Identifying the Drought Impact Factors and Developing Drought Scenarios Using the DSD Model

Droughts have become increasingly severe and frequent due to climate change. Droughts cause water scarcity and various socio-economic issues. Therefore, it is necessary to address drought conditions water resource management policies. The drought characteristics within each administrative division need to be closely analyzed to support effective policy-making. A drought impact factor analysis using a drought scenario development (DSD) model, is presented in this study. The DSD model identifies the drought impact factor for each administrative division through the relationship between various hydrometeorological factors and drought indices, and constructs a drought scenario. The proposed model was applied to 167 administrative divisions in South Korea. Hydrological factors significantly impact droughts than meteorological factors in South Korea. Identified drought impact factors are analyzed based on spatio-temporal variability to recognize the features in various aspects of droughts. Changes in water demand and severe drought periods were considered for temporal variability. Administrative divisions were grouped into four zones based on the type of water demand, and zonally analyzed to examine spatial variability. Finally, a drought scenario based on the identified drought impact factor was constructed to present the probable drought conditions. Components of the drought scenario represent the constitution of water resources within an administrative division, and high and low levels of each component were combined. The constructed drought scenarios facilitate effective policy-making for managing water resources.


Introduction
Recent climate change has induced unprecedented, severe and interminable drought events (Trenberth et al. 2014). Drought is generally defined as a condition of water scarcity that originates from a complex interaction among atmospheric, hydrological, and socio-economic factors. Therefore, sufficient rainfall is key to resolving drought, and effective water resource management and policies for the entire watershed must be implemented simultaneously (Booker et al. 2005;Khan et al. 2017;Palazzo et al. 2017;Shiau et al. 2021).
Effective drought management is essential; for example, a system state index for system dynamics modeling can be used for strategic drought management (Rubio-Martin et al. 2020), and effective drought management by specific preparedness plans can be conducted using climate forecasts and crop models (Martins et al. 2018). A proactive approach incorporating technocratic support, active public participation, and institutional structure is crucial for efficient drought management (Tsakiris 2017). Additionally, the need for a proper water policy for effective water resource management is highlighted by its socioeconomic impact. Khan et al. (2017) investigated the impact of economic growth following the implementation of a water policy. Similarly, Booker et al. (2005) also highlighted the economic trade-offs of water use, regions, and drought control strategies and emphasized the importance of political and institutional jurisdictions for water resource management. In addition, Palazzo et al. (2017) stressed the necessity of considering biophysical, demographic, ecological, and economic principles for managing water resources efficiently. Therefore, water resources policies are essential for supporting human life socioeconomically.
Analyzing drought in administrative divisions is essential for managing water resources efficiently by incorporating a policy to reduce drought damage. Various studies have been conducted on the drought analysis of several watersheds (Garrote et al. 2007;Maity et al. 2013;Kallache et al. 2013;Malik et al. 2019;de Medeiros et al. 2019), but drought studies focusing on administrative divisions are insufficient, and constructing suitable policies is necessary. Policies including disaster management actions (mitigation, preparedness, response, and recovery) are adopted and applied in administrative divisions in South Korea. For example, implementing structural and nonstructural measures, and checking and quantifying drought damage costs are often completed within administrative boundaries. However, limited efforts have been devoted to identifying the critical impact factors of drought concerning administrative divisions in a country.
The severity of drought changes depending on the duration and regional characteristics. Many drought indices, including the standardized precipitation index (SPI) (Mckee et al. 1993), the Palmer Drought Severity Index (PDSI; Palmer 1965), and the reconnaissance drought index (RDI) (Tsakiris et al. 2007), have been proposed and applied to quantify the severity of droughts. SPI is based on precipitation and has been utilized in many studies because of its simplicity (Kallache et al. 2013;Livada and Assimakopoulos 2007;Guttman 1998;Tigkas et al. 2019). The PDSI presents moisture conditions and is widely used to measure the cumulative water balance. However, PDSI lacks consistency in spatial distribution and is unsuitable for drought analysis, which determines spatial variability (Guttman 1998). Conversely, SPI accounts for the spatial variability of drought better, and has been widely applied to spatial analysis (Livada and Assimakopoulos 2007).
The cause-and-effect relationship between drought occurrence and hydrometeorological factors should be identified for developing proactive drought risk management. A few studies have reported a relationship between drought occurrence and hydrometeorological factors (Maity et al. 2013;Pandey et al. 2010;Pei et al. 2019). Drought impact factors (DIF) can be identified through the relationship between hydrometeorological factors and SPI. Various statistical methods such as linear discriminant analysis and autoencoders can be utilized to identify the DIF. It is noteworthy that DIF is a critical factor affecting drought.
