Drought Trend Analysis Using Standardized Precipitation Evapotranspiration Index in Cold-Climate Regions

Abstract

This study aimed to conduct a drought trend analysis using the standardized precipitation evapotranspiration index (SPEI) in two mountainous and cold-climate regions in Iran and the United States (US). The Mann–Kendall test was employed to assess the trend in the Upper Colorado River Basin (UCRB) in the US and Lorestan province. The results reveal a predominantly decreasing trend in drought occurrences across Lorestan, especially in southern and southwestern areas with lower elevations. In contrast, the UCRB showed a positive trend, indicating a wet period. The western parts of the UCRB were predominantly affected by droughts. Among the stations, the Khorram Abad station exhibited the most statistically significant trend at the 99% confidence level (Z > 2.57). A temporal trend analysis of droughts revealed more positive and negative abrupt changes in the UCRB than in Lorestan. This indicates a higher degree of small-scale variability in the UCRB compared to Lorestan. This study indicates that factors such as elevation, land use changes, and proximity to water sources may contribute to the observed variations in drought trends. Additionally, the findings highlight that rising temperatures have a significantly greater impact on drought severity than reductions in precipitation. This study provides a temperature-responsive method for drought assessments, supporting the development of adaptive strategies that address snowmelt variability, seasonal water availability, and shifting drought patterns in cold regions.

1. Introduction

Droughts have intensified in frequency and severity due to climate change, rising water demands, and threatening global water security. Declining precipitation and extreme rainfall fluctuations are among the factors exacerbating the water crisis and leading to numerous droughts in recent years [1]. A drought is one of the most significant natural disasters, defined as a prolonged period of reduced precipitation [2,3,4].

Droughts are classified into various types: meteorological droughts, agricultural droughts, hydrological droughts, and socio-economic droughts [5,6]. A meteorological drought occurs first and leads to other types of droughts. A meteorological drought can result in an agricultural drought, followed by a hydrological drought with a lag [7].

One of the essential steps in drought studies is determining an appropriate index or indices for evaluating the intensity and duration of droughts. The Standardized Precipitation Index (SPI) [8] and Effective Drought Index (EDI) [9] rely solely on precipitation data. These indices are based on two assumptions. First, precipitation variations are more significant than other climatic variables, such as temperature and evapotranspiration. Second, other climatic variables, apart from precipitation, do not have temporal trends. However, many researchers emphasize the role of temperature in the occurrence of drought events [10].

Moreover, over the past 150 years, temperature has increased by approximately 0.5 to 2 °C, as shown by climate change models [11]. This increase has led to a rise in droughts and, consequently, increased water demands due to higher evapotranspiration rates [12]. To address this challenge, the standardized precipitation evapotranspiration index (SPEI) [13] was developed, which indirectly incorporates temperature by subtracting evapotranspiration from precipitation [14]. This index uses a simple water balance approach (precipitation (P)–potential evapotranspiration (PET)), allowing it to reflect moisture availability more accurately in snow-dominated regions, where temperature variations play a crucial role. SPEI has been shown to be promising for detecting long-term drought periods in the Aegean region of Turkey, but its effectiveness decreases for shorter timescales with more frequent and shorter drought events [14]. Nejadrekabi et al. (2022) [15] confirmed the effectiveness of using SPEI and the Normalized Difference Vegetation Index (NDVI) for monitoring drought conditions in Khuzestan Province, Iran.

There are drought indices which are suitable for monitoring droughts in cold-climate regions. The standardized snow water equivalent index (SWEI) was developed by considering the snow water equivalent (SWE) to assess global snow droughts [16]. The surface water supply index (SWSI) [17], snow-based hydroclimatic aggregate drought index (SHADI) [18], and aggregate drought index (ADI) [19] account for snowpack amounts in addition to other hydroclimatic variables. The standardized snowmelt and rainfall index (SMRI) [20] and hydroclimatic aggregate drought index (HADI) [21] consider snowmelt and rainfall. The trade-off between complexity and accuracy of these aggregated drought indices requires further evaluation. While the simplicity of SPEI makes it a useful tool for drought monitoring, its performance in cold-climate regions warrants further investigation. However, when Thornthwaite’s method is used to estimate PET, significant limitations arise, especially in cold climates and snow-covered regions [22]. During winter, precipitation is stored as snow rather than contributing to immediate water availability, but SPEI may misinterpret a low PET as the absence of drought. Similarly, during spring snowmelt, a large influx of water occurs despite low precipitation, but SPEI may falsely indicate a drought due to a higher calculated PET.

