1. Introduction
The Earth landscape is a result of the interaction of several parameters, such as the geomorphic process, climate and human activity. Natural hazards are physical events, able to influence the natural and human environment significantly. In this context, the morphological changes of landforms due to natural disasters can restrict human interaction with the ecosystem [
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12].
Floods have affected the natural environment, before humans’ appearance on Earth. Nowadays, they are considered natural hazards and a significant global problem threatening human life [
13,
14].
Floods have a significant impact on the development of a region. Historically, floods have large consequences for individuals and communities which include loss of human life, damage to buildings and infrastructure, destruction of crops, disruption to social affairs and loss of livestock [
15,
16].
In Europe, floods are among the most expensive types of natural disasters with costs about €4 billion per year over the period 2000–2020. Flooding is less significant in inner continental countries compared to Mediterranean countries [
17].
On the other hand, floods after a fire are typically more extensive than before wildfires. Worldwide, fires burn more than 300 Mha h
−1 per year, and in the Euro-Mediterranean region, wildfires burn about 450,000 ha per year [
18,
19,
20]. Greece in particular experienced unusually high burned areas in 2007 and 2018. In the aftermath of wildfires, all surrounding and impacted areas are at risk of flooding for several years. Damaged areas are left without the vegetation that can help absorb rainfall. Wildfires dramatically change the landscape and the hydrologic response of a watershed and thus, a simple storm can cause extreme flash floods [
21,
22]. Another invisible threat is the deposited ash layer which contains potentially toxic elements [
23]. However, much work should be performed on the topic of the impact of wildfire ash on human health and ecological terrestrial receptors [
23,
24,
25].
Burned areas, along with human activities, influence flooding. Proactive planning is among the powerful tools to minimize losses and reduce the economic impact that accompanies floods’ occurrence. Moreover, accurate spatial data related to the distribution of floods are a powerful tool for planers and engineers to decide whether or not the land of an area is safe for urban development [
3,
13,
20,
21,
22,
26,
27].
Flood peaks, depths, volumes and mapping of flood hazard areas may be estimated by applying hydrologic and hydraulic–hydrodynamic models. Their produced flood hazard maps linked to the probability of occurrence of a flooding phenomenon in a certain period of time [
28,
29,
30]. On the other hand, these models involve data that in many cases are often unavailable. Alternatively, several researchers have used deterministic methods to define flood-prone areas [
27,
28,
31,
32]. A deterministic method, proposed by [
33] is the analytical hierarchy process (AHP) method, which is a multi-criteria decision-making technique. It is a weight evaluation procedure which uses qualitative and quantitative criteria to evaluate alternative solutions for a particular problem, among which the most suitable solution is selected [
34,
35]. Besides, the AHP can easy couple with a geographical information system (GIS) to estimate natural hazard maps. However, the AHP method cannot identify the uncertainty associated with spatial outputs [
36,
37].
This study’s main scope is to propose a method to produce a flood hazard map in burned and urban areas. Hence, a coastal part of the northeastern Attica peninsula in central Greece has been selected as the case study. The study area is a part of the Athens metropolitan area and has been affected by extreme summer fires in the recent past.
To achieve this goal, geomorphologic, geologic and land use data affecting flood events were considered. These factors were evaluated by using the AHP method and GIS capabilities and their uncertainty was assessed by using a sensitivity analysis. Previous floods which occurred in the study area were utilized for the validation of the flood hazard map. As a last step, the urban activities were evaluated with the produced flood hazard map.
2. Materials and Methods
The case study area is the municipal unit of Nea Makri and is a popular holiday resort. It is located on the coast of Marathon Bay, a bay of the South Euboean Gulf, in the northeastern part of Attica Prefecture (
Figure 1A). It is bounded on its western—southwestern part by Mount Pentelikon, while the hill Agriliki dominates its northern part. The area is a popular tourist destination with holiday houses, hotels and yacht marinas.
The study area occupies approximately 34 km
2, and it is a low altitude area, with elevation varying from 0 to 782 m above sea level. Gentle slopes characterize its morphology, while only the western and northwestern parts of the basin are mountainous (
Figure 1B).
As regards climatic conditions, the study area has a Mediterranean climate type. The mean annual temperature reaches 18 °C and the yearly precipitation is 400 mm.
Several streams drain the study area. They have a limited length and their drainage basins have small areas. The length of their mainstream channel varies from 0.2 to 5.5 km. The steams debouch northeastwards from the eastern flank of Mount Pentelikon to the South Euboean Gulf (
Figure 1B). Their flow is ephemeral, and they have torrential behavior. Although such fluvial systems in semiarid areas are frequently dry, extreme flash floods happen.
