TVDI and precipitation, supporting the improved monitoring accuracy and further indicating that the ITVDI is a reasonable indicator for drought assessment in the Heilongjiang River Basin. The optimized results with the most significant correlation are presented (Figure 7). The most pronounced optimization results were obtained for 2018; the correlation coefficient between the ITVDI and precipitation for this year was −0.512, compared with a correlation coefficient between the TVDI with precipitation of −0.083 in the same year (i.e., the absolute value of the difference was 0.429).

The ITVDI, as a vegetation–temperature coupling indicator, is constructed within the NDVI-LST feature space. The potential errors of the ITVDI arise from the combined effects of NDVI being influenced by topography, mismatch in ecological process responses, and coupling deviations in the model structure [45]. Meanwhile, although the linear correction model for LST removes large-scale thermal differentiation through H and L, its limitations stem from accuracy and scale defects of data sources, a lack of adaptation to underlying surface spatiotemporal heterogeneity, and incomplete characterization of topographic factors [46,47]. These may transmit errors to the ITVDI.

4.2. Comparison of Dry and Wet Edge Fitting Results

The traditional linear fitting method and the improved constraint line method were compared for fitting the dry and wet boundaries of scattered plot clouds. Compared with the improved constraint line method, traditional linear fitting of dry–wet boundaries was less effective. The constraint line method was more conducive to extracting scatter point clouds boundaries, with better fitting performance (Figure 8). The improved constraint line method for fitting wet and dry boundaries showed higher fitting coefficients (R²).

4.3. Analysis of Extreme Weather Events

This study revealed significant spatiotemporal variability in drought conditions within the Heilongjiang River Basin from 2001 to 2023. The proportion of area with non-drought conditions in the Heilongjiang River Basin reached a peak in 2016; these non-drought areas were primarily distributed in the eastern part of the Heilongjiang River Basin. According to the China Meteorological Disaster Yearbook and news media reports, the 2016 flood disaster in the Heilongjiang River Basin was a severe natural disaster that occurred in the summer. This flooding led to severe inundation in many places, causing farmland waterlogging and the collapse of houses. It had a serious impact on production and on the lives of local residents. According to reports from the Heilongjiang Provincial Department of Civil Affairs, over 4500 people in Shangzhi City (Harbin), Longfeng District (Daqing), and Huachuan County (Jiamusi) were affected. The affected crop area was nearly 400 hectares, with nearly 100 hectares exhibiting total crop failure and direct economic losses exceeding CNY 3 million. By contrast, a severe drought occurred in the Heilongjiang River Basin in the summer of 2017, with proportions of area with extreme and severe droughts in this year reaching their highest levels. These droughts were primarily distributed in regions such as Mongolia and Inner Mongolia. According to reports from the Inner Mongolia Meteorological Service, in the summer of 2017, most regions across Inner Mongolia were affected by lower-than-normal precipitation and higher-than-normal temperatures, and meteorological drought was persistent and intensified. Precipitation was below normal in early June, with central and eastern regions being significantly affected by drought. Severe or higher-level meteorological drought occurred in most parts of Hulunbuir City, Hinggan League, Tongliao City, and Xilingol League. This drought not only posed a significant threat to agricultural production but also exacerbated water scarcity issues.

Drought severity levels in parts of the Heilongjiang River Basin were primarily characterized by non-drought; however, there was a potential risk of an abrupt transition from drought to flooding. The ITVDI was constructed based on the feature space of the NDVI and LST. Although the index can comprehensively reflect soil moisture changes, the statistical properties of the mean value may smooth out extreme fluctuations at short time scales, failing to identify abrupt changes in moisture conditions at decadal or weekly scales. The monsoon climate characteristics of the Heilongjiang River Basin lead to highly uneven precipitation distributions, with precipitation variability becoming a key driving factor for abrupt transitions from drought to flood. Liu et al. [48] pointed out that May–June in the Songhua River Basin is a high-incidence period for abrupt transitions from drought to flood. The frequency of drought-to-flood transitions in Heihe, Nenjiang, and other regions is significantly higher than that in other areas. These spatiotemporal distribution characteristics are complementary to the mean statistics of the ITVDI; even if no drought was dominant during the study period, sudden changes in precipitation may trigger a sharp transition to flooding or drought in a short period of time. The Songhua River Basin is widely affected by abrupt drought–flood transition events and shows a significant upward trend. On average, about one-third of the spatial extent is affected by sharp drought and flood events each year [49]. Ecological vulnerability is further amplified during the alternation between drought and flood. The compound effects of abrupt drought–flood transitions far exceed the impacts of single drought or flood events. In-depth analyses of the formation mechanisms and impacts of extreme events can provide an important reference for basin drought research and regional sustainable development.

