Mapping Drought Incidents in the Mediterranean Region with Remote Sensing: A Step Toward Climate Adaptation

Abstract

The Mediterranean region, identified by scientists as a ‘climate hot spot’, is experiencing warmer and drier conditions, along with an increase in the intensity and frequency of extreme weather events. One such extreme phenomena is droughts. The recent wildfires in this region are a concerning consequence of this phenomenon, causing severe environmental damage and transforming natural landscapes. However, droughts involve a two-way interaction: On the one hand, climate change and various human activities, such as urbanization and deforestation, influence the development and severity of droughts. On the other hand, droughts have a significant impact on various sectors, including ecology, agriculture, and the local economy. This study investigates drought dynamics in four Mediterranean countries, Greece, France, Italy, and Spain, each of which has experienced severe wildfire events in recent years. Using satellite-based Earth observation data, we monitored drought conditions across these regions over a five-year period that includes the dates of major wildfires. To support this analysis, we derived and assessed key indices: the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and Normalized Difference Drought Index (NDDI). High-resolution satellite imagery processed within the Google Earth Engine (GEE) platform enabled the spatial and temporal analysis of these indicators. Our findings reveal that, in all four study areas, peak drought conditions, as reflected in elevated NDDI values, were observed in the months leading up to wildfire outbreaks. This pattern underscores the potential of satellite-derived indices for identifying regional drought patterns and providing early signals of heightened fire risk. The application of GEE offered significant advantages, as it allows efficient handling of long-term and large-scale datasets and facilitates comprehensive spatial analysis. Our methodological framework contributes to a deeper understanding of regional drought variability and its links to extreme events; thus, it could be a valuable tool for supporting the development of adaptive management strategies. Ultimately, such approaches are vital for enhancing resilience, guiding water resource planning, and implementing early warning systems in fire-prone Mediterranean landscapes.

1. Introduction

The Mediterranean region is increasingly acknowledged as a critical climate change hot spot, where rising temperatures, decreasing precipitation, and heightened climatic variability are intensifying the occurrence and severity of extreme environmental events [1,2]. Among these, droughts represent one of the most pressing challenges [3], both as a standalone hazard and as a key driver of secondary impacts such as wildfires. Droughts can be understood as extreme disruptions in the continental water cycle, placing significant stress on all dependent natural and human systems. The nature of processes that lead to drought contributes to their inherent complexity, which in turn makes them particularly challenging for the scientific community to quantify and forecast [4,5] and assess in terms of potential impacts. This complexity is expected to deepen under future climate change conditions, as the factors contributing to drought become more variable and difficult to predict [6]. In recent years, countries including Greece, France, Italy, and Spain have experienced devastating wildfire events, many of which were preceded by prolonged drought periods [7]. Understanding the linkages between drought dynamics and wildfire outbreaks has thus become a focal point of climate impact research in Mediterranean ecosystems [8].

Accurate and timely drought monitoring is essential for mitigating such risks and informing environmental management strategies. Traditional ground-based methods, while valuable, often lack the spatial and temporal coverage required for regional-scale assessment [9,10,11]. As a result, the use of satellite-based Earth Observation (EO) data has gained prominence, offering consistent, repeatable, and high-resolution insights into vegetation health, soil moisture, and hydrological stress over time [12,13,14,15,16]. In this context, spectral indices such as the Normalized Difference Vegetation Index (NDVI) [17,18,19,20], the Normalized Difference Water Index (NDWI) [21,22,23], and the Normalized Difference Drought Index (NDDI) [3,9,24,25] have been widely applied to evaluate drought conditions and vegetation stress across diverse landscapes. For example, Chou et al. [26] in their study calculated NDVI and NDWI from the geostationary satellite Himawari-8 for drought monitoring by assessing vegetation health and hydrological conditions. Their aim was to provide a comprehensive framework for early drought detection and monitoring, particularly during Taiwan’s Spring 2021 drought. Patil et al. [24] used LANDSAT-8 images and analysis of historical rain data in the Parbhani district in India, demonstrating that vegetation indices effectively complement traditional methods by reflecting drought conditions based on rain patterns and dry spells. Bhaga et al. [27] assessed the extent to which Landsat-8 OLI and Sentinel-2 MSI satellite data can be used to characterize and monitor the impacts of drought on water resources in the Western Cape, South Africa. The study found that Sentinel-2-derived NDVI was the most suitable index for mapping surface waterbodies, while NDWI was also utilized. Both indices effectively contributed to monitoring drought impacts, correlating well with climate data during the 2016–2018 drought period.

With the intention to connect the incidence of wildfires and the drought conditions of the affected areas, Chavez et al. [28] introduced a probabilistic methodological approach to evaluate extreme drought conditions prior to the occurrence of a wildfire in Central Chile, analyzing monthly the Enhanced Vegetation Index (EVI), and evaluated the post-fire severity over the affected area with the calculation of the Normalized Burning Index (NBR). Filipponi in Italy [29] used Sentinel-2 data to produce medium-high spatial resolution burned area mapping, with no a priori knowledge about wildfire occurrence or burned areas’ spatial distribution. Ma et al. [30] studied the spatial and temporal distribution patterns of high temperatures, droughts, and forest fires in Northeast China, using meteorological data such as the daily maximum surface temperature from MODIS data and the Keetch–Byram Drought Index, and highlighted the strong correlation between high temperatures and droughts and forest fires.

While previous studies have acknowledged the strong link between extreme weather events and wildfires [31,32], the extent of their influence on fire occurrence varies considerably across different regions [30,33]. Prior research has not yet addressed the systematic examination and comparison of fire-affected areas across multiple Mediterranean regions using consistent indices and temporal frameworks. Towards this direction, the present study leverages satellite-based Earth observation data and advanced cloud computing through Google Earth Engine (GEE) to monitor and analyze drought dynamics in Mediterranean regions that are highly vulnerable to climate change and wildfire risk. By employing vegetation and moisture-related indices, NDVI, NDWI, and NDDI, over a five-year period, the study offers a robust, scalable, and temporally consistent approach to detect spatial patterns of drought and their relationship to wildfire events. This study is guided by three primary research questions: (i) How do drought conditions evolve spatially and temporally in Mediterranean regions that are prone to wildfires? (ii) What is the relationship between pre-existing drought conditions and the occurrence of major wildfire events, and does a consistent pattern emerge across the regions studied? and (iii) How can remote sensing-based drought monitoring support the development of early warning systems and adaptive management strategies in the Mediterranean? To the best of our knowledge, comparative research examining drought dynamics across multiple Mediterranean countries remains limited. Therefore, the methodological approach applied in this study offers a valuable and replicable tool for advancing regional drought assessment and supporting climate resilience efforts in this region.

2. Materials and Methods

2.1. Study Area and Data

Four Mediterranean locations were selected in order to implement our research methodology, Greece, France, Italy, and Spain, as presented in Figure 1. Each selected area has experienced severe wildfire events in recent years.

In detail, the selected Greek area was the Evros region. Evros is situated in the northeastern part of Greece (41°10′ N 26°05′ E) and serves as a critical ecological and geopolitical zone. Bordered by Turkey to the east and Bulgaria to the north, it encompasses diverse landscapes, including forests, wetlands, and agricultural lands. A prominent feature of this region is the Dadia–Lefkimi–Soufli Forest National Park, renowned for its rich biodiversity and as a habitat for all four European vulture species [34].

