Drought Analysis Methods: A Multidisciplinary Review with Insights on Key Decision-Making Factors in Method Selection

Abstract

Drought is one of the most complex natural hazards, characterized by its slow onset, persistent nature, diverse sectoral impacts (e.g., agriculture, water resources, ecosystems), and dependence on meteorological, hydrological, and socioeconomic factors. Over the years, significant scientific effort has been devoted to developing methodologies that address its multifaceted nature, reflecting the interdisciplinary challenges of drought analysis. However, previous reviews have typically focused on individual methods, while this study presents a unified, multidisciplinary framework that integrates multiple drought analysis methods and links them to key factors guiding method selection. To address this gap, five widely used methods—index-based, remote sensing, threshold-level methods (TLM), impact-based methods, and the storyline approach—are critically evaluated from a multidisciplinary perspective. In addition, the study examines spatial and temporal trends in scientific publications, illustrating how the application of these methods has evolved over time and across regions. The primary objective of this review is twofold: (1) to provide a holistic, state-of-the-art synthesis of these methods, their applications, and their limitations; and (2) to evaluate and prioritize the critical decision-making factors, including drought type, data type/availability, study scale, and management objectives that influence method selection. By bridging this gap, the paper offers a conceptual decision-support framework for selecting context-appropriate drought analysis methods. However, challenges remain, including the vast diversity of methods beyond the scope of this review and the limited consideration of less influential factors such as user expertise, computational resources, and policy context. The paper concludes with insights and recommendations for optimizing method selection under varying circumstances, aiming to support both drought research and effective policy implementation.

1. Introduction

Droughts are among the most devastating natural hazards, causing widespread economic losses, displacing populations, and threatening food and water security [1]. Unlike rapid-onset disasters such as forest fires or flooding, droughts evolve gradually, often spanning months to years. Their creeping nature strains ecosystems, agriculture, and livelihoods long before their impacts are formally recognized [2]. Climate change has further intensified drought severity and frequency, with prolonged dry spells worsening water scarcity as well as socioeconomic vulnerabilities [3]. Consequently, droughts have led to substantial financial costs. For instance, since 1980, drought-related losses have exceeded $360 billion in the U.S. alone and €9.1 billion annually in Europe, with figures expected to rise under future climate scenarios [4,5]. Moreover, recent events, including the 2018–2019 drought in Germany (€20.7 billion in damages) and the 2022 drought in the Horn of Africa (over 37 million people affected), underscore the growing severity of drought impacts [6,7]. Understanding drought development and its intensification drivers is therefore critical for effective risk management.

Droughts typically begin as a prolonged precipitation deficit in a region, resulting in negative impacts on social, economic, and environmental systems [8]. However, additional climate factors such as high temperatures, strong winds, and low relative humidity can further exacerbate their severity [9]. Drought will most likely turn to a natural disaster if not managed properly, as seen in recent megadroughts (e.g., the 2022 drought in the Horn of Africa). Consequently, timely and accurate drought analysis is vital for informed decision-making and risk reduction.

Over the years, diverse methods have been developed to investigate droughts, each aligned with specific drought types and data requirements. For instance, meteorological droughts are analyzed using drought indices, agricultural droughts rely on soil moisture and vegetation health data, and hydrological droughts are evaluated through threshold-based methods (e.g., river discharge, reservoir levels). Remote sensing offers large-scale monitoring capabilities, while impact-based methods integrate socioeconomic data to assess localized consequences. Meanwhile, storyline approaches contextualize drought scenarios using climate models and Shared Socioeconomic Pathways (SSPs). These methods originate from diverse research disciplines, including hydrology, statistics, remote sensing, and social studies, each contributing unique strengths and limitations. A multidisciplinary review that strategically combines methods from these fields enhances the accuracy and relevance of drought investigation, ultimately improving preparedness, mitigation, and adaptation strategies.

Despite significant progress in drought research, existing literature reviews tend to focus on individual methods rather than integrating them within a comparative framework. Reviews on drought indices (e.g., [10,11]) have primarily focused on refining indicators for drought investigation. Meanwhile, remote sensing-based studies (e.g., [12,13]) highlight advancements in satellite technology but also underscore limitations in data consistency and temporal coverage. The impact-based drought analysis approaches—including impact-based forecasting (IbF) (e.g., [14,15])—have gained attention for their potential in integrating vulnerability data, yet their application remains regionally biased. Similarly, the threshold-level method (e.g., [16,17]) has been widely applied for hydrological drought assessment but requires adaptation to evolving climatic conditions [16]. More recently, the storyline approach (e.g., [18]) has emerged as a tool to stress-test and contextualize drought scenarios under climate change. Despite these advancements, there remains a gap in synthesizing these methods within a unified framework that explicitly recognizes the multidisciplinary nature of drought analysis.

This study addresses that gap by integrating five key drought analysis methods—(i) index-based, (ii) remote sensing, (iii) threshold-level, (iv) impact-based, and (v) the storyline approach—into a unified framework. It evaluates their interconnections, strengths, and limitations, and explores method selection guided by key factors such as drought type, data type/availability, study scale, and alignment with management stages. By evaluating these methods through key decision-making criteria, the study provides a decision-support tool for selecting context-appropriate methods and strengthens adaptive drought resilience by combining quantitative precision with qualitative relevance.

This paper is structured as follows:

Section 2 provides a review of five selected drought analysis methods.

Section 3 explores spatial and temporal trends in scientific publications, illustrating their evolution across regions and time.

Section 4 discusses key factors shaping method selection.

Section 5 concludes with implications and future research directions.

Drought Types

In the literature, drought has been generally classified into four types: meteorological, agricultural, hydrological, and socioeconomic [19]. This classification is widely recognized, because it reflects the complex interactions between climate drivers and sectoral impacts [20,21]. However, flash drought has recently been introduced as a subset of these four types, characterized by unusually rapid intensification, either initiating a new drought or intensifying an existing one [22,23]. Meteorological drought is commonly defined as an intensity of dryness characterized by a lack of precipitation compared to the average amount, as well as the duration of the dry period for a specific region [24]. Agricultural drought links precipitation deficits to elevated evapotranspiration and soil moisture depletion, leading to crop failure [25]. Hydrological drought results from a prolonged meteorological drought, causing declines in surface and subsurface water resources that fail to meet the demands of a specific water management system [26]. Socioeconomic drought arises when water shortages disrupt economic activities, such as agriculture, energy, tourism, and municipal supply systems [10]. Figure 1 illustrates these drought types, their key drivers, and the complex interlinkages among meteorological, hydrological, agricultural, and socioeconomic droughts.

2. Review of Drought Analysis Methods

2.1. Index-Based Method

2.1.1. Definition and Introduction

Drought indices are standardized metrics that simplify complex hydroclimatic data to objectively assess drought characteristics such as severity, duration, and spatial extent [27]. They are essential for identifying vulnerable regions and understanding the temporal and spatial patterns of drought. Therefore, it is essential to rigorously test the suitability of drought indices across diverse geographic locations to ensure their effective and context-specific application [28]. Evaluating correlations between indices and observed drought impacts helps determine their regional relevance, acknowledging the diverse nature of drought effects [27].

Drought indices are calculated using climatic and hydrometeorological variables, such as precipitation, temperature, streamflow, and soil moisture, commonly referred to as indicators [29,30]. Since different drought types reflect different parts of the hydrological cycle, no single index can capture all drought conditions across sectors or climates [31,32]. Even the same type of drought can be assessed using different drought indices or their combination [33]. Several factors influence the choice of an appropriate index or combination of indices, including drought type, climate conditions, data availability, affected sectors, regional characteristics, and desired temporal and spatial scales of analysis [32,34,35].

2.1.2. Classification of Drought Indices

Drought indices are broadly classified based on two principles: (1) drought type and (2) methodology [30]. The first category includes indices designed for specific drought types—meteorological, hydrological, agricultural, and socioeconomic—using relevant indicators. The second is based on computation method and data source, including standardized, remote sensing, composite, and model-based indices.

