Drought sensitivity analysis of meteorological and vegetation indices in Dak Nong, Vietnam

The consumption of fossil fuels in production systems such as crop cultivation, livestock farming, and industries is one of the primary drivers of climate change, as highlighted by multiple studies (Elahi et al. 2022; Abbas et al. 2023a, 2023b; Elahi et al. 2024). To meet the increasing demand for agricultural productivity and ensure food security, these systems are becoming more reliant on energy from fossil fuels to operate machinery and pump irrigation water. However, this heightened reliance on fossil fuels increases greenhouse gas emissions, further exacerbating climate change. Conversely, climate change has become a critical concern impacting food supply and security. Abnormal weather patterns, such as rising temperatures and prolonged droughts, reduce crop productivity and necessitate emergency irrigation measures to save crops. This situation leads to even greater reliance on fossil fuels for pumping water, emitting more greenhouse gases and creating a negative feedback loop.

Understanding the impacts of changing weather patterns and increased drought occurrences on crop productivity has become imperative in this context. This heightened significance is reflected in extensive worldwide research efforts dedicated to addressing drought using a variety of approaches. These studies encompass the examination of various drought indices, including the standardized precipitation index (SPI), standardized precipitation evapotranspiration index (SPEI), temperature vegetation dryness index (TVDI), and the vegetation health index (VHI), among others. This diversity underscores the multifaceted nature of drought assessment and monitoring systems, as highlighted by multiple studies (Chen et al. 2015; Javanmard 2017; Pervez et al. 2021). The selection of appropriate drought indices tailored to each region is vital, as it significantly contributes to the accuracy of drought monitoring, as emphasized by other researchers (Eshetie et al. 2016; Shi et al. 2020).

Moreover, multi-temporal MODIS NDVI data have proven to be valuable for monitoring drought conditions (Vicente-Serrano et al. 2010). The growing trend of incorporating satellite data and remote sensing technologies in drought assessment is increasingly important, playing a critical role in monitoring and evaluating drought conditions. This trend highlights the continuous progress in harnessing modern technology for drought monitoring (Kloos et al. 2021; Zeng et al. 2022, 2023) – numerous studies investigating drought using various indices and remote sensing techniques in different regions and climates. For instance, Abbas et al. (2021) have explored trends in vegetation productivity related to climate change in China's Pearl River Delta, emphasizing the importance of vegetation indices (VIs) in assessing environmental changes.

Similarly, Alshaikh (2015) has utilized space applications to assess drought in the Wadi-Dama region of Saudi Arabia, highlighting the role of satellite data in monitoring drought conditions. In Turkey, Anli (2017) has investigated the temporal variation of reference evapotranspiration and regional drought estimation using the SPEI method for the semi-arid Konya closed basin. This research underscores the significance of drought indices in semi-arid regions.

Additionally, Bae et al. (2018) have analyzed drought intensity and trends in South Korea using the modified SPEI, demonstrating the versatility of this index in different geographical contexts. Furthermore, studies like the one conducted by Catchment (2018) have employed satellite-based multi-indices to evaluate agricultural droughts in dynamic tropical catchment areas, similar to the context of the Central Highlands in Vietnam. Chen et al. (2015) investigated soil moisture estimation under different tree species using the TVDI. This research highlights the applicability of VIs in understanding soil moisture dynamics. In addition to VIs, some studies have emphasized the importance of combined indices for drought monitoring. For instance, Chu et al. (2019) have explored the relationship between the TVDI and normalized difference vegetation index (NDVI) for drought monitoring in the Greater Changbai Mountains. Their research showcases the effectiveness of combining multiple remote sensing-based indices. As for drought monitoring in semi-arid regions, Chuai et al. (2013) have examined temperature, vegetation, and precipitation changes in Inner Mongolia, China. Their findings shed light on the impact of these climate variables on drought in arid and semi-arid regions.

Additionally, Çuhadar & Atiş (2019) have conducted a drought analysis in the Ceyhan Basin, emphasizing using SPI to assess drought conditions. Their research demonstrates the application of meteorological indices for drought monitoring. For the Northeastern Highlands of Ethiopia, Chere (2023) has employed the earth observation-derived SPEI and VHI to model agricultural drought. This study emphasizes the importance of remote sensing and indices for assessing drought in regions facing agricultural challenges. Furthermore, Hanadé Houmma et al. (2022) have explored the spatial heterogeneity influence on the concordance of remote sensing drought indices in agrosystems in Morocco. Their research underscores the complexity of drought assessment in semi-arid regions.

