Comprehensive risk assessment of drought disasters from the perspective of multi-source geospatial big data: evidence from China's grain production bases Open Access

Drought is a persistent phenomenon of water shortage caused by the disruption of the balance between water receipts and expenditures or supply and demand, evolving into a drought disaster when it reaches a certain level and causes adverse social or ecological impacts (Nongsheng 2012). Globally, among various types of natural disasters, drought disasters occur with the highest frequency, affect the widest areas, last the longest, and result in the greatest economic losses, making them one of the most severe natural disasters worldwide (Vishwakarma & Goswami 2022; Pei et al. 2023). Approximately 70% of the economic losses from global natural disasters are due to meteorological events, with droughts contributing about half of these losses (Kaur et al. 2022; Zhan et al. 2023). Since the 20th century, the global area affected by droughts has been expanding at a rate of approximately 1.74% per decade (Lu et al. 2017). The increasing disparity between water supply and demand has led to significant economic losses, amounting to about $6–8 billion annually, posing a serious threat to agricultural production, ecological environments, and economic development (Lu et al. 2017; Li et al. 2019). To scientifically describe droughts and support various sectors of society, droughts are commonly categorized into meteorological, hydrological, agricultural, and socio-economic types. Droughts have complex internal mechanisms, and their occurrence and development are closely linked to factors such as the natural environment and human activities. The integration of multiple factors related to drought is essential for developing a comprehensive index that effectively represents meteorological, hydrological, and agricultural droughts. Integrating multiple drought-related factors is critical for developing a comprehensive index that effectively reflects meteorological, hydrological, and agricultural droughts. Quantitative research on drought disasters is essential for formulating effective global strategies to prevent and mitigate disaster impacts, particularly in regions like China, where drought poses a significant threat to agriculture and socio-economic stability.

Like other natural disasters, droughts combine natural and social attributes, resulting from the combined effects of disaster-causing factors, disaster-inducing environments, and disaster-bearing bodies (Yang et al. 2023). At present, the international community has paid great attention to the disaster risk caused by climate change and actively promotes a global shift from post-disaster response to comprehensive disaster risk prevention. Natural disasters are the products of the combined effects of disaster-causing factors, disaster-conceiving environments, and disaster-bearing bodies on the Earth's surface. A drought disaster system was constructed by combining the theory of the natural disaster system. The disaster-causing factor is the main trigger of drought, the disaster-conceiving environment is the contact surface between the disaster-causing factor and the disaster-bearing body, and it is the background environment for the occurrence of drought. The degree of exposure of the disaster-bearing body is the necessary factor for drought to evolve into a drought disaster, which generally means that drought is formed when the disaster-bearing body is damaged. The greater the hazard, vulnerability, and exposure, the greater the risk of drought disaster; conversely, the smaller the three, the lower the risk of drought disaster (Li et al. 2015). In recent years, machine learning and deep-learning algorithms have increasingly been applied by scholars to the field of drought disaster risk assessment. These models are advantageous in processing large datasets and capturing the complex nonlinear relationships among drought-related factors. For instance, Zhang et al. (2023) and Safwan et al. (2024) have demonstrated the effectiveness of machine-learning algorithms, such as random forest and XGBoost models, in enhancing the accuracy of drought risk prediction. By integrating these advanced AI methods with traditional statistical techniques, researchers can develop more accurate spatial distribution models of drought risk and more precisely predict future trends.

Scholars both domestically and internationally have conducted extensive research on drought disaster risk assessment and zoning. Based on natural disaster risk theory, Li et al. (2021) constructed a drought disaster risk assessment model to evaluate Shanxi Province, noting that the region's drought risk is generally characterized by a ‘high in the north and low in the south’ distribution; Zhao et al. (2012) established a drought disaster risk evaluation model using an entropy weight combination method and a weighted comprehensive analysis method to assess the drought disaster risk in northern Henan Province. Han et al. (2016) constructed a drought disaster risk model based on hierarchical analysis, combining precipitation, vegetation greenness, river network density, and soil indicators, and applied it to analyze the risk and spatial distribution of drought disasters in Southwest China. The results showed that the risk in the eastern part of the study area was extremely high, with the northern part being generally at greater risk than the southern part, and the risk increasing from the southwest to the northeast. However, most assessments are conducted only for specific years, and few studies have evaluated inter-annual variations in drought disaster risk.