The second area under study is Gironde and Landes located in Southwest France (44°50′ N, 0°40′ W). This area encompasses extensive pine forests, sandy plains, and a densely populated coastline that is also a major tourist destination. It includes the Landes Forest, one of the largest man-made forests in Europe, and a mosaic of natural and semi-natural landscapes. The third area is the Montiferru region in Sardinia, Italy (40°08′ N, 08°36′ E), which is characterized by mountainous terrain, Mediterranean scrubland, and a dry, hot climate during the summer months. The Montiferru area in particular includes forested slopes, agricultural lands, and rich biodiversity. Finally, the fourth area is the Benahavis region in Spain (36°31′ N, 5°02′ W). The dense coniferous trees and sclerophyllous vegetation of the well-known Sierra Bermeja Natural Park in this mountainous area are combined with the Mediterranean climate that defines the hot and dry summers, and the wet and mild winters of the area. All of these render the region vulnerable to drought and fire phenomena.

2.2. Wildfire Events Overview

In August 2023, Evros experienced a catastrophic wildfire (Figure 2a), marking one of the most severe environmental disasters in recent European history [35]. The blaze scorched approximately 93,560 hectares (935.6 square kilometers) over 17 days, making it the largest recorded wildfire in the European Union since 2000 [35]. The fire predominantly affected the Dadia Forest, a protected area known for its ecological significance [36].

The summer of 2022 brought unprecedented heat and drought to Europe, with Southwest France (Figure 2b) experiencing its worst wildfires since 1949 [37]. Three intense heatwaves swept across France in June, July, and August, with extreme conditions persisting from mid-July. These harsh climatic events severely impacted the regional ecosystem, igniting massive wildfires in the Landes Forest that consumed over 30,000 hectares. This was an extraordinary event in both scale and intensity for the region, which is not only densely populated, over 86,000 residents across 1.4 million hectares, but also a major tourist destination [37]. The population has doubled since 1970, and the summer months see a sharp influx of visitors. The fires caused devastating ecological and socio-economic consequences: 50,000 people were evacuated, five campsites were destroyed, peak tourist activity was heavily disrupted, and local timber production suffered major losses. To combat the disaster, 3000 firefighters were deployed from across France and seven other European nations [7].

The Montiferru region in Sardinia experienced another extreme wildfire event (EWE) in July 2021 (Figure 2c). Notable for both its scale, approximately 13,000 hectares burned, and its unpredictable behavior, the fire began at noon on 24 July in the Oristano province and rapidly spread along the southeastern slopes of the Montiferru, enhanced by strong southeasterly winds. Although the majority of the area burned on 24 and 25 July, the fire was only officially extinguished on 14 August. It caused extensive damage to farmland, forests, and pastures, making it the most devastating wildfire in Sardinia in the past 30 years [38].

The region of Benahavis, which is a municipality in Andalucia of south Spain in the Malaga province, is the fourth wildfire-affected area under study (Figure 2d). The fire on 7th June 2022 was caused by extreme heat and extremely windy conditions, sprawling rapidly at around 30 m per minute. The fire afflicted the municipalities of Benahavis, Pujerra, and Juzcar. Finally, the wildfire affected approximately 5068.35 hectares, 3000 people evacuated these areas, and the damages were mostly found in coniferous trees, sclerophyllous shrublands, olive groves, and fruit trees [39]. This wildfire followed another one that occurred one year earlier in Sierra Bermeja, highlighting the intense fire risk in the region [40].

Figure 2. Spatial distribution of fire-affected areas in the four regions under study, where the mapped extents illustrate the severity and spatial extent of the burned areas. In detail: (a) Map presenting the burnt area (orange color) in the region of Evros in Greece. Source: © European Union’s Copernicus program—Official Emergency Management Service (EMS)—Mapping [41]. (b) Map presenting the evolution of the burnt area (blue, green, and red colored area), based on Copernicus Emergency Management Service—CEMS data, close to Gironde and Landes in France on the basis of the three consecutive CEMS products. Source: © European Union’s Copernicus program—Official Emergency Management Service—Mapping [41], and chart showing the total area (in hectares) affected by wildfires in France from 2009 to 2024, with 2022 standing out as the most severely impacted year. (The year 2024 is marked with an asterisk and shown in grey, as the data available at the time of analysis did not encompass the full calendar year.) Source: © [42]. (c) False-color Copernicus Sentinel-2 image, acquired on 30 July 2021, just after the Montiferru fires in Sardinia Italy. Burnt areas (dark brown) are quite evident in the image. It was Sardinia’s most destructive fire in the last 30 years. Source: © [43]. (d) Map depicting the wildfire delineation and grading in Pujerra, Andalusia, Spain, using post-event VHR satellite images. The analysis resulted in 5068.35 hectares burnt, of which 960.32 ha have been only slightly damaged, 1214.61 hectares moderately damaged, 1885.56 hectares highly damaged, and 1007.86 hectares destroyed. Source: © [39]. For a detailed description of the maps and legends, please refer to the official source websites provided above for every map.

Figure 2. Spatial distribution of fire-affected areas in the four regions under study, where the mapped extents illustrate the severity and spatial extent of the burned areas. In detail: (a) Map presenting the burnt area (orange color) in the region of Evros in Greece. Source: © European Union’s Copernicus program—Official Emergency Management Service (EMS)—Mapping [41]. (b) Map presenting the evolution of the burnt area (blue, green, and red colored area), based on Copernicus Emergency Management Service—CEMS data, close to Gironde and Landes in France on the basis of the three consecutive CEMS products. Source: © European Union’s Copernicus program—Official Emergency Management Service—Mapping [41], and chart showing the total area (in hectares) affected by wildfires in France from 2009 to 2024, with 2022 standing out as the most severely impacted year. (The year 2024 is marked with an asterisk and shown in grey, as the data available at the time of analysis did not encompass the full calendar year.) Source: © [42]. (c) False-color Copernicus Sentinel-2 image, acquired on 30 July 2021, just after the Montiferru fires in Sardinia Italy. Burnt areas (dark brown) are quite evident in the image. It was Sardinia’s most destructive fire in the last 30 years. Source: © [43]. (d) Map depicting the wildfire delineation and grading in Pujerra, Andalusia, Spain, using post-event VHR satellite images. The analysis resulted in 5068.35 hectares burnt, of which 960.32 ha have been only slightly damaged, 1214.61 hectares moderately damaged, 1885.56 hectares highly damaged, and 1007.86 hectares destroyed. Source: © [39]. For a detailed description of the maps and legends, please refer to the official source websites provided above for every map.