This dual categorization provides a systematic framework for analyzing the diverse methodologies and applications of drought indices. Standardized indices, such as the Standardized Precipitation Index (SPI) [36], are widely used and derived from historical data fitted to a probability distribution and transformed to standard normal scores, allowing comparison across time and space [37]. Negative values indicate drought, while positive values indicate wet conditions. Figure 2 illustrates threshold classifications as an example for the SPI index. Similar methods have produced other standardized indices, such as the following:

• Standardized Precipitation Evaporation Index (SPEI) [38];
• Standardized Streamflow Index (SSFI) [39];
• Standardized Runoff Index (SRI) [40];
• Standardized Soil Moisture Index (SSMI) [41];
• Standardized Groundwater level Index (SGI) [42];
• Standardized Snow Melt and Rain Index (SMRI) [43];
• Standardized Relative Humidity Index (SRHI) [44];
• Standardized Terrestrial Water Storage Index (STI) [45];
• Standardized Water Supply and Demand Index (SWSDI) [46];
• Multivariate Standardized Drought Index (MSDI) [29].
• Although model-based drought indices, such as the Palmer Drought Severity Index (PDSI) [47], can also be considered as standardized indices, their derivation is based on a different modeling approach [16]. Remote sensing indices utilize satellite data as indicators and will be discussed in detail under Remote sensing (Section 2.2).

Figure 2. SPI classification scheme based on drought severity.

Figure 2. SPI classification scheme based on drought severity.

Given the multivariate nature of droughts, relying on a single index often provides an incomplete assessment [30]. Composite or multivariate indices, such as the Aggregate Dryness Index (ADI) [48] and Joint Drought Index (JDI) [49], integrate multiple variables (e.g., precipitation, streamflow, etc.) and offer broader analysis of drought. In the literature, such indices have been described by various terms, including hybrid, aggregated, composite, multi-scalar, combined, or comprehensive (e.g., [50,51]).

Over time, drought analysis has evolved through diverse methodologies, shifting from simple, single-variable indices (e.g., SPI) to complex, multi-source frameworks (e.g., CDI, GIDMaPS, SPESMI), integrating satellite and socioeconomic data for real-time and predictive monitoring [27,52]. Figure 3 chronologically maps this evolution over the last six decades, categorizing indices by ease of use—green for simple indices (e.g., SPI), yellow for moderately difficult indices (e.g., ADI), and red for complex indices (e.g., model-based ETDI)—based on data accessibility, computational demands, and operational adoption [27].

2.1.3. Application Selection Criteria of Indices

The suitability of a drought index depends on its indicators as inputs, which determine its sensitivity to specific drought types. Meteorological indices use indicators like precipitation, temperature, and evapotranspiration, while hydrological indices incorporate surface and subsurface water levels, and agricultural indices include soil moisture and vegetation health. However, the adaptability of a particular drought index across timescales enables its application in assessing multiple drought types. For example, the SPI, though originally designed for meteorological drought, is commonly applied to other types depending on its timescale. SPI-3 (3-month) often correlates with agricultural drought, while SPI-6 and SPI-12 align more closely with hydrological drought [53,54,55,56,57,58]. The groundwater drought, a subcategory of hydrological drought, caused by a prolonged period of low precipitation, high evapotranspiration, and high demand for groundwater, has shown a promising correlation with the SPI-12 [59,60,61,62].

This flexibility has been demonstrated in real-world applications. For instance, Sutanto et al. assessed the potential of using SPI-6 forecasts to predict hydrological droughts in streamflow and groundwater across diverse catchments, emphasizing the influence of catchment properties on drought propagation [57]. Wang et al. applied SPI together with relative yield data to evaluate maize vulnerability to drought in northwest China, showing how SPI-derived meteorological drought indicators relate to agricultural impacts [58]. Similarly, Samantaray et al. compared SPI with the newly developed rainfall pattern drought index (RPDI) in India, demonstrating how alternative indices that incorporate rainy-day frequency provide a more nuanced assessment of drought severity and spatial variability, particularly under future climate scenarios [56]. These examples illustrate how index selection, timescale adaptation, and incorporation of additional rainfall characteristics can support nuanced and integrated drought assessments across sectors and regions.

2.1.4. Limitations, Challenges and Future Directions

Conventional drought indices face notable limitations that restrict their effectiveness under contemporary environmental conditions. A key issue is their fundamental assumption of stationarity (e.g., in standardized indices), which presumes stable statistical properties of variables over time. In reality, climate change and human activities have disrupted historical precipitation, temperature, and groundwater patterns, diminishing the reliability of traditional probability models for capturing non-stationary trends [63]. In response, recent efforts have developed non-stationary frameworks, such as the time-dependent SPI (SPIt), which offer improved drought assessments but still grapple with challenges like parameter uncertainty and long data requirements [64,65,66].

Moreover, most indices reduce drought to single-variable metrics, overlooking complex interactions among climatic drivers (e.g., atmospheric circulation shifts) and anthropogenic factors (e.g., land-use change, water extraction). Data limitations—especially the lack of long-term, high-quality hydroclimatic records—also hinder calibration, particularly in data-scarce regions (e.g., where >30 years of data are recommended [67]). Additionally, the monthly resolution of many indices fails to capture flash droughts, which require finer temporal granularity [68].

In light of these limitations, emerging trends emphasize AI and machine learning (ML) to enhance drought index performance and forecasting. Integrating indices into ML-based models improves early warning systems by capturing non-linear patterns across diverse datasets [69,70]. A prominent example is the ANYWHERE [71] Drought Early Warning System (AD-EWS), which fuses multiple standardized indices (SPI, SPEI, SRI, SGI) with satellite and climate model data, delivering seasonal forecasts at 5 km resolution with lead times of up to seven months [72]. Compared to earlier systems like the EDO-DEWS, which provided shorter-term monitoring and forecasts (e.g., up to three months for meteorological drought and seven days for soil moisture anomalies), AI-enhanced platforms offer greater spatial detail and temporal foresight. Several recent case studies have further demonstrated the potential of AI for drought investigation. For instance, in northwestern Iran, wavelet-hybridized AI models, including multilayer perceptron neural networks (W-MLPNN) and support vector regression (W-SVR), were applied to forecast the SPEI at 1-, 6-, and 12-month timescales using historical precipitation and temperature data (1987–2018). The study found that W-MLPNN performed best for short-term forecasts (1 and 6 months), while W-SVR was more effective for longer-term forecasts (12 months), illustrating the value of hybrid AI models for improving drought prediction across varying temporal scales [73]. In Australia, AI-driven drought indices developed using models such as Decision Trees, Generalized Linear Models, Support Vector Machines, and Random Forests outperformed conventional drought indices, achieving stronger correlations with hydrological indicators like soil moisture and runoff [74]. These examples illustrate how AI can contribute to more accurate, localized, and actionable drought forecasting and risk management.

Collectively, these limitations highlight the need for adaptive drought assessment frameworks that incorporate non-stationary dynamics, multi-source data integration, and interdisciplinary metrics. AI-driven methods show strong potential to meet these demands by enhancing index flexibility, predictive accuracy, and decision-making relevance under evolving climate conditions.

2.2. Remote Sensing Method

2.2.1. Definition and Introduction

When it comes to investigating drought with extensive geographical coverage, real-time monitoring, and high temporal resolution, remote sensing is a favored choice in climatic research [75]. Its capability to capture dynamic environmental parameters such as vegetation health, soil moisture, and groundwater storage makes it indispensable for drought assessment and early warning systems. Remote sensing provides continuous and consistent data for drought characterizations at both global and regional scales, making it especially valuable for regions with no or sparse in situ data [76,77].

2.2.2. Remote Sensing’s Contribution to Drought Monitoring

Numerous studies underscore the direct and indirect contributions of remote sensing to drought analysis and management [12]. It utilizes data from active and passive satellite sensors across optical, thermal, and microwave domains to capture a wide range of environmental indicators relevant to drought conditions [78].

• Direct contribution

Active sensors (e.g., radar, LiDAR) emit their own energy and are capable of operating in all weather conditions, making them suitable for monitoring surface deformation and vegetation structure [79]. Passive sensors, reliant on external energy (e.g., solar energy), include optical sensors that capture visible and infrared light to monitor vegetation and land surface characteristics, and thermal sensors that measure emitted heat, providing insights into water stress and surface temperatures. This combination of sensor types offers a comprehensive approach to drought monitoring, providing valuable data on environmental changes and water availability [80].

