Analysis of the Spatiotemporal Trends of Standardized Drought Indices in Sicily Using ERA5-Land Reanalysis Data (1950–2023)

Abstract

In this study, a spatiotemporal analysis of drought occurrence and trends across Sicily using ERA50-Land continuous gridded data is carried out. We first use the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI) to evaluate drought conditions at various time scales from 1950 to 2023. Then, the Modified Mann–Kendall test was employed to detect trends and Sen’s slope estimator was used to quantify their magnitude. An analysis of the historical series confirms that 2002 was the most severe drought year, impacting all time scales from short-term to long-term. The spatial analysis revealed that the western regions of Sicily experienced the highest severity and frequency of drought events. In contrast, the northeastern regions were less severely affected compared with the other parts of the island. The analysis detects significant increasing trends in SPI values in the eastern coastal areas of the island, which are related to a possible historical increase in precipitation. On the other hand, the analysis of the SPEI indicates significant decreasing trends in the western part of the island, which are mainly related to increased evapotranspiration rates. These results are partially consistent with previous analyses of future climate change scenarios, where changes in the SPEI values in the island are projected to be way clearer than changes in SPI values.

1. Introduction

Drought is one of the most significant hydroclimatic risks for water resources, agricultural practices, and food security [1,2]. The frequency, duration, and intensification of droughts are also rising in relation to climate change [3,4,5]. These concurrent and intensive droughts result in losses in crop productivity [6,7], water and food insecurity, and socioeconomic impacts [8,9]. Drought risks not only vary in duration, frequency, and severity, but also in their trends and spatial distribution worldwide. Climate model results show that in both the 20th and 21st centuries, the most drought-prone areas have been in, and will continue to increase across most of Africa, southern Europe, the Middle East, the Americas, Australia, and Southeast Asia [10].

Drought spatiotemporal trend analysis is essential to understand drought severity and frequency to implement appropriate mitigation and adaptation measures. Globally, in relation to climate change, drought intensity is projected to increase, and the impacts are anticipated to worsen significantly [11].

Analyzing the spatiotemporal drought trends is crucial for early warning systems and climate change adaptation, as it ensures drought assessments dynamically account for potential non-stationarities over time, enhancing mitigation strategies [11]. The Mediterranean region is one of the most prone to climate change extremes, such as drought. Future climate change scenarios also project that the Mediterranean region will experience an increased frequency, duration, and severity of droughts [12,13]. The above-mentioned studies showed that future drought risks will severely impact rain-fed agriculture and the hydrology of the region, posing devastating socioeconomic challenges in the region [14].

Using the Standardized Precipitation Index (SPI), a spatiotemporal agricultural drought trend analysis was conducted for the Calabria region of southern Italy from 1925 to 2007 [15]. The results showed an increasing drought magnitude and a negative SPI trend throughout nearly the entire region. The only exception was during the summer (SPI2-Sep and SPI3-Sep), when 50% or more of the stations showed positive trends. In contrast, the precipitation, SPI, and Standardized Deficit Index (SDI) in central Italy showed an increasing drought trend and a decreasing precipitation trend, specifically during the wet season [16]. Similarly, using the Precipitation Concentration Index (PCI), the result showed that the main amount of rainfall is declining during the wet season in southern Italy, specifically on the Tyrrhenian side of the peninsula from 1917 to 2007 based on the analysis of time series from 559 stations. Conversely, there was a widespread increase in rainfall during the summer season across the entire investigated territory [17]. The Cheliff Basin in northwestern Algeria, part of the Mediterranean region, also exhibited a decrease in rainfall and experienced a very high drought intensity and duration from 1974 to 1980 [18]. Another Mediterranean region, the Oued Sebaou Basin in northern central Algeria, has also shown an increase in drought since 1980. This study applied the PCI and the Modified Fourier Index (MFI) for 23 stations using SPI drought indices. Over 50% of the stations exhibited moderate to severe dry events over the time span from 1986 to 2001 [19].