Vietnam is one of the top five countries most affected by climate change. According to the Global Climate Risk Index 2021, Vietnam ranks among the top five countries most impacted by extreme weather events over the past two decades (Eckstein et al. 2021). These changes include increased frequency and intensity of tropical storms, rising sea levels, higher temperatures, and prolonged drought periods. Khoi & Nhi (2021) conducted a study focusing on the Central Highlands and investigated the impact of drought on the Srepok River basin using three drought indices: SPI, standardized runoff index, and standardized soil wetness index. The research calculated meteorological and hydrological drought levels using data on soil moisture and streamflow from the SWAT hydrological model to study meteorological and agricultural droughts. The results indicated an expected increase in drought frequency in the Srepok River basin in the future, with a decrease in the duration and severity of drought. Ha et al. (2021) focused on assessing drought risk in Vietnam by integrating exposure to drought and vulnerability based on specific socio-economic conditions in the Central Highlands and Southern Vietnam. Although the study did not provide details on the methods used, it employed a comprehensive approach combining multiple data sources and analytical techniques to assess drought risk in the Central Highlands and Southern Vietnam.

Dinh et al. (2020) concentrated on Dak Nong Province, where natural conditions for coffee cultivation are favorable. However, current climate changes have led to extended dry seasons with high temperatures and low rainfall during the rainy season. The study applied remote sensing and Geographic Information System (GIS) methods to analyze coffee's water interaction in the Dak Nong basin. The results showed varying water consumption for coffee crops across districts in Dak Nong, influenced by climate and surface water supply, coffee varieties, and irrigation practices. The average water consumption for coffee in Dak Nong was found to be 5,231.37 m³ per ton of product. This study offered conclusions and recommendations for effective water resource management and sustainable development for the province. Thi et al. (2019) conducted research focusing on the Central Highlands region of Vietnam, an area heavily impacted by drought, leading to severe agricultural losses and severe water shortages. The study surveyed five farming communities experiencing drought in Dak Nong Province, one of Vietnam's most vulnerable regions. The survey of over 250 households collected data on socio-economic conditions, livelihoods, social networks, health status, food and water security, drought conditions, and climate change in the region. The study results showed that the Quang Phu community was the most vulnerable, followed by Nam N'dir, Dak Nang, Duc Xuyen, and Dak D'ro in decreasing order of vulnerability. Water supply capacity and livelihood strategies were the most critical variables in determining the five communities' vulnerability levels. As part of a broader effort to assess drought, Hang & Trang (2010) have analyzed drought conditions in Central Vietnam. Their study aligns with the focus on drought analysis in the Central Highlands of Vietnam. These studies collectively demonstrate the diversity of methods and indices employed in drought research across different regions, and they highlight the crucial role of remote sensing and meteorological data in this context.

Climate change has intensified the need for effective drought research, especially in agriculture-dependent regions like Vietnam. The proliferation of drought indices necessitates careful selection of monitoring tools. While the SPEI has shown efficacy in regional analysis, index suitability varies with crop types and local climate. Our research in Dak Nong province, Vietnam, explores the intricate relationship between meteorological conditions and vegetation responses to water stress. We aim to enhance our understanding of local drought dynamics by analyzing spatial and temporal drought patterns. This knowledge will inform targeted strategies for improving agricultural productivity and water management, boosting drought resilience in Dak Nong and similar tropical regions. Our study bridges the gap between theoretical indices and practical applications, offering insights into resource management in vulnerable agricultural areas.

This study aims to enhance drought monitoring and management in Dak Nong province through a comprehensive analysis of meteorological and VIs. We aim to identify the most practical combination of these indices for drought assessment, determine their optimal time scales, delineate drought-sensitive areas, analyze seasonal drought impacts across vegetation types, and provide evidence-based recommendations for improving regional drought resilience.