Analyzing drought trends in cold-climate regions can provide insights into its effectiveness in identifying droughts in such environments. The Lorestan province, located in western Iran, is highly vulnerable to droughts, which impact agricultural productivity, water availability, and wildfire occurrences. Several studies have assessed the region’s drought trends using SPEI and other drought indices. Adib et al. (2024) [23] evaluated meteorological drought impacts on rainfed wheat yields in the Karkheh watershed, southwestern Iran, using the SPI and SPEI indices from 1981 to 2016. Their study analyzed drought trends across 34 synoptic stations and assessed the correlation between drought indices and wheat yields using Pearson’s correlation coefficient. Results show that SPEI indicated milder but longer droughts compared to SPI. A spatial analysis revealed a clustered drought pattern, with milder droughts in the north and more severe droughts in the south. Given the role of temperature in meteorological droughts, their study recommends incorporating SPEI alongside SPI for improved drought predictions in the region. Beiranvand et al. (2024) [24] investigated long-term drought variability in the broader of region of Lorestan province in the central Zagros Mountains of western Iran using tree-ring chronology from 1802 to 2022. Their study found a strong positive relationship between the tree-ring width and the SPEI-48 index for March–September, with a two-year lag. Safdary et al. (2024) [25] investigated the spatiotemporal relationship between droughts and fire occurrences in Lorestan province, Iran, using drought indices (SPI and PDSI) from 2000 to 2021. Their study found a significant correlation between winter and spring droughts and increased fire frequency in spring and summer, with June and July experiencing the highest fire occurrences. Notably, fire incidents showed an increasing trend, peaking in 2019, even during wet conditions, due to abundant plant biomass. The findings suggest that droughts contribute to fire spread in the Zagros region. Lornezhad et al. (2023) [26] analyzed precipitation and drought trends in Lorestan province, Iran, using a modified Mann–Kendall method. Annual and monthly rainfall data were examined to assess historical trends, while the SPI and Mann–Kendall test were applied to evaluate drought conditions. The results indicate a significant negative trend in precipitation across all stations, with extreme drought conditions expected to persist in the future. Predictions until 2032 suggest a continued decline in rainfall, emphasizing the need for sustainable water resource management to mitigate the impacts of droughts on agriculture and water availability in the region.

The Upper Colorado River Basin (UCRB) is similarly a cold-climate region with a critical water source for the western United States, where drought conditions significantly impact water availability, wildfires, and ecosystem health. Studies incorporating SPEI have contributed to a better understanding of drought dynamics in these regions. Han et al. (2024) [27] identified SPEI as a key predictor of wildfire occurrences in the UCRB. Their analysis demonstrated that SPEI effectively captured drought-induced fire risks, reinforcing its utility in fire prediction models. Their study highlights the importance of integrating SPEI into wildfire management strategies to enhance preparedness and mitigation efforts. Tang and Piechota (2009) [28] analyzed the spatial and temporal variability in soil moisture in the UCRB using the VIC-3L hydrological model over a 50-year period. A wavelet analysis showed a strong correlation between deep soil moisture, PDSI, precipitation, and streamflow, indicating its potential as a drought indicator. A spatial analysis identified regions more vulnerable to droughts and wet conditions. A temporal analysis of four major droughts (1953–1956, 1959–1964, 1974–1977, and 1988–1992) revealed that droughts persisted until soil moisture anomalies turned positive for at least two consecutive years. McCabe et al. (2020) [29] identified winter precipitation deficits as the primary cause of droughts in the Upper Colorado River Basin (UCRB), rather than elevated temperatures. Between 1901 and 2014, eight drought periods were recorded, with the most severe one occurring from 1901 to 1904 and the longest one spanning from 1943 to 1956. Miller et al. (2011) [30] found that the timing of the last day of the snow season strongly correlates with the runoff volume during the peak flow season (April–July), whereas the start of the snow season shows little correlation. Comparisons of the standardized precipitation and evapotranspiration index (SPEI) and the Standardized Precipitation Index (SPI) reveal similar spatial and temporal trends in the UCRB, though SPEI, which incorporates temperature, exhibits more extreme drought magnitudes. These findings highlight the importance of incorporating both temperature and snow-based drought indices in UCRB drought assessments.

The reviewed studies underscore the importance of SPEI in drought analyses across both Lorestan province and the UCRB. While Adib et al. (2024) [23] and Beiranvand et al. (2024) [24] demonstrated the effectiveness of SPEI in capturing long-term drought trends and agricultural impacts in Iran, Han et al. (2024) [27] and Tang and Piechota (2009) [28] highlighted its relevance in fire risk assessments and soil moisture variability in the UCRB. Additionally, the studies in Lorestan [25,26] emphasized the increasing drought severity and its implications for wildfire management and water resource planning. Overall, the findings highlight the importance of incorporating SPEI in regional drought assessments, as it effectively accounts for temperature influences. A comparative study of these two cold-climate regions could help generalize the findings and enhance broader applicability. This study evaluates the effectiveness of the SPEI in monitoring drought conditions in cold, mountainous climates by conducting a drought trend analysis in two hydroclimatically similar yet geographically distinct regions. Given the unique hydrological dynamics of snow-dependent areas, where seasonal snow accumulation and melting play a critical role in water availability, traditional drought indices may not fully capture drought onsets and severity. By assessing the accuracy and responsiveness of SPEI against observed drought trends and historical hydroclimatic data, this study aims to determine its reliability in regions where snowmelt is a primary water source. This paper is organized as follows: Section 2 describes the study areas and methodology for the calculation of SPEI and the Mann–Kendall test. Section 3 provides the results of the SPEI, drought, and trend analyses based on the SPEI in the two regions and discusses the results. Finally, Section 4 presents the conclusions.