In Sicily, droughts were also studied using different drought indices such as the SPI [20,21,22,23], the SPEI [23,24], and the Palmer Hydrological Drought Index [25]. Ref. [20] analyzed the SPI-based spatial extent of a drought using observational data, while Ref. [24] applied the SPEI using ERA5-Land reanalysis data in Sicily to estimate the spatial extent of drought severity and frequency. Additionally, Ref. [23] analyzed droughts using both the SPI and the SPEI, considering future climate change scenarios. However, there is still no clear comprehension of drought trends in the Sicily region. This knowledge would be invaluable for water resource management and climate mitigation strategies in the region.

Due to the continuity and higher spatial and temporal resolution capability of drought monitoring, ERA5-Land reanalysis has higher compatibility. This enables the easy detection of drought-prone areas with continuous spatial and temporal coverage [26,27,28]. Thus, in Sicily, there is no satellite or ERA5-Land reanalysis-based analysis for drought detection and trends over the island. Considering these gaps, this study aimed to analyze the spatiotemporal trends in droughts using the ERA5-Land reanalysis datasets across Sicily. This study addressed the seasonal magnitude and identified the areas that are most prone to drought in the region. Recently, Sicily has also experienced intensified droughts, which have significantly impacted agriculture and irrigation in the region. Additionally, as projected by the IPCC, the region will continue to address such issues going forward. Given the historical experience of droughts, it is essential to analyze and understand the intensity and frequency of droughts in different spatiotemporal contexts to assess their magnitude over time. These analyses are imperative for effective water resource management and for developing climate mitigation and adaptation measures in the region. Additionally, we examined the severity, frequency, and duration of droughts, providing invaluable insights for the most critical areas in terms of agricultural water resource management and climate mitigation. The findings are particularly crucial for irrigation and agricultural practices, which are essential for Sicily.

2. Description of the Study Area and Data

The study area delineated all over the island of Sicily, in southern Italy (Figure 1). It ranks among the vastest islands in the Mediterranean Sea, with an area of over 25,000 km². Sicily experiences a hot summer Mediterranean climate and is classified as semi-arid, with an average annual precipitation of around 700 mm, although there is significant variability from year to year. [29,30]. At a yearly scale, rainfall in Sicily spans from 400 to 1400 mm, and the evaporation is typically between 900 and 1000 mm [31,32].

The study used the ERA5-Land reanalysis monthly precipitation and potential evaporation (PET) data from the Copernicus ECMWF ERA5-Land program, the Copernicus Climate Change Service, which is based in Reading, United Kingdom [33]. ERA5-Land hourly data were obtained from 1950 to the present, via the Copernicus Climate Change Service (C3S) Climate Data Store (CDS) (Accessed on 10 April 2024). These data were provided in a gridded format, with a cell size equal to 0.1° by 0.1° degrees, and cover the time period from 1 January 1950 to 31 December 2023.

3. Methods

3.1. Standardized Precipitation Index (SPI)

The Standardized Precipitation Index (SPI) is used to estimate precipitation deficits or surpluses over specific periods, helping to identify droughts and wet periods at various time scales [34]. The computation of the SPI involves several steps. First, a Gamma distribution is fitted to the long-term precipitation data [21,35]. The Gamma distribution is defined by its shape (α) and scale (β) parameters, which are estimated using the maximum likelihood method, though the method of L-moments is generally a suitable choice as well [36]. The probability density function of the Gamma distribution is:

$$g\left(x\right)={\displaystyle \frac{1}{{\beta }^{\alpha }\Gamma \left(\alpha \right)}} x^{\alpha -1}{e}^{-x/\beta }$$

(1)

where x > 0, α > 0, β > 0, and Γ(α) is the Gamma function. The cumulative distribution function is then transformed into the standard normal distribution, plugging in the probability of the observed values, computed by the Gamma distribution, into the inverse CDF. To this purpose, numerical approximations of the inverse CDF can be adopted [37].

The suitability of the Gamma distribution for monthly precipitation data has been tested at the 5% significance level using the Lilliefors test [38,39] with corrected critical values from [40]. The null hypothesis of data coming from a Gamma distribution has been rejected only for a few cells (max 6.5% for June). The computation of the SPI has been carried out in R version 4.2 using the SPEI software v. 1.8.1 (https://cran.r-project.org/web/packages/SPEI/index.html) (Accessed on 22 April 2024).