2.3. Data Collection and Processing

The Google Earth Engine (GEE) platform provides access to a wide range of satellite datasets, including imagery from the USGS Landsat missions and the European Copernicus Sentinel-2 program. For this study, Sentinel-2A surface reflectance products were used due to their atmospheric correction and suitability for land cover classification. Sentinel-2 offers global coverage, a five-day revisit cycle under consistent viewing conditions, and high spatial resolution across 13 spectral bands in the visible, near-infrared, and shortwave infrared ranges [44]. In extensive land cover mapping projects, like those involving the four chosen regions, handling vast amounts of data and acquiring cloud-free images can be major obstacles. GEE offers a practical solution by enabling users to process and analyze remotely sensed imagery directly through its web-based Integrated Development Environment (IDE), eliminating the need to download data onto a local machine. Therefore, as an initial step in GEE, four Sentinel-2A image collections were created for the designated study areas, specifically using the harmonized image collection available through the “COPERNICUS/S2_HARMONIZED” dataset in GEE. To ensure data quality and reduce atmospheric artifacts, only images with a cloud cover percentage below 20% were retained. The areas of interest were defined according to bounding boxes covering the four selected Mediterranean areas and each image in the collections were spatially clipped to the respective area of interest for focused analysis.

For each Sentinel-2 image, three key vegetation and moisture-related indices were computed. The Normalized Difference Vegetation Index (NDVI) was calculated as the normalized difference between the near-infrared (B8) and red (B4) bands, reflecting vegetation health:

NDVI = (NIR − RED)/(NIR + RED),

(1)

where NIR: near-infrared band (B8) and RED: red band (B4).

The Normalized Difference Water Index (NDWI) was derived using the green (B3) and near-infrared (B8) bands to estimate surface moisture:

NDWI = (GREEN − NIR)/(GREEN + NIR),

(2)

where GREEN: green band (B3) and NIR: near-infrared band (B8).

The interpretation of NDVI and NDWI values plays a fundamental role in assessing vegetation health and water availability [18,23,45], especially in drought monitoring. NDVI values typically range from −1 to 0.9 and provide a clear distinction between land cover types. Values from −1 to 0 indicate non-vegetated features such as water bodies, clouds, or snow, while values between 0 and 0.2 are generally associated with barren land, including rock, sand, or urban surfaces with minimal vegetation. Intermediate values (0.2 to 0.4) correspond to sparse vegetation, while higher values (0.4 to 0.6) and above (up to 0.9) reflect moderate to dense, healthy vegetation, such as forests or well-developed crops [19,46] (

Table 1. NDVI value classification.

NDVI Value	Interpretation
<0.2	Very sparse vegetation
0.2–0.4	Sparse vegetation
0.4–0.6	Moderate vegetation
0.6–0.9	Dense vegetation
0.9–1	Healthy vegetation

). NDWI, on the other hand, assesses the water content in vegetation or the presence of open water; values between 0.2 and 1.0 typically indicate water surfaces, while values from 0.0 to 0.2 may suggest high moisture levels or recent flooding. Negative values, particularly from −0.3 to 0.0, are associated with moderate drought or dry, non-aqueous surfaces, whereas values below −0.3 denote severe drought conditions and extensive dryness (

Table 2. NDWI value classification.

NDWI Value	Interpretation
<−0.3	Extensive dryness
−0.3–0.0	Moderate drought
0.0–0.2	High moisture levels
0.2–1.0	Water surfaces

). Together, these indices provide a robust framework for land surface monitoring, enabling a detailed spatial analysis of vegetation stress and water scarcity across varied landscapes.

From these two indices, the Normalized Drought Difference Index (NDDI) was computed as follows:

NDDI = (NDVI − NDWI)/(NDVI + NDWI),

(3)

By integrating these two indices into NDDI, we were able to derive a more comprehensive indicator that reflects both vegetation condition and surface moisture [7,8,9]. This enabled a spatially detailed and relatively rapid assessment of drought severity, highlighting zones with potential water stress or resilience in each region.

2.4. Seasonal Compositing

Since our analysis framework was intended to capture seasonal patterns in drought patterns, all seasons (Spring, Summer, Autumn, and Winter) were defined using fixed calendar boundaries. To handle the transitional nature of the winter season (spanning two calendar years), the script was configured to extend winter into the following year (December to February), allowing for continuous temporal representation.

The temporal scope of this study covers a five-year period for each of the four selected regions, tailored to capture the conditions leading up to significant wildfire events. Although the overall analysis spans the years 2019 to 2024, the specific five-year window for each area was defined based on the timing of major wildfire occurrences. In the case of the Evros region in Greece, the analysis covers from 2020 to 2024, as a large-scale wildfire occurred in August 2023. Similarly, for the region in France, the period from 2020 to 2024 was selected to encompass the conditions leading up to the severe wildfires during the summer of 2022. For Italy, the timeframe was set from 2019 to 2023 to include the July 2021 fire event. Lastly, in Spain, the analysis also spans from 2020 to 2024 to capture the drought dynamics prior to the wildfires that broke out in July 2022. This region-specific temporal approach enabled us to proceed with a focused examination of drought evolution and vegetation stress in the years leading up to each major fire.

For each year and season, the filtered Sentinel-2 images were aggregated by computing their mean, creating a representative seasonal composite image. This composite reflects the average vegetation and moisture conditions over a consistent temporal window, minimizing noise from single-date anomalies. The NDVI, NDWI, and NDDI were recalculated for these seasonal mean images to ensure consistency in index derivation and to produce robust inputs for subsequent drought classification.

2.5. Drought Level Determination

To translate NDDI values into actionable insights, a threshold-based classification scheme was implemented. Using a series of conditional expressions, each pixel was assigned a drought severity level ranging from 1 to 5, corresponding to five classes (

Table 3. Drought level determination.

No.	NDDI Value	Drought Severity Class
1	<−2	Very low
2	−2–0.7	Low
3	0.7–1.25	Moderate
4	1.25–3	High
5	>3	Very high

): very low, low, moderate, high, and very high drought. The thresholds are based on established ranges of NDDI values [47], as follows:

The drought level map was generated by classifying each pixel’s NDDI value into discrete drought severity classes. The mean drought level for the region was then calculated by averaging these classes across all pixels. This mean does not always correspond to the drought class of the region’s mean NDDI due to the nonlinear nature of classification. However, this approach preserves spatial heterogeneity and provides valuable information on the distribution and prevalence of drought severity across the landscape. Averaging the classified map yields a representative index of the region’s overall drought status, while the map itself is essential for spatially explicit drought monitoring

This approach enables a discrete, interpretable classification of drought severity that is both spatially and temporally scalable. By applying the same logic across multiple seasons and years, drought dynamics can be compared within and across the studied Mediterranean regions.

2.6. Visualization and Export

The applied script in GEE included detailed visualization components to enhance interpretation. NDVI, NDWI, NDDI, and drought levels were rendered on the map using predefined color palettes that highlighted vegetation health and water stress.

Additionally, the script generated seasonal time-series charts for each index using area-averaged statistics. These charts were color-coded for clarity (e.g., green for NDVI, blue for NDWI, red for NDDI) and allowed for the exploration of temporal patterns. Finally, in order to support further analysis and comparison between the examined areas, we computed the mean seasonal values for each index and drought level using a reducer over the areas of interest. These values were further compiled into a feature collection, summarizing data by year and season, and were exported as a *csv file. This provided us with a structured dataset for downstream analysis for a comparison between regions and correlation with wildfire events.

