Assessment of Drought Indices Based on Effective Precipitation: A Case Study from Çanakkale, a Humid Region in Türkiye

Abstract

This study investigates the influence of different effective precipitation (P_e) estimation methods on drought index performance in a humid region of Türkiye. The standard precipitation index (SPI) and the reconnaissance drought index (RDI) were compared with their effective precipitation-based counterparts, Agricultural Standardized Precipitation Index (aSPI) and Effective Reconnaissance Drought Index (eRDI), using four P_e estimation methods: USBR (U.S. Bureau of Reclamation), USDA-(Simplified and CROPWAT) (U.S. Department of Agriculture), and FAO (Food and Agriculture Organization). Data from three closely located meteorological stations (Çanakkale, Bozcaada, and Gökçeada) were analyzed across multiple time scales (1-, 3-, 6-, 12-month, and annual). Statistical metrics—coefficient of determination (R²), root mean square error (RMSE), and Nash–Sutcliffe efficiency (NSE)—were used to assess the indices, and trend analyses were conducted using the Mann–Kendall and Sen’s Slope tests. The USDA-Simplified method consistently showed the highest accuracy across all stations and time scales (R² ≈ 0.99; lowest RMSE ≈ 0.09; NSE > 0.95), while the FAO method performed poorly, particularly at the 1-month scale. Drought frequency and severity were found to increase with time scale, contrary to trends observed in arid regions. Trend analysis revealed no significant changes at short time scales, but statistically significantly increasing drought severity was detected in longer scales, especially in Çanakkale, with slopes reaching up to –0.018 per year. The findings highlight the importance of selecting appropriate P_e estimation methods for accurate drought assessment, even in humid climates, and support the use of aSPI and eRDI with the USDA-Simplified method.

1. Introduction

Drought frequency and intensity are expected to rise in already drought-prone regions, such as the Mediterranean, Central Europe, southern Amazon, and southern Africa. These shifts are likely to affect ecosystems, food security, and land processes, including greenhouse gas fluxes. Droughts can become more severe due to inadequate land management practices. Additionally, urbanization contributes to an increase in extreme rainfall events either directly over cities or in areas downwind of them [1]. Extreme weather events like droughts and floods are expected to become m0ore frequent and severe, especially in vulnerable and underprepared developing regions [2]. In this context, drought studies have become even more important.

Mishra and Singh [3] emphasize that variations in hydro-meteorological conditions and socioeconomic factors, and the unpredictable nature of water demands across regions have made it challenging to establish a precise definition of drought. Drought is regarded as the most intricate and least understood of all climate extremes, impacting vast regions and populations [4].

A range of indices—such as the Standardized Precipitation Index (SPI) [5], The Standardized Precipitation Evapotranspiration Index (SPEI) [6], The China Z index (CZI) [7], Effective Drought Index (EDI) [8], Reconnaissance Drought Index (RDI) [9], Rainfall Anomaly Index (RAI) [10,11], Agricultural Standardized Precipitation Index (aSPI) [12] and Effective Reconnaissance Drought Index (eRDI) [13,14]—are utilized for monitoring and predicting drought conditions.

Among these, the SPI and RDI have been widely adopted due to their simplicity, capacity to assess drought severity and duration, and flexibility across various time scales [15,16,17]. While SPI [5] relies solely on precipitation data and was initially developed for meteorological drought, it has also been applied in hydrological and agricultural contexts. The RDI demonstrates enhanced sensitivity and global applicability for meteorological drought assessment compared to traditional indices like SPI and Deciles, likely due to its incorporation of more data-intensive parameters [9]. The aSPI and eRDI were later introduced as extensions of SPI and RDI, respectively, to improve sensitivity to agricultural drought and provide greater customization [12,13].

Although there are numerous studies in the literature involving other drought indices, comparative analyses of the aSPI and eRDI remain quite limited. Most studies have typically focused on evaluating the relationship between SPI and aSPI [18,19] or between RDI and eRDI [13,14,20], while comprehensive evaluations of all four indices—SPI, aSPI, RDI, and eRDI—are relatively scarce [21,22,23,24,25]. Moreover, in Rezaei et al. [26], the eRDI was evaluated based on different methods for calculating effective precipitation, while Zarei et al. [20] reported climate-dependent differences, recommending eRDI for arid areas and RDI for humid regions due to its simplicity and the advantage of not requiring effective precipitation calculation.

Similarly, Tigkas et al. [12] highlights that the aSPI performs better than SPI in identifying agricultural drought, particularly in semi-arid Mediterranean climates. However, the applicability of these enhanced indices in humid environments remains underexplored, especially in relation to the methods used for estimating effective precipitation.

Several studies have assessed the predictive power of these indices. Omar et al. [23] evaluated SPI, aSPI, RDI and eRDI for forecasting rainfed maize yield in Mexico (1982–2013) and found that aSPI and eRDI were the most responsive, aiding in the optimization of planting and harvesting cycles. Syed et al. [22] analyzed six drought indices—SPI, RDI, Deciles Index, Percentage Departure, aSPI, and eRDI—in Saudi Arabia (1985–2020) and revealed a strong link between droughts and the Pacific Decadal Oscillation (PDO), supporting climate-smart contingency planning. Proutsos et al. [21] analyzed drought conditions in Heraklion (1955–2022) using SPI, RDI, aSPI, eRDI, showing consistent detection of severe and extreme droughts, with some variation in identifying moderate events. Cárdenas [24] developed predictive models linking SPI, aSPI, RDI and eRDI with major climate indices (ONI, PDO, NAO, AMO) in Mexico, identifying significant correlations that can support sustainable water management. Ibrahim et al. [25] examined drought risk in Saudi Arabia’s Al-Baha region (1991–2022) using SPI, RDI, aSPI and eRDI, highlighting severe drought years, declining precipitation, rising temperatures, and projected worsening conditions under RCP4.5 and RCP8.5 scenarios.

Effective precipitation—essential for aSPI and eRDI—is commonly calculated using four approaches: USBR, USDA-SCS CROPWAT method, USDA-SCS simplified method, and the FAO method. The first three are recommended for arid and semi-arid conditions [12]. In previous studies [23,24,26], the USDA method has generally been used in arid regions. In the study by Syed et al. [22], effective precipitation was calculated using three methods: USBR and two different USDA approaches. The results obtained from all three methods were similar. Therefore, the USBR method, which is recommended for arid and semi-arid regions, was adopted in this study. The FAO method, on the other hand, follows a different approach and often results in reduced effective precipitation values compared to other methods for the same total rainfall [12].

Although widely used, the accuracy of these methods in humid climates has been questioned, with studies such as Hess [27] suggesting their inappropriateness, and Tigkas et al. [28] emphasizing the need for further evaluation of their applicability and reliability under such conditions.

Due to the recent extreme drought events and the increasing frequency and severity of droughts, numerous drought-related studies have been conducted in Türkiye, and these studies have been compiled in the review research by Soylu Pekpostalcı et al. [29]. Additionally; Öz et al. [30] analyzes drought in Türkiye using SPI, SPEI, and RDI with data from 219 meteorological stations (1991–2022). Findings reveal a trend toward drier conditions, influenced by rising temperatures, changes in precipitation patterns, and Türkiye’s Mediterranean location. The analysis highlights the growing frequency and severity of droughts, especially during positive North Atlantic Oscillation phases, and underscores the potential risks of climate change on water resources and ecosystems.