3.2. Standardized Precipitation Evapotranspiration Index (SPEI)

The Standardized Precipitation Evapotranspiration Index (SPEI) incorporates potential evapotranspiration (PET) to provide a more comprehensive measure of drought conditions [41,42]. The SPEI is calculated by first determining the difference between precipitation (P) and PET [43,44]. This difference is then fitted to a log-logistic distribution, whose probability density function is

$$f\left(x\right)={\displaystyle \frac{\beta }{\mathsf{\alpha }}} {\displaystyle \frac{{(x/\mathsf{\alpha })}^{\mathsf{\beta }-1}}{[1+{\left(x/\mathsf{\alpha }{)}^{\mathsf{\beta }}\right]}^2}}$$

(2)

where $x$ is the difference (P − PET), α is a scale parameter, and β is a shape parameter. The cumulative distribution function is then transformed into a standard normal distribution, similarly to the SPI. The SPEI has been computed using the R SPEI software v. 1.8.1 (https://cran.r-project.org/web/packages/SPEI/index.htm) (accessed on 22 April 2024).

Table 1. Classification of drought conditions according to SPI and SPEI values [34,42].

Class	SPI and SPEI
Extremely wet	>2.00
Severely wet	1.50 to 1.99
Moderately wet	1.00 to 1.49
Slightly wet	0.50 to 0.99
Near normal	−0.49 to 0.49
Mild dry	−0.99 to −0.50
Moderately dry	−1.49 to −1.00
Severely dry	−1.99 to −1.50
Extremely dry	<−2.00

shows the common classification of drought conditions based on the values of the standardized indices. We present time series and maps of drought conditions based on such a classification.

3.3. Trend Detection and Magnitude Estimation

The Modified Mann–Kendall test (MMK) is used to detect trends in time series data while accounting for autocorrelation [45,46,47]. This test modifies the traditional Mann–Kendall test by adjusting the variance to reflect the influence of serial correlation. The test statistic S is calculated as

$$S={\displaystyle \sum }_{\mathrm{i} =1}^{\mathrm{n}-1}{\displaystyle \sum }_{\mathrm{j} = \mathrm{k} +1}^{\mathrm{n}}\mathrm{Sgn}\left({\mathrm{X}}_{\mathrm{j}}-{\mathrm{X}}_{\mathrm{i}}\right)$$

(3)

where $X_i$ and $X_j$ are the sequential data points, and Sgn is the sign function defined as

$$\mathrm{Sgn}\left(X_j-X_i\right)=\left\{\begin{array}{c}1 if \left(X_j-X_i\right)>0\\ 0 if \left(X_j-X_i\right)=0\\ -1 if \left(X_j-X_i\right)<0\end{array}\right.$$

(4)

The variance of S, in the presence of tied values, is expressed by

$$\mathrm{Var}\left(S\right)={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)-{{\displaystyle \sum }}_{i=1}^m{t}_i\left(t_i-1\right)\left(2t_i+5\right)}{18}}$$

(5)

where $m$ is the number of tied groups and $t_i$ is the number of data values for the ith group.

The test statistic Z follows a standard normal distribution and is expressed as follow:

$$\mathrm{Z} =\left\{\begin{array}{c}{\displaystyle \frac{S-1}{\sqrt{Var\left(s\right)}}}, if S>0\\ 0, if S=0\\ {\displaystyle \frac{S+1}{\sqrt{Var\left(s\right)}}}, if S<0\end{array}\right.$$

(6)

When conducting trend analysis on time series data, it is important to adjust the time series when autocorrelation is present, which can influence the statistical significance of the detected trends [48,49]. If significant autocorrelation in the series is detected (as happens for the Standardized drought indexes since they are based on moving sums of hydrological variables), the variance of the Mann–Kendall test statistic S is adjusted accordingly, using the modified variance formula proposed by Hamed [50].

$$Var{\left(S\right)}^{*}=Var\left(S\right){\displaystyle \frac{n}{n^{*}}}$$

(7)

where n is the original sample size and $n^{*}$ is the effective one.