Figure 1 provides a schematic representation of this process, illustrating the progression from data collection to determining critical drought indices, appropriate accumulation periods, and geospatial areas highly sensitive to drought. This visual aid enhances understanding of how the research evaluates and selects the most suitable indices for monitoring and managing drought conditions in Dak Nong province.

In this research, a set of VIs, namely the vegetation condition index (VCI), temperature condition index (TCI), VHI, TVDI, and meteorological drought indices, SPI and SPEI, are utilized – the data span from 2000 to 2022. The VCI, TCI, VHI, and TVDI indices are derived from NDVI and land surface temperature (LST) data, which were acquired from the moderate resolution imaging spectroradiometer (MODIS) platform, accessible through the United States Geological Survey (USGS) Earth Explorer website. On the website, NDVI data are sourced from the Mod13A1 VI package, with a spatial resolution of 463 × 463 m and a temporal resolution of 16 days. LST data are obtained from the Mod11A2 package, featuring a spatial resolution of 926.6 × 926.6 m and an 8-day temporal resolution. To analyze the impact of drought on cultivated vegetation, both meteorological and VIs must share a standard temporal resolution. To meet this requirement, monthly time intervals were chosen for data calculation. The monthly average LST, calculated from the 8-day resolution Mod11A2 package, was resampled using the Bilinear method to match the spatial resolution of NDVI. Monthly NDVI data were generated from the input 16-day resolution NDVI data using the maximum value composite (MVC) method, which helps mitigate the effects of atmospheric conditions (Rahimzadeh-Bajgiran et al. 2012; Nino et al. 2014; Gong et al. 2015; Du et al. 2017; Cao et al. 2019).

The meteorological data were collected from a network of meteorological stations managed by the Vietnam Meteorological and Hydrological Administration, known for maintaining high-quality data standards. The chosen meteorological stations have relatively complete datasets, with only about 6.4% missing data, which were supplemented using a stepwise linear regression method. The crop data were obtained from local agricultural departments, ensuring accurate and relevant information for the study area.

Based on the observational data, SPI and SPEI values were computed for each station and subsequently interpolated using the inverse distance weighting method, with a grid spacing of 463 × 463 m, matching the spatial resolution of the VIs.

Drought characteristics differ across various climatic regions, and their impacts depend on the local area's environmental and socio-economic factors. It has been widely recognized in the scientific community that there is no universally applicable drought definition or index that can comprehensively describe all drought events, climatic conditions, and drought-affected sectors. In this study, specific drought indices, SPI and SPEI, were selected for analysis. To compute the SPI and SPEI in this study, researchers employed the ‘cdfgam’ subroutine, which calculates the Gamma distribution function. A time step ranging from 1 to 11 months was chosen to align with the study region's characteristics, effectively capturing seasonal variations in various indices and producing consistent results.

The SPI, developed by McKee and colleagues in 1993 (McKee et al. 1993), is a globally utilized indicator. This index is based on the distribution of precipitation and can be calculated over time scales ranging from 1 month to several years. This study selected a 1-month time scale to fulfill the requirements for short-duration drought warnings.

Let X be the rainfall corresponding to a given time scale, then SPI is calculated according to the following steps:

This study uses the gamma function in the Cdflib.f90 program from the Florida State University website to calculate the SPI: https://people.sc.fsu.edu/jburkardt/f_src/cdflib/cdflib.html

NDM was the number of days in the month, and N was the number of sunshine hours calculated based on the astronomical formula.

The SPEI was calculated for each meteorological station and then was interpolated using the inverse distance weighted (IDW) method to obtain a raster layer with a resolution of 463 × 463 m. This variable was built for a timescale from 1 to 12 months.

VCI, derived from NDVI data, has proven to be a versatile and valuable tool in assessing plant health, stress, and, most notably, drought conditions. Recent studies, including those by Baniya et al. (2019) and Van Viet & Thuy (2023) Thuy, contribute to the ongoing refinement and application of VCI, enhancing its utility across diverse geographical and climatic contexts.

The TVDI, developed by Sandholt et al. (2002), is specifically designed to evaluate drought stress in vegetation. It estimates topsoil water content under varying vegetation conditions, providing insights into soil moisture levels. TVDI is particularly beneficial in agriculture, especially in arid and semi-arid regions, for monitoring the impact of drought on vegetation growth and productivity (Bai et al. 2017; Liu & Yue 2018; Shashikant et al. 2021).