Information about the impact on the occurrence and mechanism of drought can be found by interpreting the results from diverse perspectives. An approach using spatio-temporal analysis is required to determine the drought impact due to the characteristics of the administrative division. Haslinger and Blöschl (2017) proposed a method to determine the trends and characteristics of drought accounting for spatio-temporal patterns, and identified atmospheric drought events considering drought duration, intensity, and severity. Similarly, Kallache et al. (2013) analyzed dry spells and spatial patterns using a multivariate extreme value model. The results indicated that the dry spells in subbasin are influenced by the topography and spatial distance from other subbassins. Livada and Assimakopoulos (2007) attributed the drought in Greece to spatiotemporal variability using 51 years of precipitation data. Past studies have evaluated drought trends in relation to spatio-temporal variability, but analysis of the relationship between specific spatiotemporal characteristics and drought has not been conducted.
Scenarios are useful tools for combining several types of system information to establish a plan. Particularly, scenarios can support a flexible response to overcome such variations associated with uncertainties. To establish the scenario in the context of uncertainty, we first need to recognize the problem. Next, we identify the factors that affect the problem and rank them to determine the importance and degree of uncertainty. The determined uncertainty axis can be utilized to construct the scenario and named accordingly. A previous study suggested deciding the possible future or direction of the scenario and identifying the common aspects of various scenarios (Kang and Lansey 2014). The severity of drought and changes in hydrological and meteorological factors are uncertain concerning drought scenarios. A few examples of drought scenarios construction have been introduced. Herman et al. (2016) developed a bottom-up method for synthetic scenario generation by increasing the frequency and severity of droughts. Conversely, Ghosh and Mujumdar (2007) developed a topdown method to construct a scenario using a general circulation model (GCM) to regional climate models (RCM). However, covering all hydrologic statuses in a district is needed to help decision-makers prepare for an unprecedented drought.
In this study, a drought scenario development (DSD) model is proposed to develop drought scenarios, to describe the conditions of water resources in each administrative division. Explaining future water resource conditions specifically through drought scenarios and propose effective precautions and responses is possible. The novelty of this study is the identification of DIF and the construction of drought scenarios in administrative divisions. The DIF for each administrative division was derived using principal component analysis (PCA) with hydrometeorological factors as input and SPI as the output. The results of PCA were then analyzed to determine spatiotemporal variability, and were utilized in constructing drought scenarios. Figure 1 presents the framework of the proposed DSD model with two primary components: DIF analysis (DIFA) and critical factor combination (CFC). First, the hydrologic and meteorological data of each administrative division were collected as input data for the DSD model. The critical impact factors for a drought indicator (SPI in this study) were identified in the PCA-based DIFA, where the impact level of each factor is quantified and ranked based on the principal component scores obtained. The highly ranked factors are then combined in the CFC block to construct a drought scenarios (output of the DSD model). The aforementioned processes were performed at individual administrative divisions, and thus, the results can be compared across administrative divisions. The following subsections describe the methodologies used: PCA, ranking method, factor combination, and scenario generation.

Drought Impact Factor Analysis (DIFA)
PCA is an unsupervised, machine learning techniques for dimensionality reduction and feature extraction. In PCA, a linear combination of the variables (Eq. 1), which maximizes the variation (Eq. 2) is found, and the principal component, which is the factor with the most impact, is extracted.
(1) Fig. 1 A schematic of the proposed DSD model with its input and output where x is the given data, y is the transformed data of x, w is the weight vector, and n is the number of variables. Therefore, PCA can be utilized to extract input factors that affect output factors and is generally applied to determine explanatory variables, processes for factor analysis, and clustering. PCA has been applied to drought analyses in several studies (Maity et al. 2013;Tijdeman et al. 2018;Santos et al. 2010). However, most of the studies focused on identifying the regional drought characteristics, while exploration of studies identifying the relationship between hydrometeorological factors and the drought index, remains limited.
Policies to reduce the damage caused by drought can be constructed through the relationship between hydrometeorological factors and the drought index. PAC is performed using hydrometeorological factors as the input and SPI as the output to establish the relationship, and the procedure is as follows: First, the covariance matrix of existing data was calculated. Second, the eigenvalues and eigenvectors of the covariance matrix were computed. Third, we listed eigenvectors in the order of magnitude of the eigenvalues. Fourth, we selected an eigenvector that could describe the majority of the total. Finally, the principal component score was calculated using the dot product with the corresponding eigenvector and data. In a higher chance, the factor with a larger principal component score will have a greater impact on drought. Hence, hydrometeorological factors were ranked according to the principal component score. The rank index (RI) is utilized to interpret the final result comprehensively. RI was calculated by averaging the rank of hydrometeorological factors for the total administrative divisions and taking the inverse of the average value. The value ranged from 0 to 1, and a larger value indicated a greater impact on the drought.