The ratio ${\displaystyle \raisebox{1ex}{$n$}\!\left/ \!\raisebox{-1ex}{$n^{*}$}\right.}$ is approximated with the empirical expression

$${\displaystyle \frac{n}{n^{*}}}=1+{\displaystyle \frac{2}{n\left(n-1\right)\left(n-2\right)}} {\displaystyle \sum }_{i=1}^{n-1}\left(n-i\right)\left(n-i-1\right)\left(n-i-2\right){\rho }_S\left(i\right)$$

(8)

where $n$ is the original sample size and ${\rho }_S\left(i\right)$ is the sample autocorrelation function of the associated ranks of the time series, namely the lag-i significant auto-correlation coefficient of rank I of the time series. Specifically, the formula is extended just for the lags with statistically significant autocorrelation.

This adjustment ensures that the trend analysis accounts for the influence of autocorrelation, leading to more accurate significance testing.

To assess trend magnitude, Sen’s slope estimator, which provides a robust estimate of the trend slope in time series data [33,51], has been used. It is computed by considering the median of all possible pairwise slopes between data points. For n data points, the slope Q between any two points $(X_{i, } X_j$) is

$$Q=Median\left({\displaystyle \frac{X_j-X_i}{j-i}}\right)$$

(9)

for all of them, 1 ≤ i < j ≤ n. This method is resistant to outliers and provides a reliable trend estimate.

Figure 2 provides a diagram summarizing the various main steps of the illustrated methodology.

4. Results

4.1. Spatiotemporal Analysis of Drought in Sicily

Figure 3 shows that different scales of the SPI and the SPEI detect different drought event episodes. The mean values of the Sicily drought pattern in the SPEI and the SPI indicate consistency in drought occurrences over the island at various times. The most significant drought event occurred in 2002, affecting both short scales (e.g., one- and two-month scales, implying agricultural droughts) and long-term scales (e.g., six-, twelve-, and twenty-four-month scales, implying hydrological droughts). Hence, 2002 is considered to be the main drought season in Sicily from 1950 to 2023. As shown in Figure 3, in addition to 2002, which was the driest year on both the SPI and SPEI long-term scales (12- and 24-month scales), the SPEI at the 24-month scale also identified 1953 and 1963 as additional drought episodes. Meanwhile, the SPI at the 24-month scale highlighted 1991 and 1963 as two other drought episodes.

Following the identification of the drought season in 2002, we analyzed the spatial variation of drought conditions during this event. Figure 4 shows that the western parts (Trapani province), southeastern and southern parts like the south and east of the Catania province, the Syracuse province, and the Ragusa province experienced severe and extreme droughts on 24-month scales of the SPI/SPEI. Conversely, the northeastern part, specifically the province of Messina, and the east of the Palermo province were the least severely affected.

The frequency of drought months across the region using different SPI/SPEI scales is analyzed in Figure 5. In particular, the frequences of severe drought conditions (SPI/SPEI ≤ −1.5) are shown at different scales of the SPI/SPEI across the island. For long-term scales, such as the 12- and 24-month SPI/SPEI scales, up to 60 months are counted to have experienced severe drought conditions, while for the short-term scales, like the 1-, 3-, and 6-month SPI/SPEI scales, months amount up to 50. Regarding the spatial distribution of the severe drought months, for short-term SPI/SPEI scales, spatial variation is low, while for long-term scales, like the 12- and 24-month SPI/SPEI scales, spatial variations are more pronounced. These long-term spatial variations suggested that the central areas of the region suffered the greatest number of months in severe drought conditions, while the eastern, southeastern, and southern parts of Sicily experienced fewer months in severe drought conditions in comparison to the other parts of the region.

Extreme drought conditions occurred with varying frequency across the island as well. Figure 6 shows that in the 12- and 24-month scales of the SPI, up to 40 months in extreme drought were detected. The highest frequency for SPEI values is observed for the 12-month timescale, with 25 extreme drought months in the region. As Figure 5 demonstrates, both the SPI and the SPEI confirmed relatively similar patterns in detecting frequencies of severe drought. However, differences were observed in extreme drought frequency between the SPI and the SPEI, with a maximum of 40 extreme drought events detected by the SPI and 25 detected by the SPEI.