2. Materials and Methods

This study compares Lorestan and UCRB due to their shared characteristics of being cold, mountainous regions with seasonal precipitation variability. While Lorestan represents a mid-sized basin with a strong winter precipitation regime, UCRB spans a large area with more spatial heterogeneity but remains an important reference for snowpack-dependent hydrology.

2.1. Study Area

Lorestan in Iran and UCRB in the United States are mainly characterized as mountainous, snow-dominated, and cold-climate regions. Figure 1 illustrates the geographical location of UCRB within the United States and Lorestan province in Iran.

Lorestan province, located in western Iran, spans an area of 28,064 km² and is largely covered by the Zagros Mountains. The UCRB, covering over 279,720 km² across seven states, provides water to 26 million people within and beyond the basin. Elevation in Lorestan ranges from 500 m in the southernmost areas to 4050 m at its highest point. Similarly, elevation in the UCRB varies from 900 m in the southwest to 4300 m in the northeast.

Lorestan has a long-term average annual precipitation of 498.6 mm, with the highest monthly rainfall of 86.4 mm occurring in April. Of its total precipitation, 43% falls in winter, 29.5% in autumn, and 27% in spring. In contrast, approximately 64% of the UCRB is classified as arid or semi-arid, receiving an average annual precipitation of 370 mm and maintaining a mean annual temperature of 6 °C [30]. Temperature plays a vital role in shaping droughts and wet conditions in the Colorado River Basin [31]. Despite the basin’s vital significance, its limited water resources have encountered significant challenges due to climate change [32] and human activities [33,34]. Various climate change projections indicate reductions in late-summer precipitation and mean annual streamflow [35], leading to more severe droughts [36] and an increased risk of wildfires in the UCRB [33].

The long-term average maximum and minimum temperatures in Lorestan are 22.6 °C and 7.7 °C, respectively, with Poldokhtar recording the highest average temperature (22.8 °C) and Nourabad the lowest (11.9 °C). August is the hottest month, while January is the coldest. February experiences the highest relative humidity (66.2%), making it the wettest month, whereas August is the driest (25.1%). The land cover in the UCRB consists mainly of rangeland (65%) and evergreen forests (25%).

For this study, data from 26 stations in UCRB [37] (McAfee et al., 2019) were analyzed, with details presented in

Table 1. Geographical characteristics of stations in the UCRB and the climate divisions (CDs).

CD Name	CD ID	Station Name	State	Elevation (m)	Longitude	Latitude
Green and Bear Drainage	4803	Green River	WY	1852.3	−109.4767	41.5314
		Pinedale	WY	2193	−109.8642	42.8797
		Rock Springs AP	WY	2055	−109.0653	41.5942
Colorado Drainage Basin	502	Collbran	CO	1822.7	−107.9631	39.2425
		Dillon 1E	CO	2763	−106.0353	39.6261
		Fruita	CO	1378.9	−108.7331	39.1653
		Gunnison 3SW	CO	2323.8	−106.9675	38.5258
		Montrose #2	CO	1764.5	−107.8792	38.4858
		Steamboat Springs	CO	2094	−106.8233	40.4883
		Telluride 4WNW	CO	2643.2	−107.8733	37.9492
Uinta Basin	4206	Duchesne	UT	1682.5	−110.395	40.1678
		Ft. Duchesne	UT	1539.2	−109.8611	40.2842
		Vernal 2SW	UT	1668.5	−109.5531	40.4269
Northern Mountains	4205	Scofield-Skyline Mine	UT	2655.4	−111.2047	39.6847
Southeast	4207	Green River Aviation	UT	1240.5	−110.1544	38.9906
		Thompson	UT	1554.2	−109.7167	38.9667
		Moab	UT	1242.7	−109.5458	38.5744
		Hanksville	UT	1313.1	−110.7153	38.3706
		Blanding	UT	1854.7	−109.4847	37.6131
		Bluff	UT	1318	−109.5578	37.2828
South Central	4204	Escalante	UT	1770.9	−111.5978	37.7686
		Salina 24 E	UT	2304.3	−111.4161	38.9139
Northwestern Plateau	2901	Aztec Ruins NM	NM	1720.3	−108.0006	36.835
		Dulce	NM	2070.5	−107	36.9358
Northeast	202	Canyon de Chelly	AZ	1709.9	−109.5394	36.1533
		Lees Ferry	AZ	978.4	−111.6022	36.8644

.

Table 2. Geographical and climatic characteristics of synoptic stations in Lorestan province.