TCI has emerged as a preferred metric over LST in assessing the impact of temperature on plant health. Developed by Kogan in 1995, TCI normalizes LST values over time, providing a comprehensive view of plant stress dynamics. The scale of TCI ranges from 0 to 100, with higher values indicating more significant plant stress during periods of low LST. Various studies have been conducted to compare the effectiveness of the VCI and TCI in evaluating land cover conditions. However, these studies have produced conflicting results (Chu et al. 2019; Mlenga et al. 2019). To further understand these discrepancies, it is crucial to consider findings from various relevant literature (Lawal et al. 2021; Hanadé Houmma et al. 2022; Van Viet & Thuy 2023).

LST_i is the LST value of pixel i, and LST_max and LST_min are the maximum and minimum LST values in the period included in the analysis, respectively. TCI has a value between 0 and 100. TCI is maximum when LST is the smallest. This index is significant in assessing vegetation stress related to temperature.

VHI is an integrative index that combines VCI and TCI. It is widely used to evaluate vegetation growth, assess drought conditions, and predict crop yields (Kloos et al. 2021). This index plays a pivotal role in remote sensing and agriculture. The VHI incorporates information from NDVI and LST, with a typical weight parameter ‘α’ set at 0.5, to provide an overall assessment of vegetation health. A VHI value below 40 indicates plant stress, while a value exceeding 60 signifies healthy vegetation. VHI offers valuable insights into the impact of temperature and vegetation growth on overall plant health. Monthly calculations of these VIs were conducted from 2000 to 2,022 using specific formulas. You can refer to our other study for more detailed information on calculating these indices (Van Viet & Thuy 2023).

where ∝ is a weight usually taken as 0.5. This index shows the effect of temperature on vegetation growth. A VHI less than 40 indicates the presence of vegetation stress, while a value greater than 60 indicates that the vegetation is in good condition.

In Dak Nong, Vietnam, the landscape is characterized by various land types, each with distinct vegetation cover and ecological significance.

Perennial agriculture (PeA) land is dedicated to cultivating perennial crops, prominently featuring fruit trees and industrial crops, with a pronounced emphasis on coffee production in Dak Nong Province. These areas are pivotal in the province's agricultural landscape, substantially contributing to economic development and local livelihoods. However, the escalating prevalence of drought poses significant challenges to the agricultural sector, particularly impacting coffee cultivation. Prolonged periods of drought can severely damage perennial crop areas, adversely affecting yield and product quality. During extended droughts, the heightened water demand for coffee and other perennial crops exerts pressure on water sources, intensifying competition among various water users.

Productive forest land (PdF) is widespread in Dak Nong and primarily concentrated in highland areas. It plays a pivotal role in protecting headwaters, preventing erosion, and maintaining the ecological balance. The Productive Forest Land in Dak Nong is a sanctuary for valuable timber species such as Dalbergia, Lim, and Acacia confusa, prized for their economic significance and routinely harvested for industrial and commercial purposes. However, despite its importance, PdF in Dak Nong confronts formidable challenges, notably climate change and unsustainable logging practices.

Annual crops land (AnC) areas are allocated explicitly for cultivating annual crops, including seasonal rice, maize, potatoes, and other crops with a yearly growth cycle. These areas are pivotal in sustaining local agricultural production and ensuring food supply. As annual crops are integral to the region's food security, responsible and sustainable practices in AnC areas are paramount to maintaining a reliable and resilient agricultural system.

Protective forest land (PtF) is strategically conserved and managed to preserve forest resources, safeguard the environment, and sustain vital ecosystems. PtF is pivotal in safeguarding precious forest resources, including timber, plant species, and wildlife. These PtF areas serve as crucial components for preserving and restoring the natural ecosystem, particularly in the face of increasingly intricate climate change scenarios.

Special-use forest land (SpF) areas are specifically designated for conserving wildlife and plant species, maintaining water sources, and protecting crucial ecosystems. By safeguarding clean water, SPF areas are pivotal in providing backup water resources for local communities and agriculture.