The trend analysis of drought indices has been a central focus in numerous studies within the literature. Among the most commonly applied methods are the Mann–Kendall test and Sen’s Slope estimator, widely used in meteorological, hydrological, and oceanographic research. These methods have been employed in various studies [31,32,33,34,35,36]. Additionally, several other works [18,21,37,38,39,40,41,42] have also conducted trend analyses of drought indices using these techniques, confirming their prevalence in drought-related research.

This study aims to evaluate the influence of different effective precipitation calculation methods on drought assessment by comparing precipitation-based drought indices (SPI and RDI) with effective precipitation-based indices (aSPI and eRDI), computed using four distinct approaches across multiple time scales (1-, 3-, 6-, 12-month, and annual). To the best of our knowledge, this is the first study to systematically compare these four methods for estimating effective precipitation in humid regions within the context of drought analysis. Furthermore, the study investigates whether the choice of effective precipitation method leads to significant differences in the trend behavior of the resulting drought indices. The central hypothesis of this study is that the choice of the effective precipitation estimation method significantly influences drought index performance, particularly in humid regions where traditional precipitation-based indices may underestimate drought persistence. Accordingly, the objective is to identify the most reliable Pe estimation method for accurate drought characterization across multiple temporal scales.

2. Materials and Methods

2.1. Study Area and Data

Çanakkale is located in the northwest of Türkiye. This region lies between the latitudes of 39° 30′ N and 40° 42′ N and longitudes of 25° 35′ E and 27° 45′ E (Figure 1). The region lies along the Dardanelles Straight, which connects the Aegean Sea to the Sea of Marmara. The total area of Çanakkale province is approximately 9817 km², including two major islands: Bozcaada and Gökçeada. Elevation in the region ranges from sea level to about 1767 m, with the highest elevations found in the Kaz Mountains. According to long-term meteorological records (1929–2023), the region has an annual average temperature of 15.2 °C and average annual total precipitation of 625.3 mm [43].

Monthly total precipitation and monthly mean temperature data for three stations—Çanakkale, Bozcada and Gökçeada—were obtained from Turkish State Meteorological Service (TMSS). The locations of the stations within the study area are shown in Figure 1.

Meteorological data from these stations were analyzed for the period from October 1967 to September 2023. These three stations were selected because they provide the longest and most consistent precipitation and temperature records (56 years) in the region. The aridity index for each station was calculated based on the UNEP method [44], and the characteristics of the stations are summarized in

Table 1. Characteristics of the meteorological stations in the study area.

Station	Elevation (m)	Annual Precipitation (mm)				Monthly Precipitation Obs. < 110 mm	Aridity Index	Climate
		Mean	Max	Min	Std Dev.
Çanakkale	6	592.4	843.6	332.8	134.2	89%	0.73	Humid
Bozcaada	30	466.8	812.8	124.4	148.7	92%	0.58	Dry-Subhumid
Gökçeada	79	716.5	1287.9	419.3	195.6	82%	0.88	Humid

.

The Çanakkale region holds notable agricultural importance, as approximately 76% of its cultivable land consists of arable fields [45]. This high proportion of agricultural land, together with its humid climatic setting and rainfall variability, underscores the region’s hydrological relevance and makes it an appropriate case for investigating how different effective precipitation estimation methods influence drought index performance in humid environments.

Data availability of the precipitation is greater than at least 94.5% for all stations. The missing values were estimated using a combination of simple imputation (nearest neighbor) and EM-MCMC (EM Monte Carlo Markov Chain) method [46]. Missing data were addressed through a two-step process. First, the nearest neighbor method was applied selectively during the summer months when no precipitation was recorded at any of the nearby stations, which, in this study, refers to the three stations that are geographically close to each other. This approach was preferred because applying the MCMC method directly in such cases produced non-zero rainfall estimates for periods known to be completely dry, which would be unrealistic in the context of a drought analysis. Using the nearest neighbor method in this way ensured that true dry-season conditions were preserved. For the remaining missing values outside the dry season, the MCMC method was used, as it was identified by Yozgatlıgil et al. [46] as the most accurate approach for completing meteorological station data in Türkiye, and was therefore preferred in this study.

Annual totals for precipitation, annual total potential evapotranspiration (PET), and annual average temperature values for each station are presented in Figure 2. Based on the 56-year dataset, Bozcaada has the lowest aridity index, whereas Gökçeada receives the highest annual precipitation. Notably, Gökçeada is also the highest elevation among the three stations. In the last two decades (2004–2023), a decreasing trend in aridity index values has been observed at all stations, indicating a potential shift toward more humid conditions.

2.2. Drought Indices

In this study, four drought indices—SPI, aSPI, RDI and eRDI—were calculated for time scales of 1, 3, 6, and 12 months and annually. All calculations of indices were performed using the DrinC software (version 1.7), which is specifically designed for drought analysis [12,28,47].

DrinC utilizes the hydrological year—commonly starting in October for Mediterranean regions—as its reference period. This framework facilitates structured evaluation of water shortages and proves especially useful for hydrological analyses, drought surveillance, and the development of early warning systems [48].

2.2.1. Standardized Precipitation Index (SPI)

The Standardized Precipitation Index (SPI), which is the most commonly used index in drought analysis studies, calculates precipitation deficits over various time scales (e.g., 1 to 24 months) by fitting cumulative precipitation data to a statistical distribution, which is then normalized. This standardization simplifies drought classification, with negative SPI values indicating dry periods and positive values signifying wet conditions [5]. The index’s adaptability and ease of calculation have made it a globally recognized method for studying droughts in diverse contexts. It is particularly effective in capturing temporal variability in precipitation, aiding in the identification of both short-term and long-term drought patterns. The theoretical foundation of the SPI is extensively detailed in the literature, such as [18,23,47,49,50].

2.2.2. Agricultural Standardized Precipitation Index (aSPI)

The absence of a soil water balance in SPI limits its application in vegetation-related studies. By incorporating effective precipitation (P_e) instead of total precipitation, aSPI becomes more suitable for characterizing agricultural droughts and assessing their impacts on vegetation [12]. The aSPI, by estimating P_e through various methods, retains the SPI’s core advantage of relying solely on precipitation data, making it particularly valuable in data-scarce regions [12].

For detailed computation procedures of aSPI, refer to [12,18,23,28]. The specific methods used to estimate effective precipitation in this study are described in Section 2.3.

2.2.3. Reconnaissance Drought Index (RDI)

The Reconnaissance Drought Index (RDI) was introduced for evaluating meteorological drought by considering cumulative precipitation (P) and potential evapotranspiration (PET) [9].

According to Vangelis [51], the RDI is a reliable index for evaluating drought severity and remains unaffected by the method employed for PET calculation.

In this study, the Thornthwaite method, which uses mean temperature data, was employed to calculate PET. This method, proposed by Thornthwaite [52], is one of the most commonly used for PET [51].

2.2.4. Effective Reconnaissance Drought Index (eRDI)

According to Tigkas et al. [14], P_e provides a more precise measure of the water available for productive use by crops compared to total precipitation. The proposed eRDI aims to better correlate drought severity with agricultural production losses, making it more suitable for assessing drought impacts on rainfed agricultural systems. Therefore, the eRDI is calculated by using effective precipitation instead of total precipitation in the RDI calculation. The normalized (eRDIn) and standardized (eRDIst) versions of the RDI are computed using procedures similar to those outlined for the RDI [13,14].