Climate Types (Emberger)	Station Name	Mean Precipitation (mm)	Tmean	Tmax	Tmin	Elevation (m)	Longitude	Latitude
Cold Semi-Dry	Aleshtar	444.7	13	19.32	6	1567	48°15′	33°49′
	Boroujerd	456.4	16.75	22.42	10.03	1629	48°45′	33°35′
	Azna	411.6	13.73	22.17	3.30	1872	49°25′	33°27′
	Poldokhtar	360.1	23.3	29.07	16.73	713	47°43′	33°09′
	Koohdasht	366.7	15.04	21.73	8.61	1198	47°39′	33°31′
Cold Dry	Aligoudarz	387.3	16.64	24.87	7.18	2022	49°42′	33°24′
	Doroud	631.7	14.13	20.37	4.72	1527	49°14′	33°29′
Cold Wet	Khorram Abad	500.1	12.93	19.35	4.77	1148	45°17′	33°26′
Hot Dry	Nourabad	467.9	17.32	25.32	9.09	1860	48°19′	34°03′

provides detailed information about the latitude, longitude, elevation, long-term climatic data, and climatic classification (based on the Emberger method) for nine synoptic stations in Lorestan, whose geographical distribution is shown in Figure 1.

2.2. Data and SPEI Calculation

Figure 2 shows the methodology utilized in this study. For a drought analysis using SPEI, precipitation and temperature data (mean, minimum, and maximum) were obtained from nine stations in Lorestan and 26 stations in UCRB. The data period for UCRB covered 65 years (1950–2014). For Lorestan, the data periods varied by station, ranging from 2007–2022 (Azna station) to 1958–2022 (Khorram Abad station). To compare drought conditions between the two regions, the overlapping period of 2007–2014 was considered. To calculate SPEI in UCRB, we utilized input data sourced from the Global Land Data Assimilation System (GLDAS) [38]. GLDAS integrates satellite and ground-based observations with advanced land surface modeling and data assimilation techniques to produce high-quality land surface state and flux estimates [38]. These datasets are available globally at spatial resolutions ranging from 2.5 degrees to 1 km, providing near-real-time data from 1948 to the present. However, for this study, the dataset was limited to 2014. The required data for calculating SPEI in Lorestan was obtained from the Meteorological Organization and the Ministry of Energy of Iran [39]. No missing data and outliers were detected, since reconstructed and reanalyzed monthly data were used.

The SPEI was calculated at a monthly timescale using a simple water balance equation that measures the difference between precipitation (P) and potential evapotranspiration (PET). The Thornthwaite approach was used for estimating potential evapotranspiration, as it requires only air temperature and the station’s latitude [13,40]:

$$D_i=P_i-{PET}_i$$

(1)

The values of D at different timescales were calculated using Equation (2) [13]:

$$D_n^k={\sum }_{n=0}^{k-1}{P}_{n-1}-{PET}_{n-i}$$

(2)

where k is the selected timescale in months, and n is the number of months. A three-parameter distribution function was applied to account for negative values of D. Among the tested distributions, the log-logistic distribution showed the best fit for time-series data.

The cumulative probability function for the D data series was calculated as follows:

$$F(x)={\left[1+\left({\displaystyle \frac{\alpha }{x-\gamma }}\right)\right]}^{-1}$$

(3)

where α, β, and γ are the scale, shape, and location parameters, respectively, for the values of D. After calculating the cumulative distribution function and converting it to normalized values, the SPEI was extracted. The 1-month SPEI was used due to its high sensitivity to temperature spikes and evaporation rates. A drought begins when the SPEI falls below −1 and ends when it becomes positive. The classification of SPEI values is shown in

Table 3. Classification scale of SPEI values [8] (McKee et al., 1993).

Class	SPEI Values
Extreme drought	<−2
Severe drought	−1.99 to −1.49
Mild drought	−1 to −1.49
Normal	−0.99 to 0.99
Mild wet	1 to 1.49
Severe wet	1.5 to 1.99
Extreme wet	>2

.

2.3. Mann–Kendall Test

Originally introduced by Mann (1945) [41] and further refined by Kendall (1975) [42], this test is particularly effective for time-series analysis. In this test, the null hypothesis states that there is no trend in the data, meaning that the observations are independent and randomly distributed over time, while the alternative hypothesis (H1) suggests a significant increase or decrease in the trend. In this method, the S statistic for the g-th month and the k-th station was calculated as follows:

$${\mathrm{S}}_{\mathrm{gk}}={\sum }_i^{n-1}{\sum }_{j=i+1}^{n-1}sgn ({\mathrm{X}}_{\mathrm{jgk}}-{\mathrm{X}}_{\mathrm{igk}}), \forall \le \mathrm{i}<\mathrm{j}\le \mathrm{n}$$

(4)

where n is the number of series data, sgnθ is a function of the sign, and θ is the difference between the two observations in each of the parameters in different years I and j, which are as follows:

$$\mathrm{S}\mathrm{g}\mathrm{n}(\theta )=\left\{\begin{array}{c}1 if \theta >0\\ 0 if \theta =0\\ -1 if \theta <0 \end{array}\right.$$

(5)

When n ≥ 10, the S statistic is distributed almost normally and has a mean of 0 and the following standard deviation:

$${\left({\sigma }_{\mathrm{g}\mathrm{g}}\right)}_{\mathrm{k}}={\displaystyle \frac{\left[n\right(n-1\left)\right(2n+5)-\sum d(d-1\left)\right(2d+5\left)\right]}{18}}$$