Liaoning Province is one of the main grain-producing areas in China, characterized by a temperate continental monsoon climate and high susceptibility to drought disasters due to climatic conditions and geographic environment (Cao et al. 2021). As one of China's grain production bases, Liaoning Province's vulnerability to drought poses significant risks to its agricultural output and overall economic stability. Drought has a complex internal mechanism, with its occurrence and development closely related to factors such as the natural environment and human activities (Adjah et al. 2022). Previous indicators for drought disaster risk assessment have included precipitation, temperature, and evaporation. However, these indicators do not adequately describe the extent of drought disaster. The aridity index has been used to quantify drought, and the combination of the aridity index and run theory can determine factors such as the duration, severity, and peak value of drought (Wang et al. 2019; Zuo et al. 2019). Additionally, many researchers have demonstrated that a drought index combining multiple elements provides better monitoring of drought (Vyas et al. 2015; Bayissa et al. 2019; Cheng et al. 2023).

Bayissa et al. (2019) used principal component analysis (PCA) to construct a comprehensive drought index (CDI-E) based on rainfall, land surface temperature, soil moisture, and normalized difference vegetation index (NDVI). However, the interaction between drought-related variables is highly complex, and although the drought index constructed using the weighting method is simple and fast, its results are highly dependent on the weighting coefficients, introducing a level of subjectivity. Moreover, PCA often assumes a linear relationship between drought variables and is therefore unable to capture their nonlinear impact characteristics (Suo et al. 2024)). Thus, better methods are required to reveal the inherent relationships between the different related variables in drought events. The Copula function can construct multivariate joint distribution functions with different marginal distributions, objectively reflecting the nonlinear characteristics of variables, and many scholars have applied it to the construction of drought monitoring models (Cheng et al. 2023). Kavianpour et al. (2018) combined the standardized precipitation index (SPI) and the standardized discharge index (SDI) using the Copula function to construct a multivariate drought index, evaluating it as more accurate than SPI and SDI.

While similar approaches have been employed, the novelty of this study lies in its integration of multi-year geospatial datasets and its focus on a region that has not been extensively studied in terms of comprehensive drought risk. Furthermore, the use of the Copula function to model the joint distribution of NDVI, precipitation, and temperature offers a more robust framework for understanding drought occurrence and severity over time. This method provides a more precise assessment of drought severity and duration, especially in agricultural regions such as Liaoning Province, where vegetation response plays a critical role in the impact of drought. Therefore, this paper constructs a drought index (CDI) based on the Copula function, integrating precipitation, air temperature, and NDVI to assess the risk of drought disaster in Liaoning Province. This assessment, based on the natural hazard system theory, explores possible future evolutions to provide scientific references and a decision-making basis for regional drought disaster risk management and water resource allocation. In this paper, the term ‘geospatial big data’ refers to an integrated collection of large-scale, high-resolution spatial datasets. These datasets come from a variety of sources, such as satellite imagery, meteorological observations, and socio-economic data. The vast amount of data requires advanced computational methods for processing and analysis. By using geospatial big data to capture fine spatial changes over time, drought risk assessments can be more precise and comprehensive. The expected outcomes of this study include a comprehensive drought risk map for Liaoning Province, an assessment of future drought trends, and validation of the proposed risk assessment model through comparison with actual drought impacts. This is of practical significance for improving agricultural production and ensuring food security in Liaoning Province.