4.4. Recommendations and Regulatory Strategies

The Heilongjiang River Basin, as an important agricultural region, also serves as a crucial safeguard area for food security. Drought in the basin has far-reaching impacts on agricultural production. In severe cases, it can even lead to complete crop failure, threatening food security and hindering the sustainable development of the agricultural economy. Thus, it is necessary to address the spatiotemporal differentiation characteristics of drought and its potential impacts on the agroecosystem in the China–Mongolia–Russia transboundary Heilongjiang River Basin. From the perspective of zoned management and transboundary collaboration, the following strategies are proposed to safeguard regional food security and ecological sustainability.

(1)

In the Heilongjiang River Basin Region of China, technology-driven and food security guarantees should be implemented. There was significant spatial heterogeneity in drought severity across Northeastern China and Inner Mongolia. As a core region for national food production, this area should prioritize technological innovation as the core approach to address drought risks. Strengthen capacity building for agricultural disaster mitigation and prevention in the black soil region. Continuously improve the construction level of farmland water conservancy facilities and disaster resistance capacity. Establish a precision agricultural meteorological service system covering the entire growth cycle from sowing to harvesting. Reduce climate disaster risks in agricultural production. In response to the ecological vulnerability of the Inner Mongolian grassland region, accelerate the genetic improvement and large-scale cultivation of drought-tolerant forage varieties. Construct a synergistic grass–livestock–water resilient ecosystem.

(2)

In the Heilongjiang River Basin Region of Mongolia, efforts should focus on ecological restoration and sustainable pasture management. In the grassland region of Mongolia, extreme drought occurs frequently, with scarce non-drought areas. Ecological restoration is necessary as the core to stabilize the foundation of animal husbandry. It is recommended to designate seasonal grazing ban areas and rotational grazing demonstration areas. Implement natural vegetation restoration and overseeding of drought-resistant grass species to enhance grassland productivity. Construct decentralized rainwater harvesting facilities in settlements of pastoral households, alleviating the shortage of drinking water for livestock. Establish a drought insurance mechanism for animal husbandry. Enhance climate adaptation training for pastoralist communities. Promote grass–animal balance management techniques. Promote the transition from traditional nomadic pastoralism to sustainable intensification.

(3)

In the Heilongjiang River Basin Region of Russia, resource optimization and cross-border data collaboration are necessary. The drought risk in the Russian Far East is relatively low, and local moderate-to-severe droughts may impact agricultural exports. It is recommended to optimize crop allocation in agricultural zones, such as Amur Oblast and Primorsky Krai. Reduce the proportion of water-intensive crops grown. Priority should be given to the development of drought-tolerant crops (buckwheat and oats) and water-saving dry farming. In the area of cross-border cooperation, establish a shared China–Mongolia–Russia drought data platform, and jointly develop cross-border river basin drought prediction models. Drought management in cross-border river basins requires breaking through administrative barriers. Construct a “region-specific strategies–collaborative response” framework, and provide scientific support for synergistic regional drought responses.

5. Conclusions

The transboundary Heilongjiang River Basin faces long-term challenges from frequent agricultural droughts and floods. This study used DEM data to correct LST and integrated the constraint line method to improve the TVDI, with the improved index being referred to as the ITVDI. The MODIS normalized vegetation index product (MOD13A1) and land surface temperature product (MOD11A2) were used to evaluate spatiotemporal variation in drought in the Heilongjiang River Basin during summer (June-August) from 2001 to 2023. There was substantial spatiotemporal variation in drought in the Heilongjiang River Basin. Drought grade exhibited obvious geographical zonation, following a differentiation pattern of being high in the west and low in the east. The western region, Mongolian Plateau, and grassland zones (such as Inner Mongolia and Zabaykalsky Krai) were identified as drought hotspots. The eastern region, including the Far East Asia forest belt (Jewish Autonomous Oblast, Khabarovsk, and Primorsky Krai), was relatively humid, with an overall lower drought risk. The central region (Heilongjiang Province, Jilin Province, Amur Oblast, etc.) exhibited variation in drought characteristics. From the perspective of cross-national differences, drought patterns in China, Mongolia, and Russia were distinct. Drought severity in Northeast China and Inner Mongolia exhibited marked spatial heterogeneity. In Mongolia, regional drought levels exhibited a notable trend toward homogenization, with a higher proportion of areas classified as extreme drought than in other areas. In contrast, the overall drought risk in the Russian Far East was relatively low, with the potential for extreme drought and flooding events. A Sen–MK trend analysis showed the drought changes in the Heilongjiang River Basin as a whole, showing a trend toward mitigation in the west and intensification in the east. Most areas of change were relatively stable, with few areas showing significant change. Areas with slight drought intensification accounted for 52.29% of the total study area, while those with slight drought mitigation accounted for 35.58%. Areas with significant drought intensification were mainly concentrated around some cities in the Chinese part of the Heilongjiang River Basin and some regions in Russia. Against the background of climate change, the Heilongjiang River Basin needs to coordinate regional governance and improve cross-border collaboration, building a drought-adaptive system through technological innovation-driven approaches, ecological resilience restoration, and optimized resource allocation. These findings provide a scientific basis for food security and ecologically sustainable development within the river basin.