Paddy rice (PdR) land is allocated explicitly for cultivating rice and other crops, playing a pivotal role in agricultural production and food supply in Dak Nong. However, the escalating impact of drought poses a growing threat to rice productivity in Dak Nong province. The increasing severity of drought emphasizes the need for sustainable water resource management practices, innovative agricultural techniques, and proactive measures to mitigate the adverse effects of water scarcity on paddy rice cultivation.

In light of our research and two key references, one focusing on the integration of VHI and SPEI to assess agricultural drought in Morocco (Moutia et al. 2021) and another by Hanadé Houmma et al. (2022), emphasizing the spatiotemporal coherence of drought indices, we outline a procedure for selecting the most appropriate indices for monitoring drought conditions in Dak Nong. The specific steps are as follows:

• Data collection: Initiate gathering meteorological drought indices and VI data within the research area. Ensure data spans from January to November and compute grid-cell mean values without distinction for VIs (VCI, VHI, TCI, and TVDI) and meteorological drought indices (SPEI and SPI).

• Calculate correlation coefficients: compute correlation coefficients between pairs of meteorological drought indices and VIs. Identify pairs with high correlation levels.

• Compare VIs: evaluate different vegetation index pairs to determine which is more suitable for monitoring crop status based on water scarcity levels.

• SPI vs. SPEI comparison: conduct a comparison between SPI and SPEI to ascertain whether SPI can be utilized as a substitute for SPEI without significant differences and provide a rationale for the choice.

Finally, conclusions will be presented based on the analysis of the results, and the most appropriate indices for monitoring drought conditions in the Dak Nong region will be proposed. This concluding section will also explain why these indices were chosen and their significance. The outcome of this process will enhance our understanding of selecting the most suitable indices for drought monitoring and provide a foundation for improved drought management in the Dak Nong region.

To identify the most suitable accumulation period for MDI*, the correlation between MDI* and VI* (VIs) is analyzed as follows:

• Compile data for selected MDI* and VI* covering 1–11 months over 23 years.

• Gather information on vegetation cover types, including area coverage and details.

• Compute correlation coefficients between MDI* and VI* for each vegetation cover type.

• Calculate these coefficients for various temporal scales to determine the most suitable scale.

• Visualize results on a graph to illustrate correlation coefficients at various temporal scales.

• Identify the vegetation cover type with the lowest correlation coefficient, indicating less drought sensitivity.

• Determine the most appropriate temporal scale for vegetation types susceptible to drought.

• Recommend the temporal scale with the highest correlation coefficient for monitoring drought conditions.

To determine the most drought-sensitive geographic regions, the following method is employed:

• Compile data for MDI* and VI* within the research area.

• Identify drought months and wet months.

• Compute the divergence in VI between wet and drought months.

• Identify regions with high correlation coefficients between MDI* and VI* and significant divergence in VI* between wet and drought months.

• Present the results by displaying the geometric mean of the correlation coefficient and VI* divergence over the sensitive region.

• Cross-reference results with geographical characteristics such as terrain slope and river/stream density.

This methodology helps identify and prioritize the regions most sensitive to drought conditions, which can be crucial for effective drought management and preparedness.

Figure 3 reveals that only four pairs of relationships, namely TCI_SPI, TCI_SPEI, VHI_SPI, and VHI_SPEI, exhibit significantly high correlation coefficients (p = 0.001) across all temporal scales of meteorological indices. This suggests that these pairs demonstrate higher correlations compared to others and are suitable for drought monitoring.

Compared to TCI, VHI demonstrates a slightly higher correlation with meteorological indices, highlighting its potential as a better choice for monitoring crop conditions based on water scarcity. Comparing the pair VHI_SPI and VHI_SPEI reveals that SPEI is slightly more suitable for monitoring drought conditions than SPI. This difference can be attributed to the factors involved in calculating SPI and SPEI and the equations used. Previous studies (Yang et al. 2016; Anli 2017; Bezdan et al. 2019) have also indicated that SPEI is more sensitive than SPI in drought monitoring. Figure 3 also shows no significant difference in the correlation coefficients between the VHI_SPI and VHI_SPEI groups. This suggests that SPI can be used as a substitute for SPEI in drought monitoring without substantial differences. The simplicity of SPI data compared to SPEI, which only requires precipitation data, may contribute to this minor difference.