3. Results

3.1. Greece—Evros Region

3.1.1. Pre-Fire Period (Spring 2020–Spring 2023)

Leading up to the fire event in August 2023 (Summer 2023), the area was subjected to persistent and severe drought conditions (

Table 4. Mean seasonal values per index for the study area of Greece.

Drought Level	NDDI	NDVI	NDWI	Season	Year	Drought Level Explained
4.13	7.14	0.41	−0.36	Spring	2020	Very high
3.55	4.07	0.44	−0.40	Summer	2020	Very high
3.29	2.98	0.35	−0.33	Autumn	2020	High
3.91	5.73	0.27	−0.23	Winter	2020	Very high
4.09	6.80	0.38	−0.35	Spring	2021	Very high
3.68	4.88	0.46	−0.42	Summer	2021	Very high
3.39	2.49	0.34	−0.31	Autumn	2021	High
3.74	4.85	0.28	−0.25	Winter	2021	Very high
3.84	5.39	0.39	−0.36	Spring	2022	Very high
3.75	4.88	0.45	−0.41	Summer	2022	Very high
3.24	1.13	0.34	−0.32	Autumn	2022	Moderate
3.81	4.60	0.26	−0.23	Winter	2022	Very high
3.94	5.50	0.41	−0.36	Spring	2023	Very high
3.39	3.71	0.41	−0.39	Summer	2023	Very high
2.63	0.08	0.27	−0.28	Autumn	2023	Moderate
3.66	3.50	0.22	−0.19	Winter	2023	Very high
4.17	8.73	0.39	−0.35	Spring	2024	Very high
2.80	1.19	0.36	−0.37	Summer	2024	Moderate
2.71	0.16	0.28	−0.28	Autumn	2024	Moderate
3.69	4.40	0.23	−0.20	Winter	2024	Very high

and Figure 3). In Winter 2022, the drought level was classified as very high, with an NDDI of 3.8, an NDVI of 0.26, and an NDWI of −0.23. These values reflect dry vegetation and low water content. Moving into Spring 2023, conditions remained equally severe, with a slightly higher NDDI of 3.94 and an NDVI increase to 0.41, indicating more vegetation biomass but under increasing drought stress (NDWI = −0.36). These parameters suggest that the landscape had accumulated sufficient dry biomass, creating highly flammable conditions just before the fire outbreak. The continued presence of vegetation (as seen in the NDVI values) combined with low water content heightened the fire risk.

3.1.2. Fire Season (Summer 2023)

The fire occurred during Summer 2023, which aligns with a period of very high drought intensity as recorded in the data. The NDDI was 3.39, and NDVI was 0.41, still indicating green vegetation cover, but the NDWI dropped further to −0.39, highlighting extremely low water availability. This combination of dry conditions and vegetative presence provided the fuel necessary for wildfire ignition and propagation. The persistently high NDDI values during this time confirm that the fire occurred under severe drought stress, reinforcing the strong link between drought conditions and fire risk.

3.1.3. Post-Fire Period (Autumn 2023–Winter 2024)

Following the fire, Autumn 2023 showed a noticeable shift in drought conditions. The drought level dropped to moderate, with the NDDI decreasing to 2.63 and the NDVI plummeting to 0.27, the lowest among the years recorded, suggesting significant vegetation loss, likely as a direct consequence of the fire. NDWI remained low at −0.28, indicating that soil moisture and water availability had not improved significantly. In Winter 2023, the drought level sharply increased again to very high (NDDI = 3.66), although NDVI dropped slightly from 0.27 to 0.22. NDWI, however, was still quite low (−0.19). By Spring 2024, there was a remarkable rebound in vegetation (NDVI = 0.39), but drought remained at the same levels (NDDI = 4.17, NDWI = −0.35). This pattern of vegetation recovery under persistent drought continued into Summer and Autumn 2024, with moderate drought levels but still low NDWI values (−0.37 and −0.24, respectively), suggesting vegetation was regrowing but under water stress. Finally, in Winter 2024, the drought level rose again to very high, possibly indicating a return to pre-fire risk levels.

Figure 4 presents the derived seasonal drought level maps for Evros, covering Spring, Summer, and Autumn.

3.2. France—Gironde Region

3.2.1. Pre-Fire Period (Spring 2020–Spring 2022)

During the pre-fire period, vegetation in southwest France showed moderate to sparse coverage (

Table 5. Mean seasonal values per index for the study area of France.

Drought Level	NDDI	NDVI	NDWI	Season	Year	Drought Level Explained
4.59	8.07	0.47	−0.41	Spring	2020	High
4.77	8.25	0.51	−0.43	Summer	2020	High
4.85	8.61	0.48	−0.39	Autumn	2020	High
4.63	8.37	0.42	−0.34	Winter	2020	High
4.16	8.62	0.43	−0.38	Spring	2021	High
4.89	9.12	0.55	−0.45	Summer	2021	High
4.84	8.75	0.53	−0.43	Autumn	2021	High
3.61	5.36	0.36	−0.29	Winter	2021	High
4.93	10.58	0.44	−0.39	Spring	2022	High
4.63	8.77	0.51	−0.44	Summer	2022	High
3.68	6.48	0.39	−0.32	Autumn	2022	High
3.49	6.53	0.36	−0.29	Winter	2022	High
4.63	8.03	0.41	−0.35	Spring	2023	High
4.85	9.14	0.53	−0.44	Summer	2023	High
4.78	9.57	0.51	−0.41	Autumn	2023	High
4.80	6.67	0.40	−0.31	Winter	2023	High
5.00	10.84	0.52	−0.45	Spring	2024	High
4.84	9.27	0.56	−0.46	Summer	2024	High
4.86	7.75	0.49	−0.39	Autumn	2024	High
3.26	5.71	0.37	−0.29	Winter	2024	High

and Figure 5). NDVI values were generally in the 0.4 to 0.6 range, indicating moderate vegetation. The highest NDVI values occurred in Summer 2021 (0.55) and Autumn 2021 (0.53), suggesting healthy vegetation before the fire. In contrast, Winter 2021 showed the lowest NDVI (0.36), pointing to sparse vegetation during colder months. NDWI values during this period were consistently negative, which indicates drought conditions. The most severe NDWI value was in Summer 2021 (−0.45) and Autumn 2021 (−0.43), both within the “extensive dryness” class (<−0.3). The least severe was in Winter 2021 (−0.29), still indicating moderate drought. NDDI values were consistently high, peaking at 10.58 in Spring 2022. The drought level values fluctuated between 4.15 and 4.88, with critical highs in Autumn 2020 (4.85), Summer 2021 (4.88), and Spring 2022 (4.63), showing escalating drought severity leading up to the fire.

3.2.2. Fire Season (Summer 2022)

In Summer 2022, the NDVI reached 0.51, ranking among the highest values of the 2020–2024 period. This indicates moderate vegetation cover, bordering dense vegetation according to scientific thresholds. NDWI, however, was −0.44, the lowest of that year, placing it firmly in the “extensive dryness” class (<−0.3). The NDDI measured 8.77, slightly lower than Spring 2022’s peak of 10.58 but still indicating high vegetation drought stress. Drought level was reported at 4.67, the highest possible in the dataset. For Spring (4.93), Autumn (3.68), and Winter (3.69) of 2022, the drought level remained at the maximum value seen during the entire study period.