Figure 4 illustrates how these sensors capture reflected sunlight and emitted thermal radiation, providing crucial data on vegetation health, soil moisture, and water stress.

Among optical-based indices, Vegetation Indices (VIs) such as the Normalized Difference Vegetation Index (NDVI) [81] and Enhanced Vegetation Index (EVI) [82] are widely applied to monitor vegetation vigor. EVI improves upon NDVI by minimizing atmospheric and soil background influences [83]. However, both indices are sensitive to ecosystem variability and may not reliably detect early drought stress [84]. To address this limitation, Kogan [85] developed the Vegetation Condition Index (VCI), which compares current NDVI/EVI values to historical observations from the same period in previous years [86,87]. Alongside the NDVI, the Normalized Difference Water Index (NDWI) [88], derived from NIR and shortwave infrared (SWIR) bands, is effective for detecting soil and canopy moisture, especially during early crop development stages [89].

Thermal infrared (TIR) data provide land surface temperature (LST), which responds more quickly to water stress than Vegetation Indices alone [90]. Neglecting temperature effects and solely relying on vegetation data will not be able to capture drought dynamics on time. Therefore, integrating thermal channels data can provide more detailed insights into drought conditions [91]. Rising leaf temperatures, driven by stomatal closure and decreased latent heat flux during water stress, signal early drought onset [92,93]. Similar to NDVI, due to its high sensitivity to regional characteristics, direct use of LST for drought assessment is limited. To overcome this, the Temperature Condition Index (TCI) [91] was developed through the normalization of LST data. Several other LST-based drought indices are applied for drought analysis, such as the Crop Water Stress Index (CWSI) [94] and Normalized Difference Temperature Index (NDTI) [95].

Due to the limitations of single-factor indices like VCI and TCI that focus solely on vegetation (e.g., NDVI) or temperature (e.g., LST), combined indices such as the Vegetation Health Index (VHI) [91] have been developed integrating VCI and TCI, providing a more holistic measure. To consider the role of precipitation, Rhee et al. [96] introduced the Scaled Drought Condition Index (SDCI) which combines the scaled NDVI, LST (from MODIS), and the Tropical Rainfall Measuring Mission (TRMM)-based precipitation data using weighted integration for improved drought monitoring. More recent remote sensing-based drought indices include the Probabilistic Precipitation Vegetation Index (PPVI) [97], merging SPI and VHI to assess meteorological and agricultural drought, and the Type Response-Aided Drought Index (TRADI) [98], which accounts for vegetation-type-specific responses to drought stress.

The practical value of these optical and thermal indices has been demonstrated in several regional applications. For example, Le Page and Zribi analyzed the combined use of optical and microwave satellite for drought severity mapping in Northwest Africa, highlighting the consistency between these indicators in quantifying dry conditions over a decade [99]. Similarly, Ramat et al. developed a novel drought severity index combining optical and Synthetic Aperture Radar (SAR) data to monitor drought patterns and water status in agricultural areas of Egypt and Tunisia, demonstrating the benefits of sensor synergy for improved drought characterization [100].

• Indirect contribution

Complementing its direct applications, remote sensing provides crucial variables that serve as critical inputs for broader drought investigation frameworks. While optical and thermal indices are primarily used to monitor droughts associated with land cover changes—such as vegetation stress and surface water dynamics—microwave remote sensing offers advantages for detecting drought types less influenced by surface conditions, such as meteorological drought [101]. For example, the Microwave Integrated Drought Index (MIDI) [102] uses microwave signals, which are less affected by atmospheric interference, to improve meteorological drought monitoring.

Satellite-based data enhance the spatial resolution of indices such as SPI and SPEI, especially in data-sparse regions [103], with numerous datasets available for precipitation, temperature, soil moisture, and evapotranspiration (Appendix A). Multi-satellite products like TRMM and the Global Precipitation Measurement (GPM) address rain gauge scarcity [104], while reanalysis datasets (Appendix B) integrate observations and models to estimate precipitation and other meteorological variables [105]. However, their performance varies by topography, climate zone, and station density [106], necessitating validation for regional applications.

Beyond precipitation, gravity-based remote sensing—especially through the Gravity Recovery And Climate Experiment (GRACE) satellites—offers a unique means to monitor changes in total terrestrial water storage (TWS), including surface water, soil moisture in the root zone, and groundwater [107]. GRACE-derived Terrestrial Water Storage Anomalies (TWSAs) [108] have proven particularly valuable in hydrological drought studies, offering broader spatial coverage than traditional well-based groundwater monitoring [109,110]. TWS data are available from 2002 to the present, except for a gap during 2017–2018 [111], and are widely used in groundwater drought assessments [61,112,113,114,115].

Soil moisture is another key variable for agricultural drought analysis, as it complements vegetation-based indices by capturing early or delayed plant responses to water stress [116,117,118]. Remote sensing has significantly improved soil moisture monitoring by enabling large-scale and continuous assessments. Microwave remote sensing, through missions like the Advanced Microwave Scanning Radiometer for the Earth Observation System (AMSR-E) [119], and the Soil Moisture and Ocean Salinity (SMOS) [120], provide surface soil moisture estimates (0–5 cm), though it is limited by shallow penetration depth [121].

Recent advancements in data assimilation allow the extrapolation of surface data to estimate root-zone moisture, enhancing the value of remote sensing in hydrological and agricultural drought contexts. The integration of remote sensing with land surface models, as seen in systems like the Global Land Data Assimilation System (GLDAS) [122], has revolutionized our ability to monitor soil moisture globally, offering unprecedented insights into soil–water dynamics and aiding in better drought management.

Moreover, the European Center for Medium-Range Weather Forecasts Interim Reanalysis (ECMWF) [123], the National Aeronautics and Space Administration (NASA) Modern-Era Retrospective Analysis for Research and Applications [124] and several other reanalysis datasets incorporate various observations and models to provide global soil moisture estimates. For a more comprehensive review, please see the work by Petropoulos et al. [125].

2.2.3. Limitations, Challenges and Future Directions

Despite its strengths, remote sensing faces several limitations in drought monitoring. Cloud cover frequently obstructs optical sensors, particularly in humid and semi-humid regions, reducing data availability during critical periods [126]. However, radar-based sensors (e.g., Sentinel-1, ALOS PALSAR) offer an all-weather alternative, and data fusion techniques are increasingly mitigating this issue. For instance, combining soil moisture datasets using methods like Triple Collocation and Linear Weight Fusion has enhanced agricultural drought monitoring in China [127], while advanced weighted averaging techniques such as ORNESS-OWA have been successfully applied to fuse multiple precipitation datasets for improved meteorological drought estimation [128].

A key challenge remains the trade-off between spatial and temporal resolution. Coarse-resolution datasets like GRACE (~300 km) are less suited for localized assessments, though downscaling methods and improvements in its follow-on mission, GRACE-FO, have improved the resolution [129,130]. New missions such as the Surface Water and Ocean Topography (SWOT) (launched in 2022) offer ~1 km resolution for surface water monitoring, albeit without subsurface sensitivity. In contrast, high-resolution sensors (e.g., Sentinel-2, Landsat) capture detailed spatial information but lack the spectral capabilities (e.g., microwave bands) required to directly monitor soil moisture and groundwater, limiting their ability to monitor soil moisture and groundwater [131]. Additionally, the high cost of commercial satellites (e.g., PlanetScope, WorldView) limits their accessibility for routine drought analysis. Another significant challenge is validation due to limited and inconsistent ground-truth data, particularly in developing regions. However, integrating multi-source datasets, crowdsourced observations, and reanalysis products (e.g., ERA5, GLDAS) is improving validation frameworks.

Future efforts should prioritize advanced data integration strategies, the development of machine learning techniques for automated drought classification, and stronger synergies between satellite observations and in situ measurements. Notably, recent studies have demonstrated meaningful progress in these directions, illustrating how AI and remote sensing can be effectively combined to enhance drought monitoring and forecasting. For example, in Bangladesh, Al Kafy et al. applied Cellular Automata-Artificial Neural Networks (CA-ANNs) using Landsat-derived indices (e.g., NDVI, MNDWI, VHI) to assess and predict agricultural drought vulnerability, projecting an increase in extreme drought conditions from 7% in 2021 to 18% by 2031 [132]. Similarly, in Türkiye, Akcapınar and Çakmak combined MODIS-based drought indices (VCI, TCI, VHI, VSWI) with machine learning models to predict wheat yield under drought conditions, achieving high predictive accuracy (R² up to 0.97) and demonstrating the potential for early warning applications in agriculture [133]. These examples highlight the promising role of AI-integrated remote sensing frameworks in addressing key challenges in drought monitoring and preparedness.