(6)

where d is the same number of data in the time series. In this method, Sgk was normalized as follows:

S′_gk = S_gk − sgn(S_gk)

(7)

Then, the standardized test statistic, or Z, which has a standard normal distribution with a mean of 0 and a variance of 1, was obtained as follows:

$${\mathrm{Z}}_{\mathrm{g}\mathrm{k}}={\displaystyle \frac{{S^{\prime }}_{gk}}{\left({\sigma }_{gg}\right)}}1/2$$

(8)

A positive Z-value (Z > 0) indicates a positive trend (an increasing trend in the time series), and a negative Z-value (Z < 0) indicates a negative trend (a decreasing trend in the time series). If the value of Z is greater than ±1.96, the data has a trend, and the null hypothesis is rejected; otherwise, it has no trend. Z is a standard normal distribution statistic and is used in a two-domain test depending on the confidence levels of the item. The test can take different values, and S is a parameter of the Mann–Kendall method, which is calculated. The value of the Z statistic for 95% and 99% confidence levels are considered to be 1.96 and 2.58, respectively. If the p-value of the Z statistic is greater than the significance level (e.g., 0.05), then H0 cannot be rejected, meaning there is no significant trend. Conversely, if the p-value is less than the significance level, H0 is rejected, indicating that there is a statistically significant trend in the direction indicated by Z.

3. Results and Discussion

Drought conditions in Lorestan and UCRB were evaluated and compared using the SPEI index. The temporal variations in this index were analyzed and compared at a monthly timescale for each station.

3.1. Drought Trends in Lorestan

The time-series charts in Figure 3 illustrate the monthly SPEI variations for nine stations in Lorestan. Based on these charts, drought variability over time is significant. However, drought conditions remained within the “normal” range at five stations: Aleshtar, Aligoudarz, Azna, Boroujerd, and Doroud. In contrast, drought conditions deviated from the normal range at four stations: Khorramabad, Koohdasht, Nourabad, and Poldokhtar. For example, droughts were recorded in Koohdasht in 2022; Khorram Abad in 1970, 1980, and 2022; and Poldokhtar in 2005 and 2022. Conversely, wet periods were observed in Khorramabad in 2000 and Nourabad in 2005. The results indicate that the southern and southwestern regions of Lorestan, which are predominantly lowland areas, have experienced more severe droughts.

3.2. Drought Trends in UCRB

The time-series charts in Figure 4 and Figure 5 show the monthly SPEI variations for 26 stations in UCRB. Across all 26 stations during the study period, despite fluctuations in the drought index over various years, drought conditions generally remained within the “normal” range.

Examining the overlapping period between the two regions (highlighted by a dashed line in the charts) revealed that in Lorestan, the SPEI index showed decreasing trends at the Azna, Khorram Abad, and Nourabad stations, while increasing trends were observed in other areas, indicating a tendency toward wetter conditions. Our findings align with those of Lornezhad et al. (2023) [26], who analyzed drought trends in Lorestan province using the SPI and Mann–Kendall test. Their results reveal a significant downward trend in precipitation across all stations, with extreme drought conditions expected to persist. Forecasts through 2032 predict a continued decrease in rainfall. In contrast, UCRB exhibited a consistent pattern across all 26 stations.

3.3. A Temporal Trend Analysis Using the Mann–Kendall Test

The Mann–Kendall test was applied to assess temporal trends in the SPEI-based drought time series for both Lorestan and the UCRB. The test results are summarized in

Table 4. The Mann–Kendall statistics values for Lorestan and UCRB.

Lorestan	Z	Poldokhtar	Nour Abad	Koohdasht	Khorram Abad	Doroud	Boroujerd	Azna
		−0.39	−0.69	−0.43	3.66	−0.44	−0.54	−0.33
		Aligoudarz	Aleshtar
		0.43	0.48
UCRB	Z	Telluride 4WNW	Steamboat Springs	Montrose #2	Gunnison 3SW	Fruita	Dillon 1E	Collbran
		0.47	−1.13	−0.44	−0.7	−0.3	2.08	0.73
		Lees Ferry	Canyon de Chelly	Hanksville	Green River Aviation	Ft. Duchesne	Escalante	Duchesne
		1.13	−0.3	0.04	0.73	0.53	0.24	1.13
		Bluff	Blanding	Dulce	Aztec Ruins NM	Rock Springs AP	Pinedale	Green River
		0.07	−0.21	0.6	0.4	1.82	2.05	2.05
		Vernal 2SW	Thompson	Scofield-Skyline Mine	Salina 24 E	Moab
		0.4	1.07	0.6	0.73	0.44

, while Figure 6a,b illustrate the spatial distribution of Mann–Kendall Z-statistics and their normalized values across the UCRB and Lorestan, respectively.

In Lorestan, the Khorramabad station exhibited a statistically significant increasing trend in drought conditions. The computed Z-value of 3.66 exceeds the critical threshold of 2.57, indicating a highly significant trend at the 99% confidence level. This finding was further supported by a confidence band analysis, which revealed that the observed Z-value lies well above the upper limit (±1.33), as shown in Figure 7. The strong upward trend at Khorramabad may be attributed to its location within a cold wet climate zone and its status as the station with the highest annual mean precipitation in the province.

