GIS-Based Integrated Multi-Hazard Vulnerability Assessment in Makedonska Kamenica Municipality, North Macedonia

Abstract

This study presents a comprehensive analysis of natural hazard susceptibility in the Makedonska Kamenica municipality of North Macedonia, encompassing erosion assessment, landslides, flash floods, and forest fire vulnerability. Employing advanced GIS and remote sensing (RS) methodologies, hazard models were meticulously developed and integrated to discern areas facing concurrent vulnerabilities. Findings unveil substantial vulnerabilities prevalent across the area, notably along steep terrain gradients, river valleys, and deforested landscapes. Erosion assessment reveals elevated rates, with a mean erosion coefficient (Z) of 0.61 and an annual erosion production of 182,712.9 m³, equivalent to a specific erosion rate of 961.6 m³/km²/year. Landslide susceptibility analysis identifies 31.8% of the municipality exhibiting a very high probability of landslides, while flash flood susceptibility models depict 3.3% of the area prone to very high flash flood potential. Forest fire susceptibility mapping emphasizes slightly less than one-third of the municipality’s forested area is highly or very highly susceptible to fires. Integration of these hazard models elucidates multi-hazard zones, revealing that 11.0% of the municipality’s territory faces concurrent vulnerabilities from excessive erosion, landslides, flash floods, and forest fires. These zones are predominantly located in upstream areas, valleys of river tributaries, and the estuary region. The identification of multi-hazard zones underscores the critical need for targeted preventive measures and robust land management strategies to mitigate potential disasters and safeguard both human infrastructure and natural ecosystems. Recommendations include the implementation of enhanced monitoring systems, validation methodologies, and community engagement initiatives to bolster hazard preparedness and response capabilities effectively.

1. Introduction

Natural processes and human impacts on the environment cause natural hazards. For example, excessive deforestation may accelerate soil erosion and landslides during heavy rainfall [1]. They profoundly impact health, economy, and society, leading to casualties, infrastructure damage, income loss, and altered consumer behavior [2,3,4,5,6,7,8,9]. While natural hazards may not always present the most significant threat to humanity [10], their impact and consequences can vary greatly, with different levels of vulnerability [11,12,13,14,15,16].

Excessive erosion represents a significant threat, especially to agricultural production. Although our study employs the erosion potential model (EPM) for soil erosion assessment, various other methodologies are available for this purpose. It is also worth noting that most of the global soil erosion assessments were carried out using the Universal Soil Loss Equation (USLE) or its revised versions (e.g., RUSLE) [17,18,19,20,21]. While these methods offer comprehensive approaches to erosion evaluation, we specifically chose the EPM due to its perceived advantages in the context of our study. The EPM’s simplicity and applicability to small municipalities, like ours, align with our research objectives focused on localized erosion risk assessment within specific geographic boundaries. Also, the erosion map of Macedonia was also created using the erosion potential method; it was created in the mid-seventies of the last century, and was published together with an interpreter in 1993. Because of the above, the results of this research are comparable to the results of previous research.

Mapping landslide-prone areas through Landslide Susceptibility Zonation (LSZ) is pivotal for addressing landslide vulnerability [22,23,24,25,26,27,28,29]. While methodologies such as the Analytical Hierarchy Process (AHP) are widely used in landslide susceptibility assessment, the frequency ratio (FR) method offers distinct advantages for our study. FR was chosen for its capability to handle categorical data effectively and its simplicity, which aligns well with our objectives and the available data.

Recent studies [15,30,31,32] emphasize Europe’s vulnerability to hydro-meteorological hazards, particularly flash floods, which threaten environmental stability, population centers, infrastructure, and socio-economic aspects. In SEE, including North Macedonia, flash floods caused by intensive rainfalls are a significant hazard [33,34], and pose a serious threat to both people and the environment. The Flash Flood Potential Index (FFPI) is widely regarded as one of the most effective methods for the assessment of flash flood (torrent) areas worldwide [35]. In addition to the FFPI, several other methodologies have been employed in various studies to analyze flash flood occurrences, such as the Hydrological River Basin Environmental Assessment Model (Hydro-BEAM) [36] and the Flood Intensity Index (Iw) [37]. However, in this study, we opted for FFPI due to its comprehensive evaluation of flash flood potential, considering terrain characteristics, lithological composition, and land cover attributes. GIS is a valuable tool for integrating various factors influencing flash flood hazard susceptibility [38]. While climate is the primary factor of floods, the hydrological response depends on physical geographic characteristics such as relief slope, soil texture, and land use. Deforestation, in particular, can significantly impact slope water propagation [39]. Numerous studies (e.g., [40,41,42,43,44,45]) have further explored the FFPI method for flash flood analysis at the municipal or regional level. Improving land use patterns, restoring forests and arable lands, and enhancing municipal functions are essential to mitigate the threats of flash floods, soil erosion, and landslides. These efforts require comprehensive studies and research to generate relevant information, which can then inform land use decisions and spatial planning in the watershed [46].

Access to forest fire vulnerability zone mapping is crucial to addressing forest fire challenges and devising appropriate solutions effectively. Remote sensing enables qualitative analysis of ecosystems, including forests, across various geographic and spatial scales. Understanding the factors contributing to forest fire-prone environments and fire behavior is crucial for effective forest fire management [47]. The Mediterranean coasts, characterized by a Mediterranean climate and fire-sensitive tree species, face significant forest fire vulnerability. In this study, the objective is to assess fire vulnerability zones in North Macedonia using a GIS-based AHP methodology [48,49,50,51]. While the AHP methodology is employed in this study for forest fire vulnerability zone mapping, other methodologies are also utilized in forest fire research. These include the use of statistical models like logistic regression and machine learning algorithms such as random forest. However, the AHP methodology stands out for its ability to incorporate expert knowledge and prioritize factors based on their relative importance, providing a transparent and systematic approach to vulnerability assessment.