The data sources utilized in this study are delineated in Supplementary Table S1, encompassing three primary categories: meteorological data, statistical data, and geographic big data. Precipitation and temperature data were sourced from the CRU TS v4.06 dataset, courtesy of the Climatic Research Unit (CRU) at the University of East Anglia, UK (Harris et al. 2020). This dataset has undergone rigorous quality-control procedures, ensuring no missing measurements. It features extensive temporal coverage dating back to 1901 and high spatial resolution of 0.5 × 0.5°. The MOD13C2 vegetation index data comprises a gridded, cloud-free spatial composite derived from 16-day, 1 km MOD13A2 images, with a spatial resolution of 0.05 × 0.05° and a monthly time-scale (Cheng et al. 2023). Subsequently, the data were standardized to a spatial resolution of 0.05 × 0.05° using the bilinear interpolation method, resulting in 6,279 valid rasters at the regional scale of Liaoning Province. Population size, grain output per unit area, crop-sown area, and gross regional product data represent yearly statistics of municipal administrative districts in Liaoning Province spanning from 2001 to 2021. Additionally, crop drought damage and affected area data are yearly statistics for Liaoning Province over the same period. The spatial resolution of Liaoning Province's elevation and slope data is 80 × 80 m. ArcGIS software was employed to extract Liaoning Province's elevation and slope data and standardize the spatial resolution of the aforementioned datasets to 0.05 × 0.05° using the bilinear interpolation method. Bilinear interpolation has been widely employed in the resampling of climate data to achieve finer spatial resolution (Mishra & Singh 2010), as it provides a balance between maintaining spatial continuity and minimizing interpolation-induced artifacts. Furthermore, this method has been shown to preserve the integrity of large-scale climatic patterns, which are essential for drought risk analysis in regional studies (Fang et al. 2020). ArcGIS software was utilized to calculate the river network density value for each city in Liaoning Province. Total water resources data were sourced from the Liaoning Province Water Resources Bulletin, comprising yearly statistical data for each municipal administrative district spanning from 2001 to 2021.

Normal, gamma, Rayleigh, lognormal, logistic, GEV, and Weibull probability distribution functions were chosen to fit the precipitation, NDVI, and temperature variables in Liaoning Province. The normal distribution (Alzaatreh et al. 2021) is typically used to model continuous variables, particularly when the data are symmetrically distributed around the mean. For variables that meet the normality assumption, this distribution was applied. The gamma distribution (Altun et al. 2021) is often used to model non-negative, skewed data, such as precipitation. The Rayleigh distribution (Han et al. 2018) is commonly used to model the magnitude of two-dimensional vectors, such as wind speed or wave height. The lognormal distribution (Cohen & Whitten 1980) is widely used to model positively skewed data, where negative values are not possible, such as temperature and NDVI. Its versatility makes it suitable for environmental variables. The logistic distribution (Lv et al. 2024) is useful for modeling data with similar characteristics to the normal distribution but with heavier tails, providing a better fit for certain variables. This method was selected because it captures more extreme values. The generalized extreme value (GEV) distribution (Thorarinsdottir et al. 2018) is typically used for analyzing extreme weather events, such as maximum precipitation or temperature. This distribution was applied to model extreme climatic events in the study region. The Weibull distribution (Olivera & Heard 2019) is commonly used to model survival time and other environmental variables. In this study, it was applied to capture variability in extreme climate data. Parameter estimation was conducted using the maximum likelihood method. Concurrently, the Kolmogorov–Smirnov hypothesis test (K-S test) was employed to assess the consistency level through significance testing. Upon passing the K-S test at the 0.05 significance level, the distribution yielding the minimum K-S test value was identified as the optimal marginal distribution function for a specific variable, and cumulative probability calculations were carried out (Chen et al. 2023).

The subjective assignment is entirely dependent on the experience and judgment of experts, potentially introducing significant subjective bias. By integrating objective assignment methods, this bias can be mitigated, leading to a more scientifically and logically grounded weight distribution (Song & Chen 2022). In this study, a combination of the analytic hierarchy process (AHP) and the entropy weighting method is employed to ascertain the weight coefficients of each assessment indicator. This approach enables a more precise reflection of the importance of the assessment indicators. For detailed calculation procedures, please refer to the literature (Dou 2023). The entropy weight method formula is as follows:

• (1) The initial data were standardized to obtain the data matrix⁠:

  

  (2)

  

  (3)

  

  (4)

Here, represents the standardized value for the i indicator, and and are the actual maximum and minimum values, respectively. When an evaluation indicator (e.g., drought duration) is positively correlated with drought disaster risk, Equation (2) is used. Conversely, when an evaluation indicator (e.g., total water resources) is negatively correlated with drought disaster risk, Equation (3) is applied.