2.3. Threshold-Level Method (TLM)

2.3.1. Definition and Introduction

TLM defines hydrological droughts by comparing streamflow, reservoir, or groundwater levels to predefined thresholds, often derived from historical percentiles or operational targets [134]. Its independent applicability in drought monitoring systems (e.g., low-flow alerts, reservoir management) and its flexibility in analyzing diverse hydrological variables justify its role as a core methodology in hydrological drought research. Unlike comprehensive hydrological models such as SWAT, VIC, and HBV, which simulate general water balance processes, TLM focuses on low-flow conditions, making it particularly suitable for direct drought investigation. Streamflow simulations from models like SWAT or VIC can be post-processed using TLM to quantify future drought frequency [135,136].

Two distinct approaches of the TLM method are applied in hydrological drought analysis—the fixed threshold method (constant over time) and the variable threshold method (seasonally adjusted)—as demonstrated in previous studies [137,138,139,140]. Unlike standardized indices, TLM uses raw time-series data, capturing actual water deficits rather than normalized deviations [137], which provide information about the real water deficit rather than relative deviations from a long-term statistical mean.

Figure 5 illustrates the key components of this method.

2.3.2. Application Selection Criteria

The TLM method, first introduced by Yevjevich [17] and later elaborated by Zelenhasić & Salvai [142], is commonly used for hydrological drought monitoring, particularly in low-water conditions, using time-series data of river discharge, reservoir levels, and groundwater [16]. While TLM is predominantly applied in hydrological contexts, a few studies have extended its use to other drought types. For example, Santos et al. [143] applied the TLM method to rainfall data to monitor meteorological drought, and this approach was later extended by Portela et al. to detect meteorological drought at early stages of development in southern America [144]. Moreover, the TLM method has been used for agricultural drought assessment, integrating soil moisture, precipitation, surface water, and groundwater data through various hydrological models [145,146,147,148]. Several regional studies have demonstrated the practical use of the TLM method in drought monitoring and analysis. For example, Heudorfer and Stahl applied constant and variable threshold-level methods to observed precipitation, streamflow, and groundwater data across Germany to evaluate drought propagation characteristics, highlighting how threshold choice influences drought statistics [137]. Sarailidis et al. used fixed and variable thresholds for streamflow drought analysis in a semi-arid basin in Cyprus, demonstrating the sensitivity of threshold selection for severity–duration–frequency assessments [139]. Similarly, Rivera et al. applied TLM to develop a streamflow drought climatology in the Central Andes of Argentina, linking drought patterns to large-scale climate drivers like La Niña and the Pacific Decadal Oscillation [149].

The threshold selection for perennial rivers in the TLM method may not be suitable for non-perennial rivers. This is because the TLM method, as defined, tends to overestimate drought duration in drier regions by incorrectly classifying periods of zero runoff as drought events. In non-perennial rivers, these zero runoff periods may be a natural part of the river’s flow regime rather than indicative of drought conditions [134]. Thresholds between the 70th and 90th percentiles are appropriate for perennial rivers, based on their exceedance frequencies [16]. Upon evaluating the exceedance frequency of relevant variables during the control period, it is concluded that the 80th percentile (representing the value exceeded 80% of the time) is the most common choice [137,150]. However, a recent study by Cammalleri et al. suggests the 85th percentile of exceedance (Q85) to estimate drought predictions under both current and future climate scenarios [151].

2.3.3. Advantages, Limitations and Future Directions

The TLM method has some advantages over drought indices. It can be applied on shorter timescales, detecting drought occurrences on a daily basis, unlike many drought indices that primarily rely on monthly aggregations [139,152]. Additionally, TLM enables flexible threshold adaptation, allowing thresholds to be adjusted based on specific sectoral needs, such as irrigation, industry, drinking water supply, navigation, and ecological flow preservation [153]. This flexibility makes TLM effective for localized drought monitoring, especially at the catchment level, and enables the computation of deficit volume, crucial for drought management [145].

However, TLM faces challenges that limit its broader application. The absence of a standardized definition for threshold levels leads to inconsistencies in drought categorization across regions [146]. Additionally, threshold determination can be subjective, varying based on hydroclimatic conditions, data availability, and local water management policies. This subjectivity can lead to inconsistencies in drought analysis, particularly in regions with limited hydrological data [154]. Another challenge in drought research is the impact of climate change, which alters hydrological regimes and introduces significant variability in river temperature dynamics. This variability makes fixed threshold levels less reliable over time, as the interactions between climate and local hydrological factors can lead to unpredictable changes in river conditions [155]. Addressing these challenges requires the following:

• Developing standardized methods for threshold selection across different hydroclimatic regions.
• Leveraging machine learning techniques for adaptive, dynamic threshold setting based on real-time hydrological data.
• Incorporating probabilistic and non-stationary approaches to ensure that threshold levels remain relevant under climate change.
• Enhancing data assimilation techniques by combining ground observations, satellite data, and reanalysis datasets to improve TLM’s applicability in data-scarce regions.

2.4. Impact-Based Method

2.4.1. Conceptual Foundation and Introduction

Effective drought risk mitigation requires understanding how hazards translate into tangible socioeconomic and environmental impacts. According to the IPCC [156], disaster risk is determined by three inter-related components: hazard, exposure, and vulnerability. Hazards refer to potential damaging events (e.g., drought), exposure involves the population or assets that may be affected, and vulnerability indicates the susceptibility of these elements to harm due to socioeconomic or environmental factors. The relationship between these three determinants can be mathematically expressed as follows [8,157]:

Risk = Hazard × Exposure × Vulnerability

(1)

This equation underscores that disaster risk is not solely a function of the hazard itself but is also shaped by the extent of exposure and the degree of vulnerability. Conventional drought assessments predominantly address the hazard dimension by quantifying hydroclimatic anomalies (e.g., precipitation deficits, soil moisture depletion). However, they often overlook the diverse impacts resulting from drought events [15].

To better address these impacts, integrating exposure and vulnerability into drought analysis is crucial. For example, a region frequently affected by droughts (high hazard) may still suffer minimal losses if it has resilient infrastructure and water management systems (low vulnerability), while another region with limited resources (high exposure) might experience substantial impacts from moderate droughts. This underscores the importance of shifting from hazard-focused to risk-informed drought analysis.

Building on this concept, impact-based forecasting (IbF) represents a recent paradigm shift in early warning services, moving beyond conventional physical forecasts to predict what droughts will do rather than just when and where they will occur [14,15]. The World Meteorological Organization (WMO) recommends a multi-hazard IbF approach to improve preparedness and reduce the impacts of concurrent hazards [14]. The risk theory-based IbF facilitates the integration of diverse methodologies and encourages collaboration across disciplines and stakeholders [15].

Drought impacts span agriculture, water security, energy production, forest health, greenhouse gas emissions, and employment, particularly in agriculture [34,158], see Appendix C for a network analysis of drought impacts. Addressing these impacts effectively requires context-specific analysis, considering geographical, sectoral, and temporal characteristics [159]. To address this, researchers increasingly rely on diverse data sources, including agricultural statistics, impact bulletins [160] and media reports [161,162,163,164,165,166].

This conceptual shift is exemplified by several case studies. For instance, Chen et al. investigated the 1928–1931 drought in China, which documented hierarchical and cascading drought impacts across natural, support, and human systems, revealing the non-linear decline of social resilience over time [164]. Similarly, Sodoge et al. advanced impact monitoring through AI-driven analysis of drought-related newspaper reports, enabling high-accuracy detection of sector-specific impacts and their spatial–temporal evolution across Germany [166]. Furthermore, Erfurt et al. contextualized two centuries of droughts in southwestern Germany, showing how societal vulnerability and adaptive responses have evolved to mitigate impacts on critical sectors like food supply [167].