The drought classes for the RDI-eRDI are provided by Tigkas et al. [14], while the drought classes for the SPI-aSPI are outlined by Tigkas et al. [12]. Additionally, Tigkas et al. [28] define the drought classes for the aSPI and eRDI. Building on previous classifications, as similarly outlined by Ibrahim et al. [25], the drought categories employed in this study for the SPI, aSPI, RDI, and eRDI are detailed in

Table 2. Drought category based on SPI-aSPI-RDI-eRDI values [25].

aSPI-eRDI Value	Category
≥2.00	Extremely wet
1.5 to 1.99	Severely wet
1 to 1.49	Moderately wet
0.5 to 0.99	Mildly wet
−0.49 to 0.49	Normal
−0.50 to −0.99	Mild drought
−1.00 to −1.49	Moderate drought
−1.50 to −1.99	Severe drought
≤−2.00	Extreme drought

.

2.3. Effective Precipitation for aSPI and eRDI

Effective precipitation (P_e) is defined as the portion of total precipitation that is available for plant use after accounting for losses such as surface runoff, deep percolation, and evaporation. For the computation of aSPI and eRDI, accurate estimation of P_e is crucial, especially in agricultural drought assessments.

In the DrinC program, four different methods are used to calculate effective precipitation using monthly precipitation data:

• USBR (U.S. Bureau of Reclamation) method;
• USDA-SCS CROPWAT (Soil Conservation Service/U.S. Department of Agriculture–CROPWAT version) method;
• USDA-SCS (simp.Soil Conservation Service/U.S. Department of Agriculture–simplified version) method;
• FAO (U.N. Food and Agriculture Organization) method.

The USBR method is a temperature-based method that estimates effective precipitation as a function of mean monthly temperature and precipitation. It is one of the simplest and most commonly used approaches for arid and semi-arid regions [12]. The method assumes that precipitation becomes less effective at higher temperatures due to increased evapotranspiration. In this method, effective rainfall is calculated based on rainfall classification using the criteria presented in

Table 3. Estimation of effective precipitation based on total monthly precipitation Classes [12,28,53].

Total Monthly Precipitation Class (mm)	Effective Precipitation Class (%)
0–25.4	90–100
25.4–50.8	85–95
50.8–76.2	75–90
76.2–101.6	50–80
101.6–127	30–60
127–152.4	10–40
>152.4	0–10

.

Under different soil and climatic conditions of the USA, and based on the analysis of long-term data series, the US Department of Agriculture (USDA-SCS) developed the following equation (Equation (1)), which considers total precipitation (P), crop evapotranspiration (ETc), and soil water storage factor (SF) [27,54,55].

$${\mathrm{P}}_{\mathrm{e}}=25.4 \mathrm{S}\mathrm{F}\left(0.04931 {\mathrm{P}}^{0.82416}-0.11565\right){10}^{0.000955 {\mathrm{E}\mathrm{T}}_{\mathrm{C}}}$$

(1)

In this equation, where all units are in millimeters (mm), SF is calculated using Equation (2):

$$\mathrm{S}\mathrm{F}=0.531747+0.011621\mathrm{D}-8.943\times {10}^{-5}{\mathrm{D}}^2+2.321\times {10}^{-7}{\mathrm{D}}^3$$

(2)

where D represents the available water storage in the soil. For the root zones of plants, this value is typically taken to be in the range of approximately 40–60%.

In the study area, when ETc is assigned a constant value and SF is set to 1 (equivalent to 75 mm of available water storage in the soil), the simplified form of this method (USDA-Simp.) is obtained. Although this value may vary depending on soil texture, depth, and vegetation characteristics, no regional calibration was applied in order to evaluate the performance of standardized methods under local humid conditions.

Another method calculates effective precipitation using the CROPWAT model, considering the monthly total precipitation [56]. In this method, effective precipitation is calculated using the following Equation (3):

$$\begin{gathered} {\mathrm{P}}_{\mathrm{e}}=\mathrm{P}(125-0.2\mathrm{P})/125\hspace{1em}\hspace{1em}\mathrm{P}\le 250\mathrm{mm} \\ {\mathrm{P}}_{\mathrm{e}}=0.1\mathrm{P}+125\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\mathrm{P}>250\mathrm{mm} \end{gathered}$$

(3)

The FAO method, calculated using a simple empirical formula, is used in areas with a slope of up to a maximum of 4–5%. For this method, effective precipitation is calculated using Equation (4) [57]:

$$\begin{gathered} {\mathrm{P}}_{\mathrm{e}}=0.6\mathrm{P}-10\hspace{1em}\mathrm{P}<70 \mathrm{m}\mathrm{m} \\ {\mathrm{P}}_{\mathrm{e}}=0.8\mathrm{P}-25\hspace{1em}\mathrm{P}\ge 70 \mathrm{m}\mathrm{m} \end{gathered}$$

(4)

The monthly effective precipitation values calculated from the monthly total precipitation using the four mentioned methods are presented in Figure 3 [12]. It is observed that for monthly total precipitation up to 110 mm, which is commonly seen in arid and semi-arid regions, the USBR, USDA-CROPWAT, and USDA-Simple methods generate similar results. When the total precipitation exceeds 110 mm, the USBR method estimates the effective precipitation to be approximately 100 mm. For total precipitation up to 150 mm, both USDA-CROPWAT and USDA-Simp. methods calculate similar effective precipitation value. At higher precipitation levels, the USDA-Simp method maintains a linear increase, while the CROPWAT version has a much lower slope [12].

Compared to the other methods, the FAO method calculates lower effective precipitation for the same total monthly precipitation, yielding different results. However, this trend reverses for precipitation amounts greater than 150 mm [28]. Syed et al. [22] calculated the same effective precipitation using the first three methods (USBR, USDA-CROPWAT, USDA-Simp.) in their study in arid regions.

These methods are commonly used under arid and semi-arid conditions [53,58,59]. However, in some studies, such as Hess [27], these methods are generally considered unsuitable for humid climates, while Tigkas et al. [28] suggest that their reliability should be tested in humid environments.

2.4. Statistical Parameters

Statistical parameters, including the correlation coefficient (R²), Root Mean Square Error (RMSE) and Nash–Sutcliffe Efficiency (NSE), were calculated to compare drought indices with each other and effective precipitation with precipitation. These parameters are widely used in hydrological and climatological research for model validation and comparative analysis. A lower RMSE value indicates a better fit and closer match between datasets. The R² and NSE values approaching 1 indicate a closer match between data.