• (2) Calculation of information entropy ⁠:

  

  (5)

  

  (6)

If ⁠, it is defined that ⁠.

• (3) Determination of evaluation indicator weights ⁠:

  

  (7)

• (4) Combination of weights:

Drought duration and intensity were selected as indicators for assessing the hazard risk of disaster-causing factors in Liaoning Province. Elevation, river network density, slope, and total water resources were chosen as indicators to assess the vulnerability of the environment to disasters. Grain yield per unit area, crop planting area, population density, and per capita GDP were selected as indicators to assess the exposure of vulnerable entities. An improved AHP was used to rank the hazard risk assessment indicators, environmental vulnerability indicators, and exposure indicators based on their relative importance, determine the scale, construct the comparison matrix, and solve for the weight vector to calculate the weight coefficient. Finally, the combined weight coefficients of each evaluation index are calculated.

The risk of disaster-causing factors refers to the degree of variability that may result in disasters. For instance, the greater the intensity of the disaster and the longer its duration, the more significant the loss to human society, resulting in a higher risk. Conversely, a lower intensity and shorter duration correspond to a lower risk. In this study, drought duration and drought intensity are selected as indicators to assess the risk of disaster-causing factors in Liaoning Province.

According to the Liaoning Provincial Meteorological Disaster Bulletin and Figure 4(b), from July to August 2018, 40 cities and counties in the province, except for Chaoyang City and the Huludao area, experienced severe drought. Shenyang, Dalian, Anshan, Benxi, Dandong, Jinzhou, and other areas reached exceptional drought levels in severe cases. In July 2020, the overall CDI of Liaoning Province ranged from −2.32 to −1.61. Droughts were more severe in the north-central region. In the west, Fuxin City and Jinzhou City experienced higher drought intensity, while the east and south had lower intensity. This is consistent with the records of the Liaoning Provincial Meteorological Disaster Bulletin. In summary, the CDI closely aligns with the actual drought conditions. The drought index (CDI), constructed from precipitation, NDVI, and air temperature, effectively reflects the drought situation in Liaoning Province.

In summary, this paper comprehensively considers the hazard-causing factors, the vulnerability of the disaster-conceiving environment, and the exposure degree of the disaster-bearing body, referring to commonly used assessment indicators both domestically and internationally, while taking into account the available data and information. Finally, ten assessment indicators were selected to construct a drought disaster risk assessment indicator system for Liaoning Province. To comprehensively, reasonably, and accurately reflect the importance of the assessment indicators, the results of the hierarchical analysis method and the entropy method were synthesized to obtain the final combined weight coefficients. Subsequently, the Liaoning drought disaster risk assessment model was constructed, and the results are shown in Supplementary Table S2.

As shown in Figure 6(a)–6(u), there are significant inter-annual differences in the hazard across various regions of Liaoning Province from 2001 to 2021, with uneven spatial and temporal distributions. The overall severity of drought occurrences in the study area was high in 2001 and 2002, with the vast majority of the province in the high-hazard zone. In 2007, nearly 70% of the province was in the medium-hazard zone, while about 20% was in the high-hazard zone. In 2014, nearly 50% of the province was in the second-highest-hazard zone, and only about 20% of the province was in the second-lowest to low-hazard zone. From 2017 to 2019, consecutive droughts with second-highest to high-hazard levels were widely distributed. In contrast, in 2003, 2010, 2011, 2012, and 2013, the hazard levels were low. From 2010 to 2013, over 60% of the province was in the low-hazard zone.