Differences are also present in the spatial distribution of extreme drought episodes according to the SPI and the SPEI. The SPI denotes that the western side of Sicily characterized the highest frequency of extreme droughts compared to other parts of the region. Specifically, in the SPI at 12 and 24-month scales, the highest frequency of extreme drought is significantly concentrated in the western and northwestern parts of Sicily. In contrast, the SPEI did not show consistent concentrations in specific areas and shows sensible spatial variation across the region. For instance, the SPEI indicated that the northeastern, southern, and southeastern parts exhibited the highest frequency of extreme droughts compared to other areas of the island.

4.2. Spatiotemporal Trend Analysis

Significance and magnitude of temporal trends in the SPI and the SPEI from 1950 to 2023 in Sicily are shown in Figure 7 and Figure 8, respectively. Application the MMK test reveals that most of the island has no trends at a 5% level of significance. There are instead subregions where significant trends are present. The SPI exhibited an increasing trend in all months in the eastern part of Sicily, especially in the province of the Catania area. Few cells with a decreasing trend are indicated, especially in the southern part. The SPEI showed a decreasing trend mainly in the western part of Sicily and also partially in the Messina province (the northeastern part). Some cells in the southern central part of Sicily are also present.

Sen’s slope estimator was also computed over Sicily to determine the magnitude of drought indices across the island. Similar to the MMK trend, Sen’s slope estimator exhibited different magnitudes across the island throughout the last 74 years. Figure 8 reveals that the coastal parts of the western, southern, and northern regions, as well as most of the western part of Sicily, displayed higher magnitudes in the drought of the Sen’s slope results for both the SPI and the SPEI across all time scales. In contrast, Mount Etna and the eastern part of Sicily showed the lowest Sen’s slope compared to other parts of the region in the SPI/SPEI analysis across all time scales.

5. Discussion

Drought is one of the primary hydroclimate hazards in the Mediterranean lands and is also expected to become more challenging in future climate change projection scenarios [3,52]. Sicily has historically experienced various drought events [23], with the 2002 drought being the most extreme [24]. ERA5-Land reanalysis data presents some biases but their quality seemed adequate to detect the main drought events, such as for the 2002 event which, for the 6, 12, and 24 -months timescales, was confirmed as the one with the highest intensity in the considered period. In a previous study, leveraging the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis data, mapping SPI-24 showed that Sicily experienced the most severe drought event in 2002 in comparison to the rest of the European continent [53]. By using the Global SPEI Database (GSD) and integrating the 6-month SPEI scale, the results showed that 2002 was the most severe drought episode, according to both the M5P algorithm and the measurement-based analysis [54]. In Calabria, situated in the south of Italy, both short and long time scales of the SPI exhibited extreme drought conditions in 2002 [55].

Frequency of extreme drought events in the Mediterranean region are increasing continuously and is much higher compared to the rest of Europe [56]. Using the E-OBS daily gridded data, which were produced by the Royal Netherlands Meteorological Institute (KNMI), De Bilt, Netherlands, and integrating the SPI and the SPEI, the study found that Southern Europe, specifically the Mediterranean region including Sicily, experienced significantly higher instances of severe and extreme droughts compared to the rest of Europe [57]. Additionally, using the Global Precipitation Climatology Centre (GPCC) which operates under the Deutscher Wetterdienst (DWD), Offenbach, Germany for SPI computation, in the Mediterranean region, including Sicily, it has been found that after the 1990s, significant parts of the region experienced both severe and extreme drought episodes [58]. In Turkey, i.e., another Mediterranean area, the SPI revealed more severe and extreme droughts compared to the SPEI [59].

In terms of trend analysis, some discrepancies exist between the results derived from ERA5-Land data and those obtained from observed data. Our analysis indicates an increasing trend in the SPI, i.e., fewer drought events related to precipitation. This somehow contrasts with the global reduction in total annual rainfall that was found for the entire region using observations [60]; also, the application of the Mann–Kendall test to an observational dataset from 1921 to 2012 highlighted a general decrease in precipitation in Sicily [32]. These different trends may be related to the selection of the analysis period, which can influence the outcomes when assessing changes in drought conditions, an issue also discussed in relation to climate change impact assessments based on climate model future projections [23]. Also, the presence of biases in ERA5-Land data may in this case significantly affect the outcomes of the trend analysis.