2.4.2. Global Drought Impact Databases

To support comprehensive impact evaluations, dedicated drought impact databases have been developed. The Drought Impact Reporter (DIR), launched by the National Drought Mitigation Center [24], offers a structured summary of drought impacts in the United States. In Europe, the European Drought Impact Report Inventory (EDII) [168] was developed under the Drought R&SPI project to collect and categorize drought impact reports, initially focused on project partner countries and key drought events. An updated version [169] expanded its temporal and geographic coverage, capturing environmental, economic, and social drought impacts across Europe since 1900. The EDII has supported numerous impact-based studies [159,170,171,172,173]. The EDII was later extended to cover the Alpine region [174].

These databases enable more robust risk assessments and highlight the need to incorporate impact-based practices into operational drought early warning systems [15], reinforcing the global scientific imperative to systematically assess drought impacts.

2.4.3. Methodological Workflow and Data Sources

Impact-based drought investigation employs diverse methodologies to link socioeconomic impacts with hydrometeorological drivers. Typically, studies begin with data collection from structured sources (e.g., DIR, EDII), media reports, agricultural statistics, and historical archives such as tambora.org. These data are categorized by sectors such as agriculture, water supply, and energy and then analyzed using statistical (e.g., correlation, regression) or machine learning techniques (e.g., Random Forest). Proxy indicators such as impact frequency or economic losses often serve as substitutes for direct impact measurements.

Appendix D summarizes the methodologies and datasets used in recent studies, highlighting a concentration of research in Europe and the U.S. due to the availability of comprehensive databases. Studies in other regions typically rely on ad hoc data sources [161], resulting in less consistent and comparable findings. Agriculture, water supply, and energy remain the most commonly assessed sectors due to their high vulnerability to drought. Analytical workflows vary significantly across studies. Statistical approaches—such as logistic regression by Blauhut et al. [170] and correlation analysis by Bachmair et al. [175]—have been used to quantify sector-specific impacts and validate hazard–impact relationships. In contrast, machine learning methods (e.g., [176,177]) are employed to capture complex interactions between variables. Fuzzy logic addresses uncertainties in regions with sparse data [159]. Drought indices (e.g., SPI, SPEI) commonly support validation by identifying impact thresholds or forecasting impacts in advance [177].

2.4.4. Advantages, Limitations and Future Directions

Despite growing use, impact-based drought analysis faces key challenges that hinder standardization and broader application. These discrepancies arise from the following fundamental challenges:

• Data granularity: Sources range from structured databases to narrative reports, leading to inconsistencies in impact quantification;
• Impact categorization: Sector-specific versus multi-sectoral classifications limit cross-regional comparisons;
• Spatiotemporal scales: Studies vary from event-specific (e.g., 2018–2019 Central European drought; [159]) to multi-decadal analyses (e.g., 19th-century droughts; [167]), limiting generalizability.

Methodologically, the approaches also present limitations. Fuzzy logic offers a solution for uncertainty but lacks cross-scale consistency. Machine learning techniques provide predictive capacity but depend heavily on quality data for training, often requiring region-specific calibration. Furthermore, reliance on text-based reports (e.g., media reports) introduces subjectivity and biases, particularly in regions with sparse reporting.

Despite these challenges, impact-based methodologies share a common goal: forecasting and mitigating drought impacts by applying historical hazard–impact relationships to future scenarios. However, their fragmented nature—ranging from region-specific calibrations [176] to continental-scale databases [169]—highlights the absence of a universal framework [178]. Future advancements should focus on three strategic priorities:

• Standardization: Expand or adapt existing systems (e.g., EDII, DIR) and develop new protocols tailored to under-represented regions like Africa and South Asia.
• AI Integration: Utilize advanced AI tools—such as (BERT) [179] for automated text mining of drought impacts from news and reports—to enhance impact detection and forecasting. Notably, recent studies have demonstrated the potential of AI and machine learning in impact-based drought analysis. For example, Zhang et al. applied an explainable machine learning framework (XGBoost with SHAP) to predict complex drought impacts across the U.S., improving both the accuracy and interpretability of multidimensional impact forecasts [180]. Similarly, Sodoge et al. developed an automated approach combining natural language processing and machine learning to detect and map drought impacts from 40,000 newspaper articles in Germany, achieving classification accuracy exceeding 90% [166].
• Global Collaboration: Promote data sharing, expand global drought impact databases, and implement unified metadata standards to achieve comprehensive, equitable impact representation.

2.5. Storyline Approach

2.5.1. Conceptual Framework and Introduction

The storyline approach has emerged as an innovative method for analyzing and communicating climate risks, especially for extreme events like droughts [181,182]. By constructing plausible narratives of how climate events might unfold, it complements probabilistic methods, offering clearer, more relatable insights for public understanding and stakeholder engagement [183].

In climate science, the concept of “storyline” has evolved across disciplines. The IPCC AR5 defines storylines in scenario-based approaches as qualitative descriptions of plausible climate futures, shaped by key drivers and interactions [156]. The IPCC AR6 refines this with Physical Climate Storylines (PCS), emphasizing the physically consistent unfolding of past or future extreme events [184]. Unlike traditional probabilistic approaches, storylines are particularly user-relevant and effective in supporting decision-making, stress-testing, and policy development [156].

Baulenas et al. [185] identified three main types of storyline approaches:

• Scenario-Based (SB)—focusing on long-term developments and drivers;
• Discourse-Analytical Approach (DAA)—examining how risks are framed and communicated;
• Physical Climate Storylines (PCSs)—focusing on physical processes behind extreme events.

Among these, PCS, introduced by Hazeleger et al. [186], has gained prominence for investigating droughts by simulating how they might evolve under varying climate conditions [187,188,189]. PCS facilitates analysis of event-specific scenarios based on physical plausibility rather than statistical probability [184,190,191].

Event-based PCS storylines are especially useful for “stress-testing”, a practice in disaster risk management that uses hypothetical yet plausible scenarios to evaluate system vulnerabilities and resilience [191]. In drought studies, this often involves re-running historical droughts in climate models under future warming scenarios to assess changes in intensity, duration, and frequency [192]. This ‘what-if’ methodology helps researchers enhance understanding of future drought risk in a changing climate.

2.5.2. Integration of GCMs, SSPs, and RCPs with the Storyline Approach

The storyline approach integrates General Circulation Models (GCMs), Shared Socioeconomic Pathways (SSPs), and Representative Concentration Pathways (RCPs) to contextualize drought events within human and emissions scenarios [193]. GCMs (e.g., CMIP6 ensembles) offer comprehensive quantitative projections of future climate conditions, such as precipitation deficits, temperature anomalies, and soil moisture trends, which are essential for assessing the probability and severity of future extreme events, including droughts. For instance, Aalbers et al. [187] used climate models to simulate the 2018 drought under 1.5 °C, 2 °C, and 3 °C warming, showing how its severity and frequency could increase in a warmer world.

SSPs complement this by incorporating the socioeconomic context into these narratives. For example, SSP3 reflects limited cooperation and low adaptive capacity, whereas SSP5 illustrates high economic growth and increased water demand. Studies such as Bjarke et al. [194] reveal that drought risks shift dramatically across these pathways, from moderate impacts under SSP1–2.6 to prolonged and severe events under SSP5–8.5.

Meanwhile, RCPs represent a range of greenhouse gas concentration trajectories and associated radiative forcing levels, which shape the magnitude of climate impacts. For instance, RCP2.6 describes a low-emission pathway requiring significant and rapid reductions in greenhouse gas emissions, while RCP8.5 reflects a high-emission scenario with continued increases in emissions [195]. Applying RCP8.5 to historical droughts, as Van Tiel et al. [192] demonstrated, leads to intensified low-flow conditions, especially in glacier-dependent river basins.

Altogether, this integration of GCMs (physical climate), SSPs (socioeconomics), and RCPs (emissions) enables a multidimensional analysis of future drought risks that supports decision-making in both science and policy.

2.5.3. Methodological Workflow and Applications

The storyline approach has been used in a variety of studies with two main objectives: understanding how extreme droughts could change under future climate conditions, and stress-testing systems to inform adaptation planning [184,186]. This method focuses on a historically significant drought event or series of events with the contribution of future global warming scenarios to understand how the severity, frequency, and impacts of the event may intensify [196].