The multi-year average hazards of hazard-causing factors in Liaoning Province are illustrated in Figure 6(v). The hazard levels in Chaoyang City, Huludao City, the southern part of Dalian City, and the sporadic plains crossed by the mainstream of the Liaohe River are generally high, with most areas falling in the second-highest to high-hazard zones. Tieling City, Shenyang City, Jinzhou City, Panjin City, Fuxin City, and Yingkou City are mostly in the medium to second-highest hazard zones. Fushun City, Liaoyang City, Benxi City, Anshan City, and Dandong City have lower drought hazard indices overall, with most areas falling in the second-lowest to low-hazard zones.

From 2001 to 2021, the multi-year changes in the vulnerability of the disaster-conceiving environment in Liaoning Province were minimal, with little inter-annual variation (Figure 7(a)–7(u)). Among these years, the vulnerability of Liaoning Province was the highest overall in 2014, with only Anshan City and Shenyang City having a vulnerability index below 0.46. Conversely, the overall vulnerability level was lower in 2010, with most areas having a vulnerability below 0.46, particularly in the central part of the study area. Additionally, Chaoyang City remained in the high vulnerability zone, with the vulnerability index consistently above 0.65 from 2001 to 2021. In contrast, the vulnerability index of Shenyang City and Panjin City remained below 0.46, indicating overall low vulnerability.

The multi-year average vulnerability of the disaster-conceiving environment in Liaoning Province is shown in Figure 7(v), indicating that Chaoyang, Fushun, and Benxi Cities have higher vulnerability levels, placing them in the high vulnerability zone. Yingkou and Huludao Cities follow, being in the second-highest vulnerability zone. Tieling, Fuxin, Liaoyang, Anshan, and Dandong Cities have an overall environmental vulnerability index between 0.46 and 0.57, placing them in the medium vulnerability zone. Shenyang, Jinzhou, Panjin, and Dalian Cities have lower overall vulnerability levels, placing them in the low to second-lowest vulnerability zones.

Overall, from 2001 to 2021, the spatial and temporal distribution patterns of the environmental vulnerability to disasters in Liaoning Province were clearly evident, with minimal inter-annual changes across cities. Additionally, Fushun City remained in a highly vulnerable category throughout the years, and trend results indicate that vulnerability risks may further intensify in the future.

From 2001 to 2021, there were relatively minor multi-year changes in the exposure of disaster-bearing bodies in Liaoning Province, with little variation between years (Figures 8(a)–8(u)). In 2011, Liaoning Province had the highest overall exposure level, with only Fushun City and Dandong City having an exposure index below 0.24, categorizing them as low-exposure zones, while most areas fell into the medium to second-highest-exposure zones. In 2014, Liaoning Province had a lower overall exposure level, with an exposure index of 0.38 or lower in most areas. Additionally, from 2001 to 2021, the exposure index of Shenyang City and Tieling City has consistently been above 0.50, placing them in the second-high to high-exposure zones, whereas the exposure index of Fushun City and Dandong City has remained below 0.38, indicating a low exposure risk.

The multi-year average exposure of disaster-bearing bodies in Liaoning Province, depicted in Figure 8(v), shows higher exposure mainly in the central plains and the northern hilly areas. Shenyang City has had an exposure index above 0.61, classifying it as a high-exposure area, while Tieling City falls into the second-highest-exposure area. Jinzhou City, Panjin City, Yingkou City, Fuxin City, Liaoyang City, and Anshan City have exposure indices between 0.38 and 0.50, placing them in the medium-exposure zone. Chaoyang, Fushun, Benxi, Huludao, Dandong, and Dalian generally have low exposure indices, categorizing them as low to second-lowest zones.

From 2001 to 2021, significant inter-annual variations in the comprehensive risk of drought disasters were observed across regions, with uneven spatial and temporal distributions (Figure 9(a)–9(u)). The overall risk level in Liaoning Province was elevated in 2001, 2002, 2014, 2017, and 2018, with most regions having a risk index higher than 0.15, placing them in the medium-risk zone or above. During 2001 and 2002, the distribution of the drought disaster composite risk index was similar, with higher levels in the central plains and western hilly and mountainous areas. Conversely, risk levels were lower in 2010 and 2013. From 2001 to 2021, the composite risk of drought disasters in Fushun, Dandong, Panjin, and Benxi cities remained at a low level.