In this study, these three complementary criteria were jointly applied to comprehensively evaluate both the accuracy and predictive capability of each method. R² represents the strength of linear association, RMSE quantifies the magnitude of deviation, and NSE measures overall model efficiency; together, they form a robust statistical framework for cross-validating the performance of drought indices under different Pe estimation methods. Mathematical equations of each of the statistical parameters mentioned are given in Equations (5) to (7). By combining these three performance criteria, a comprehensive assessment of drought index behavior and P_e estimation accuracy is achieved, providing insight into the reliability and applicability of different calculation methods across stations and time scales.

$${\mathrm{R}}^2={\displaystyle \frac{{\left[\sum _{\mathrm{i}=1}^{\mathrm{N}}\left(\mathrm{R}\mathrm{D}\mathrm{I}-\overline{\mathrm{R}\mathrm{D}\mathrm{I}}\right)\left(\mathrm{e}\mathrm{R}\mathrm{D}\mathrm{I}-\overline{\mathrm{e}\mathrm{R}\mathrm{D}\mathrm{I}}\right)\right]}^2}{\sum _{\mathrm{i}=1}^{\mathrm{N}}{\left(\mathrm{R}\mathrm{D}\mathrm{I}-\overline{\mathrm{R}\mathrm{D}\mathrm{I}}\right)}^2\sum _{\mathrm{i}=1}^{\mathrm{N}}{(\mathrm{e}\mathrm{R}\mathrm{D}\mathrm{I}-\overline{\mathrm{e}\mathrm{R}\mathrm{D}\mathrm{I}})}^2}}$$

(5)

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{1}{\mathrm{N}}}\sum _{\mathrm{i}=1}^{\mathrm{N}}{(\mathrm{R}\mathrm{D}\mathrm{I}-\mathrm{e}\mathrm{R}\mathrm{D}\mathrm{I})}^2}$$

(6)

$$\mathrm{N}\mathrm{S}\mathrm{E}=1-\left[{\displaystyle \frac{\sum _{\mathrm{i}=1}^{\mathrm{N}}{\left(\mathrm{R}\mathrm{D}\mathrm{I}-\mathrm{e}\mathrm{R}\mathrm{D}\mathrm{I}\right)}^2}{\sum _{\mathrm{i}=1}^{\mathrm{N}}{\left(\mathrm{R}\mathrm{D}\mathrm{I}-\overline{\mathrm{R}\mathrm{D}\mathrm{I}}\right)}^2}}\right]$$

(7)

2.5. Statistical Tests for Trend Analysis

In this study, Mann–Kendall and Theil–Sen methods were used for trend analysis, both of which are non-parametric techniques that are robust against outliers in time series data. We used the non-parametric Mann–Kendall test specifically for assessing trend significance. The Mann–Kendall test, introduced by Mann [60] and Kendall [61], is a statistical tool designed to identify monotonic trends—either increasing, decreasing, or no trend—within time series data. This method is particularly useful for datasets that do not follow a normal distribution, making it suitable for a wide range of fields such as hydrology, climate science, environmental monitoring, and economics [62]. It is particularly appropriate for hydro-meteorological data, which often deviate from normality and variance homogeneity [63]. The test evaluates the data without assuming a specific distribution, which makes it highly versatile and reliable for analyzing non-parametric data [64].

In conjunction with Mann–Kendall, the Theil–Sen estimator [65,66] is often employed to estimate the slope of the trends. This method is specifically designed to determine the rate of change over time. Positive slopes indicate an increasing trend, negative slopes suggest a decreasing trend, and near-zero slopes imply stability or no significant change. The Theil–Sen estimator is particularly favored for its resilience to outliers and is widely applied in hydrometeorological and oceanographic studies [34,35,67].

Both methods are integral for analyzing long-term trends in data where traditional parametric assumptions may not hold, making them invaluable tools in environmental and climate research. In this study, they were applied to all indices (SPI, aSPI, RDI, and eRDI) across multiple time scales to determine whether the use of different effective precipitation methods leads to significant changes in the direction or magnitude of drought trends.

As the mathematical formulas for these methods are well-established and frequently cited, they are not explicitly included here. Readers interested in the computational details are referred to previous studies such as [18,36,68,69].

3. Results and Discussion

3.1. Effective Precipitation

The average of annual total precipitation (P) and effective precipitation (P_e) obtained with four different estimation methods, calculated for all stations is presented in

Table 4. Average of annual total precipitation and effective precipitation obtained using different methods at three stations.

		Çanakkale	Bozcaada	Gökçeada
P (mm)		592.4	466.8	716.5
Pe (mm)	USBR	476.8	385.4	531.5
	USDA-CROPWAT	498.2	399.2	572.2
	USDA-Simp.	489.9	388.6	577.0
	FAO	288.0	212.6	381.7

. Among the three locations, Gökçeada shows the highest values for both total and effective precipitation, consistent with its relatively high aridity index (0.88; see Table 1). Conversely, Bozcaada, which has the lowest aridity index (0.58), records the lowest annual totals.

Figure 4 presents monthly total precipitation and effective precipitation calculated based on different methods in Çanakkale station. These visualizations allow for an assessment of how each method responds to seasonal variations in precipitation. Notable differences arise in the estimation of P_e, particularly during months with high rainfall, underscoring the sensitivity of each method to precipitation magnitude and distribution.

To quantitatively assess the applicability of each method, correlation coefficients (R²) and root mean square error (RMSE) values were computed between monthly total precipitations and estimated effective precipitation (

Table 5. Correlation coefficient (R²) and Root Mean Square Error (RMSE) between monthly precipitation and monthly effective precipitation with different methods.

	R2			RMSE (mm)
	Çanakkale	Bozcaada	Gökçeada	Çanakkale	Bozcaada	Gökçeada
USBR	0.90	0.90	0.86	23.55	19.69	36.07
USDA-CROPWAT	0.97	0.97	0.95	17.42	14.64	26.81
USDA-Simp.	0.99	0.99	0.99	14.83	12.34	20.51
FAO	0.97	0.97	0.98	30.87	27.15	34.49

). The USDA-Simplified method consistently demonstrates the highest predictive agreement with total precipitation across all three stations, achieving R² values of 0.99 and the lowest RMSE values. This suggests a high degree of consistency in P_e estimation despite its simplified formulation. However, the USDA-Simplified method uses a fixed soil water storage value (SF = 1, ≈75 mm) [12], which may not represent all humid regions, as soil water capacity depends on factors like texture, depth, and land use [70]. This approach, while suitable for our comparative and transferability assessment, has a recognized limitation: the use of a fixed SF value may underrepresent the high spatial variability of soil moisture storage capacity found in humid areas with diverse land cover and vegetation.

• Nevertheless, the primary objective of this study was not to calibrate these parameters locally, but rather to assess the transferability and performance of the predefined methods embedded in the DrinC software when applied to a humid region such as northwestern Türkiye.

Conversely, the FAO method yields the highest RMSE values at the Çanakkale (30.87 mm) and Bozcaada (27.15 mm) stations, indicating reduced accuracy in estimating effective precipitation in humid settings. This result is consistent with previous findings [27,28], which caution against the application of FAO-based empirical methods in non-arid regions due to their tendency to underestimate P_e under moderate to high precipitation regimes.

The USBR method, although commonly employed in arid and semi-arid regions, produces relatively high RMSE values across all stations in this study. This outcome likely reflects the method’s calibration for low-precipitation environments, which may not adequately capture the hydrological dynamics of humid regions such as Çanakkale, Bozcaada, and Gökçeada.

3.2. Assessment of Drought Indices

To assess the consistency and responsiveness of drought indices under varying effective precipitation estimation methods, time series of SPI, aSPI, RDI, and eRDI were analyzed at multiple temporal scales (1-, 3-, 6-, and 12-month and annual). Figure 5 displays representative time series plots for the Gökçeada station—the wettest among the three—at 6- and 12-month scales. These plots illustrate the temporal evolution of SPI-aSPI and RDI-eRDI pairs under each method.

For same time scale, the SPI/aSPI and RDI/eRDI produce similar results, which is consistent with the findings of Cheraghalizadeh et al. [71], who reported no significant differences among precipitation-based and PET-based indices (SPI, RDI, and SPEI) in cold and humid basins due to the limited influence of evapotranspiration on drought occurrence under such climatic conditions. Similar spatial variability in the relationship between SPI and RDI was also reported by Zarch et al. [72], who found that differences between these two indices were more pronounced in humid and sub-humid regions of Iran, while arid and semi-arid areas exhibited stronger agreement between them. This supports the present finding that the divergence between precipitation-based and PET-based drought indices tends to increase under humid climatic conditions. Khalili et al. [73] similarly emphasized that RDI provides a more representative measure of climatic variability, particularly for agricultural applications, underscoring its advantage over purely precipitation-based indices.