3.2.3. Post-Fire Period (Autumn 2022–Winter 2024)

NDVI values immediately dropped post-fire, with 0.39 in Autumn 2022 and 0.36 in Winter 2022, both within the sparse vegetation range. Recovery began in Summer 2023 with a rise to 0.53, peaking at 0.56 in Summer 2024. However, NDVI declined again during Winter, reaching 0.40 in 2023 and 0.37 in 2024. NDWI remained negative throughout, indicating persistent drought. The lowest post-fire NDWI was −0.29 in Winter 2024, slightly better than pre-fire lows, now within the moderate drought range rather than severe. NDDI remained elevated, with 9.57 in Autumn 2023 and 10.84 in Spring 2024, suggesting continued drought stress. The drought level varied between 4.49 and 4.85, remaining consistently high.

Figure 6 presents the derived seasonal drought level maps for Gironde region, covering Spring, Summer, and Autumn.

3.3. Italy—Montiferru Region

3.3.1. Pre-Fire Period (Spring 2019–Spring 2021)

During the pre-fire period, vegetation in the Sardinia study area exhibited moderate to dense coverage, as reflected in NDVI values ranging between 0.34 and 0.63 (

Table 6. Mean seasonal values per index for the study area of Italy.

Drought Level	NDDI	NDVI	NDWI	Season	Year	Drought Level Explained
4.90	12.22	0.56	−0.48	Spring	2019	High
2.78	2.18	0.43	−0.43	Summer	2019	Low
3.49	2.43	0.34	−0.32	Autumn	2019	Moderate
4.92	9.40	0.51	−0.41	Winter	2019	High
4.92	10.13	0.58	−0.49	Spring	2020	High
2.63	0.43	0.39	−0.40	Summer	2020	Low
4.88	6.68	0.47	−0.40	Autumn	2020	High
4.94	10.16	0.56	−0.46	Winter	2020	High
4.88	11.62	0.63	−0.53	Spring	2021	High
2.53	−0.12	0.35	−0.37	Summer	2021	Low
2.91	0.50	0.30	−0.29	Autumn	2021	Low
4.80	8.43	0.44	−0.35	Winter	2021	High
4.89	11.05	0.59	−0.50	Spring	2022	High
2.15	−1.66	0.35	−0.39	Summer	2022	Low
4.68	5.06	0.39	−0.34	Autumn	2022	High
4.92	9.89	0.52	−0.42	Winter	2022	High
4.91	10.52	0.54	−0.46	Spring	2023	High
2.75	1.68	0.40	−0.40	Summer	2023	Low
2.97	0.54	0.35	−0.34	Autumn	2023	Low
4.92	9.32	0.54	−0.43	Winter	2023	High

and Figure 7). Notably, the highest NDVI was recorded in Spring 2021 (0.63), indicating healthy and dense vegetation cover. Other relatively high NDVI values were observed in Winter 2019 (0.51), Winter 2020 (0.56), and Spring 2020 (0.58), suggesting favorable vegetation conditions prior to the fire. Conversely, the lowest NDVI was noted in Autumn 2019 (0.34), pointing to a seasonal decline in vegetation. NDWI values during this period were consistently negative, confirming the presence of drought stress. The most severe NDWI readings were observed in Spring 2021 (−0.53) and Spring 2020 (−0.49), indicating conditions within the extensive dryness category (<−0.3). Less severe yet still concerning values were noted in Autumn 2019 (−0.32) and Autumn 2020 (−0.40), both suggesting continued drought pressure. NDDI values fluctuated significantly, ranging from 2.78 in Summer 2019 to a peak of 11.62 in Spring 2021, aligning with increased drought stress in the lead up to the fire. These elevated values, particularly in Spring 2021, support the presence of critical vegetation water stress conditions. The drought level metric further verifies this pattern, with high levels recorded in Spring 2019 (4.90), Spring 2020 (4.92), Winter 2020 (4.94), and Spring 2021 (4.88), signaling escalating drought severity across consecutive growing seasons.

3.3.2. Fire Season (Summer 2021)

In the fire season (Summer 2021), NDVI dropped to 0.35, the lowest summer value across the five-year span, indicating sparse vegetation during the period of ignition. This suggests reduced vegetation health or early degradation possibly due to advancing drought conditions. NDWI recorded −0.37, placing it within the extensive dryness threshold, further affirming the region’s vulnerability. Interestingly, the NDDI sharply declined to a negative value (−0.12), which might be linked to anomalies in data or abrupt vegetation loss. Despite the lower NDDI, the drought level was reported as low (2.53), potentially reflecting a transitional period of acute environmental stress not fully captured by individual indices.

3.3.3. Post-Fire Period (Autumn 2021–Winter 2023)

Following the fire, vegetation conditions showed variability. NDVI values remained low in Autumn 2021 (0.30) and Winter 2021 (0.44), within the sparse to moderate vegetation range. Recovery signs began in Spring 2022 (0.59), one of the highest post-fire values, before declining again in Summer 2022 (0.35). A secondary increase was observed in Winter 2022 (0.52) and Spring 2023 (0.54), suggesting a partial regrowth phase. However, Summer and Autumn 2023 saw another decline in NDVI, indicating inconsistency in vegetation recovery. NDWI remained persistently negative throughout the post-fire period, with values ranging from −0.50 in Spring 2022 to −0.34 in Autumn 2023. These readings signify continuous drought conditions, although slight improvements are seen compared to the peak pre-fire drought levels. NDDI values remained elevated post-fire, peaking at 11.05 in Spring 2022, aligning with ongoing vegetation stress. Similarly, drought level remained predominantly high, notably in Spring 2022 (4.89), Autumn 2022 (4.68), Winter 2022 (4.92), Spring 2023 (4.91), and Winter 2023 (4.92), all indicating sustained critical drought conditions in the aftermath of the fire.

Figure 8 presents the derived seasonal drought level maps for the Montiferru region, covering Spring, Summer, and Autumn.

3.4. Spain—Benahavis Region

3.4.1. Pre-Fire Period (Spring 2020–Spring 2022)

During the pre-fire period, the region experienced high drought stress and moderate vegetation cover (

Table 7. Mean seasonal values per index for the study area of Spain.