Meanwhile, the SPEI outcomes of this study indicated a decreasing trend in drought across different time scales.

Generally, the significance of trends (the outcome of MMK analysis) for both SPI and SPEI trends present some clustering. This may be related to spatial correlation which tends to be present in hydro-meteorological data produced by reanalysis models. Reanalysis models consider large-scale atmospheric and land processes, but they often struggle with capturing local-scale variability accurately due to their coarse spatial resolution and the simplifications necessary for representing complex processes, even though observations are assimilated. As a result, reanalysis models can produce an output that is overly smooth, leading to an overestimation of spatial correlation compared to observational data.

Finally, the study’s outcomes corroborate those observed in our previous research based on climate projections, where a decrease in the SPEI is clear, while the same cannot be stated for the SPI, for which an increase is indicated in some future scenarios and temporal horizons [23].

6. Conclusions

Drought poses significant risks to water resources, agricultural productivity, and socioeconomic stability, particularly in the Mediterranean region, which is strongly susceptible to climate change. Despite previous studies on drought in Sicily, a clear understanding of its frequency, severity, and trends still remained elusive. This study aimed to analyze the spatiotemporal trends of drought in Sicily, utilizing the ERA5-Land continuous gridded data, which is more effective for spatial continuity and coverage over the island Sicily, thereby providing critical insights for agricultural water resource management and climate mitigation.

The study utilized ERA5-Land reanalysis precipitation and potential evapotranspiration (PET) data covering January 1950 through December 2023. The SPI and the SPEI were employed to quantify drought conditions at various time scales. The Modified Mann–Kendall (MMK) test and Sen’s slope estimator were applied to identify and evaluate trends in the drought data, accounting for autocorrelation.

In spite of the biases present in ERA5-Land reanalysis data, the SPI and the SPEI computed by the reanalysis correctly identify the most significant drought event that occurred in Sicily, i.e., the event started in 2002. In 2002, which was the driest year on both the SPI and SPEI long-term scales (12- and 24-month scales), the SPEI at the 24-month scale also identified 1953 and 1963 as additional drought episodes. Meanwhile, the SPI at the 24-month scale highlighted 1991 and 1963 as two other drought episodes. Western, southern, and southeastern Sicily, including the provinces of Trapani, Catania, Syracuse, and Ragusa, experienced severe and extreme droughts. The northeastern part of the island, particularly the province of Messina and the east of Palermo, experienced fewer drought episodes.

While trends in the SPI and the SPEI were found to be non-significant over most of the island, significant trends were found for some subregions of Sicily. Specifically, the SPI showed mainly increasing trends across different time scales, while the SPEI exhibited mainly decreasing trends, indicating a general trend towards drier conditions. The SPI exhibited increasing trends in the eastern part of Sicily (the province of Catania). The SPEI exhibited a consistent decreasing trend in the western part of Sicily and partially in the Messina province. Sen’s slope analysis revealed that the coastal areas, especially the western, southern, and northern regions, displayed higher magnitudes of drought severity. In contrast, Mount Etna and the eastern part of Sicily showed lower drought magnitudes. Clustering in the areas subject to trends may, however, indicate limitations of the ERA5-Land reanalysis dataset related to a possible overestimation of spatial correlation.

Severe drought episodes (SPI/SPEI ≤ −1.5) were mostly evenly distributed across short-term scales but showed spatial variation on long-term scales. Central Sicily experienced the largest monthly count with severe drought conditions, while the eastern, southeastern, and southern parts experienced fewer episodes. Extreme drought episodes (SPI/SPEI ≤ −2) were more frequent in the western and northwestern parts of Sicily, with the SPI indicating up to 40 extreme drought events and the SPEI showing up to 25.

The findings highlight the need for targeted and adaptive measures to mitigate drought impacts in Sicily. Determining drought-sensitive areas provides a basis for prioritizing water resource management and agricultural practices to enhance resilience and sustainability in the face of climate change. The study underscores the importance of continuous monitoring and advanced modelling techniques to better predict and manage drought risks in the Mediterranean area.