Critical Factor Combination (CFC) for Scenario Development
The scenario predicts probable circumstances in an uncertain situation. During drought, severity is uncertain, and hydrological drought status in the administrative division changes accordingly. Therefore, establishing a drought scenario based on the water scarcity in administrative division helps to establish the uncertainties. The hydrologic drought status of an administrative division can be expressed using three components: the inflow volume of water into the administrative division, the water stored within the administrative division, and the volume of outflow. Drought scenarios were established by combining the different levels of these components. The process for determining the level of the component for the drought scenario varied by component (Fig. 2). The inflow volume affects the drought significantly, and the water scarcity condition of the administrative division can be represented by a no-rainfall period, and severe drought occurs when the no rainfall period increases. Therefore, a no rainfall period was sets as an element of water inflow in the drought scenario. To consider the extreme conditions of drought, no rainfall for 6, 9, and 12 months were used for the CFC. Additionally, outflow can be determined by the amount of water used within the administrative division and water demand. High and low levels of water demand are critical factors for water outflow. Water demand in the administrative division can be controlled by policies that encourage people to reduce water usage. Finally, water storage is related to stored water in the administrative division and is described by hydrological factors, which are utilized as input for the DSD model. The first DIF among the hydrological factors in the administrative division was identified using the DSD model. The critical factor of the first DIF was determined through frequency analysis, and tercile points under the average were set as the criteria. Figure 2 presents an example of the frequency analysis. The first to second tercile criteria and second to third tercile criteria are set as critical factors of water inflow to drought scenarios. Drought scenarios were established using the CFC of each component.

Other Measures
Drought damage occurs when the supply is lower than the demand, and the input factors for PCA are related to the water supply. Therefore, an analysis of the demand is necessary. An increase or decrease in population in the administrative division or construction of a plant, indicates the temporal variation of demand. Therefore, temporal variation should be considered for DIF analysis of water demand change. The absolute percent error (APE) was calculated using Eq.
where RI 1 is the RI value from the former period and RI 2 is from the latter period. This index implies a change in the impact of drought due to water demand, and a higher value of APE indicates that the factor is sensitive to water demand variability. Furthermore, the percentage of rank changing factors (PRCF) was then used to determine the influence of water demand on the administrative division (Eq. 4). Through the PRCF, it was possible to determine the administrative division that most effects the water demand variability.
Drought has been analyzed for the following two severe drought events: one with the lowest precipitation drought and one with the historically severe droughts. The latter case is often considered a reference point in South Korea in setting the water supply date index for drought management. Identifying the characteristics of severe drought was possible through the analysis. Following the same process of PCA in applying the DIF, the period (4) PRCF = Numberofrankchangingfactors Totalnumberoffactors × 100(%) Fig. 2 Diagram of CFC for drought scenario of input and output data is distinct depending on the severe drought period. Determining the features of severe drought was possible by comparing the results with the total period. Additionally, drought is related to the spatial characteristics of the administrative division, and it depends on the type of water used. administrative divisions were categorized into three zones to analyze drought based on spatial variability and compare the DIF to identify the characteristics of drought by zone. The three zones were urban, industrial, and agricultural zones, determined by the purpose of water usage. The sum of residential, industrial, and agricultural water demand was set as the total water demand, and the percentages of residential, industrial, and agricultural water demand were calculated to determine the water demand characteristics of each administrative division. Based on the results of the frequency analysis for each type of water demand percentage of the total administrative divisions, criteria were set based on the corresponding upper tercile of the total administrative divisions.