Figure 6 outlines a generalized methodological workflow, informed by studies summarized in Appendix E. The process begins with selecting a historical drought event, followed by defining future climate scenarios using global warming levels (e.g., +1.5 °C, +2 °C). Key drivers are then identified, including emission trajectories (RCPs) and socioeconomic trends (SSPs). These drivers help generate modeled drought scenarios, which are validated using historical data. Finally, the outcomes are transformed into stakeholder-relevant storylines (e.g., “Under RCP8.5–SSP5, drought frequency may be tripled by 2050”), enabling more accessible communication of complex risks.

Several recent studies demonstrate the versatility of the storyline approach in assessing drought risks under diverse climate conditions. For example, van Tiel et al. applied stress-test storylines to model future low-flow conditions in the Rhine basin, highlighting how glacier retreat could exacerbate extreme low flows during drought events under RCP8.5 scenarios [192]. Similarly, Gessner et al. generated multi-year drought storylines for Europe and North America using an iterative ensemble resampling method, creating physically consistent yet extreme drought scenarios to assess soil moisture recovery times and stress-test adaptation strategies [183]. In the UK context, Chan et al. leveraged seasonal hindcasts to produce alternative drought storylines during the 2022 event, offering valuable insights for water resource planning by exploring plausible worst-case hydrological scenarios for subsequent seasons [197]. These studies illustrate how storyline approaches can inform adaptation strategies and stress-test water management systems under complex climate and socioeconomic futures.

2.5.4. Limitations, Challenges and Future Directions

While the storyline approach provides detailed, event-specific insights into drought risk under climate change, it faces several methodological challenges. One primary issue is the lack of standardized protocols across studies [200]. Diverse frameworks—such as PCS, ensemble resampling, and pseudo-global warming experiments—complicate cross-comparisons and hinder generalization [18,183,187]. Moreover, global-scale analyses often rely on CMIP6 projections, which, despite their comprehensiveness, can oversimplify interactions between climate variables and may fail to capture regional dynamics accurately [194,198].

Another methodological challenge lies in the heterogeneity of focus variables. Some studies prioritize soil moisture and streamflow [192], while others focus on precipitation or atmospheric drivers [183], making standardization difficult. Downscaling GCM data to capture localized impacts and interpreting complex societal developments embedded in SSPs also remain unresolved challenges [185,194].

To enhance its robustness, future research should explore hybrid frameworks that combine storylines with probabilistic models or machine learning [197]. Integrating quantitative climate modeling with qualitative socioeconomic narratives can improve policy relevance, as shown by Fossa Riglos et al. [199]. Furthermore, involving stakeholders in co-developing storylines can better align scientific outputs with local adaptation planning.

Advancements in ensemble resampling (e.g., [183]) can produce more plausible extreme drought scenarios while reducing dependence on rigid assumptions.

Addressing these limitations will enhance the credibility, applicability, and global relevance of the storyline approach for drought risk management in a changing climate.

3. Spatiotemporal Trends in Drought Method Studies

This section analyzes the spatial and temporal distribution of scientific publications on drought analysis methods reviewed in this paper, covering trends from 1981 to 2024.

3.1. Data Source and Search Strategy

To evaluate publication patterns, we used Scopus as the primary data source, chosen for its wide coverage of peer-reviewed literature [201]. The search strategy was developed following recommendations from [31], with screening based on titles, abstracts, and keywords. We applied separate queries for each method discussed in the review, restricting results to English-language publications. Full query codes are available in Appendix F. Moreover, the spatial analysis relies on automated extraction of country names from publication metadata, which may omit studies lacking explicit location references. Manual review, though more accurate, is beyond this paper’s scope. Excluding global studies may also reduce regional representativeness.

3.2. Temporal Trends

The temporal trends demonstrate a marked increase starting from 2004, peaking in 2024 (Figure 7). Index-based methods dominate the literature throughout, followed by remote sensing approaches. Both trends reflect the overall growth in drought research. In contrast, impact-based, threshold-level methods, and the storyline approach appear less frequently, reflecting their more recent emergence in the field.

3.3. Spatial Distribution

The spatial distribution of drought research varies significantly across countries, with China and the United States leading the overall literature (Figure 8a). This trend is consistent across specific drought analysis methods, including index-based, remote sensing, and the threshold-level method, where China is particularly dominant (Figure 8b,c,e). Conversely, European countries are prominent in impact-based and storyline-related studies (Figure 8d,f). This shift can be linked to factors such as Europe’s well-maintained historical drought impact databases, like the EDII, which aids in the development of impact-based drought analysis. Additionally, Europe’s strong tradition of interdisciplinary research and its diverse climate zones make it an ideal region for testing narrative-based approaches like the storyline approach [203].

4. Significant Decision-Making Factors in Method Selection

Given that droughts are most effectively characterized by a combination of climatological and hydrological parameters [10], it is evident that no single approach can universally address all drought-related challenges. Through a synthesis of insights from multiple sources in the drought research literature discussed earlier in the paper, we have identified four critical factors (primary drivers) that serve as a framework for selecting appropriate drought analysis methods, thereby facilitating the development of effective drought investigation strategies. However, secondary factors, such as user expertise, computational resources, and policy context, may play a role in specific contexts, but are generally less decisive. The subsequent sections provide a detailed description of each of these factors, highlighting their significance in drought analysis and investigation.

4.1. Drought Type

The selection of drought analysis methods is inherently linked to drought type and sectoral focus of a study, as each method offers distinct strengths tailored to specific drought characteristics. As introduced earlier, droughts are commonly categorized into four types: meteorological, agricultural, hydrological, and socioeconomic. The following section discusses each type in relation to the analytical methods employed for their analysis.

Meteorological drought: Characterized by precipitation deficits, temperature anomalies, and atmospheric dynamics, this is effectively analyzed through a combination of complementary methods. Index-based methods are most commonly used due to their ability to track drought conditions based on meteorological variables. Widely used indices such as SPI, PDSI, and SPEI have evolved over time to capture variations in precipitation and temperature. This is supported by Kchouk et al. [31], who found that indices are more frequently applied to meteorological drought in the literature than for other drought types. Remote sensing also supports meteorological drought analysis by providing satellite-derived climate data—including precipitation, temperature, and evapotranspiration (e.g., CHIRPS, MODIS)—which serve as essential inputs for drought indices [183,187,200]. The storyline approach is particularly valuable for meteorological drought, as it examines historical drought events by isolating key climate variables (e.g., precipitation deficits, temperature anomalies) and projects how these meteorological drivers might evolve under future climate conditions, capturing both their physical mechanisms and sequential drought impacts [18,183,187,200]. The TLM, though inherently linked to hydrological drought, has also been applied to meteorological drought using precipitation thresholds [137]. Ultimately, the selection of methods depends on the specific objectives: indices provide universal benchmarks, remote sensing delivers spatial precision, storylines facilitate climate attribution, and TLM establishes operational triggers for drought response.

Agricultural drought: Driven by soil moisture deficits and vegetation stress, studying this type of drought similarly benefits from an integrated methodological approach. Index-based methods (e.g., SSMI) provide standardized metrics to quantify soil moisture availability, a critical factor in crop health and drought severity. Remote sensing plays a key role in agricultural drought monitoring, using satellite imagery to monitor vegetation stress and land cover changes, enabling large-scale agricultural drought monitoring [12,13,204]. The storyline approach extends these insights by integrating historical soil moisture and climate data to project how future warming could amplify agricultural drought impacts, such as yield loss [18,187]. Finally, the TLM operationalizes soil moisture thresholds (e.g., percentile-based) to define drought onset and duration, offering a practical framework for agricultural risk management [146,205]. Here, method selection is guided by distinct needs: indices establish baseline conditions, remote sensing enables dynamic monitoring, storylines inform long-term adaptation strategies, and TLM provides thresholds for immediate action.

Hydrological drought: Characterized by deficits in streamflow, reservoir levels, and groundwater availability, studying this type requires context-specific methodological approaches to analyze its complex dynamics. Index-based methods provide a foundational framework, with standardized indices (e.g., SSI, SRI, SGI) enabling systematic comparisons of surface and subsurface water deficits. For operational applications, the TLM offers a pragmatic solution by defining drought events based on critical thresholds (e.g., Q85 for streamflow), quantifying key characteristics such as onset, duration, and cumulative deficit, essential for water resources management [16]. Remote sensing complements these by providing spatially continuous data on water bodies’ dynamics, river discharge anomalies, and snowpack variability, bridging gaps in ground-based monitoring networks [206]. The storyline approach further extends the analytical scope by isolating historical hydrological extremes (e.g., low-flow events) and projecting their plausibility under future climate conditions, as demonstrated by Chan et al. [197]. Together, these methods form a hierarchical toolkit, where selection depends on the specific objectives, whether for real-time monitoring (TLM, remote sensing), long-term trend assessment (indices), or scenario-based planning (storylines).