The study area was severely hit by fire on 23 July 2018. The fire started on the Pentelikon Mountain, spread very quickly to the seaside towns of Mati and Neos Voutzas (
Figure 1B) and grew into one of the largest urban wildfires in recent European history, which left 102 people dead and 164 injured. Seventy percent of Mati was utterly burnt from the wildfires, while the remaining 30% was either partially burned or contains burnt areas. Houses and vehicles were totally burned, leaving Mati and nearby towns completely destroyed [
38,
39].
Since the early 21st century, the region has strongly urbanized and its population has greatly increased [
40]. Unchecked constructions block parts of the drainage network. The lowland areas of the towns Nea Makri, Zouberi and Mati have repeatedly been affected by flood events during extreme storm events [
41].
4. Results
The implementation of the previously described flood hazard assessment method led to the production of three flood hazard maps. The standard deviation method was used to categorize the maps. The areas of the maps were categorized into five classes, which were very high, high, moderate, low and very low flood hazard.
Figure 6 illustrates the basic flood hazard assessment map (FHb). In this map, the areas of very high and high flood hazard are located primarily in lowlands and particularly in eastern and southern parts of the study area. The moderate hazard zone is distributed mainly in the central part and the eastern and southern parts of the map. In the northern and western parts of the study area, many locations are situated in low and very low hazard areas.
Figure 7A,B demonstrates the maximum and minimum value of each pixel’s flood hazard assessment, respectively, after considering the uncertainty in the weighting coefficients. According to the flood hazard analysis, the northeastern eastern and southeastern parts of the study area host the most flood-prone regions.
The percentages corresponding to the area of each hazard zone for the three maps of
Figure 6 and
Figure 7 are tabulated in
Table 4. In the areas of the high and very high zones of the map with the upper flood hazard (FHmax), values slightly increased compared to the corresponding zones of the basic map (FH). The similar correlation proves that the spatial extents of moderate and low hazard zones decreased, while the very low hazard zone increase. Comparing the zones of the map with the lower flood hazard values (FHmin) to those of the basic map (FH), the areas of very high and high hazard zones showed a little increase, the moderate and low hazard areas slightly decreased, while the area of very low hazard zone remained unchanged.
Table 5 tabulates the estimated frequency ratio of the flood events for each flood hazard zone. According to [
54,
55,
56], when the ratio values are more than 1, then there is a strong relationship between events and the given natural hazard zone.
Table 5 shows that flood events were significantly correlated to very high and high hazard zones. In the very high hazard area, the frequency ratio was higher than 2, demonstrating a very high probability of flooding. Similarly, the ratio value was observed to be 1.7, in the high hazard zone, showing a high likelihood of flood occurrences. Because the ratio value was lower than 1 in the moderate hazard zone, the probability of floods is low. Since the ratio value was found to be equal to 0 in the low and the very low hazard zones, the likelihood of flooding is minimal.
Moreover, the urban area and the road network were superimposed on the basic flood hazard map. In this context, the area of the urban fabric and the length of the road network that overlaps into each flood hazard zone were estimated (
Table 6).
The results demonstrate that the vast majority of the urban areas are located within very high and high flood hazard zones. Half of the road network’s total length is situated within very high and high flood hazard zones. In contrast, small parts of the urban fabric and the other half of the road network are located within very low, moderate and low landslide hazard zones.
5. Discussion
The urban area constitutes about 33% of the entire study area. The study area has suffered from urban fires in 2018, resulting in about 25% of the area being burned. According to Fox [
21], Euro-Mediterranean countries have common post-fire risk characteristics. They have relatively high population densities, and much of the coastal lowland area is developed. Hence, even small fires can be hazardous, and after fire these coastal areas are vulnerable to flooding. So since the study area is a coastal area and more than half of it is occupied by urban and burned areas, is at high risk of flooding.
In the present study, a method is presented to assess the flood hazard in burned areas and it can be applied even outside the localities affected by the fire. It provides an estimation of flood-prone areas where data are not available, which are typically used more often in connection with floods. Six physical and anthropogenic factors were considered as the most relevant parameters affecting the watercourse when intense storm events cause high runoff that exceeds drainage system capacity. The physical factors, related to flood activity, were slope, elevation, distance from open channel streams, and hydrolithology. The anthropogenic factors associated with floods were distance from totally covered streams, along with burnt and urban areas derived from land cover, as well as supplementary land uses.