The observed differences among SPI, RDI, aSPI, and eRDI primarily reflect the varying physical processes that each index captures. SPI depends solely on precipitation variability and thus responds directly to rainfall deficits. In contrast, RDI and eRDI integrate both precipitation and potential evapotranspiration (PET), providing a more complete representation of climatic water balance.

• This distinction becomes particularly important in humid regions, where evapotranspiration significantly influences water availability. Moreover, the inclusion of effective precipitation (Pe) in aSPI and eRDI further refines their response by accounting for the portion of rainfall that contributes to actual soil moisture and runoff, explaining their greater sensitivity at longer timescales.

For a more comprehensive comparison,

Table 6. Statistical Parameters between indices at Çanakkale station.

ÇANAKKALE			R2	RMSE	NSE			R2	RMSE	NSE
1	SPI-aSPI	USBR	0.97	0.158	0.973	RDI-eRDI	USBR	0.98	0.152	0.975
		USDA-CROP	0.99	0.082	0.993		USDA-CROP	0.99	0.082	0.993
		USDA-simp	0.93	0.271	0.910		USDA-simp	0.93	0.263	0.917
		FAO	0.68	0.615	0.478		FAO	0.68	0.616	0.493
	SPI-RDI		0.89	0.322	0.888
3	SPI-aSPI	USBR	0.95	0.236	0.944	RDI-eRDI	USBR	0.95	0.230	0.947
		USDA-CROP	0.99	0.121	0.985		USDA-CROP	0.99	0.123	0.985
		USDA-simp	0.99	0.115	0.986		USDA-simp	0.99	0.115	0.987
		FAO	0.92	0.286	0.908		FAO	0.92	0.295	0.902
	SPI-RDI		0.92	0.28	0.922
6	SPI-aSPI	USBR	0.90	0.318	0.899	RDI-eRDI	USBR	0.90	0.313	0.902
		USDA-CROP	0.97	0.169	0.971		USDA-CROP	0.97	0.170	0.971
		USDA-simp	0.99	0.092	0.992		USDA-simp	0.99	0.093	0.991
		FAO	0.96	0.201	0.959		FAO	0.96	0.212	0.955
	SPI-RDI		0.96	0.213	0.955
12	SPI-aSPI	USBR	0.87	0.372	0.862	RDI-eRDI	USBR	0.87	0.368	0.865
		USDA-CROP	0.96	0.200	0.960		USDA-CROP	0.96	0.201	0.959
		USDA-simp	0.99	0.095	0.991		USDA-simp	0.99	0.097	0.991
		FAO	0.97	0.177	0.969		FAO	0.96	0.195	0.962
	SPI-RDI		0.95	0.222	0.951
An	SPI-aSPI	USBR	0.86	0.384	0.852	RDI-eRDI	USBR	0.86	0.378	0.857
		USDA-CROP	0.96	0.207	0.957		USDA-CROP	0.96	0.208	0.957
		USDA-simp	0.99	0.095	0.991		USDA-simp	0.99	0.097	0.991
		FAO	0.97	0.178	0.968		FAO	0.96	0.194	0.963
	SPI-RDI		0.95	0.219	0.952

,

Table 7. Statistical Parameters between indices at Bozcaada station.

BOZCAADA			R2	RMSE	NSE			R2	RMSE	NSE
1	SPI-aSPI	USBR	0.98	0.139	0.978	RDI-eRDI	USBR	0.98	0.134	0.979
		USDA-CROP	0.99	0.074	0.994		USDA-CROP	0.99	0.074	0.994
		USDA-simp	0.93	0.263	0.908		USDA-simp	0.94	0.260	0.911
		FAO	0.60	0.703	0.246		FAO	0.61	0.701	0.258
	SPI-RDI		0.94	0.238	0.936
3	SPI-aSPI	USBR	0.96	0.205	0.958	RDI-eRDI	USBR	0.96	0.204	0.958
		USDA-CROP	0.99	0.111	0.988		USDA-CROP	0.99	0.112	0.987
		USDA-simp	0.98	0.139	0.980		USDA-simp	0.98	0.139	0.980
		FAO	0.84	0.412	0.789		FAO	0.84	0.416	0.786
	SPI-RDI		0.95	0.223	0.95
6	SPI-aSPI	USBR	0.94	0.255	0.935	RDI-eRDI	USBR	0.94	0.256	0.935
		USDA-CROP	0.98	0.142	0.980		USDA-CROP	0.98	0.144	0.979
		USDA-simp	0.98	0.134	0.982		USDA-simp	0.98	0.133	0.982
		FAO	0.93	0.265	0.926		FAO	0.93	0.269	0.923
	SPI-RDI		0.97	0.165	0.973
12	SPI-aSPI	USBR	0.92	0.278	0.923	RDI-eRDI	USBR	0.93	0.275	0.924
		USDA-CROP	0.98	0.155	0.976		USDA-CROP	0.98	0.154	0.976
		USDA-simp	0.99	0.084	0.993		USDA-simp	0.99	0.083	0.993
		FAO	0.97	0.173	0.970		FAO	0.97	0.180	0.968
	SPI-RDI		0.98	0.141	0.98
An	SPI-aSPI	USBR	0.93	0.268	0.928	RDI-eRDI	USBR	0.93	0.264	0.930
		USDA-CROP	0.98	0.148	0.978		USDA-CROP	0.98	0.146	0.979
		USDA-simp	0.99	0.081	0.993		USDA-simp	0.99	0.079	0.994
		FAO	0.98	0.152	0.977		FAO	0.98	0.155	0.976
	SPI-RDI		0.98	0.128	0.984

and

Table 8. Statistical Parameters between indices at Gökçeada station.

GÖKÇEADA			R2	RMSE	NSE			R2	RMSE	NSE
1	SPI-aSPI	USBR	0.96	0.194	0.958	RDI-eRDI	USBR	0.97	0.178	0.965
		USDA-CROP	0.99	0.101	0.989		USDA-CROP	0.99	0.096	0.990
		USDA-simp	0.94	0.244	0.925		USDA-simp	0.95	0.231	0.935
		FAO	0.73	0.551	0.571		FAO	0.74	0.542	0.594
	SPI-RDI		0.91	0.295	0.905
3	SPI-aSPI	USBR	0.92	0.287	0.918	RDI-eRDI	USBR	0.92	0.277	0.923
		USDA-CROP	0.98	0.152	0.977		USDA-CROP	0.98	0.152	0.977
		USDA-simp	0.99	0.110	0.988		USDA-simp	0.99	0.110	0.988
		FAO	0.93	0.263	0.922		FAO	0.93	0.269	0.919
	SPI-RDI		0.93	0.263	0.931
6	SPI-aSPI	USBR	0.88	0.345	0.881	RDI-eRDI	USBR	0.89	0.343	0.882
		USDA-CROP	0.96	0.190	0.964		USDA-CROP	0.96	0.191	0.963
		USDA-simp	0.99	0.088	0.992		USDA-simp	0.99	0.089	0.992
		FAO	0.97	0.172	0.970		FAO	0.97	0.180	0.967
	SPI-RDI		0.96	0.19	0.964
12	SPI-aSPI	USBR	0.86	0.375	0.859	RDI-eRDI	USBR	0.86	0.376	0.859
		USDA-CROP	0.95	0.214	0.954		USDA-CROP	0.95	0.215	0.954
		USDA-simp	0.99	0.088	0.992		USDA-simp	0.99	0.090	0.992
		FAO	0.98	0.147	0.978		FAO	0.98	0.157	0.975
	SPI-RDI		0.97	0.162	0.974
An	SPI-aSPI	USBR	0.85	0.391	0.847	RDI-eRDI	USBR	0.86	0.387	0.850
		USDA-CROP	0.95	0.220	0.951		USDA-CROP	0.95	0.218	0.952
		USDA-simp	0.99	0.094	0.991		USDA-simp	0.99	0.094	0.991
		FAO	0.98	0.154	0.976		FAO	0.97	0.161	0.974
	SPI-RDI		0.97	0.164	0.973

have been prepared, presenting statistical parameters for all stations, including a comparison between SPI and aSPI, as well as between RDI and eRDI.