General Information and Past Drought Events
South Korea is located in the middle latitudes of East Asia ( Figure SI 1) and lies in a temperate region, experiencing the four seasons. Depending on the seasonal variability, meteorological factors, such as temperature, precipitation, and humidity, change. The weather is sunny and dry during spring and fall because of the migratory anticyclone, and drought is likely to occur. Cold arid continental high pressure, temperatures, and humidity are low during winter. Summer experiences high temperatures influenced by the hot and humid North Pacific anticyclone with monsoon season generally from June to July. In addition, most of the rainfall is concentrated during June to August, increasing drought vulnerability.
Water-related risk has socio-economic implications and depletes physical and human resources. Additionally, South Korea experiences water shortage, and resulting in increased drought occurrences (Nam et al. 2015;Kim et al. 2014). Drought analysis from an administrative perspective is necessary by establishing a policy for water resources conservation. According to the Korea Meteorological Administration (KMA), precipitation in South Korea in 2015 was 80% of the annual average.
The country is composed of 167 administrative divisions, and possesses diverse geographical characteristics. In the administrative division population of the country, 19% and 0.017% live in Seoul and Ulleung, respectively. Additionally, Hongcheon and Guri account for 1.8% and 0.033%, respectively. The characteristics of each administrative division vary (Table SI 1) and affect water resource management.

Hydrometeorological Data and SPI
In this study, monthly data of six meteorological and five hydrological factors from January 2009 to December 2019, are utilized as a PCA input factor. Meteorological factors are of two types: automated synoptic observing system (ASOS) data with 102 meteorological sites, and automatic weather station (AWS) data with 510 meteorological sites ( Figure  SI 1). AWS data are weather observations to prevent natural disasters due to atmospheric phenomena. Monthly precipitation, average temperature, and average wind speed were used in this study. Hydrological factors provide information about dams from the WAter resources Management Information System (WAMIS), including average low water level, average inflow, average discharge, average floodgate discharge, and catchment average rainfall. These factors can be classified into natural (e.g., average catchment rainfall and all meteorological factors) and artificial factors (e.g., average floodgate discharge, average discharge, average inflow, and average low water level). It is possible to identify whether drought is artificially or naturally affected by the characteristics of the administrative division, by analyzing the classification of artificial and natural factors. SPI 3, SPI6, and SPI 9 were applied in this study depending on the cumulative month of precipitation. The SPI of 167 administrative divisions was obtained from the KMA.

Identify Nationwide DIF
The impact of hydrometeorological factors on drought was identified using nationwide DIF analysis. RI was utilized by administrative division to determine the comprehensive result of DIF analysis, and the results of 167 administrative divisions were integrated (Fig. 3). As shown in Fig. 3, SPI3, 6, and 9 all exhibit similar trends. Hydrological factors such as average discharge, average inflow, and average floodgate discharge were placed at the top, indicating that they were related to drought. Additionally, monthly precipitation was the highest among the meteorological factors.
For each administrative division, the impact on drought was identified using the value of the principal component score for hydrometeorological factors (Figure SI 2). A larger value in the chart indicates a greater impact on drought. Average discharge, floodgate discharge, and inflow were ranked first, second, and third, respectively, for Seoul. In Boryeong, drought has occurred frequently, and based on the results of the IDF analysis, the average low level, average discharge, and average floodgate discharge scored high. Unlike the Seoul and nationwide results, the average low water level significantly impacted drought occurrences, influenced by the Boryeong dam's waterway operation, and including low water level fluctuations. The results also imply that the Boryeong dam's water reserve rate was under 30%, the lowest rate among all dams in South Korea; this significantly affected the frequent drought in Boryeong. Busan has the top-ranked factor for average inflow, Fig. 3 Integration of nationwide DIF analysis results catchment average rainfall, and monthly precipitation. The significant feature of Busan's result is that it has meteorological factors for the third rank, the largest DIF among the meteorological factors in the nationwide results. Additionally, the impact of the average floodgate discharge in Busan was lower than that of other administrative divisions, attributed to the fact that 93% of the water intake is from downstream of the Nakdong River. Gwangju, a metropolitan city located in southwest South Korea, has a DIF of the first rank for average floodgate discharge, second rank for average discharge, and third rank for average inflow. Despite the different water resources, population, and size of the administrative division, the radar chart shape looks similar to that of Seoul. Based on the results of major administrative divisions and administrative divisions with frequent droughts, determining that major administrative divisions highly impact hydrological factors is facilitated, and DIF reflects the characteristics of the administrative division.