Other stations in Lorestan, including Azna, Boroujerd, Doroud, Koohdasht, Nourabad, and Poldokhtar, exhibited decreasing trends in the SPEI. However, these trends were not statistically significant, as their Z-values fell within the ±1.96 range. For example, Nourabad, which has the lowest annual mean temperature (11.9 °C) among the Lorestan stations and moderate precipitation (467.9 mm), recorded the lowest Z-value of −0.69. Its classification as a hot dry climate may partially explain this result. These variations suggest that localized factors—such as elevation, land use change, and proximity to water sources—play an important role in influencing drought trends across the region. Similar conclusions were drawn by Nejadrekabi et al. (2022) [15], who found temperature increases to be more influential than precipitation reductions in driving drought severity.

In the UCRB, trend patterns varied across the 26 stations. Stations such as Canyon de Chelly, Fruita, Gunnison 3SW, Montrose #2, Blanding, and Steamboat Springs showed decreasing trends in SPEI values, although these were not statistically significant. Conversely, stations including Dillon 1E, Green River, and Pinedale exhibited statistically significant increasing trends, with Z-values exceeding the 1.96 threshold at the 95% confidence level. These stations are primarily located in the northern and central parts of the UCRB, as represented by yellow zones in the spatial distribution map (Figure 6a).

Overall, the Mann–Kendall test results reveal spatial heterogeneity in drought trends within both regions. The statistically significant upward trend at Khorramabad and other localized differences underscore the need for region-specific drought monitoring and interpretation, particularly in topographically complex and climatically diverse areas.

3.4. Analysis of SPEI Trend Charts

The trend charts for the SPEI in both regions are shown in Figure 5 and Figure 6, which align with the Z-statistic results. In Lorestan (Figure 8), a sharp rise in drought trends occurred in Aleshtar (2006, 2010, and 2016); Doroud (2014); Khorram Abad (1991); Koohdasht (2009, 2013, and 2018); and Nourabad (2012, 2014, and 2016). A sharp drop in trends of SPEI were observed in Aleshtar (2007), Khorramabad (2015), Koohdasht (2018), and Nourabad (2013 and 2020). These abrupt changes significantly influenced the overall drought trends during the study period.

In the UCRB (Figure 9 and Figure 10), a sharp drop occurred in Steamboat Springs (2000), which corresponded to a period of severe droughts across the western United States, attributed to La Nina-induced precipitation deficits, Ft. Duchesne (2008 and 2012), and Salina 24 E (2011). Sudden spikes were observed in Green River (1954), Pinedale (1955), and Rock Springs AP (1954). The wet anomalies in Green River and Pinedale align with documented El Nino events that brought above-average precipitation to the region. These significant fluctuations highlight the role of precipitation and temperature changes in drought variability across the UCRB.

3.5. Comparison of Lorestan and UCRB

The results of the drought analysis using the SPEI revealed distinct patterns between Lorestan and the Upper Colorado River Basin (UCRB), despite some similarities in their geographic and climatic characteristics. The UCRB is much larger and spans multiple climatic zones, whereas Lorestan covers a more localized region. One key difference is UCRB’s strong reliance on snowpack, while Lorestan has a mix of rain-fed and snow-dependent regions. Additionally, the UCRB contains extensive semi-arid and arid areas, whereas Lorestan generally experiences a wetter climate. In terms of drought trends, Lorestan exhibited a generally declining trend, indicating an increase in drought frequency, particularly in southern and southwestern areas with lower elevations. Conversely, the UCRB showed increasing trends, suggesting wetter periods in most regions, although drought-affected areas were mainly concentrated in the western part of the basin. Moreover, the temporal variability in droughts was higher in the UCRB, with more pronounced positive and negative jumps observed throughout the study period. This suggests that the UCRB experiences greater small-scale variability in drought conditions compared to Lorestan, likely due to its larger spatial extent and diverse climatic influences. This suggests greater small-scale variability in the UCRB compared to Lorestan.

3.6. Seasonal and Annual Drought Analyses in the Upper Colorado River Basin and Lorestan Province

This section examines the temporal variability in droughts in the UCRB and Lorestan, using the SPEI. By identifying the driest and wettest years and seasons at various stations across both regions, we provide insights into historical drought patterns and their seasonal distribution.

Table 5. The results of seasonal and annual drought analyses for Lorestan.

Station	Year		Season
	Driest	Wettest	Driest	Wettest
Aligoudarz	1997	2016	Autumn 2002	Summer 1994
Aleshtar	2022	2018	Spring 2015	Winter 2011
Azna	2017	2013	Summer 2007	Autumn 2016
Boroujerd	1996	2000	Spring 2015	Autumn 2010
Doroud	2017	2011	Summer 2007	Autumn 2013
Khorram abad	2022	1989	Summer 1973	Autumn 2001
Koohdasht	2022	2009	Summer 2008	Winter 2013
Nourabad	2020	2013	Autumn 2017	Winter 2013
Poldokhtar	2005	2013	Autumn 2017	Spring 2020

and

Table 6. The results of seasonal and annual drought analyses for UCRB.