Figure 4 illustrates the significant relationship between rainfall and evapotranspiration in Dak Nong Province, Vietnam. Figure 4(a) shows that within the period from November to July, the 3 months with the highest rainfall variability are May, June, and April. These months can significantly impact drought and soil moisture with substantial rainfall variability. Figure 4(b) depicts evapotranspiration during the same period. The peak of evapotranspiration occurs in the late dry season (April) and early wet season (May). This indicates substantial soil moisture variability during this time frame. The high rainfall and evapotranspiration in these months set the foundation for critical considerations regarding their relationship. The connection between meteorological and VIs becomes more pronounced during these months.

In summary, this analysis's results emphasize the importance of low-rainfall months, especially April and May, in providing water to crops and monitoring drought conditions. It also discusses using SPI as a simpler alternative to SPEI with user-friendly precipitation data.

Assessing the correlation between SPEI and VHI aims to pinpoint the most suitable accumulation period for SPEI in drought monitoring. Following the previous step, wherein VHI and SPEI emerged as the most appropriate indices for tracking drought effects on vegetation, this analysis focuses on elucidating the sensitivity of different vegetation types and the optimal time scale for meteorological drought monitoring.

The study area comprises six primary vegetation types, as described in Table 1. Notably, perennial agriculture (PeA), production forests (PdF), and annual crops (AnC) dominate the land cover in this region, accounting for 76.5% of the area. The average VHI value for each vegetation type was computed over a 276-time series (12 months × 23 years) to determine the correlation coefficient.

Crop vegetation types, including agriculture and afforestation (PeA, PdR, Pdf, and AnC), exhibit higher correlation coefficients than natural forests. Among these, PeA (perennial agriculture) has the highest correlation coefficient. PeA includes coffee, cashews, pepper, and macadamia, with robusta coffee being the primary crop. Coffee is sensitive to drought and thrives in high-temperature conditions, often coinciding with drought periods. Furthermore, the water supply for coffee is typically rainfed, contributing to the high correlation between SPEI and VHI.

Similarly, PdR (paddy rice) is a water-demanding crop with limited irrigation capabilities during drought conditions. PdF (production forests) have deep root systems that make them less susceptible to drought, but their susceptibility increases due to steep terrain and low river/stream density (Table 2). AnC (annual crops) in the region mainly include corn and cassava, constituting 60% of the AnC area, while beans and various vegetables occupy 18%. These crops generally exhibit better drought tolerance than PeA and PdR, resulting in a lower correlation coefficient.

The VHI variations between wet and drought months, as determined by the SPEI, identify the time frames during which crops are significantly affected by drought conditions. In this analysis, only the SPEI and VHI pairs are considered.

As illustrated in Figure 6, there is no significant difference in VHI variations between drought and wet conditions. Similar to Figure 3, this figure also highlights that the impact of drought is less pronounced in natural forest regions than in other crops. Regarding the SPEI time scale, just like in Figure 3, this figure reveals the most pronounced influence between 3 and 5 months, particularly at the 4-month time scale. Throughout the year, the impact of drought varies across months. VHI exhibits the most significant variations between wet and dry SPEI conditions from November to May, especially from February to March. These are months with very low rainfall (Figure 3(a)), making VHI highly sensitive to precipitation deficits. From June to October, VHI variations are negligible, corresponding to periods with ample rainfall meeting crop water requirements, resulting in low VHI fluctuations. As analyzed earlier, VHI and SPEI4 are the most suitable indices for monitoring drought in this region. Therefore, the analysis focuses on this time frame from January to May when VHI variations between drought and wet months are most distinct (Figure 5).

Comparing Figure 10(b) and 10(a), it is evident that river and stream density in the area with a correlation coefficient higher than that in Figure 7(a) is lower than the overall area. This difference is most prominent in the coffee-growing regions. In the coffee cultivation area in Figure 10(a), 50% of the area has river and stream densities above 0.4 km/km². In Figure 10(b), this value is only 11%. River and stream density reflects the potential for surface water irrigation. When meteorological drought occurs, its impact on vegetation in areas with low river and stream densities and high water requirements becomes substantially evident.