Across all time scales and stations, the SPI and aSPI demonstrate a high degree of similarity, particularly when the aSPI is calculated using the USDA-based methods. Likewise, RDI and eRDI show strong alignment when effective precipitation is estimated using the USDA-Simplified method. This suggests that USDA methods yield effective precipitation values that most closely resemble total precipitation in relative variation, especially at longer time scales. The USDA-Simplified method shows the best overall performance, with the highest R² values (up to 0.99) for all-time series beyond the 3-month scale, along with the lowest RMSE and highest NSE values, reinforcing its robustness. At the 1-month time scale, however, the USDA-CROPWAT method consistently exhibits slightly higher R² values across all stations—reaching 0.99—indicating marginally better agreement for short-term drought index calculations. This finding is consistent with Muratoğlu et al. [74], who observed that higher humidity increased the accuracy of the USDA-SCS method. In the study of Muratoğlu et al. [74], the USDA-SCS approach was presented in its general form (Equations (1) and (2)), encompassing both the USDA-Simplified (USDA-Simp) and USDA-CROPWAT methods. Some researchers have described the CROPWAT method as a simplified adaptation of the USDA-SCS approach [75], whereas others—particularly in the context of water footprint assessments—have regarded it as essentially identical to the USDA-SCS method [27,76,77]. Moreover, [75] highlighted that the CROPWAT method generally provides higher and more consistent estimates of effective precipitation than the USDA-SCS approach, which supports its reliability for drought index calculations in humid regions.

The coefficient of determination (R²) values between SPI and RDI have shown a strong agreement for all time scales. The lowest agreement between SPI and RDI, with an R² value of 0.89, was observed at the Çanakkale station for the 1-month time scale, while the highest agreement, with an R² value of 0.98, was found at the Bozcaada station for the 12-month and annual time scales.

The radar charts presented in Figure 6 offer a clear and concise visual summary of the statistical performance of different methods across all indices for the Gökçeada station. By simultaneously displaying key metrics such as R², RMSE, and NSE, they allow for an integrated comparison that highlights performance differences between methods. This visualization helps identify approaches that maintain consistently high accuracy and low error across metrics, thereby informing the selection of the most suitable drought assessment methods for similar contexts.

As shown in Table 6, Table 7 and Table 8 and Figure 6, the FAO-based aSPI and eRDI diverge significantly from the other methods at the 1-month scale, exhibiting notably lower R² and NSE values and higher RMSE, consistent with findings from arid basin studies reported in the literature. In arid regions, it has been noted that a high percentage of zero values for effective precipitation calculated using the FAO method can make calculating indices challenging, leading to inaccurate results on the 1-month time scale [26].

A similar pattern is observed in this study, where under the humid conditions of the Çanakkale region, the FAO method exhibits limited drought sensitivity at the 1-month time scale. This is consistent with previous research [28] indicating that the FAO method—along with other empirical effective precipitation estimation methods—is mostly considered suitable for arid and semi-arid conditions, with limited credibility in humid environments. Bokke and Shoro [58] stated that the FAO (dependable rain) method is more suitable for regions with adequate water availability and for use in small-scale irrigation schemes. In addition, in the FAO method, for a given amount of precipitation, the estimated effective precipitation is generally lower than in the USBR and USDA methods, which can amplify differences at shorter time scales [26]. Furthermore, the FAO method can be applied mainly for plain areas with a maximum slope of 4–5% [28]. While these factors may explain part of the divergence at 1-month scale, the difference diminishes with increasing temporal aggregation, and at longer time scales (3-month, 6-month, 12-month, and annual) the FAO-based indices align more closely with those derived from the other methods.

Conversely, the USBR method tends to produce increasing RMSE values for aSPI and eRDI as the time scale lengthens, particularly beyond the 3-month scale. This indicates a cumulative mismatch in long-term water deficits when using USBR-derived P_e in humid settings. This limitation is consistent with previous findings, as Ali and Mubarak [78] emphasized that although the USBR method can be used for broad planning purposes, its accuracy is generally low and it may lead to under- or over-estimation of effective precipitation depending on rainfall distribution, making it less suitable for detailed drought assessments.

The USDA-Simplified method, by contrast, maintains the most stable and accurate index behavior across all time scales and stations.

These findings collectively underscore that both the choice of effective precipitation method and the temporal scale significantly influence the magnitude and variability of drought indices. Among the methods tested, the USDA-Simplified approach appears to offer the most reliable and consistent basis for drought assessment in humid environments such as the Çanakkale region.

Table 9. Number of drought periods for all time scales.