Station	Year		Season
	Driest	Wettest	Driest	Wettest
Canyon de Chelly	1958	1965	Spring 1980	Autumn 1972
Lees Ferry	1958	1957	Spring 1980	Autumn 1972
Collbran	1958	1957	Summer 1985	Autumn 2013
Dillon 1E	1958	1957	Spring 1995	Autumn 1957
Fruita	1958	1957	Spring 1980	Autumn 1972
Gunnison 3SW	1958	1957	Summer 1980	Autumn 2007
Montrose #2	1974	1957	Spring 1985	Autumn 2007
Steamboat Springs	1983	1984	Spring1993	Autumn 1951
Telluride 4WNW	1974	1957	spring 1985	Winter 1972
Aztec Ruins NM	1958	1965	Spring 1980	Autumn 1972
Dulce	1958	1957	Spring 1980	Autumn 1972
Blanding	1958	1957	Spring 1980	Autumn 1972
Bluff	1958	1957	Spring 1972	Autumn 1980
Duchesne	1958	1957	Spring 1980	Autumn 1951
Escalante	1958	1957	Spring 1980	Autumn 1978
Ft. Duchesne	1974	1957	Spring 1980	Winter 1978
Green River Aviation	1984	1983	Spring 2005	Autumn 1981
Hanksville	1974	1957	Spring 2005	Autumn 1972
Moab	1958	1965	Summer 1972	Winter 1980
Salina 24 E	1958	1957	Summer 1980	Autumn 1978
Scofield-Skyline Mine	1958	1957	Summer 1980	WINTER 1951
Thompson	1958	1957	Summer 1993	Autumn 1972
Vernal 2SW	1958	1957	Summer 2005	Winter 1981
Green River	1958	1957	Summer 1971	Autumn 1994
Pinedale	1974	1957	Summer 1998	Autumn 1980
Rock Springs AP	1984	1983	Summer 1971	Autumn 1994

summarize the seasonal and annual drought conditions observed in Lorestan province and the UCRB, respectively.

In Lorestan province, the most severe droughts have occurred in recent decades, with 2022 marking the driest year for several stations, including Aleshtar, Khorramabad, and Koohdasht. A seasonal analysis indicated that droughts in Lorestan are more frequent during summer and autumn. Notably, significant dry spells were observed in summer 2007 and autumn 2017. In contrast, the wettest years varied by station, with Khorramabad recording a peak in 1989 and Aleshtar in 2018. Wet conditions were generally concentrated in the winter and autumn seasons.

For the UCRB, the most severe droughts occurred in the mid-20th century, with 1958 standing out as the driest year across numerous stations. The wettest years were typically 1957 and 1965. Seasonal drought variability was evident, with spring and summer being the driest in most areas, whereas autumn consistently emerged as the wettest season. Some stations, such as Green River Aviation and Rock Springs AP, also experienced prolonged drought periods in the 1980s, indicating episodic multi-year dry conditions.

These patterns confirm the findings from the Mann–Kendall trend analysis and highlight contrasting temporal drought dynamics between the two regions. While Lorestan has faced more frequent and intense droughts in recent years, the UCRB’s most severe events were concentrated in earlier decades. This temporal divergence reflects underlying regional differences in drought-driving mechanisms.

In the UCRB, winter precipitation is a critical determinant of drought conditions. McCabe et al. (2020) [29] emphasized that winter deficits reduce snowpack, leading to diminished spring runoff and subsequent spring and summer droughts. This relationship is evident in our analysis, which shows a concentration of drought events in these seasons. The extreme drought of 1958 aligns with a prolonged dry period from 1943 to 1956, reinforcing the historical importance of snowpack variability in the UCRB. Additionally, Miller et al. (2011) [30] found that the timing of snowmelt—especially the last day of the snow season—is a strong predictor of peak runoff, further supporting the role of winter hydrology in shaping drought patterns. Conversely, Lorestan’s droughts appear more influenced by rising temperatures and changing precipitation seasonality. The increased frequency of summer and autumn droughts, particularly in recent years, suggests a shift in the timing and distribution of rainfall, possibly linked to regional climate change. Unlike the UCRB, Lorestan’s droughts are less dependent on snowpack and more sensitive to thermal and rainfall dynamics. This study contributes to addressing a key knowledge gap by evaluating how drought patterns and trends diverge across two cold, mountainous regions with differing hydrological regimes. While previous studies have applied SPEI locally, few have directly compared its responsiveness in snowpack-dependent versus temperature-sensitive climates. Our findings demonstrate that SPEI captures divergent temporal drought signals in these regions, highlighting the need for regional calibrations or complementary indices in transregional drought monitoring. This contributes to improving drought assessment frameworks in a changing climate, particularly in snow-fed and semi-arid basins where climate sensitivities differ markedly.