A flood hazard assessment map was created via the AHP method and a GIS. According to Hervás and Bobrowsky [
57], flood hazard refers to the probability of flood occurrence within a reference period of time. In the present study, the link of the flood hazard assessment to return periods was not possible. So, a flood susceptibility map was produced which refers to the spatial likelihood or probability for a flood to occur in the future. This map represents the potential flood hazard in the study area, and was further categorized by using the standard deviation method with five classes (
Figure 6) of hazard level: very low, low, moderate, high and very high.
The areas of flood hazard varying between very high and high were recorded mostly in the eastern and southern parts of the Nea Makri area. The area of these two hazard zones covers almost 40% of the whole area (
Table 4). In Greece, extreme flash flood events of low frequency and high magnitude occur in many fluvial systems with a relatively small extent drained by ephemeral streams [
58]. Thus, the small drainage basins, along with the torrential behavior of the streams, influence the flood manifestations in the study area. Moreover, the lowland morphology, gentle slope, the covered streams and intense urbanization in the eastern part of the municipality unit of Nea Makri increase the runoff and produce encouraging locations for flood occurrences. According to Shakesby [
59], in Euro-Mediterranean regions wildfires in forests leave the terrain bare and vulnerable to storms which can cause severe runoff and soil erosion. The runoff increases often in the upper parts of watersheds, and thus poses flooding risks in urban zones located downstream. Thus, the burned area in the southern part of the study area creates favorable conditions for flooding.
In contrast, low and very low flood hazard areas are situated in the mountainous northern, northwestern and western parts of the municipality unit of Nea Makri. These areas are not urbanized, contain open stream channels and do not include burned areas. Therefore, the hazard of flooding in these areas is very low. The spatial distribution of the moderate hazard zone is observed primarily in the central, eastern and southern parts of the municipality unit of Nea Makri. These three hazard zones cover almost 60% of the whole area (
Table 4).
The AHP method is a powerful and valuable tool for flood hazard analysis [
32,
36,
46,
47], although it shows several constraints in assessing the impact of uncertainties [
60,
61]. This fact was evaluated by calculating two additional scenarios that present the upper and lower flood hazard values after accounting for the uncertainty in the weighting coefficients (
Table 3). The findings of the uncertainty analysis showed that the two additional maps (FHmin and FHmax) illustrate a low variation of the presence of flood hazard zones (
Table 5) concerning the basic flood hazard map (FHb). Consequently, the analysis proved slight differences in the spatial and quantitative distribution of the flood hazard zones. This fact indicates robust behavior for the predictions of the applied method.
The combination of several different maps can create a map that does not contain an area’s actual hazard level [
62]. Thus, the basic flood hazard map was verified by means of frequency ratio and flood events which affected the study area over the past 20 years. The findings established that the vast majority of the floods, almost 90% (
Table 5), were located within the limits of the high and very high flood hazard zones. Moreover, the frequency ratio was higher than 1 in the high and very high hazard zones, indicating that these hazard zones are strongly correlated with flood occurrences (
Table 5). The flood hazard intensity rises as the frequency ratio values increase.
The applied method brings many advantages; first of all, that of offering completeness of the analysis of the territory and methodological homogeneity. It is simple and easy to use where primary data are scarce. The results provide reasonable values, which could lead to its further application. The used approach is, therefore, useful in synergy with methods linked to the attribution of a probability.
Comparing the existing urban fabric and road network to the proposed hazard areas of the basic flood hazard map, it is recorded that the vast majority (77%) of the urban fabric and 50% of the road network’s length are situated within the boundaries of high and very high flood hazard zones (
Table 6). This result demonstrated that they have been constructed in areas prone to flood. In these areas, proactive planning and selection of the proper construction rules are essential for the prevention and mitigation of the consequences of flood hazards.
Detailed works relating to the condition of events and evolution of natural hazards and the vulnerability of the exposed elements at risk (i.e., buildings) will help propose the proper protection measures [
46]. The application of the procedure facilitates identifying those sites from the already existing urban fabric and road network situated in non-safe areas. Thus, the appropriate mitigation or hazard reduction approaches can be more effectively designed and applied. Furthermore, the awareness of the non-safe regions related to flood hazard might be helpful in emergency preparedness planning.
Therefore, scientists, engineers, stakeholders, planners and decision makers may utilize the proposed approach in forthcoming spatial planning projects [
63,
64,
65,
66,
67,
68]. Additionally, the local authorities may use the produced map to guide the adoption of measures and strategies aiming towards flood hazard mitigation and post-fire management.