		Number of drought months (%)			Max severity of drought			Occurrence date of max severity
Time	Index	ÇAN	BOZ	GÖK	ÇAN	BOZ	GÖK	ÇAN	BOZ	GÖK
1	SPI	27.7	26.6	26.6	−3.59	−3.63	−3.65	Jan.89	Feb.89	Jan.89
	aSPI-USBR	26.0	25.9	25.1	−4.11	−3.96	−4.31
	aSPI-USDA-CROP	26.0	25.9	26.0	−3.91	−3.89	−4.06
	aSPI-USDA-simp	25.1	21.3	24.1	−3.20	−2.10	−3.74	Apr.06		Feb.23
	aSPI-FAO	17.0	12.2	19.5	−1.80	−1.24	−2.10	*A	*B	Mar.08
	RDI	27.5	26.8	28.1	−3.40	−3.40	−3.42	Jan. 89	Feb.89	Dec.15
	eRDI-USBR	26.6	25.9	25.6	−3.80	−3.57	−3.91
	eRDI-USDA-CROP	27.1	25.9	26.0	−3.65	−3.54	−3.72
	eRDI-USDA-simp	25.7	21.9	25.1	−3.00	−2.10	−3.17	Apr.06		Feb.23
	eRDI-FAO	16.8	12.2	18.8	−1.80	−1.24	−2.10	*A	*C	Mar.08
3	SPI	29.1	29.6	31.0	−3.18	−3.55	−2.96	Oct.69	May.20	Oct.69
	aSPI-USBR	28.4	27.8	28.5	−3.45	−3.97	−3.45			Mar.90
	aSPI-USDA-CROP	28.8	28.4	30.3	−3.40	−3.87	−3.19
	aSPI-USDA-simp	29.7	27.9	29.3	−2.92	−4.84	−3.26	Mar.90		Jul.99
	aSPI-FAO	26.6	23.6	25.7	−3.11	−3.35	−3.37		Jan.08	Mar.90
	RDI	28.5	29.1	31.2	−3.15	−3.32	−2.94	Oct.69	May.20
	eRDI-USBR	28.2	28.8	29.4	−3.39	−3.70	−3.27
	eRDI-USDA-CROP	28.5	28.8	30.7	−3.35	−3.61	−3.12
	eRDI-USDA-simp	29.0	28.8	30.0	−3.01	−4.59	−3.26	Mar.90		Jul.99
	eRDI-FAO	26.1	24.8	25.5	−3.24	−3.14	−3.41		Jan.08	Mar.90
6	SPI	30.1	28.3	32.1	−3.34	−3.95	−3.06	Oct.69	Jun.20	Sep.85
	aSPI-USBR	28.6	27.3	30.1	−3.65	−4.39	−3.27
	aSPI-USDA-CROP	28.8	27.0	31.2	−3.59	−4.27	−3.21
	aSPI-USDA-simp	29.1	25.9	31.5	−3.87	−5.38	−3.62		Aug.20
	aSPI-FAO	31.6	26.8	29.7	−2.79	−2.52	−2.88	Jun.08	Apr.20	Jan.74
	RDI	30.6	27.7	31.3	−3.29	−3.94	−3.02	Oct.69	Jun.20	Sep.85
	eRDI-USBR	28.5	27.7	31.2	−3.58	−4.39	−3.23
	eRDI-USDA-CROP	29.2	26.8	31.2	−3.52	−4.26	−3.17
	eRDI-USDA-simp	28.9	25.8	30.7	−3.82	−5.31	−3.58		Aug.20
	eRDI-FAO	31.3	27.4	29.2	−2.88	−2.60	−2.77	Jun.08	Apr.20	Jan.74
*A: March 2001–2003			*B: January-89/90/92/00/06/20					*C: January 89/90
		Number of drought months (%)			Max severity of drought			Occurrence date of max severity
Time	Index	ÇAN	BOZ	GÖK	ÇAN	BOZ	GÖK	ÇAN	BOZ	GÖK
12	SPI	33.1	27.8	31.3	−2.94	−3.28	−2.33	Jul.08	Nov.20	Dec.08
	aSPI-USBR	31.5	27.5	31.6	−3.11	−3.75	−2.40			Mar.90
	aSPI-USDA-CROP	32.1	26.8	30.6	−3.09	−3.60	−2.35
	aSPI-USDA-simp	32.1	27.2	31.3	−3.05	−3.60	−2.33			Dec.89
	aSPI-FAO	31.9	29.2	32.2	−2.74	−2.78	−2.62		Sep.20	Dec.08
	RDI	31.5	30.1	33.6	−2.95	−3.34	−2.34		Nov.20	Oct.16
	eRDI-USBR	29.0	28.9	32.2	−3.05	−3.80	−2.59
	eRDI-USDA-CROP	30.3	30.4	31.8	−3.05	−3.66	−2.53
	eRDI-USDA-simp	31.6	29.8	32.4	−3.03	−3.65	−2.42
	eRDI-FAO	31.0	30.3	33.1	−2.77	−2.86	−2.60		Sep.20	Dec.08
An	SPI	33.9	25.0	32.1	−2.21	−3.16	−1.77	2008	2020	1990
	aSPI-USBR	32.1	23.2	28.6	−2.19	−3.62	−2.32			1985
	aSPI-USDA-CROP	32.1	23.2	28.6	−2.21	−3.47	−2.01
	aSPI-USDA-simp	32.1	25.0	28.6	−2.19	−3.45	−1.82			1990
	aSPI-FAO	33.9	28.6	32.1	−2.29	−2.78	−1.91			1993
	RDI	33.9	26.8	30.0	−2.23	−3.28	−1.98	2023		2023
	eRDI-USBR	30.4	26.8	28.6	−2.31	−3.74	−2.22			1985
	eRDI-USDA-CROP	28.6	28.6	30.4	−2.29	−3.59	−2.01			2023
	eRDI-USDA-simp	32.1	28.6	30.4	−2.25	−3.57	−1.97	2008
	eRDI-FAO	33.9	32.1	30.4	−2.32	−2.86	−1.94

presents drought characteristics derived from all indices and time scales, including the percentage of drought months, the maximum severity of drought, and the corresponding dates of occurrence. Drought periods were identified as months when the index value dropped below −0.5.

In the humid conditions of the Çanakkale region, a clear pattern emerges: the percentage of drought months increases with longer time scales, from 1-month up to annual, across all stations and indices. Similar findings were also reported for the Marmara region by Soydan Oksal [79], who noted that drought events tended to persist for longer durations and become more pronounced at extended timescales, which aligns with previous studies [6,80]. This finding contrasts with results from arid region studies, such as Rezaei et al. [26], which reported stronger model performance and more distinct drought signals at shorter time scales (1- and 3-month) and where longer aggregation periods typically reduce drought frequency. In the present study conducted in a humid region, however, the number of drought months increases as the time scale extends, highlighting a fundamental difference in drought behavior across climatic zones. It is well established in the literature that the choice of timescale depends on the purpose of the study: Mishra and Singh [3] emphasized that monthly and annual periods are the most commonly used in drought assessments, and Panu and Sharma [81] highlighted that shorter timescales (1–3 months) are more appropriate for agricultural and water supply problems. The results of this study provide further evidence that timescale selection substantially affects drought assessments, particularly in humid climates where longer aggregation periods yield stronger drought signals.

At the 1-month time scale, the aSPI-USBR method yielded the highest maximum drought severity at all three stations. At 3 months, the highest severity was also produced by aSPI-USBR for Çanakkale and Gökçeada, and aSPI-USDA-Simp for Bozcaada. At 6 months, aSPI-USDA-Simp generated the highest severity across all stations. At the 12-month scale, the maximum drought severity was calculated by aSPI-USBR in Çanakkale, and by eRDI-USBR in Bozcaada and Gökçeada.

The aSPI-FAO and eRDI-FAO indices display a notably different pattern from the other indices, particularly at the 1-month time scale, characterized by a lower frequency of drought events and reduced maximum drought severity.

This result is consistent with findings from arid regions such as Rezaei et al. [26], where eRDI-FAO values were found to be higher due to the FAO method’s tendency to estimate lower effective precipitation compared to USBR and USDA methods. In arid and semi-arid climates, as well as in humid regions during dry periods, total precipitation is often minimal or zero. As the FAO method frequently assigns zero effective precipitation under such conditions, it can lead to computational difficulties and underestimate drought severity, particularly at short time scales. Consequently, this method has limited reliability for drought assessment at the 1-month scale.

When comparing index types, SPI and aSPI generally result in greater maximum drought severities than RDI and eRDI, particularly at shorter time scales (e.g., 1 month). This suggests that SPI-based indices are more responsive to abrupt precipitation deficits, while RDI-type indices, which account for PET, exhibit a more buffered response.