It is also important to note several methodological limitations of this study. The SPEI was calculated using the Thornthwaite method for estimating potential evapotranspiration (PET), which is based solely on temperature and may introduce biases in cold or snow-dominated regions where factors such as wind speed and solar radiation also affect PET [43,44]. Additionally, this study used a 1-month SPEI timescale, which is effective for short-term drought monitoring but may not fully reflect long-term hydrological drought trends. Data-related limitations also impact the analysis. This study relied on station-based precipitation and temperature records, which may not capture spatial variability in large and topographically complex basins, such as the UCRB. Differences in data coverage between the regions—Lorestan (1958–2022) and the UCRB (1950–2014)—further complicate direct trend comparisons. Standardization techniques, such as the automated normalization of large-scale datasets [45], could improve comparability in future analyses. Another limitation is that SPEI does not incorporate snowpack dynamics, a crucial hydrological component in cold, mountainous regions. Drought indices that include snow water equivalents, such as the standardized snow water equivalent index (SWEI) or the Snow-based Hydroclimate Aggregate Drought Index (SHADI), may provide more accurate assessments in such regions. Furthermore, uncertainty was not explicitly addressed in this study. Future work should incorporate uncertainty quantification techniques, including geostatistical interpolation, quantile regression, or hybrid modeling approaches [46], to enhance the reliability of drought assessments.

Wind speed and direction significantly influence air temperature, especially in cold-climate regions [43,44]. This study utilized the Thornthwaite method for PET estimation and SPEI computation, which may introduce biases in cold climates where PET is influenced by factors beyond temperature, such as wind speed and solar radiation. SPEI was calculated at a 1-month timescale, which is useful for short-term drought monitoring but may not fully capture long-term hydrological drought trends. When it comes to data, this study relied on station-based precipitation and temperature data, which may not fully represent spatial heterogeneity in large basins, such as the UCRB. While efforts were made to standardize datasets, differences in data availability between Lorestan (1958–2022) and the UCRB (1950–2014) influence trend comparability. To remove this challenge, other standardization methods can be utilized. For example, using automated normalizing large-scale data [45] is suggested. Besides methodological and data constraints, SPEI does not account for snowpack accumulation and melting, which are crucial hydrological components in cold, mountainous regions. Other drought indices incorporating snow water equivalents would provide additional insights. The uncertainty in this study was not evaluated. Various uncertainty approaches, including intelligent quantile regression, traditional geostatistical interpolation algorithms, and the hybrid approach [46], can be utilized for visualizing and post-analyzing the generated maps for drought analyses.

4. Conclusions

This study conducted a comparative drought trend analysis using the standardized precipitation evapotranspiration index (SPEI) in two cold and mountainous regions: Lorestan province in Iran and the UCRB in the United States. While both regions share key hydrological features, such as elevation-driven snow accumulation and strong seasonal precipitation, their drought dynamics differ significantly. Lorestan exhibited more frequent and intense drought events in recent decades, particularly in the southern and southwestern lowland areas, whereas the UCRB displayed a generally wetter trend, with 20 out of 26 stations showing positive, albeit mostly non-significant, SPEI trends.

Importantly, the shared study period of 2007–2014, though limited in duration, highlighted distinct volatility patterns. Lorestan showed sharper temporal fluctuations in drought severity which align with recent warming trends and the shifting seasonal rainfall distribution across the Zagros region. In contrast, the UCRB’s more stable drought behavior during this window is consistent with its reliance on long-term snowpack and a larger hydroclimatic buffering capacity. This analysis reinforces the need to contextualize SPEI-based assessments with regional climatic drivers and landscape characteristics, such as land use, elevation gradients, and proximity to snow-dominated zones.

However, this study also underscores the limitations relying on SPEI, especially with the Thornthwaite method for PET estimations, where the effects from solar radiation, wind speed, and snowmelt dynamics are not accounted. Moreover, SPEI’s insensitivity to snow accumulation and release introduces uncertainties in mountainous regions where snowpack is the dominant water storage mechanism. These challenges are compounded by the differing data coverage and quality across both Lorestan and the UCRB.

To address these gaps, future studies should consider the following:

• Employ more physics-based PET estimation methods, such as Penman–Monteith.
• Analyze multi-timescale SPEI metrics (3, 6, and 12 months) to better distinguish between meteorological, agricultural, and hydrological drought conditions.
• Integrate snow-based indices, such as SWEI and SHADI, to improve drought detection in snow-dominated basins.
• Assess the influence of large-scale climate oscillations, such as El Niño–Southern Oscillation (ENSO), North Atlantic Oscillation (NAO), and Pacific Decadal Oscillation (PDO), to disintegrate local vs. global drought drivers.
• Incorporate remote sensing and machine learning approaches for high-resolution spatial mapping, especially in data-scarce regions.
• Evaluate the uncertainty of droughts using hybrid statistical and geospatial modeling frameworks.

Finally, expanding the analysis to include additional hydrological variables, such as soil moisture, streamflow, groundwater levels, and snow water equivalents, will provide a more robust, multi-dimensional understanding of drought behavior in cold regions. This will strengthen the predictive capability of drought assessments and support the development of adaptive water management strategies under climate variability and change.