Soil types do not directly correlate with the formation of drought-sensitive regions. In the sensitive area, 80.6% of the area consists of ‘Rhodic Ferralsols’ (FRr). This soil type has average nutrient levels and good water retention capabilities (available water content – AWC) and is prevalent in PeA and AnC regions.

The correlation coefficient between SPEI4 and VHI, based on the data in Figure 11, is 0.6, which is significant at p = 0.001. However, relying on SPEI4 for crop condition forecasting may need to be higher. By considering data only from January to May (when VHI variations between wet and drought conditions are most distinct, Figure 6), the correlation coefficient rises to 0.72. This indicates that meteorological drought indices can be used for crop condition forecasting in low-rainfall months. The relatively low correlation coefficient is partly due to this area's low density of weather stations.

This study provides significant insights into drought monitoring in Dak Nong Province, Vietnam, contributing to the growing literature on drought assessment in Southeast Asia. Our findings underscore the complex interplay between meteorological and VIs in assessing drought conditions, aligning with previous research that emphasizes the importance of combining multiple indices for more accurate drought monitoring (Moutia et al. 2021; Hanadé Houmma et al. 2022).

The key results of our study include (1) identifying VHI and SPEI as the most suitable indices for drought monitoring in the region, (2) determining SPEI4 as the optimal time scale for drought assessment, (3) delineating 10.8% of Dak Nong's area as drought-sensitive, and (4) establishing the relationship between topographic characteristics, river density, and drought sensitivity.

Our study found that SPEI4 exhibited the highest correlation with VHI compared to other time scales (1–11 months), with a consistently strong correlation coefficient (r > 0.7, p < 0.001). This finding aligns with Vicente-Serrano et al. (2010), who highlighted SPEI's effectiveness in capturing drought impacts on vegetation by incorporating precipitation and PET. The reliability of SPEI in drought assessment is further supported by recent studies (Abbas et al. 2022a, 2022b, 2023a, 2023b), emphasizing the importance of accurate precipitation data in drought monitoring.

Comparing our results with previous studies in Vietnam, we find both consistencies and advancements. Van Viet & Thuy (2023) used VHI for coffee yield forecasting in Dak Lak, confirming its effectiveness in assessing crop conditions in the Central Highlands. Our study extends this application by combining VHI with SPEI4, providing a more comprehensive approach to drought monitoring. While Khoi & Nhi (2021) used multiple indices (SPI, and Standardized Soil Wetness Index) for drought assessment in the Central Highlands, our approach simplifies the process by demonstrating the effectiveness of the VHI-SPEI4 combination. VHI's ability to integrate temperature information and detect vegetation stress makes it particularly suitable for Vietnam's tropical monsoon climate, as demonstrated by its higher correlation with meteorological indices compared to TCI. SPEI4's incorporation of precipitation and evapotranspiration data provides a more comprehensive measure of drought conditions, especially relevant in the context of climate change, where both rainfall patterns and temperatures are changing.

The identification of drought-sensitive regions within Dak Nong is a critical outcome of this research, with implications for water resource management and land use planning. These regions, characterized by steep terrain and limited river density, are predominantly occupied by perennial agriculture (PeA), annual crops (AnC), and productive forest land (PdF). This finding builds upon the work of Dinh et al. (2020), who highlighted the water footprint of coffee production in Dak Nong. Our results provide a spatial context to their findings, identifying areas where water management is most critical.

However, our study has limitations that should be addressed in future research. The reliance on remote sensing data and the low density of weather stations in the study area may affect the precision of our results, a challenge also noted by Khoi & Nhi (2021). Additionally, our focus on physical and meteorological indices leaves room for a more in-depth analysis of socio-economic impacts, which Thi et al. (2019) explored in their assessment of drought vulnerability in Dak Nong's farming communities.

Looking ahead, future research should focus on integrating multi-source data, including high-resolution remote sensing data and ground-based observations. Recent studies (e.g., Abbas et al. 2022) suggest that the application of advanced technologies like machine learning and artificial intelligence could enhance drought prediction and response capabilities. Furthermore, expanding the spatial and temporal scope of the study would provide a more comprehensive view of drought conditions in the region.