The multi-year average drought disaster risk in Liaoning Province exhibits an overall pattern of ‘high in the west, low in the east, high in the north, and low in the south’. Notably, the drought disaster risk index is higher in the western mountainous and hilly areas, with Chaoyang City, particularly serious, in the high-risk zone, and Tieling City, which has the second-highest risk level, in the second-highest-risk zone across a wide area. Aside from Panjin City, most of the central plains of Liaoning Province fall into the medium-risk zone. Fushun, Benxi, Panjin, Dandong, and Dalian cities have low drought disaster risk levels, primarily in the low to second-lowest-risk zones.

Exposure overall shows a trend of weakening risk in the center and increasing risk in the surrounding area (Figure 10(c)). Specifically, Fuxin City, Chaoyang City, Jinzhou City, Liaoyang City, Dandong City, and Tieling City showed a trend of increasing exposure. In contrast, Benxi City, Shenyang City, Yingkou City, Huludao City, Anshan City, and Dalian City showed a trend of decreasing exposure. Most regions in Liaoning Province showed a trend of weakening exposure to the combined drought hazard. Only a few regions showed a trend of increasing exposure (Figure 10(d)), such as a small area in Fuxin City.

Overall, the trends of the three metrics are consistent, with a fluctuating decline from 2001 to 2005, followed by a gradual increase. Both crop drought-affected areas and disaster-affected areas remained low from 2010 to 2013, coinciding with a period of low drought risk. In 2014, there was a sharp increase in crop drought-affected areas, reaching the highest level in the past two decades. This was followed by an unstable phase with fluctuating drought risk. This relationship between high risk and high loss underscores the validity of the evaluation results presented in this study. The sharp rise in crop drought-affected areas in 2014 highlights the vulnerability of agricultural systems to sudden drought events (Guga et al. 2023; Hu et al. 2023), and the subsequent fluctuations in drought risk emphasize the need for ongoing monitoring and adaptive strategies to manage the dynamic nature of drought risk (Burgan et al. 2013; Hu et al. 2023).

This study constructed a CDI by integrating precipitation, normalized difference vegetation index (NDVI), and temperature using the Copula function, demonstrating significant potential in drought risk assessment. Compared with single-factor drought indices, such as the SPI (Marini et al. 2019) or the standardized precipitation evapotranspiration index (SPEI) (Liu, J. et al. 2024; Liu, Q. et al. 2024), which primarily focus on individual factors like precipitation or temperature, CDI offers a more comprehensive assessment by integrating multiple factors. This multidimensional approach more accurately reflects the actual drought conditions in Liaoning Province, enabling the monitoring of both meteorological and agricultural droughts. Our findings corroborate previous research on multidimensional drought indices (Chattopadhyay et al. 2020), demonstrating that indices integrating multiple drought factors outperform single-factor indices in capturing the complexity of drought events. Similarly, studies by Liu, J. et al. (2024), Liu, Q. et al. (2024) and Xu et al. (2023) also emphasize that multifactor drought indices significantly enhance the accuracy of drought identification and detection.

CDI exhibited a stronger correlation with affected areas, outperforming SPI, SPEI, VHI (vegetation health index), and VCI (vegetation condition index), with a correlation coefficient of 0.7. This finding aligns with similar studies in drought-prone regions by Bravo et al. (2021), who found that extracting and integrating multiple factors to construct indices provided more reliable estimates of drought risk impacts. Additionally, CDI demonstrated the ability to capture drought events across different temporal scales, addressing the limitations of other indices that often focus on only one aspect of drought. For example, SPI primarily reflects precipitation deficits, while VHI combines vegetation health and temperature (Bouras et al. 2020). In contrast, CDI integrates critical drought factors, providing a more robust tool for comprehensive drought assessment. This comprehensive approach is also supported by studies from Bravo et al. (2021) and Xu et al. (2023), who concluded that combining multiple indicators allows for more accurate and reliable descriptions of drought severity across different climatic regions and provides precise estimates of drought duration.