Europe, especially southeastern Europe (SEE), faces significant hydro-meteorological hazards, resulting in environmental, social, and economic damage. Climate change exacerbates these risks, with projections indicating worsening conditions and increased damage [52,53,54,55,56]. As the global population grows and urbanization expands into disaster-prone areas, the vulnerability to natural hazards increases rapidly [57,58]. Given the substantial casualties and damage caused by natural hazards globally, there’s a growing interest in analyzing and assessing their risks [59,60,61,62,63,64,65]. Given the region’s favorable natural conditions and historical human impact, North Macedonia faces significant vulnerability to various natural hazards. Previous studies on natural hazards in the area have been conducted (e.g., [28,29,34,66,67]). However, the importance of employing multi-hazard techniques for comprehensive hazard analysis is emphasized [46,68,69,70,71]. Integrating probabilistic and deterministic stochastic processes can provide further insights.

This study contributes original insights by conducting a comprehensive assessment of natural hazard vulnerability in North Macedonia, particularly focusing on Makedonska Kamenica municipality. Through the development of hazard maps and identification of hot-spot areas, the research aims to facilitate effective mitigation measures at both regional and national scales. Additionally, the study emphasizes the importance of integrating probabilistic and deterministic stochastic processes to gain further insights into hazard analysis. These efforts are informed by rigorous research findings aimed at facilitating informed decision-making and spatial planning processes.

2. Materials and Methods

2.1. Study Area

Makedonska Kamenica municipality (in further text: MKM) is located in the northeastern part of North Macedonia. It is situated at the foot of the Osogovo Mountains on its southern slope. On the northwest are borders with Kriva Palanka municipality, on the west with Kočani municipality, on the south with the municipalities of Vinica and Delčevo, and on the north-east is the country border with the Republic of Bulgaria. It occupies a total area of 190 km². The region encompasses rocks dating from the Precambrian, Paleozoic, Pliocene, and Quaternary periods. Dominantly, it features Precambrian gneisses and Paleozoic schists. Pliocene clastic lake sediments and Quaternary River sediments are prominently located along the rivers and Kalimanci Lake. A small portion comprises Precambrian amphibolite, Paleozoic compact volcanic rocks, and Paleozoic granitoids [72].

The topography of MKM ranges between 401 m and 2191 m asl. The lowest is the southern part (Kalimanci Lake), and the highest is the mountain area (Osogovo Mountains) in the northern part of the municipality (Figure 1). The municipality’s average slope measures 18.9°. The south aspect is dominant, with the second being eastern, closely connected with the direction of the highest Osogovo Mountain ridges and the Kamenica valley between them. The terrain ruggedness index (TRI) in the specified area ranges between 0 and 173 (m/km²), and the average value of 65.8 m/km² is relatively high, indicating a significantly rugged and mountainous landscape. This index quantifies the terrain’s roughness or heterogeneity by measuring the variability in elevation within a given area [73].

For the examination of temperature patterns, the ERA-5 dataset was employed due to its superior resolution in temperature measurements. In contrast, for the analysis of precipitation, WorldClim data were selected for their unparalleled resolution in precipitation data, ensuring meticulous accuracy and reliability, especially in regions characterized by pronounced variability. Thus, according to ERA-5 reanalysis data for 1999–2023, the average temperature ranged from 3.8 to 10.6 °C. The average precipitation in the specified period (1980–2020) using the WorldClim 2 dataset [74] ranged from 603 to 816 mm (Figure 2).

Hydrologically, the main watercourse in the study area is the Kamenica River, a right tributary of the Bregalnica River (Figure 3). Its source originates from the Osogovo Mts. at an elevation of 1780 m and through the Kalimanci Lake (reservoir) flows into the Bregalnica River on 510 m asl. Notable left tributaries include the Selška and Vlainska Rivers, while on the right, it is joined by the Gatešnica and Sušica Rivers. Another significant watercourse within the MKM is the Kosevička River, originating to the south of the Šamska Čuka (1514 m); it is also known as Lukovička Reka, ultimately emptying into the Bregalnica River via the Kalimanci Lake. The Bregalnica River (225 km), the second-longest river in North Macedonia, traverses the southern part of the municipality. These rivers are all part of the Bregalnica catchment area. Lake Kalimanci is located at 510 m in the Istibanja Gorge, where the Kamenica River meets Bregalnica.

According to the CORINE Land Cover [75], the predominant land cover types consist of areas characterized by complex cultivation patterns and land primarily dedicated to agriculture. Notably, significant portions are covered by natural vegetation (29.4%), broad-leaved forests (22.4%), and transitional woodland-shrub habitats (19.9%). Mixed forests cover an area of 9.5%, followed by pastures (8.6%) and coniferous forests (8.0%). The forests have undergone significant degradation due to past destruction, likely stemming from various human activities such as deforestation, logging, and agricultural expansion. Water bodies, including parts of Kalimanci Lake, account for 1.4% of the land cover, while artificial surfaces comprise 0.8%.

According to the soil map of Macedonia [76], the municipality’s largest area is covered by Cambisols and a complex of Cambisols and Leptosols, predominantly situated in the northern (mountainous) part. The second largest soil type in the area is Ranker, which also prevails in higher elevations. In the southern part, a dominant presence has a complex comprising Ranker, Regosol, and Leptosol, along with Chromic Luvisols on saprolite. Additionally, Albic Luvisol, Regosol, Ranker, and Leptosol complexes are characterized. The soils are quite eroded and sandy (due to geology). Fluvisols are found adjacent to rivers, particularly at their downstream parts toward Kalimanci Lake.

By socio-economic overview, the municipality comprises nine settlements, eight of which are rural: Todorovci, Lukovica, Kostin Dol, Kosevica, Moštica, Dulica, Sasa, and Cera, with the town of Makedonska Kamenica as the municipal center. According to the last census in 2021 [77], the Municipality of Makedonska Kamenica has a population of 6439, which is significantly less than the census of 2002 (8110 inhabitants), with the highest decrease in the agricultural population. Thus, the municipality has a predominantly rural settlement pattern with an underdeveloped road network, limiting accessibility and economic development. Also, it lies on an important road connecting the eastern and central parts of the republic. Even though the local roads are not great, this main road helps connect the region. All these aspects are important, and the hazard vulnerability assessment focuses on analyzing the causal interaction between natural and anthropogenic systems within the landscape.