In summary, this study contributes to the academic knowledge on drought monitoring and provides practical tools and information for decision-making and policy planning in Dak Nong and similar regions in Vietnam. By offering a more nuanced understanding of drought patterns and vulnerable areas, our findings can significantly enhance drought resilience and support sustainable agricultural development in the face of changing climatic conditions.

Our research has yielded crucial insights into drought sensitivity in Dak Nong Province, Vietnam. Among the various meteorological and VIs, the VHI has emerged as a robust indicator, exhibiting stronger correlations with meteorological drought indices than the TCI. This finding underscores the importance of VHI as a preferred choice for monitoring crop conditions under water scarcity.

The consistently higher correlation values within VHI_SPEI pairs compared to VHI_SPI emphasize the superiority of the SPEI over the SPI, particularly during the dry months from December to May in Dak Nong. Our research identifies the 4-month time scale, SPEI4, as the optimal choice for drought monitoring in this region, which is significant for various crops such as coffee, rice, and other agricultural plants and production forests.

Identifying a drought-sensitive area, covering approximately 10.8% of Dak Nong's territory, is equally crucial. This region's steep topography and reduced river and stream density make it inherently more vulnerable to drought, regardless of soil composition or land cover diversity. Our findings establish a robust scientific framework for predicting and monitoring drought in Dak Nong.

Our research results align with previous studies on drought assessment in Vietnam and Southeast Asia while extending the existing knowledge by combining VHI and SPEI4 to provide a more comprehensive approach. However, the study's limitations related to remote sensing data and the low density of weather stations in the area need to be addressed in future research by integrating multi-source data and expanding the study's spatial and temporal scope.

Based on our research findings, we propose a coherent set of policy recommendations to address the critical challenge of drought in Dak Nong, including (1) establishing a robust drought monitoring and alert network based on the SPEI4 index, (2) enhancing water resource management, (3) empowering and supporting farmers through training and guidance, (4) developing drought preparedness plans, and (5) continuous research and monitoring to adapt to changing climatic conditions.

In conclusion, this study contributes to the academic knowledge of drought monitoring. It provides practical tools and information for decision-making and policy planning in Dak Nong and similar regions in Vietnam. Our findings can significantly enhance drought resilience and support sustainable agricultural development in the face of climate change.

We thank the Dak Nong province for generously providing the essential data required for this study. Detailed population data for each commune within Dak Nong province were collected and utilized. The valuable soil and land use data were also acquired from the Department of Agriculture and Rural Development of Dak Nong province. The research also leveraged satellite data, which were sourced from the United States Geological Survey website (accessible at: https://earthexplorer.usgs.gov/) and were retrieved on October 25, 2022. Our heartfelt thanks go out to all these organizations and entities for their indispensable contributions, without which this study would not have been possible.

Van Viet Luong completed his Ph.D. in Environmental Management at Vietnam National University Ho Chi Minh City in 2011. Currently serving as an assistant professor at the Industrial University of Ho Chi Minh City in Vietnam, he has contributed to the academic field with over 30 journal papers, multiple books, and a book chapter. His research primarily examines the effects of climate change and variability on water resources and crops.

Tran Thi Thu Thuy obtained her Ph.D. in Environmental Science and Management from the University of Liege, Belgium, in 2019. She also holds a Master's degree in Applied Economics and Public Policy Analysis from Fulbright Vietnam University, which she earned in 2022. She is a lecturer at the Industrial University of Ho Chi Minh City in Vietnam. She has published five papers on climate change and air pollution.

T.T.T.T. and V.L.V. conceptualized the whole article; T.T.T.T. and V.L.V. developed the methodology; V.L.V. arranged the software; V.L.V. and T.T.T.T. validated the data; V.L.V. rendered support in formal analysis; V.L.V. investigated the data; V.L.V. and T.T.T.T. arranged the resources; V.L.V. wrote the original draft; T.T.T.T. wrote the review and edited the article; V.L.V. visualized the article; V.L.V. supervised the article; V.L.V. rendered support in project administration; V.L.V. rendered support in funding acquisition. All authors have read and agreed to the published version of the manuscript.

All relevant data are available from an online repository or repositories.

The authors declare there is no conflict.