This study utilized the Copula function to integrate multiple drought factors, improving the classification of drought disaster risk levels and enhancing the detection intensity and accuracy of various drought types. As emphasized by Suo et al. (2024), this approach aids in capturing the joint behavior of drought factors, enabling a more refined analysis of drought disaster risk assessments. Compared with traditional drought monitoring models, the CDI approach offers a promising direction for improving drought disaster prevention and management. Additionally, Copula-based models are gaining increasing attention in drought disaster risk modeling because they allow for the analysis of correlations between multiple variables (Bazrafshan et al. 2021). Due to the complexity of drought formation and the dynamic nature of changing social and environmental conditions (Guga et al. 2023), the drought disaster risk varies significantly across regions each year. Compared with previous studies, this research employs multi-source geographic big data and a multifactor drought index to provide a more detailed drought disaster risk assessment.

CDI has demonstrated strong performance in drought monitoring; however, like other drought indices, it also has its limitations. Due to data limitations and anthropogenic factors in some urban areas, this may reduce the accuracy of environmental vulnerability and exposure assessments. As noted by Ghazaryan et al. (2020), this is a common challenge in regional drought risk assessments, where the resolution of remote-sensing data often limits the construction of comprehensive indices and the accuracy of risk evaluations. The model could be further improved by incorporating high-resolution satellite data, enabling more detailed assessments of drought impacts. Research by Gholinia & Abbaszadeh (2024) suggests that combining traditional methods with drought monitoring based on modern high-resolution remote-sensing data significantly enhances the evaluation and monitoring of drought events, especially in complex terrains.

Future research could address these uncertainties by incorporating more reliable datasets and improving the Copula function (Won et al. 2020) to enhance drought index and risk assessment efforts. Additionally, future studies should aim to refine the CDI method by incorporating more environmental and socio-economic factors, ultimately providing a more comprehensive and reliable drought risk assessment model. The impact of human activities is another aspect to be considered in future studies. As noted by Wu et al. (2024), accounting for land-use changes and human interventions on drought dynamics could improve drought monitoring models.

This paper combines multi-source geographic big data to construct a comprehensive drought disaster risk assessment model by integrating hazardous factors, the vulnerability of the disaster-bearing environment, and exposure of the disaster-bearing body. The model was developed by combining the AHP with the entropy weight method. Furthermore, it explores possible future evolution based on the median trend of Sen. The findings indicate that:

• (1) The drought disaster risk in Liaoning Province exhibits the distribution pattern of ‘high in the west and low in the east, high in the north and low in the south’, with significant variations among different areas. Higher risk levels are observed in the southern regions of Chaoyang City, Huludao City, and Dalian City, while Chaoyang City, Fushun City, and Benxi City demonstrate higher vulnerability levels. The central plains and northern hilly areas exhibit elevated exposure levels, with Shenyang City being the most exposed.

• (2) The majority of regions in the province exhibit a trend of decreasing drought disaster risk, although Fushun City is projected to experience a further increase in vulnerability risk in the future. Additionally, Chaoyang City and Dandong City are expected to face heightened exposure risk. Despite this, the perennial average exposure risk of disaster-bearing bodies remains low.

• (3) The comprehensive risk assessment results align closely with actual drought losses, underscoring the study's contribution to a comprehensive and scientific understanding of drought disaster risk. Moreover, the findings offer valuable data support for drought disaster prevention and mitigation efforts.

This work was funded under the auspices of the Natural Science Foundation of Shandong Province (ZR2024ME171, ZR2024QD207) and the National Natural Science Foundation of China (Grant Nos. 41471160, 42377077).

We consent to the publication of our research and manuscript.

Q.Z.: methodology, writing – original draft, formal analysis, conceptualization. W.W.: writing – review & editing, conceptualization, formal analysis. Y.C.: methodology, formal analysis. Q.W.: formal analysis. Q.S.: formal analysis.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.