2.2. Methodology for Soil Erosion Assessment

The research procedure for the GIS-Based Integrated Multi-Hazard Vulnerability Assessment in MKM, North Macedonia involves a comprehensive approach utilizing various models to assess the most pronounced hydro-meteorological and climatological hazards within the case study. The procedure encompassed gathering the relevant geospatial data, utilizing the erosion potential model (EPM) to assess the susceptibility of the area to erosion, quantifying the landslide susceptibility index (LSI) to identify areas prone to landslides, implementing the Flash Flood Potential Index (FFPI) to evaluate the likelihood of flash floods in the area, assessing the risk of forest fires using a dedicated model that considers variables such as vegetation type, temperature, humidity, and historical fire occurrence data, and integrating the outputs of the individual hazard models into a comprehensive multi-hazard assessment. This model combines the results of the EPM, LSI, FFPI, and forest fires model to identify areas vulnerable to multiple hazards simultaneously in a spatial-temporal context. In the final step, validation of the models and assessment results using ground-based data and expert knowledge was performed to acquire the information necessary for refinements of the models and methodologies based on feedback and additional data if necessary.

The flow chart in Figure 4 presents all procedures and approaches utilized for this research.

For the assessment of soil erosion intensity and sediment yield, the erosion potential model (EPM), also known as the Gavrilović method [78], is widely utilized in the Balkans and worldwide, including Serbia, Croatia, Slovenia, Germany, Italy, Argentina, Belgium, North Macedonia, Greece, etc. (e.g., [29,79,80,81,82,83,84]). It reliably assesses soil erosion rate, mean annual soil loss, sediment yield, erosion control works, and torrent regulation on a regional scale. While convenient for areas with limited erosion data, the EPM does not delve extensively into the underlying physics of erosion processes. The method relies on the following equation:

Wy = T ∙ H ∙ π ∙ √Z³ ∙ F

(1)

where W represents the average annual sediment yield in m³; T is the temperature coefficient, calculated as T = (0.1 ∙ t + 0.1) ^ 0.5, with t being the annual mean air temperature; H indicates the annual precipitation in mm; Z is the erosion coefficient, ranging from 0.1 to 1.5 and beyond; and f is the studied area in km². Among these factors, the coefficient Z holds the utmost importance, as it combines rock erodibility/erosion resistance (Y), land cover index (Xa), index of visible erosion processes (φ), and mean slope of the catchment (J) in the ratio:

Z = Y ∙ Xa ∙ (φ + √J)

(2)

Contrary to the traditional method, the GIS approach for implementing the erosion potential model (EPM) heavily relies on deriving most parameters from digital elevation models and satellite images [85,86,87,88]. For the Y coefficient, a previously digitized geological (100 k) [72] and soil map (50 k) [76] was rasterized to 15 m to correspond to the DEM resolution used, and erodibility values were added according to Gavrilović’s approach [78]. These values generally range from 0.1 for very resistant rocks to 2.0 for non-resistant rocks and soils. However, accurately quantifying coefficient Y poses challenges because of the different soil and rock compactness of the field. Thus, a fitting procedure in the form of rooting Y = sqrt (Y1) is used [82]. The land cover index, Xa, is derived from the CORINE Land Cover model [75], with values ranging from 0.1 (indicating dense forests) to 1.0 (indicating bare rocks), and they are based on suggestions from the original model. The values for coefficient φ (visible erosion processes) are determined from the spectral band 4 of suitable cloud-free Landsat 8 scene in the form φ = a/255, where a is the greyscale pixel value. The results are between 0 and 1, where the lower values indicate areas without visible erosion processes, while higher values show sites with severe erosion processes. Recently, for the coefficient φ, some researchers have used the so-called BSI (Bare Soil Index), based on Landsat 8 or Sentinel-2 imagery [45]. However, further research will show the eventual advantage of this approach. The average terrain slope (J) was calculated from a 15 m digital elevation model (DEM), expressed in decimal percentage. It is recognized that higher slopes correspond to reduced stability, increased erosion potential, and increased susceptibility to torrential floods, as noted by [89]. Using this, the GIS-calibrated coefficient Z is determined by the equation:

Z = √(Y) × φ × ((Xa + φ) × log(a +1) + √(a/57.3))

(3)

Climate parameters (T and H) were derived from ERA-5 average monthly data for precipitation and air temperature, as well as MODIS land surface temperature (LST) via Google Earth Engine. These data, spanning from January 1999 to December 2023, produce raster models of average air temperatures and precipitation for the MKM, forming the basis for T and H coefficients in the erosion potential model (EPM). Their accuracy is verified against measurements from nearby meteorological stations (Kočani, Berovo, and Delčevo), with a high accuracy level of 97.5%. Multiplied, climate potential and erosion coefficient Z determine the mean annual erosion loss (Wy) for the catchment. The GIS-based EPM approach has been tested in various regions in North Macedonia [29,34,82,90], and other countries as well [79,91,92,93,94,95,96]. In assessing the accuracy of the EPM, a common approach involves comparing its results with the measurement of sediment deposition in reservoirs [97,98,99]. This comparison consistently reveals strong correlations between the model’s outputs and measured and observed data. Field investigations were conducted to check and correct the obtained coefficient Z.

2.3. Methodology for Landslide Susceptibility Assessment

Determining landslide-prone areas on a regional scale is a multifaceted endeavor influenced by many natural and anthropogenic factors. Previous research on areas of similar size has provided valuable insights into landslide susceptibility assessment (LSA) [28,100]. Moreover, the work of Milevski and Dragićević [101] underscores evaluating six primary triggering factors: lithology, slope angle, land cover, terrain curvature, distance from rivers, and distance from roads. These factors are the foundation for various LSA methods [102]. They are typically categorized into internal and external factors, with internal factors affecting the mass’s shear strength and external factors surpassing its strength. Topography, lithology, and land use are considered essential factors for GIS analysis [23], although other factors like tectonic features and rainfall distribution can also be influential [103]. In this research, following the study of Milevski et al. [28], we specifically consider six causative factors: lithology, slope, plan curvature, land use, distance from streams, and distance from roads (elaborated further in the following

Table 1. Data sources for landslide susceptibility analysis (LSA).