DIF Analysis Regarding the Water Demand Variability
DIF analysis was performed using data from 2009 to 2019, over a period of 132 months. Water demand fluctuations due to temporal variability were identified to analyze the DIF response. The trend in water demand (million m 3 /year) change over time for each administrative division, was discovered by the summation of the residential, industrial, and agricultural water demand. To determine the effect of demand in the DIF analysis, the total period is divided into the following two periods: 2009 to 2013 as the high-demand period, and 2014 to 2019 as the low-demand period. Agriculture land use and water demand have decreased due to urbanization and recent droughts.
The APE was calculated for each factor (Fig. 4) to identify the change in DIF, due to high and low water demand periods. The overall APE trends for SPI 3, 6, and 9 were similar, and SPI 9 had the largest value for the sum of all factors' APE (110%). Additionally, the average values of SPI 3, 6, and 9 for each factor were high for average discharge (26%) and average floodgate discharge (20%) and low for average pressure (2%). Therefore, the average discharge and average floodgate discharge significantly affect the temporal variability due to the water demand change, and the average pressure has a lower effect. Additionally, the average APE value for hydrological factors was larger than meteorological factors by approximately 3.2 times, and hydrological factors may be more affected by temporal variability due to water demand change than meteorological factors. This implies a changing operation of the dam, depending on the fluctuation of the water demand.
The PRCF based on DIF from the high and low demand periods was calculated to determine the variability of the DIF rank in the administrative division. The total administrative divisions were classified into four groups based on PRCF values. The total administrative divisions are categorized into four groups for SPI 6, and the percentage of the administrative divisions for PRCF under 25% was 8% of administrative divisions, PRCF from 25 to 50% was 11%, PRCF from 50 to 75% was 31%, and PRCF over 75% was 51% ( Figure SI 3).

DIF Analysis During Severe Drought Period
Two periods of severe drought were chosen to analyze the characteristics of severe drought in Korea. From June 1994 to July 1995, droughts were recorded as severe droughts, utilized as the water supply date index of the water resources plan in South Korea. Additionally, the 2015 drought was recorded as a severe drought with the lowest precipitation. The DIF from June 1994 to July 1995 (drought period 1) and 2015 (drought period 2) were used to determine the characteristics of the severe drought periods.
The impact on drought was examined using the value of RI during drought periods 1 and 2, using the same method as the total period DIF analysis. Unlike the results of the total period in which hydrological factors had a significant impact, meteorological factors contributed to severe drought. The average RI value was calculated to compare the impacts of hydrological and meteorological factors on drought. For the total period, meteorological factors had a value of 0.17, and hydrological factors had 0.33, approximately double that of the meteorological factors. Conversely, during drought period 1, meteorological and hydrological factors had values of 0.33. and 0.31, respectively, and drought period 2 had meteorological and hydrological values of 0.19, and 0.23, respectively. Thus, during the severe drought period, there were no significant differences in the impacts of hydrological and meteorological factors. Additionally, DIF was analyzed based on the top three ranked factors, depending on temporal variability (Table 1). During the total period, the top three DIFs were all artificial factors; however, during drought period 1, natural factor ranked first, and drought period 2 included two natural factors in the top three of the DIF. This implies that during the severe drought period, the impact of artificial factors decreased, and natural factors significantly influenced drought.