Socioeconomic drought: Characterized by water scarcity, this type affects both human systems and economic drivers—including agricultural, energy, and employment trends—and requires methods that link physical drought conditions with socioeconomical impacts. The impact-based method is particularly well-suited for this purpose, drawing on impact databases (e.g., EDII, DIR), media reports, and stakeholder inputs to assess impacts like crop failure, energy disruptions, and food insecurity [158]. Index-based methods also contribute by using indices (e.g., SWSDI, WSI) to quantify human impacts and economic indicators, although their role is generally secondary. While remote sensing does not directly measure socioeconomic impacts, it provides critical support. For instance, satellite data can improve agricultural production monitoring [207], support food security policy, and enables early prediction of crop losses, such as through JPSS-1’s VIIRS sensor, which supports yield prediction two months before harvest [208,209]. Together, impact-based, index-based, and predictive remote sensing convert physical drought indicators into practical tools that support informed policy and economic decision-making.

Figure 9 provides a visual summary of the contributions of drought analysis methods to drought types, with symbol ratings reflecting their applicability and sectoral focus.

4.2. Data Type and Availability

The selection of an appropriate drought analysis method is heavily influenced by data type and availability, necessitating careful alignment of methodological choices with data characteristics [210]. As drought analysis methods have evolved significantly to address the complexity of this phenomenon, their data requirements have diversified, ranging from long-term climatological records to stakeholder-driven narratives [211,212].

In situ hydroclimatic data: In situ measurements refer to ground-based observations collected directly at specific locations [213], such as meteorological stations or hydrological monitoring points. For drought analysis, it typically includes precipitation, temperature, river discharge, groundwater levels, and soil moisture data. Both index-based and TLM methods rely heavily on long-term, standardized in situ records. For instance, the SPI requires at least 30 years of monthly precipitation data (50–60 years being optimal) to establish statistical baselines, while TLM derives drought thresholds from hydrological time-series (e.g., streamflow percentiles). The accuracy of these methods hinges on data duration, quality, and alignment with local hydroclimatic conditions which makes them impractical for regions with sparse or inconsistent observational networks.

Satellite-derived remote sensing data: Remote sensing data, collected by satellite-based sensors, provide large-scale, spatially continuous coverage of environmental conditions. Key drought monitoring variables include Vegetation Indices (e.g., NDVI), land surface temperature, soil moisture, precipitation, and other hydroclimatic indicators. This data type is central to remote sensing methods and also supports index-based methods, especially in regions lacking long-term in situ measurements. Remote sensing uses real-time earth observations to detect anomalies in vegetation health, temperature, and soil moisture. Additionally, satellite-based and reanalysis datasets serve as inputs for drought indices, compensating for gaps in ground data. Their broad coverage makes them valuable in remote or data-poor areas, though challenges like cloud cover and complex terrain may reduce accuracy, emphasizing the need for ground-truthing where feasible [214].

Impact datasets: Impact data capture the real-world effects of drought, including agricultural losses, water shortages, and economic damage. These datasets can be structured (e.g., DIR, EDII) or unstructured (e.g., media reports, agricultural statistics). The impact-based drought analysis method primarily relies on such data to link drought hazards with socioeconomic and environmental consequences. In impact data-rich regions like the U.S. and Europe, structured inventories enable consistent and robust analyses. In contrast, in data-scarce regions lacking systems like DIR or EDII, the method depends on heterogeneous sources that require preprocessing to address biases, such as the media data focus on extreme events only. These ad hoc sources need to be carefully validated against hydrometeorological indices to ensure reliability [159,173,177].

GCMs and Scenario-based narrative data: GCMs provide essential climate variables—such as precipitation, temperature, and soil moisture—that support drought assessment through index-based methods. These datasets, particularly from CMIP6 ensembles, are commonly used to derive drought indices (e.g., SPI, SPEI) for future climate scenarios, especially in long-term planning and regional vulnerability assessments where observed data are limited or insufficient [198]. Beyond their role in index-based methods, GCMs are central to the storyline approach when integrated with SSPs and RCPs. In this context, GCMs define the physical basis of future drought conditions, while SSPs add socioeconomic narratives (e.g., governance, land use, adaptive capacity) and RCPs describe greenhouse gas concentration trajectories. Together, they construct plausible drought scenarios under different warming levels and policy choices [187,192,194].

In summary, method selection needs to align with the type and availability of data, whether ground-based, satellite-derived, impact-focused, or model-based. Figure 10 illustrates how different data types correspond to specific methods, helping researchers balance precision, scalability, and contextual relevance.

4.3. Scale of the Study

The spatial scale of a drought study directly influences the choice of analytical methods, balancing trade-offs between broad applicability and local relevance. Two distinct approaches—top–down and bottom–up—define climate change adaptation plans for regions vulnerable to drought. At regional or global levels, drought investigations often adopt the top–down approach, relying on large-scale data, models, and centrally developed policies [215]. While offering broad coverage, these approaches may overlook local conditions and social dynamics, resulting in uncertainties and limited stakeholder engagement [216].

On the other hand, the bottom–up approach emphasizes local knowledge, community-level data, and decentralized decision-making [217], making it more responsive to site-specific needs. It supports context-specific drought responses and complements top–down strategies by facilitating implementation at the ground level [217,218]. Typically, top–down methods are suited for global or regional applications, while bottom–up approaches are implemented at local scales, such as catchments or municipalities.

Each drought analysis method discussed in this study falls along a spectrum between top–down and bottom–up approaches. Understanding their positioning helps researchers select suitable methods based on their study’s scale and institutional context.

Top–down: Index-based method aligns predominantly with the top–down approach. This method relies heavily on large-scale climatic data, making it particularly suitable for monitoring drought dynamics across extensive geographical areas. Its standardized approach offers wide applicability but may overlook local variables and detailed community impacts unless complemented by local and ground-based data. However, the literature provides evidence of index-based method contributing to bottom–up approaches by considering the socioeconomic impacts [46,219,220,221,222]. Despite this advancement, the index-based method remains predominantly top–down due to its reliance on large geographical coverage and centralized implementation systems.

Remote sensing methods typically adhere to the top–down approach, as they rely heavily on satellite and aerial imagery, enabling vast geographical coverage and centralized data processing. While not inherent to the method itself, remote sensing data can be integrated with the local knowledge to create more comprehensive drought investigation and support the bottom–up approach. For example, high-resolution remote sensing imagery, such as data from Sentinel-2 or UAVs (drones), has been used to monitor drought impacts on specific crops at the farm level, helping farmers optimize irrigation strategies and mitigate losses [223].

Bottom–up: The TLM, which is primarily employed for hydrological drought analysis utilizing river discharge data, combines elements of both bottom–up and top–down approaches, with a bias toward the bottom–up side. TLM is often applied at the catchment level, focusing on small-scale geographical areas and potentially incorporating stakeholders’ input in threshold determination, as threshold selection is not inherent to the method and can vary in practice [16]. These characteristics align TLM with the localized nature of the bottom–up approach. However, it also contains top–down elements, as the method typically relies on historical data and its implementation is often overseen by government or institutional agencies. The balance between its localized focus, stakeholder involvement, and institutional oversight places TLM in a hybrid position, leaning towards the bottom–up end of the spectrum.

The impact-based method leans strongly towards the bottom–up approach. By using local data sources—such as agricultural statistics, media, and impact reports—this method offers detailed and context-specific drought analysis. However, from the implementation perspective, this method still relies on governmental, research institutes, and other centralized bodies responsible for conducting the research, indicating the involvement of the top–down elements. Despite this top–down aspect, the method’s emphasis on local impact data and stakeholder’s involvement places it more towards the bottom–up end.