Data Type	Source	Details	Relevance
Lithology	Digitalized Geological Map (100 k)	Focused on seven distinct lithological units within the municipality area.	Provides lithological composition data for landslide analysis.
Slope and Plan Curvature	15 m Digital Elevation Model (DEM)	Derived from a 15 m DEM.	Key for understanding terrain morphology and its influence on landslides.
Land Use	CORINE Land Cover 2018 Classification Hierarchy	Extracted land cover types within the study area.	Aids in assessing landslide susceptibility based on land cover types.
Distance from Streams	Topographic River Networks	Distances were calculated with buffer zones (20 m increments) and converted to raster format.	Proximity to streams influences landslide susceptibility.
Distance from Roads	Open Street Map Road Networks	Distances were calculated with buffer zones (20 m increments) and converted to raster format.	Proximity to roads influences landslide susceptibility.

).

After the preceding steps, the next crucial phase was selecting a method for assessing landslide susceptibility (LS). Statistical analysis and frequency ratio methods are effective in extensive study areas lacking comprehensive landslide inventories. This involves correlating landslides’ spatial distribution with conditioning parameters to determine the LS approach [104]. This study evaluates landslide susceptibility occurrence with conditioning factors (e.g., slope, lithology, land cover) through the landslide susceptibility index (LSI). This involves calculating the index for each factor category and estimating susceptibility based on individual characteristics. Weighting factors compare calculated landslide density with the overall density in the study area. This comparison helps assess the relative significance of each variable concerning landslide susceptibility [105]. It can be formulated as:

W_ij = 1000(f_ij − f) = 1000(A_ij × /A_ij × A × /A)

(4)

This model indicates areas with varying levels of susceptibility to landslides, considering multiple factors. Weight values for the LSA model factors are provided as follows:

• W_ij: the importance of a certain class (i) of parameter (j).
• f_ij: the landslide density within the class (i) of the parameter (j).
• f: the overall landslide density across the entire model.
• A_ij*: the landslide susceptibility in a specific class (i) of the parameter (j).
• A_ij: the surface area of a particular class (i) of the parameter (j).
• A*: the total area of landslides across the entire model.
• A: the total area of the entire model.

As per [28], a methodology akin to the previous one is adopted for reclassifying LSI values into distinct susceptibility zones and conducting map validation. Each factor is assigned a weighted value, with 30 for slope, 15 for lithology, 10 for land cover, 8 for curvature, and 2 for stream and road distance, totaling 67. These weights signify the relative importance of each factor in triggering landslides within the catchment, with slope being the most influential factor, followed by lithology, land cover, planar curvature, and distance to streams and roads. The cumulative sum of all six parameters determines landslide susceptibility, computed by aggregating values for each grid cell across all layers. The resulting model is classified into five distinct susceptibility classes using natural breaks classification, ranging from very low to very high susceptibility.

2.4. Methodology for Flash Flood Potential Index (FFPI)

The Flash Flood Potential Index (FFPI) is a widely utilized statistical method for identifying areas vulnerable to flash flooding [34,35,43,45,106,107,108,109]. It integrates various factors such as slope, vegetation cover, land use, and soil type (in this case, represented by lithological type due to data constraints) to provide a quantitative susceptibility assessment. Flash floods often occur in municipality areas with reduced soil infiltration rates, typically due to bare terrain, steep slopes, deforestation, urbanization, or industrialization. The combination of intense precipitation, snowmelt, terrain characteristics, soil properties, and vegetation density (including forests, shrubs, and grasses) influences susceptibility to flash flooding (e.g., [110,111,112,113,114,115]). Less vegetated areas generally have a higher vulnerability.

To enhance the accuracy of this study, GIS and RS analysis were conducted to identify areas prone to flash floods within the MKM. These analyses employed the Flash Flood Potential Index (FFPI), a statistical method used to assess flash flood susceptibility. Due to limited soil data from the available soil map of Macedonia (50 k) [76], lithological type was considered instead of soil type. Thus, the model incorporated weighting factors, including slope, vegetation cover, lithology, and land use. The FFPI method is formulated as follows [35]:

FFPI = (M + S + L + V)/4

(5)

The Flash Flood Potential Index (FFPI) incorporates several factors, specifically terrain slope (M), lithology (S), land use (L), and vegetation index (V).

• Terrain slope (M) was derived from a 15 m digital elevation model (DEM). Slope percentages were calculated, and the formula M = 10n/30 was applied, where n represents the average slope. If n is 30% or higher, M is set to 10. The municipality’s average slope is 32.1%, resulting in M = 10.
• For lithology (S) analysis, a digital lithological map based on the geological map of Macedonia (100 k) [72] was utilized (basic geological map). Rocks were classified on a scale from 1 to 9 based on erosion susceptibility.
• Land use (L) data were obtained from the CORINE Land Cover CLC2018 database [75]. Classes were assigned values from 1 to 10 based on their impact on flash flood occurrence.

The Bare Soil Index (BSI) served as the calculation’s vegetation density index (V). Sentinel-2 satellite images from Google Earth Engine script, spanning the period from 1999 to 2023, were utilized for BSI derivation, offering a robust method for assessing erosion rates, often associated with flash floods. Remote sensing, particularly in erosion and flood assessment, provides significant advantages. The BSI was calculated using the formula:

BSI = ((Red + SWIR) − (NIR + Blue))/((Red + SWIR) + (NIR + Blue))

(6)

where Red represents the red spectral channel, SWIR is the short-wave infrared spectral channel, NIR is the near-infrared spectral channel, and Blue is the blue spectral channel. BSI values range from −1.9 to 0.7, with an average of −0.9. To prevent negative values in the vegetation index (V) formula, a constant value of 1 was added. Thus, V = 7.68 × ln(BSI + 1) + 8.