Drought Level	NDDI	NDVI	NDWI	Season	Year	Drought Level Explained
4.93	10.00	0.48	−0.39	Spring	2020	High
4.79	11.64	0.42	−0.36	Summer	2020	High
4.89	8.86	0.42	−0.33	Autumn	2020	High
4.83	8.51	0.51	−0.38	Winter	2020	High
4.93	8.55	0.45	−0.37	Spring	2021	High
4.69	12.50	0.41	−0.36	Summer	2021	High
4.86	8.39	0.41	−0.32	Autumn	2021	High
4.84	6.83	0.41	−0.31	Winter	2021	High
4.94	10.67	0.50	−0.41	Spring	2022	High
2.75	1.49	0.11	−0.11	Summer	2022	Moderate
2.27	3.11	0.11	−0.12	Autumn	2022	Moderate
2.66	−1.95	0.10	−0.08	Winter	2022	Moderate
2.01	−5.83	0.20	−0.22	Spring	2023	Moderate
1.46	9.30	0.20	−0.23	Summer	2023	Moderate
2.56	−3.15	0.21	−0.21	Autumn	2023	Moderate
3.28	0.58	0.27	−0.23	Winter	2023	High
3.22	1.73	0.34	−0.33	Spring	2024	High
1.49	11.00	0.26	−0.29	Summer	2024	Moderate
2.65	3.68	0.26	−0.27	Autumn	2024	Moderate
4.24	8.02	0.37	−0.31	Winter	2024	High

and Figure 9). NDVI values remained within the range from 0.41 to 0.51, indicating that vegetation was present but not particularly dense. The highest NDVI was observed in Winter 2020 (0.51), suggesting relatively healthy vegetation during that season. However, NDVI values varied in the range of 0.41–0.45 across most of the other seasons, reflecting moderate vegetation conditions. At the same time, NDWI values were consistently negative, ranging from −0.32 to −0.41. These figures clearly indicate moisture stress and persistent drought. The most severe NDWI was recorded in Spring 2022 (−0.41), right before the fire event. In addition, NDDI values remained elevated during the entire pre-fire period, peaking at 12.50 in Summer 2021 and 10.67 in Spring 2022. These high NDDI values signify increasing vegetation water stress. The drought level was consistently categorized as “High,” ranging between 4.69 and 4.94. This pattern underscores a buildup of environmental stress over the years leading up to the fire, with drought conditions worsening through repeated hot, dry seasons.

3.4.2. Fire Season (Summer 2022)

The fire season in Summer 2022 was marked by a decline in all vegetation and drought indicators, reflecting the immediate impact of the fire. NDVI dropped to 0.11, indicating a collapse in vegetation cover, likely due to extensive burning. NDWI also decreased to −0.11, still within the drought range but less severe than previous seasons, possibly because the destruction of vegetation reduced the vegetation water signal. The NDDI value was 1.49, a dramatic drop from the pre-fire values above 10. This drop likely reflects the lack of active vegetation rather than a relief in drought conditions. Interestingly, the drought level decreased to 2.75 and was classified as “Moderate.” This lower drought level is not indicative of improved environmental conditions, but rather a result of altered spectral responses due to the fire, as burned areas often show lower vegetation indices, which can distort drought calculations.

3.4.3. Post-Fire Period (Autumn 2022–Winter 2024)

The following period after the fire until the winter of 2024 was mainly characterized by “Moderate” to “High” drought and low values of NDDI. There were only two important increases in the NDDI, one in Winter 2023 (0.58) and one in Winter 2024 (8.02), which indicate a new more intense drought period. Even though there was a slow renaissance of the vegetation, the NDVI value progressively recovered, from 0.10 in the winter of 2022 to approximately 0.37 during the Winter of 2024. Furthermore, although NDWI values slightly increased after the fire, they remained low overall, indicating that drought conditions continued in the area of interest.

Figure 10 presents the derived seasonal drought level maps for the Benahavis region, covering Spring, Summer, and Autumn.

3.5. Statistical Analysis of Drought Metrics

To gain a deeper understanding of the interactions between drought conditions and the derived indices, a series of statistical analyses were conducted using the IBM SPSS Statistics 19 software.

3.5.1. Pearson Correlation

First, the Pearson correlation coefficient was computed for every region under study to examine the strength and direction of linear relationships among the drought level and NDDI, NDVI, and NDWI. The Pearson correlation is a statistical measure that quantifies the strength and direction of the linear relationship between two continuous variables. The coefficient ranges from −1 to +1, where +1 indicates a perfect positive linear correlation, 0 indicates no linear correlation, and −1 indicates a perfect negative linear correlation.

The results are summarized in

Table 8. Pearson correlation coefficients for the study areas.

Pearson Correlation	Drought Level	NDDI	NDVI	NDWI
	Greece—Evros region
Drought level	1.000	0.955	0.267	−0.058
NDDI	0.955	1.000	0.359	−0.179
NDVI	0.267	0.359	1.000	−0.974
NDWI	−0.058	−0.179	−0.974	1.000
	France—Gironde region
Drought level	1.000	0.798	0.775	−0.781
NDDI	0.798	1.000	0.748	−0.826
NDVI	0.775	0.748	1.000	−0.972
NDWI	−0.781	−0.826	−0.972	1.000
	Italy—Montiferru region
Drought level	1.000	0.948	0.816	−0.523
NDDI	0.948	1.000	0.931	−0.720
NDVI	0.816	0.931	1.000	−0.911
NDWI	−0.523	−0.720	−0.911	1.000
	Spain—Benahavis region
Drought level	1.000	0.550	0.872	−0.717
NDDI	0.550	1.000	0.691	−0.705
NDVI	0.872	0.691	1.000	−0.957
NDWI	−0.717	−0.705	−0.957	1.000

along with the produced heatmaps (Figure 11).

The Pearson correlation analysis revealed distinct patterns in the relationships between drought severity and spectral indices across the four Mediterranean study areas. In all regions, the NDWI showed a consistently strong negative correlation with drought level (ranging from −0.523 to −0.781), confirming its sensitivity to declining surface moisture under drought stress. NDVI and NDWI were also strongly and negatively correlated in every region (e.g., −0.972 in France and −0.974 in Greece), highlighting the inverse relationship between vegetation vitality and moisture availability.

The NDDI displayed very strong positive correlations with drought level in Italy (0.948), France (0.798), and Greece (0.955), reinforcing its reliability as a composite drought indicator. NDDI also correlated strongly with both NDVI and NDWI, reflecting its integrated sensitivity to vegetation and water stress. Overall, the analysis underscores that NDWI responds most consistently to drought, and NDDI, as a composite metric, proved effective in capturing both vegetation stress and moisture loss, particularly in post-fire contexts (more evident in Italy and Spain).

3.5.2. One-Way Analysis of Variance

An additional statistical test that was carried out was the one-way analysis of variance (ANOVA), using the classified drought levels across the four seasons (Spring, Summer, Autumn, Winter) for each of the examined regions (

Table 9. One-way ANOVA results for the calculated drought level.

ANOVA Results of Greece—Evros Region
Drought Level
	Sum of Squares	df	Mean Square	F	Sig.
Between Groups	2.695	3	0.898	12.011	0.000
Within Groups	1.197	16	0.075
Total	3.891	19
ANOVA results of France—Gironde region
Drought Level
	Sum of Squares	df	Mean Square	F	Sig.
Between Groups	2.673	3	0.874	11.950	0.000
Within Groups	1.024	16	0.069
Total	3.697	19
ANOVA results of Italy—Montiferru region
Drought Level
	Sum of Squares	df	Mean Square	F	Sig.
Between Groups	18.552	3	6.184	26.116	0.000
Within Groups	3.789	16	0.237
Total	22.341	19
ANOVA results of Spain—Benahavis region
Drought Level
	Sum of Squares	df	Mean Square	F	Sig.
Between Groups	3.214	3	1.071	0.599	0.625
Within Groups	28.597	16	1.787
Total	31.811	19

). This statistical test was applied to determine whether the mean drought level differ significantly between seasons, providing insights into the temporal dynamics of drought patterns in Mediterranean landscapes. By identifying whether certain seasons are more prone to drought stress, the analysis aims to support the understanding of seasonality in drought behavior and guide further interpretation of vegetation and moisture responses throughout the year.