The storyline approach lies almost in the middle of the spectrum between top–down and bottom–up approaches, with a slight lean towards the bottom–up. As outlined in Section 2.5, it focuses on specific observed events, aligning with the localized nature of the bottom–up approach and facilitating stakeholder engagement and risk awareness. It also incorporates GCMs and global warming scenarios to project how drought events may evolve under climate change, positioning it close to the center of the spectrum with a slight bottom–up inclination.

Figure 11 illustrates the positioning of drought analysis methods along the top–down to bottom–up spectrum, highlighting differences in centralization, spatial-scale adaptability, and institutional hierarchy. Recognizing these distinctions supports the selection of appropriate methods and enhances the effectiveness of drought resilience planning.

4.4. Management Stages

Drought analysis methods discussed in this study are further examined in terms of their purpose and application within the stage of the drought management cycle. While drought management encompasses a wide array of activities—including detection, planning, mitigation, and recovery—this study focuses on three key stages of monitoring, assessment and forecasting that directly involve drought analysis [224,225].

Monitoring: Enabling early detection and real-time tracking that forms the foundation for timely interventions [226] primarily relies on two main methods: remote sensing and index-based methods. Remote sensing is crucial for real-time monitoring, using satellite data to observe variables such as vegetation health, water body dynamics, soil moisture, satellite-based precipitation and land surface temperature [12,227]. Additionally, index-based methods (e.g., SPI, SSI) can also be utilized in near-real time through ongoing hydroclimatic data. These indices help detect climatic anomalies and track drought progression.

Assessment: This involves quantifying drought characteristics (e.g., onset, severity, and termination) and their impacts using past and present data [228]. All five methods reviewed in this study contribute to this stage, each offering different strengths.

Index-based and threshold-level methods are commonly applied in the drought assessment stage. Index-based methods support drought assessment by providing standardized quantification using long-term climate records, supporting consistent classification and trend analysis across regions. Similarly, TLM assesses drought conditions based on predefined thresholds of hydroclimatic variables, e.g., river discharge, reservoir storage, and groundwater levels for hydrological droughts; precipitation and soil moisture for meteorological and agricultural droughts.

Remote sensing also contributes to assessment through the analysis of archived satellite data, which supports retrospective evaluation of drought impacts, such as vegetation decline and soil moisture deficits [229].

Impact-based and storyline approaches are rarely used for drought assessment in the literature. The impact-based method mainly supports forecasting through IbF, linking hazard predictions to expected impacts. However, it also offers potential contributions to drought assessment by analyzing recorded impacts from structured and unstructured drought impact sources. The storyline approach is not typically used for drought assessment. Instead, it explores how past droughts might unfold under future climate and socioeconomic scenarios, making it more relevant to strategic planning than operational assessment.

Forecasting: Anticipating drought characteristics and associated risks allows stakeholders to prioritize mitigation strategies [230]. While fewer methods are directly designed for forecasting, most can be extended using predictive models or scenario-based tools. Recent advancements in impact-based methods through IbF have enhanced early warning systems by linking drought hazard predictions to anticipated socioeconomic consequences. However, these forecasting applications are still evolving and depend on robust impact databases for training and validation [177]. The storyline approach provides scenario-based projections by combining climate model outputs with historical drought analogs, offering valuable insights for long-term adaptation planning under stress-tested scenarios [187]. Remote sensing has also enhanced forecasting through data assimilation in land surface models and early detection of vegetation stress (NDVI anomalies), significantly improving the DEWS [231].

Index-based method supports forecasting when extended through statistical approaches (e.g., ARIMA) or coupled with dynamic models, providing standardized drought probabilities [211,232]. Similarly, the TLM can contribute to forecasting when integrated with hydrological and forecasting models, allowing for the prediction of future threshold breaches based on projected hydrological conditions.

In summary, each method supports different stages of the drought management cycle, with several spanning multiple roles. Index-based and remote sensing methods are central to both monitoring and assessment, while TLM primarily serves assessment. All three can support forecasting when integrated with predictive models. Impact-based methods bridge assessment and forecasting by linking hazards to socioeconomic consequences, and the storyline approach is especially suited to long-term scenario-based forecasting and planning.

Figure 12 provides a graphical representation of these relationships.

5. Conclusions

This paper provides a multidisciplinary overview of five drought analysis methods—index-based, remote sensing, threshold-level, impact-based, and the storyline approach—highlighting their strengths, limitations, and suitability in relation to drought type, data availability, study scale, and management stages. This conceptual framework supports informed method selection by aligning these approaches with key influencing factors. By comparing these methods, the paper clarifies their respective roles in addressing the complex and multifaceted nature of drought. Key conclusions are summarized as follows:

• The index-based method provides a standardized framework for drought analysis, transforming hydroclimatic variables into statistical indices. Its flexibility across drought types and spatial–temporal scales makes it suitable for monitoring, assessment, and forecasting of drought, especially when integrating diverse data sources (e.g., in situ, satellite datasets, GCMs). However, its reliance on stationarity assumptions and data quality limitations (e.g., sparse ground observations) can reduce accuracy. Future advancements should prioritize hybrid frameworks (e.g., integrating AI and machine learning) and multi-source data fusion (e.g., satellite-derived precipitation with ground measurements) [232].
• Remote sensing supports drought analysis across drought types, particularly agricultural drought through vegetation (e.g., NDVI) and soil moisture monitoring. It offers consistent, continuous data, bridging gaps in areas lacking in situ measurements. With broad spatial and temporal coverage, it supports all drought management stages: real-time monitoring, historical assessment (e.g., SMAP trends), and forecasting (e.g., integration into DEWS). However, limitations include sensor issues (e.g., cloud cover, coarse resolution) and challenges in data integration. Future improvements should focus on multi-sensor fusion, AI-based downscaling, and dynamic baselines to boost accuracy.
• TLM is fundamental to hydrological drought analysis, using site-specific thresholds to quantify absolute water deficits, critical for operational water management (e.g., reservoir alerts, low-flow measures). It supports both fixed and variable thresholds, allowing flexible use across climates. TLM excels in drought assessment (e.g., duration, intensity, severity) and contributes to forecasting when coupled with hydrological models. Limitations include potential misclassification of natural low-flow periods and dependence on stationary climate assumptions. Future improvements should emphasize dynamic threshold calibration and integration with remote sensing data (e.g., GRACE) to improve accuracy and resilience.
• Impact-based methods transform drought analysis by focusing on socioeconomic impacts, incorporating hazard, exposure, and vulnerability to assess risks across sectors. They connect drought indices (e.g., SPI) to real-world impacts (e.g., crop loss, water shortages), through IbF contributing to drought impact forecasting. It primarily focuses on regional applications and can support large-scale analysis when data is available. Future advancements should prioritize AI-driven tools to mine unstructured data (e.g., social media) and global standardization efforts to enhance equity and relevance.
• Storyline approaches combine quantitative climate projections (e.g., GCM-derived temperature anomalies, SSP-RCP scenarios) with qualitative, stakeholder-centric narratives to explore high-impact drought scenarios. It is adaptable across drought types and scalable from local to global study scales. Storyline approaches are particularly effective in assessment (e.g., evaluation of vulnerabilities in past droughts) and forecasting (e.g., projecting and stress-testing megadrought risks under SSP5-RCP8.5). Despite their advantages, limitations remain due to regional data gaps and uncertainties in GCMs, highlighting the need for better integration of remote sensing and local observations.
• The integration of AI/ML with drought analysis methods offers transformative potential for improving drought monitoring, assessment, and forecasting. AI/ML enhances the ability to capture complex, non-linear relationships across diverse data sources—including indices, in situ and remote sensing observations, and impact records—while enabling automation and improved accuracy. However, future research should focus on specialized AI architectures for drought features, better integration of multi-source datasets, and standardized protocols to ensure transparency, consistency, and fairness [233].

To conclude, no single method can fully capture the complexity of drought phenomena. Instead, integrating these approaches—tailored to specific drought types, scales, data types and management needs—may offer a more comprehensive path forward. Strengthening interdisciplinary collaboration between climate science, hydrology, remote sensing, and social sciences can significantly enhance the accuracy, reliability, and usability of drought analysis. Future advancements should prioritize hybrid frameworks, multi-source data fusion, and stakeholder engagement to support climate-resilient drought risk management in an increasingly dynamic world.