2.5. Methodology for Forest Fires

Forest fires, besides weather conditions, are influenced by various factors [116]. Human negligence and the rise in global temperature led to extensive fire outbreaks annually worldwide, including in North Macedonia. While 97% of forest fires result from human activities, natural conditions also significantly affect a region’s fire susceptibility [53]. In MKM, forest fires, especially in summer, regularly occur, adversely affecting the environment. Despite the municipality’s forest cover (39.9% of the area), the forested area diminishes yearly due to human-induced factors like forest fires and illegal logging. Conducting a vulnerability AHP analysis for forest fire occurrence could aid in implementing preventive measures to safeguard forests. The Analytic Hierarchy Process (AHP) is used for multicriteria analysis and synthesis mapping [117,118]. It prioritizes criteria by quantifying them hierarchically [119,120,121], requiring an understanding of the research domain and physical laws [122,123]. AHP’s key feature is its use of subjective judgment to assign weights, informed by previous research [124]. This enhances objectivity by reflecting expert opinions on the importance hierarchy. AHP is esteemed for expert scenario analysis and decision making, systematically evaluating objectives, criteria, sub-criteria, and alternatives. In forest fire research, criteria vary in importance, necessitating different weight coefficients. AHP uses a 9-point scale for pairwise comparisons, from 1 (equal importance) to 9 (extreme importance) [117,118]. This study simplifies the scale to 1–5 to reduce subjectivity, with 1 as the most significant and 5 as the least. SAGA GIS 9.3.0 software was used for data processing, incorporating various factors from previous research. Thus, for forest fire hazard analysis, the formula developed in one study [125] was used:

RC = 7VT + 5(S + A) + 3(DR + DS)

(7)

In this equation, RC is the numerical index of forest fire vulnerability zones where VT indicates vegetation type with 5 classes, S the slope factor with 5 classes, A the aspect variable with 4 classes, and DR and DS indicates distance factor from road and settlement [47]. Terrain slope and aspect were determined using a 15 m resolution digital elevation model (DEM). Slope values, expressed in degrees, were categorized, with steeper terrains receiving higher values due to reduced accessibility. The vegetation type index was derived from the CLC2018 geospatial database [75], assigning values from 1 to 5 based on the predominant vegetation in each area. Water surfaces and large settlements were deemed the least susceptible, as fire vulnerability in such areas is minimal [45]. To calculate the distance index from roads, settlements, and buildings, satellite images and topographic maps were digitized, creating buffer zones of suitable width. Finally, based on this analysis carried out, a fire vulnerability zone map was produced.

2.6. Methodology for Multi-Hazards

A hierarchical framework was established to model overall susceptibility to natural hazards in MKM, considering occurrence probability, frequency, and consequences of disasters. The AHP method quantifies relative priority using an appropriate value scale. With the global natural risk rising, innovative approaches are emerging to assess and zone areas susceptible to landslides, flash floods, excessive erosion, and forest fires. According to the UNDRR [126], the multi-hazard concept involves identifying major hazards and considering their potential simultaneous, gradual, or cumulative effects, including interrelated or cascading impacts.

Multi-hazard techniques are vital for hazard event analysis [29,34,46,68,69,70,71,127,128], with probability and deterministic stochastic processes offering an alternative approach. Thus, it is necessary to develop models for assessing potential erosion, landslides, flash floods, and forest fires-prone areas in the MKM, North Macedonia. This study aims to (1) assess multiple hydro-meteorological hazards using a modified erosion assessment model and systematic approaches, (2) create natural susceptibility and vulnerability maps, and (3) pinpoint high- vulnerability erosion areas for implementing prevention measures.

To identify multi-hazard areas, QGIS (v. 3.34.2) and SAGA GIS (v. 9.3.0) software were utilized with appropriate GIS tools. Values showing high vulnerability and susceptibility to natural hazards were considered, followed by reclassification and area calculations using the Grid Calculus tool. Different models were developed to enhance the analytical process.

3. Results

3.1. Erosion Hazard Modeling

A map was prepared for the MKM (Figure 5) using SAGA GIS (v. 9.3.0) to calculate the total amount of eroded material based on the previously obtained coefficient of erosion Z (erosion vulnerability).

The obtained results of the erosion coefficient (Z) reveal a substantial prevalence of areas with medium, high, and very high vulnerability (values exceeding 0.4) within the MKM (

Table 2. Erosion assessment modeling in classes.

Class Name	Class	Area in km2	Area in %
Very low	0–0.4	28.4	15.0
Low	0.4–0.8	78.4	41.4
Moderate	0.8–1.2	51.1	26.9
High	1.2–1.6	22.4	11.8
Very high	>1.6	9.3	4.9
Total		189.6	100.0

). These areas cover 82.7 km², equivalent to 43.6% of the municipality’s territory. They are prone to significant soil erosion and sediment transport, even during moderate rainfall. The land undergoes intense rain splash and erosion processes, leading to the production, transportation, and accumulation of deposited material. This phenomenon is especially pronounced during intense rainfalls with rates surpassing 0.5 mm/min or extended periods of heavy rainfall episodes. The accumulated eroded material challenges cultivated areas, roads, and other infrastructure. The mean value of the erosion coefficient Z within the MKM stands at 0.61. Several factors contribute to these elevated erosion coefficients, including the dominance of erodible lithologies, such as Precambrian mica shists, gneisses, Pliocene, and Quaternary clastic sediments, as well as the absence of vegetation and steep terrain slopes.

The model calculation suggests a total erosion production (W) of 182,712.9 m³/year and a specific erosion rate (W_spec) of 961.6 m³/km²/year. The municipality’s most erosive sites are the steep valley sides of the Kamenica River and tributaries, which signify a considerable erosion rate. These areas, especially lacking vegetation or sparse grassy cover, are highly susceptible to erosion processes exacerbated by rainfall (Figure 6).

A substantial 9.4% of the entire area is under very high erosion vulnerability, exceeding 2000 m³/km²/year, equivalent to a soil loss of 2 mm per year (

Table 3. Areas within the MKM are categorized by the extent of erosion.

Class in m3/km2/y	Area in km2	Area in %
0–500	60.7	32.0
500–1000	59.0	31.1
1000–1500	33.8	17.8
1500–2000	18.4	9.7
>2000	17.8	9.4
Total	189.6	100.0

). These high-vulnerability areas experience excessive erosion, resulting in various relief landforms, loss of fertile land, and significant sediment yield accumulating in the riverbeds of Kamenica River and its tributaries as well as in Kalimanci Lake (Figure 7). In contrast, in the northwestern, higher parts of the municipality, erosion potential remains within natural values below 500 m³/km²/year, owing to the presence of well-forested areas. To tackle degradation and mitigate the further loss of natural resources, particularly soil and water, targeted preventive erosion control works must be implemented in areas where erosion intensity exceeds 1000 m³/km²/year. These measures are vital for preserving ecosystem integrity and safeguarding the valuable resources within the region.