In Greece, the one-way ANOVA yielded a statistically significant result (F = 12.01, p < 0.001), confirming meaningful seasonal differences in drought levels. Spring showed the highest drought intensity, while Autumn had the lowest, indicating a clear seasonal gradient. Similarly, France also demonstrated significant differences (F = 11.95, < 0.001), with Spring again showing the highest drought level and Autumn the lowest, suggesting a parallel seasonal characteristic features. However, France’s post hoc test showed overlapping groups in Summer and Winter, indicating moderate differences between seasons. Italy revealed the strongest seasonal effect (F = 26.12, p < 0.001). In summary, there is a clear seasonal gradient in drought severity in these three areas, Spring > Summer > Autumn, in terms of drought level, and another finding that is highlighted is that Spring vs. Autumn shows the most statistically significant difference. In contrast, Spain showed no significant seasonal differences (F = 0.599, p = 0.625). Despite some visible shifts in drought level, notably the drop between Summer 2022 and Summer 2023, the ANOVA results suggest these changes were not statistically consistent across season. These findings reveal that Greece, France, and Italy show a strong and statistically supported seasonal pattern in drought severity, but Spain’s drought behavior appears more unpredictable. This divergence may be attributed to the smaller spatial extent of the area under analysis in the Spain region, which could make it more sensitive to localized influences. The examined area might be more affected by local climate anomalies or recent land use changes.

These observations underscore the critical role that spatial scale plays in drought and vegetation dynamics analysis. They highlight the need for monitoring approaches that are not only region-specific but also sensitive to scale, as analyses conducted over smaller areas may yield different insights compared to broader regional assessments.

3.5.3. Regression Analysis

Finally, in order to explore the relationships between drought severity and the calculated spectral indices, a regression analysis (

Table 10. Regression analysis results.

Regression Analysis Results of Greece—Evros Region Model Summary
Model		R	R Square	Adjusted R Square	Std. Error of the Estimate
1		0.986 a	0.972	0.967	0.08280
a: Predictors: (Constant), NDWI, NDDI, NDVI
Coefficients a
Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.
		B	Std. Error	Beta
1	(Constant)	2.996	0.090		33.209	0.000
	NDDI	0.133	0.014	0.673	9.271	0.000
	NDVI	9.830	1.913	1.632	5.137	0.000
	NDWI	10.588	1.940	1.648	5.456	0.000
a: Dependent Variable: Drought Level.
Regression Analysis results of France—Gironde region Model Summary
Model		R	R Square	Adjusted RSquare	Std. Error of the Estimate
1		0.857 a	0.734	0.684	0.30602
a: Predictors: (Constant), NDWI, NDDI, NDVI
Coefficients a
Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.
		B	Std. Error	Beta
1	(Constant)	2.996	0.090		33.209	0.000
	NDDI	0.133	0.014	0.673	9.271	0.000
	NDVI	9.830	1.913	1.632	5.137	0.000
	NDWI	10.588	1.940	1.648	5.456	0.000
a: Dependent Variable: Drought Level.
Regression Analysis results of Italy—Montiferru region Model Summary
Model		R	R Square	Adjusted R Square	Std. Error of the Estimate
1		0.980 a	0.961	0.954	0.23264
a: Predictors: (Constant), NDWI, NDDI, NDVI
Coefficients a
Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.
		B	Std. Error	Beta
1	(Constant)	4.025	0.535		7.522	0.000
	NDDI	0.151	0.060	0.661	2.508	0.023
	NDVI	9.922	4.851	0.915	2.045	0.058
	NDWI	13.348	4.010	0.788	3.329	0.004
a: Dependent Variable: Drought Level.
Regression Analysis results of Spain—Benahavis region Model Summary
Model		R	R Square	Adjusted R Square	Std. Error of the Estimate
1		0.956 a	0.914	0.898	0.41426
a: Predictors: (Constant), NDWI, NDDI, NDVI
Coefficients a
Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.
		B	Std. Error	Beta
1	(Constant)	2.130	0.352		6.056	0.000
	NDDI	0.007	0.025	0.029	0.282	0.782
	NDVI	21.630	2.567	2.254	8.425	0.000
	NDWI	19.612	3.687	1.461	5.320	0.000
a: Dependent Variable: Drought Level.

) was performed to assess the extent to which vegetation (NDVI), water availability (NDWI), and the combined drought index (NDDI) can predict drought level across the examined regions.

The regression analysis for the Greek region demonstrated a very strong model fit (R² = 0.972), indicating that NDVI, NDWI, and NDDI collectively explain over 97% of the variation in drought levels. All predictors were statistically significant (p < 0.001), with NDVI and NDWI showing the highest standardized coefficients that vegetation and surface moisture are dominant drivers of drought severity in the area. Similarly, in Italy, the regression model was also highly effective (R² = 0.961), showing that drought levels are strongly predicted by the three indices. In France, the model had a more moderate predictive capacity (R² = 0.734, Adjusted R² = 0.684), with only NDDI contributing significantly (p = 0.017), while NDVI and NDWI did not reach statistical significance. This may point to greater heterogeneity in land cover or delayed vegetation response to drought in the French study area. On the other hand, the Spain model showed high explanatory power (R² = 0.914), with NDVI and NDWI both highly significant predictors (p < 0.001), and NDDI not statistically significant (p = 0.782). The dominance of NDVI and NDWI in the Spanish regression suggests that vegetation condition and moisture availability were the most immediate indicators of drought severity in the post-wildfire landscape. Overall, regression results highlight regional differences in drought behavior: Greece and Italy exhibit balanced contributions from all indices; Spain is driven mainly by vegetation and water content; while France relies more on the composite NDDI metric.

4. Discussion

This study explored seasonal patterns and characteristic features in drought severity, vegetation health, and water availability in relation to fire events, using drought indices and vegetation indicators in four fire-affected Mediterranean regions by integrating remote sensing indices with statistical analysis over a multi-year period. The analysis revealed strong spatiotemporal variation in drought severity, vegetation health, and water availability, which collectively shaped fire vulnerability and post-fire ecosystem behavior. At the regional level, Greece (Evros), France (Gironde), and Italy (Montiferru) exhibited a consistently high drought level before and during wildfire events. These patterns were evident in elevated NDDI values and persistently low NDWI readings that respond to severe vegetation water stress. In the case of Greece, severe drought conditions were sustained throughout the pre-fire period, with NDVI values indicating high biomass accumulation, a combination that likely intensified wildfire risk. Similarly, the French region experienced a consistently high NDVI and extremely negative NDWI leading up to the 2022 wildfire, suggesting healthy but dry vegetation that is a critical factor in fire ignition. Italy followed a parallel pattern, with escalating NDDI and drought level leading to the 2021 fire. These observations align with previous research showing that high vegetation greenness under severe drought can paradoxically increase fire risk due to delayed wilting and the accumulation of dry, fine fuels [48,49]. Spain (Benahavis), however, presented a different pattern. While it recorded high pre-fire drought levels, the fire season itself was characterized by drops in the values of all indices, particularly NDVI and NDDI. These declines likely reflect fire-induced changes in surface reflectance, where spectral signals are dominated by burned surfaces. The divergence of the Spanish region results may be also attributed to the smaller spatial extent of the area under analysis, which could make it more sensitive to localized influences, highlighting the need for monitoring approaches that are not only region-specific but also sensitive to scale.