The dates of maximum severity identified by SPI/aSPI and RDI/eRDI often coincide, indicating consistent temporal drought signals across these index types. At the annual scale, the date of occurrence of maximum drought severity was identified as 2020 for the Bozcaada station, and 2008 and 2023 for the Çanakkale station, while different indices indicated different years for the Gökçeada station. These results are in line with previous studies reporting major drought events in Türkiye. For instance, Soydan Oksal [79] identified 1989, 1990, 2001, 2007, and 2014 as the most severe drought years in the Marmara region, while [80] highlighted the widespread and severe droughts of 1971–1974, 1989–1990, 2007–2008, and 2016–2017 across northwestern Türkiye. Similarly, Serkendiz et al. [82] reported 2001 as the most widespread drought year nationwide. The years identified in the present study (2008, 2020, 2023) therefore complement and partly overlap with these earlier findings, suggesting that the temporal distribution of severe droughts in the Çanakkale region is broadly consistent with nationwide and regional drought patterns with 2020 and 2023 underscoring the intensification of more recent drought events.

3.3. Trend Analysis of Drought Indices

A trend analysis has been conducted for all indices to determine whether the trend of the index changes based on the method used for effective rainfall calculation. The Mann–Kendall and Sen’s Slope methods were utilized to analyze the drought indices across all time scales and all stations, aiming to identify the trend values and assess their statistical significance. No significant trends were observed for the 1- and 3-month time scales at any of the stations. Trends and their significance levels for the 6-month, 12-month, and annual time scales across all drought indices are summarized in

Table 10. Sen slopes for all stations (/year).

.

Considering the variation in significance levels among stations, three confidence thresholds (10%, 5%, and 1%) were used in the evaluation. Statistically significant trends are highlighted in bold in Table 10, with significance levels additionally indicated by color coding.

Negative trend values are observed in both the SPI and RDI, as well as in the aSPI and eRDI that utilize effective precipitation, across the three given time scales. This indicates a further decrease in drought indices, suggesting a trend toward increasing drought severity for all stations. This result is consistent with drought assessments conducted in nearby regions and across Türkiye [30,37,82,83]. In particular, Soydan Oksal [79] highlighted that drought severity has increased across the Marmara region, including the Çanakkale station, further supporting the findings of this study. Furthermore, the trends calculated for the RDI and eRDI appear to be more statistically significant than those observed for the SPI and aSPI.

While the trend values in the SPI/aSPI are generally very similar across stations, more pronounced differences are observed in the RDI/eRDI. Additionally, the SPI/aSPI shows a lower increase in drought severity compared to the RDI/eRDI. In other words, the increase in drought severity is more pronounced in the RDI/eRDI than in the SPI/aSPI. This finding is consistent with Zarch et al. [17], who reported that the agreement between SPI and RDI is generally higher in arid regions and weaker in humid zones, with RDI showing stronger drying trends than SPI.

Statistically significant trends indicating an increase in drought severity in the RDI/eRDI are most prominent in Çanakkale, followed by Gökçeada and Bozcaada. In other words, the greatest increase in drought severity is expected in Çanakkale than other stations. The Çanakkale station is located within the city center, where urbanization has increased markedly over recent decades. Compared to the island stations of Bozcaada and Gökçeada, the rate of urban expansion in Çanakkale is significantly higher, and this growth has been accompanied by a reduction in agricultural land in the region [84,85]. Such changes in land use may affect local climate conditions and hydrological processes, potentially contributing to the more pronounced increase in drought severity observed at this station. The stronger drought trends identified by the RDI and eRDI can be explained by their inclusion of potential evapotranspiration (PET), unlike SPI and aSPI, which depend mainly on precipitation. At the urbanized Çanakkale station, rapid land use change and surface modification have likely intensified the urban heat island effect [86], causing higher air temperatures, reduced vegetation, and greater water loss through evapotranspiration. These factors increase local drought stress, which is better reflected by RDI and eRDI than by precipitation-based indices. In line with this, a nationwide drought assessment conducted by [82] using the SPEI identified Bozcaada as the station with the highest drought frequency in Türkiye. This external evidence further corroborates our findings, highlighting the pronounced vulnerability of Bozcaada to drought due to its Mediterranean climatic setting characterized by low precipitation and elevated temperatures.

The eRDI calculated using the FAO method shows a lower increase in drought severity compared to the RDI and eRDI values calculated using the other methods. The highest upward trends in drought severity in the region were estimated for the Çanakkale station, specifically for the 12-month time scale eRDI-USBR and the annual scale eRDI-USBR and eRDI-USDA-CROP indices, with rates of 0.018/year.

For the Bozcaada station, the lowest statistically significant upward trends were calculated for the eRDI-FAO and aSPI-USBR indices, with a rate of 0.004/year on the 12-month time scale. For the Gökçeada station, the lowest statistically significant upward trend in drought severity was calculated for the SPI on the 12-month time scale, with a rate of 0.004/year. For the Çanakkale station, the lowest statistically significant upward trends were calculated for the aSPI-USBR and aSPI-USDA-CROP indices on the 12-month time scale, and for the eRDI-FAO index on the 6-month time scale, with a rate of 0.005/year. Further, except for the indices calculated using the FAO method, no notable differences were observed in the trends of the indices calculated with effective rainfall for the same time periods.

In light of the more pronounced increase in drought severity observed at the Çanakkale station, relevant adaptation measures and policy actions—such as promoting drought- and heat-tolerant crop varieties, strengthening drought monitoring and assessment efforts, establishing drought early warning systems, and implementing agricultural drought action plans at the provincial level, as outlined in official regional and national reports [45,87,88]—should be considered for the Çanakkale region to mitigate potential impacts.

4. Conclusions

This study evaluated the performance of different effective precipitation estimation methods (USBR, USDA-Simplified, USDA-CROPWAT, and FAO) in the calculation of drought indices (aSPI and eRDI) and compared them with conventional indices (SPI and RDI) across various time scales (1-, 3-, 6-, 12-month, and annual) in a humid region of Türkiye.

The findings indicate that the USDA-Simplified method consistently yields the most reliable results (R² ≈ 0.99; lowest RMSE ≈ 0.09; NSE > 0.95) in terms of statistical accuracy across all stations and time scales, while the FAO method shows limited performance, especially at the 1-month scale. The FAO-based indices tend to underestimate drought severity and frequency due to lower effective precipitation estimates, leading to weaker drought signals—particularly evident in short-term scales. This highlights the importance of selecting physically meaningful effective precipitation methods when applying drought indices in humid environments.

Drought characteristics derived from all indices reveal that the number of drought months increases with longer time scales, which contrasts with findings from arid regions. Maximum drought severities were most frequently detected using SPI and aSPI, while the timing of maximum events generally coincided across all index types.

Trend analysis confirmed that short-term droughts (1–3 months) exhibited no statistically significant trends, whereas longer time scales (6-, 12-month, and annual) showed clear negative trends—indicating increasing drought severity—particularly for the RDI and eRDI.

These strongest trends were detected in Çanakkale, with eRDI-USBR and eRDI-USDA-CROP exhibiting slopes up to −0.018/year, highlighting a gradual intensification of drought conditions in the region.

Overall, the results highlight the importance of selecting appropriate effective precipitation methods when calculating drought indices, particularly in humid regions. The use of aSPI and eRDI in combination with physically meaningful effective precipitation methods, such as USDA-Simplified, improves the accuracy of drought detection. Additionally, trend results point to a potential increase in drought severity in the region, underscoring the need for proactive water management strategies, basin-scale analyses and further investigation under future climate scenarios. In this study, data from three closely located stations within a humid region were used. Future work should extend this analysis to basin-scale or national applications using a larger station network to further explore climatic variability and improve predictive drought modeling. Given its demonstrated accuracy, the USDA-Simplified method could be operationally integrated into drought early-warning and irrigation-scheduling systems to provide improved short- and medium-term drought forecasts for agricultural and water-management planning in humid environments.