Also, the procedure calculates the proportion of sediments reaching Kalimanci Lake (reservoir) as not all sediments exit the Kamenica River catchment. Hence, with the second part of the EPM approach, the sediment delivery ratio (Ru) is estimated using the following equation:

Ru = √(OD)/0.25 × (L + 10)

(8)

where O represents the length of the watershed border in km and D signifies the difference between the average altitude and the altitude of the catchment outlet in km [78]. The sediment yield (G) is then calculated as:

G = W × Ru

(9)

According to the calculations performed, the Ru value for the study area is 0.7 and the average annual sediment yield (G) from the MKM to the Kalimanci Lake totals 127,970 m³/y (

Table 4. Average annual deposition of eroded sediments from the area of MKM to Kalimanci Lake.

W = Annual Sediment Production from EPM	Ru = Retention Coefficient	G = Annual Sediment Yield from EPM	Measured Average Annual Sediment Yield (Kalimanci Lake)
182,713 m3/year	0.7	127,970 m3/year	214,325 m3/year

). This value aligns with one study’s echo sonar measurements of sediment deposition in Kalimanci Lake [129]. According to this study, the measured average annual sediment deposition in Kalimanci Lake is 214,325 m³/year, from which about 1/2 originate from the Kamenica and Lukovica Rivers, both in the scope of the municipality. That shows the high accuracy of our EPM approach, which needs to be further checked with additional measurements on the mentioned rivers.

3.2. Landslides Susceptibility Modeling

Based on the probability map of landslide occurrence in the MKM, generated using the frequency ratio (FR) method and the data provided in

Table 5. Landslide susceptibility areas in the MKM, categorized through the frequency ratio (FR) method.

Class	Class Name	Area in km2	Area in %
1	Very low	12.7	6.7
2	Low	21.9	11.5
3	Moderate	59.3	31.2
4	High	35.6	18.7
5	Very high	60.5	31.8
	Total	190.0	100.0

, it is apparent that class 5 prevails with the highest representation at 31.8%. This class signifies a very high probability of landslides occurring, notably observed along the steep valley sides of the municipality and the valley areas of Kamenica and Bregalnica rivers, its left and right tributaries.

Class 5, representing the largest proportion at 31.8%, predominantly occurs in areas characterized by mica shists, gneiss, clay and sand, shales, and shale granites. This distribution signifies a pronounced likelihood of landslides, particularly prevalent along the left and right banks of the Kamenica River downstream of Makedonska Kamenica, within the Lukovica River valley, and adjacent to the confluence with Kalimanci Lake. The second-largest dominant class is 3 (31.2%), which indicates average susceptibility to landslides, particularly notable on the steep valley sides of MKM, especially in the valley parts of the right Kamenica River tributaries. This area predominantly consists of mica shists and gneisses. Class 4, comprising 18.7% of the total area, signifies a high intensity of landslide occurrence, predominantly along the valley and left tributaries of the Kamenica River and in the higher mountainous part, with shists being the predominant lithological composition. Class 2, constituting 11.5% of the total area, denotes a low probability of landslide occurrences, primarily confined to a small region within the mountainous parts of the left and right tributaries of the Kamenica River. This class encompasses areas where quartz latites and amphibolites are prevalent. The first susceptibility class (class 1), which constitutes 6.7% of the total area, is defined by areas exhibiting a very low intensity of landslide occurrence. These regions are predominantly situated in the northeastern part of the municipality, particularly in the higher (source) regions of the Kamenica River’s left tributaries, where quartz latites are notably present.

Validation of the landslide susceptibility analysis (LSA) involved comparing recorded landslide locations from field surveys (totaling 20) with the LSA zonation. The comparison results are detailed in

Table 6. Landslide distribution relative to LSI classes.

Class	Class Name	Nu	%
1	Very low	/	0.0
2	Low	1	5.0
3	Moderate	2	10.0
4	High	7	35.0
5	Very high	10	50.0
	Total	20	100.0

.

Among the 20 landslides observed, 10 (or 50%) are situated within the highly susceptible LSI zone. Considering the very high (class 5) and high susceptible zones (class 4), this combination encompasses 85.0% of the recorded landslides, indicating a robust accuracy level for the model employed. No landslides were recorded in the field survey within class 1, which denotes a low probability of landslides. This validation reinforces our results’ accuracy and affirms the LSI model’s reliability.

The landslide hazard map was cross-validated with the national study of North Macedonia [28], revealing an overlap with areas highly susceptible to landslides (Figure 8). Factors contributing to landslides include the terrain slope, soils, lithology, and land use. The predominant clay composition of the soil poses a significant risk during heavy rainfall. Alterations in surrounding forest land use can exacerbate this vulnerability. Implementing preventive measures to safeguard the residents of local villages and towns is crucial.

Biosoft easyROC online platform (biosoft.erciyes), was employed to utilize the Receiver Operating Characteristic (ROC) curve as a multi-hazard validation technique, which is widely utilized in geospatial modeling. This curve allows for the visualization of the false positive fraction in relation to the true positive across all values utilized in generating the modeling results. The ROC curve graphically represents specificity on the x-axis and sensitivity on the y-axis (Figure 9). Furthermore, for validation purposes, an automatically calculated Area Under the Curve (AUC) was generated on the ROC curve. AUC values serve as indicators of the success and accuracy of a given model concerning the reference data, ranging from excellent (AUC = 0.9–1) to failed models (AUC = 0.5–0.6), with corresponding categories of good, fair, and poor models falling within specified ranges [29]. To ensure proper validation, it is recommended to have 2–3 times more false-positive landslides than true ones in the validation dataset. In this study, 60 false-positive landslides were randomly selected and thoroughly inspected, resulting in an AUC value of 0.82, showing a high level of accuracy for the employed model.

3.3. Flash Floods Hazard Modeling

The slope map, derived from the 15 m digital elevation model (DEM), plays a crucial role in hydrological processes, influencing runoff timing and infiltration rates. Infiltration rates typically decrease as the slope angle increases [130]. Across the entire area, the average relief slope is 18.9 degrees. Slope values are then converted to percentages and classified accordingly. Subsequently, the model assigns an FFPI value ranging from 1 to 10. Any slope exceeding 30 degrees is categorized with an FFPI value of 10.