DIF Analysis Based on Spatial Variability
Based on the type of water demand, administrative divisions were classified into zones to analyze the DIF depending on spatial variability. The The histogram in Fig. 5 presents Average floodgate discharge Average discharge Catchment average rainfall the residential, industrial, and agricultural water demands for 167 administrative divisions in South Korea. For residential and industrial water demand, the frequency of the lowdemand class section was high, and the frequency exponentially decreased as the water demand increased. Contrastingly, the frequency of sections at high demand was high for agricultural water demand. The criteria were set by the upper third of the water demand for type, by applying the characteristics of the water demand frequency. If the administrative division's residential water demand percentage is over 25%, it is classified as an urban zone. Similarly, the criteria for industrial water demand to be classified as the industrial area is 10%, and the agricultural area is 85%. For example, the percentages of water demand in Seoul were 87.1%, 12.2%, and 0.7% for domestic, industrial, and agricultural purposes, respectively. Based on these criteria, 34% of the administrative divisions were classified as urban zones, 19% as industrial zones, and 34% as agricultural zones (Figure SI 4). The remaining administrative divisions faced difficulty while grouping and hence were grouped as unclassified. As a result, major cities in South Korea such as Seoul, Busan, and Incheon are classified into urban zones and administrative divisions where the plants were concentrated; for example, Ansung, Ulsan, and Iksan were grouped as industrial zones. Additionally, the administrative divisions where major farm products were produced, such as Gimje, Bonghwa, and Sancheong, were classified as agricultural zones.
Based on the classification of the administrative divisions, DIF was analyzed to identify the effect due to spatial variability. There were no significant differences and showed similar trends depending on the type of SPI; hence, only the results of DIF analysis using SPI 6 were utilized. The bar chart in Fig. 5 presents the percentage of administrative divisions for the most critical DIF by zone. For urban and industrial zones, the percentage of average discharge is largest at 57% and 48%, followed by average inflow at 14% and 19%, respectively The agricultural zone had the largest percentage of average discharge (36%), followed by monthly precipitation (25%). A large percentage of the administrative division with meteorological factors was the most critical DIF compared to the other zones. Additionally, the percentage of artificial factors is approximately 6 and 5 times larger than natural factors for approximately 6 and 5.2 times for urban and industrial zones, respectively; however, industrial factors are approximately 1.4 times. Therefore, establishing the significant impact natural factors have on the agricultural zone, and that artificial factors affect urban and industrial zones, is possible.

Drought Scenario
Drought scenarios for 167 administrative divisions in South Korea were established through the CFC, utilizing the DSD model results. However, in the results of the analysis based on temporal variability due to water demand, approximately 50% of the administrative divisions in South Korea had a change in DIF rank under 75%. Therefore, water demand provides insufficient explanation for drought conditions and is not utilized in drought scenarios.
An example of a drought scenario in Seoul, Chuncheon, and Busan was presented using the suggested method ( Table 2). The most critical DIF of Seoul was the average low level of the Soyanggang and Chungju dams, and the average low water level of the Soyanggang Dam and the average inflow of the Andong dam for Chuncheon and Busan, respectively. In Fig. 6, the results of the frequency analysis of the corresponding first DIF for each administrative division and the tercile criteria represented by dashed lines, are utilized as critical factors. Drought scenarios were constructed (Table 2) using CFC, which are the key results of this study derived from the DSD model. Drought conditions with uncertainty can be examined using these scenarios.

Conclusions
This study conducted a drought analysis of administrative divisions, to understand the characteristics of drought in the administrative division. The relationship between hydrometeorological factors and the SPI was analyzed. Furthermore, an analysis based on Table 2 Drought scenario of Seoul, Chuncheon and Busan spatio-temporal variability was performed to determine the various characteristics of drought in administrative divisions. An analysis based on water demand was performed, considering temporal variability. Additionally, DIF analysis during the severe drought period was performed, and the results were compared with the total period to determine the characteristics of severe drought periods. Through the DSD model, drought characteristics of the administrative division were identified, and a drought scenario was constructed. Components of water inflow, outflow, and storage, were considered for the drought scenario. Predicting the drought conditions in each administrative division with uncertainty in drought severity, using the constructed scenarios.
The drought characteristics of drought were identified through the results of the analysis. Seoul's drought characteristics were as follows: First, the most critical DIF for Seoul was the average discharge. Furthermore, hydrological factors have a greater impact on drought than meteorological factors. However, for severe drought periods, meteorological factors have a more significant impact than hydrological factors during severe drought periods. Comparing the high-demand and low-demand periods, the value of PRCF was 55% and 125th to be affected by temporal variability based on water demand change. The first DIF among hydrological factors was average discharge, and through frequency analysis, the criteria were set as 52% and 78% of the average discharge compared to the normal year. A drought scenario was constructed by combining the criteria with six, nine, and twelve months of no rainfall.
DIF based on administrative division and various related analyses will assist in establishing effective policies for managing water resources under drought conditions. Additionally, through drought scenarios, provisions can predict future drought conditions. However, this study has scope for improvement. First, the DIF presented in this study was identified using the same time period for hydrometeorological factors and SPI. Therefore, it is unclear whether hydrometeorological factors are affected by SPI or SPI fluctuations. It may be possible to perform additional analyses based on the time difference between SPI and hydrometeorological factors. Second, the presented drought scenario was constructed by combining several levels of water resource components. It reflects the characteristics of water resources in an administrative division, so it is possible to describe probable drought conditions. However, the probability of occurrence depends on the level combination in the scenario. Additional analyses should be performed to determine the probability of occurrence.