The statistical analysis that was carried out supported and extended these observations. The Pearson correlation revealed a consistently strong negative relationship between NDWI and both NDVI and drought level across all regions, underscoring NDWI’s capability in detecting surface moisture loss. The NDDI, as a composite index, displayed the strongest positive correlations with drought level, especially in Greece, France, and Italy, validating its robustness in capturing drought stress. Regression modeling further confirmed the diagnostic utility of these indices. Greece and Italy demonstrated excellent model fits (R² > 0.96), where all three indices, NDVI, NDWI, and NDDI, contributed significantly to drought level prediction. France’s model was less predictive, relying mainly on NDDI, potentially due to complex land cover heterogeneity. Spain’s model, while statistically strong (R² = 0.914), revealed that NDVI and NDWI where more reliable predictors in drought characterization, once again revealing that drought monitoring is sensitive to scale.

Seasonality was another key finding. The one-way ANOVA confirmed statistically significant seasonal differences in drought level in Greece, France, and Italy, with Spring consistently emerging as the driest season and Autumn as the least drought-affected. This seasonal gradient (Spring > Summer > Autumn) reflects the Mediterranean climate regime, where the moisture in the colder seasons (Autumn and Winter) supports early biomass growth that becomes increasingly stressed in Spring. Spain again diverged, showing no statistically significant seasonal differences.

These findings underscore the importance of integrating spectral indices, temporal monitoring, and statistical analysis for comprehensive drought and wildfire assessment. NDDI stands out as a strong composite indicator, especially in pre-fire risk assessment, while NDWI remains critical for real-time drought monitoring. The regional disparities observed in the case of the Spanish region further highlight the necessity of developing localized drought management strategies that are also sensitive to spatial scale.

4.1. Limitations

Although our methodology revealed seasonal patterns in drought severity, it is important to acknowledge certain limitations. First, as the applied methodology relies solely on satellite imagery, the remote sensing-based drought indices (NDVI, NDWI, and NDDI) were not quantitatively validated against ground-based meteorological or hydrological observations, which may affect the robustness of the assessment. As we focused on four different regions across the Mediterranean, consistent access to reliable in situ data was not feasible, particularly due to variations in data availability, resolution, and accessibility across national monitoring systems. Our study design prioritized the exclusive use of satellite-derived data to ensure methodological consistency and comparability across these diverse landscapes and temporal windows. While this satellite-only approach allows for seamless spatial coverage and systematic monitoring, it inherently constrains the capacity to assess the absolute accuracy of the derived drought indicators. As such, our findings should be interpreted as indicative rather than definitive; they can provide a spatially informed perspective on drought severity patterns rather than serve as a direct substitute for meteorological validation. Integrating ground-based observations such as rainfall records, soil moisture measurements, and hydrological data would enhance the reliability and interpretability of our results; thus, our future work will aim to address this limitation in this direction. Incorporating available in situ datasets from local meteorological stations and national environmental monitoring networks would enable the calibration and validation of the drought severity assessments conducted in this study.

4.2. Contribution to the Field

The present study makes a substantial contribution to the field of wildfire science and environmental monitoring by demonstrating how multi-temporal remote sensing indices can reveal the compounding dynamics of drought and biomass accumulation in the Mediterranean context. By applying an integrated analysis of NDVI, NDWI, and NDDI across Greece, France, and Italy, the study captures the often-overlooked interactions between vegetation health and water stress that precede extreme wildfire events. This approach provides a critical advancement over traditional fire risk assessments that typically evaluate vegetation or drought in isolation. The use of NDDI as a composite index proved especially valuable for capturing the vegetative greenness and moisture deficit, a combination that signals heightened fire susceptibility. Thus, the study supports and extends earlier work emphasizing the utility of integrated indices for fire prediction [49]. It offers empirical validation across diverse Mediterranean ecosystems, thereby enhancing the generalizability of such indices for broader regional application. With the adoption of a satellite-only approach, our study ensures methodological consistency across diverse regions and temporal scales, particularly in contexts where in situ data are scarce or unavailable. This approach provides an efficient and scalable tool for assessing areas affected by severe wildfires and offers a novel perspective through the systematic application of spectral indices to identify recurring patterns. Furthermore, the study contributes to ongoing discourse on post-fire ecosystem resilience. By showing how NDVI often rebounds more quickly than NDWI following a fire, it highlights a potentially hazardous decoupling between vegetation recovery and water availability. This is critical for land management, as apparent vegetation regrowth may mask continued ecosystem vulnerability, a point that has been conceptually recognized but rarely quantified in long-term satellite data [50].

5. Conclusions

The proposed framework of this study advances the existing literature by integrating high-resolution temporal analysis with an accessible, cloud-based processing environment, addressing gaps in previous research that often lacked either the spatial granularity or the computational scalability to effectively assess pre-fire drought conditions across multiple countries. Furthermore, it contributes to the field by demonstrating how satellite-derived indicators can serve as early warning signals for wildfire outbreaks, thereby supporting proactive land and water resource management. By comparing four Mediterranean regions affected by wildfires, the research highlights consistent patterns in pre-fire drought stress and post-fire ecological feedbacks, confirming the diagnostic value of these indicators for early warning and resilience planning.

A key contribution lies in demonstrating how remote sensing data can bridge the gap between environmental monitoring and proactive land management. The approach enables spatial targeting of high-risk areas, supports fuel management strategies, and informs post-fire recovery with landscape-scale resilience planning. By integrating vegetation and hydrological indicators into a cohesive monitoring framework, it responds to calls in the literature for more nuanced, data-driven approaches to environmental risk management [51].

Actionable Future Directions

Building on the insights of this study, five key actions could be proposed as future directions: (a) First, utilizing more the available remote sensing indices, such as NDVI, NDWI, and NDDI, into an early-warning system. This would allow real-time detection of high-risk areas based on vegetation–moisture dynamics. (b) The fuel management should be supported by spatial and environmental data. This should be prioritized in zones where high biomass coincides with severe drought, using the satellite-derived indicators into land-use planning and fire prevention strategies. (c) Post-fire recovery should focus on interventions that take into consideration the moisture dynamics of the affected region. This could include the use of drought-resistant species, for example. (d) Promote cross-border collaboration. This joint effort should support shared protocols, coordinated monitoring, and regionally adaptive fire management strategies. (e) Finally, engaging communities through accessible risk communication tools. For example, through apps or dashboards that use color-coded risk maps could help translating scientific data into local actions, raising awareness.

In conclusion, this study lays the foundation for a more proactive wildfire management. Remote sensing indices, if translated into operational tools and planning frameworks, can bridge the gap between data and decision-making, helping governments, communities, and ecosystems prepare for a future of more frequent and severe wildfire events. Our findings not only contribute to academic understanding but also hold practical relevance for policymakers and civil protection authorities that have the important role of mitigating wildfire impacts in drought-prone regions.