In this study, the lithology index was derived from the lithological map, as documented in [72], based on five primary lithological units: clastic sediments, compact volcanic rocks, granitoids, amphibolite, gneisses, and mica schists. The susceptibility of these lithological units to torrential floods was analyzed and classified accordingly (see

Table 7. Coefficient of lithological units (S).

Rocks	Coefficient
Amphibolite	1
Granitoids	2
Gneiss	6
Compact volcanic rocks	1
Mica shists	7
Clastic sediments	9
Clastic sediments	9

). River clastic sediments received the highest coefficient (9), indicating their heightened susceptibility to flash floods. In contrast, compact volcanic rocks and amphibolite received the lowest coefficient (1), suggesting their minimal susceptibility to this natural hazard. This index holds significant importance as the composition of these units influences infiltration rates and runoff dynamics during intense rainfall events. Compact volcanic rocks, amphibolite, and granitoids are less likely to contribute to flash floods due to their erosion-resistant nature. Conversely, clastic sediments and mica schists are more prone to erosion and transport during flash floods, increasing the risk of flooding.

Table 8. Types of land use (L).

Classes of Land	Area (%)	Value
Complex cultivation patterns: Land principally occupied by agriculture, with significant areas of natural vegetation	29.4	7
Broad-leaved forest	22.4	3
Transitional woodland-shrub	19.9	5
Mixed forest; natural grasslands	9.5	3
Pastures	8.6	6
Coniferous forest	8.0	2
Water bodies	1.4	1
Discontinuous urban fabric	0.6	4
Dump sites	0.2	5

illustrates the various land use types present in the study area. The land types most vulnerable to the Flash Flood Potential Index (FFPI) are those characterized by intricate cultivation patterns and predominantly utilized for agriculture, albeit with sizable portions of natural vegetation. Conversely, areas covered by coniferous and mixed forests exhibit the lowest susceptibility to FFPI (water bodies as well). Utilizing the land use map, we generated and classified the land use index into the FFPI (Table 8).

The methodology outlined above has provided comprehensive insights into the potential for flash floods and erosion intensity across the study area. By analyzing Sentinel-2 satellite imagery and utilizing the Bare Soil Index (BSI), we have identified areas more vulnerable to flash floods with enhanced precision (in the GEE platform). Furthermore, the correlation between vegetation density and erosion rate has furnished valuable insights for formulating effective land management strategies and implementing measures to mitigate the impact of flash flood events. Integrating remote sensing techniques and BSI computation has significantly advanced our understanding of erosion dynamics and the susceptibility of the study area to flash floods. The calculated coefficient V ranges from 6.5 to 10.4, with an average of 9.6. A vegetation index is generated within a range of values from 1 to 9.

Utilizing the GIS and RS database, we analyzed the FFPI index to assess flash flood vulnerability across the MKM (Figure 10). Our calculations (

Table 9. Terrains susceptible to flash floods.

Class Name	In km2	In %
Very low	16.9	8.9
Low	17.2	9.1
Moderate	70.3	37.0
High	79.3	41.7
Very high	6.2	3.3
Total	190.0	100.0

) reveal that most of the municipality area, constituting 41.7%, falls within the class characterized by a high probability of flash floods. The second by representation is the moderate class with 70.3%, and areas with very high susceptibility to floods collectively occupy 3.3%. The very high susceptibility class primarily encompasses steep slopes along the tributaries flowing into the Kamenica River, particularly downstream towards the estuary of the Kalimanci Lake, where the Kamenica River converges with the Bregalnica River. The Flash Flood Potential Index (FFPI) yields an average value of 5.7, ranging from 2.4 to 8.1.

For the validation of the model, comparisons with the location of actual flash flood events recorded in the media and the reports of the Crisis Management Center of North Macedonia are made. Thus, in the last twenty years, frequent flash floods have been recorded in the catchments of Mostička Reka, Lukovička Reka, Sušica, and the middle part of the Kamenica River. By field prospections, huge amounts (approximately 1.5–2.0 million m³) of fresh deposits are visible in the floodplains of these rivers, indicating recent flash floods, which overlap with the model zoning. Given the limitations of basic data sources (systematic historical observation), the availability of flash flood data needed for the ROC analysis in this study is limited.

3.4. Forest Fires Hazard Modeling

Utilizing geographic information systems and remote sensing data, we conducted analysis and data processing to generate a forest fire susceptibility map for the municipality of Makedonska Kamenica. The susceptibility was categorized into four classes: low, medium, high, and very high, as illustrated in Figure 11.

Considering the extensive forested area covering 75.9 km² and the considerable terrain slope, the municipality is susceptible to forest fires (

Table 10. Forest fire susceptibility areas.

Class Name	Range	In km2	In %
Very low	0–1	17.2	9.0
Low	1–2	31.8	16.7
Moderate	2–3	35.3	18.6
High	3–4	46.0	24.2
Very high	4–5	59.8	31.5
Total		190.0	100.0

). Areas with steep slopes and southern exposure are particularly prone to such hazards. These locations typically consist of coniferous, broad-leaved, or mixed forests and are relatively proximate to roads and settlements.

The analysis of forest fire susceptibility indicates that slightly less than one-third of the municipality’s area is very highly susceptible to forest fires (31.5%). Additionally, 24.2% of the territory is highly susceptible to forest fires. Moderate susceptibility encompasses 18.6% of the municipality, and low and very low susceptibility together cover 25.7% of the municipality’s area.

Similarly to flash floods, forest fire data in this study is limited. Therefore, further research should focus on expanding the data sources for these parameters in order to improve the modeling accuracy, resulting in more precise ROC/AUC values.

3.5. Modeling the Overall Susceptibility to Natural Hazards (Multi-Hazard Modeling)

In line with the study’s objectives, a multi-hazard model was constructed by integrating erosion, flash floods, landslide susceptibility, and forest fire maps (Figure 12). Thus, regions exhibiting high erosion potential were overlaid with those highly prone to landslides, floods, and forest fires as well, within the territory of MKM. This process identified areas facing concurrent vulnerability from fourth hazards, termed multi-hazard zones.

The analysis disclosed that segments of MKM are susceptible to multiple hazards, specifically flash floods, forest fires, landslides, and excessive erosion. Designated as the “total vulnerability” areas within the study, these zones encompass around 11.0% of the municipality’s total area (refer to

Table 11. High-vulnerability natural hazard areas in the MKM.

Natural Hazard	In km2	In %
Flash floods	6.2	3.3
Landslides	27.9	14.7
Forest fires	59.8	31.5
Excessive erosion	29.4	15.5
Multi-hazards	20.9	11.0
Total vulnerability	102.4	54.0
Total	190.0	100.0

).

Located predominantly in the upstream areas of MKM, the valleys of the right tributaries of the Kamenica River, and the estuary in the Bregalnica River, these multi-hazard zones exhibit distinctive terrain characteristics. These include deforested areas, exposed soil, steep slopes, and susceptible rock formations. It is worth noting that infrastructure within these high-vulnerability zones faces potential danger. To enhance environmental management effectiveness, it is crucial to pinpoint high and very high susceptibility areas to natural disasters across all settlements. Implementing protective measures in these identified areas will mitigate the likelihood of these analyzed disasters occurring.

4. Discussion

Compared to the entire country area of North Macedonia, the MKM exhibits higher erosivity, with a coefficient Z of 0.6 compared to 0.35. Additionally, the average specific erosion rate (W_spec) for the MKM is estimated to be 961 m³/km²/year, whereas for the entire North Macedonia is reported as 681 m³/km²/year according to [82]. These results indicate that the average annual erosion in the study area is nearly twice as high as the average specific erosion rate for the entire territory of North Macedonia. Also, a comparison of the mean value of the coefficient Z with Pehčevo municipality (in the eastern part of N. Macedonia) is made, which is 0.44 [67]. Additionally, a comparative analysis was conducted with a study area of similar size, namely the municipality of Štrpce in southern Serbia. Research [45] indicated a mean Z erosion coefficient of 0.34 for Štrpce. This finding is consistent with observations from other studies (e.g., [29,34,131,132,133]) emphasizing the interconnectedness of soil erosion intensity with various factors such as natural conditions, demographic characteristics, settlement patterns, and land use dynamics.

The municipality’s total sediment yield (G) is 127,970 m³/year. Compared to [129], the annually measured sediment deposition within Kalimanci Lake totals 214,325 m³, with equitable distribution, approximately half originating from the Kamenica and Lukovica Rivers in Makedonska Kamenica River (Figure 13). The specific sediment yield (G_spec) for the area is calculated to be 673.5 m³/km²/year. Despite proposed erosion control works, this area remains critically prone to erosion, necessitating focused attention on municipality and land management strategies geared towards the conservation and protection of the land.

This study analyzed six factors influencing landslide occurrence: slope, lithology, land cover, plan curvature, distance from streams, and distance from roads. Results show that about one-third of the catchment area has a very high probability of landslides, especially in the lower part of MKM. Overall, 31.8% of the municipality is in the very high susceptibility class. In North Macedonia, regions prone to landslides are mainly hilly areas with Neogene lacustrine sediments and slopes steeper than 10° (Figure 14). The MKM shows even higher susceptibility than the national average. Urgent action is needed to mitigate landslide vulnerability in this area [28]. Exploring alternative model validation approaches, as suggested by [134,135,136], is advisable to enhance the reliability of susceptibility models. A comprehensive assessment of total susceptibility to natural hazards, advocated by [29,137], offers valuable insights for protecting vulnerable areas and cultural heritage. Validating the landslide susceptibility model is straightforward, as it only requires a basic inventory. Comparing the 20 recorded landslides with the model’s zones, we found an 85.0% overlap with the high and very high susceptibility areas. Validation was performed using the Area Under the Curve (AUC) from Receiver Operating Characteristic (ROC) curve analyses, resulting in a value of 0.82. Further research, including precise LIDAR measurements and advanced technologies such as deep learning, is needed to assess the region’s erosion and landslide vulnerability changes.

In this study, we noticed a flash flood susceptibility. Unlike studies that predominantly examine natural flash floods within municipalities [138,139,140], our study area experiences urban flash floods influenced by infrastructure such as drainage systems. The absence of certain key factors in our analysis may have impacted our findings, including the significant role of precipitation in triggering flash floods during intense rainfall. Precipitation is a key factor in flash flood validation due to its intensity and spatial spread. Integrating WorldClim 2 data [74] enhances the vulnerability assessment. Intense rainfall often triggers flash floods in the municipality, peaking in May and November. While lacking a local pluviometry station, GIS and remote sensing models helped analyze rainfall data. Areas at high vulnerability typically receive 600–815 mm of precipitation. Research confirms these findings, suggesting lower rainfall areas experience reduced vegetation and slower hazard recovery. Geological composition and deforestation also heighten erosion vulnerability. The study [34] about high-vulnerability areas in the catchment also validates this result.

The FFPI ranges from 1 to 10, with MKM’s FFPI index of 5.7, indicating a medium vulnerability for flash floods. Additional factors like SPI, TWI, TPI, and Soil Index may improve results [42]. Establishing a monitoring system is crucial due to changing natural and human factors [141]. Advanced technologies aid in data collection for predictive modeling through GIS, aiding hazard prevention [139]. Various approaches can enhance effectiveness in assessing flash flood vulnerability, and selecting the most suitable method for each situation to optimize solutions. Acquiring and analyzing databases is made simple by spatial analysis software, aiding professionals in spatial planning [142]. The model proposed in this study offers practical insights into land management and assists local authorities in mitigating flash flood vulnerability [39]. Its versatility allows implementation on a national scale, which is crucial given changing land use and increasing extreme weather events. Collaborative efforts between local, provincial, and national bodies are essential for deploying protective measures effectively. Integrating GIS and remote sensing provides a powerful tool for assessing flash flood potential, supported by the FFPI method’s accurate vulnerability reflection. Future research should explore machine learning methods to enhance susceptibility modeling, reducing subjectivity and increasing relevance for specific regional areas [44]. For more detailed validation, higher resolution historical data are needed to perform a ROC curve analysis to quantitatively assess the flash flood model’s predictive performance.

A GIS-based forest fire susceptibility map for MKM, located in a high-vulnerability forest fire area in the northeastern part of North Macedonia, evaluated factors like forest structure, topography, and proximity to vulnerable points. The map categorized forest areas into five vulnerability levels: very low, low, moderate, high, and very high. Nearly half of the forests in the study area were classified as very high or high vulnerability. This assessment helps decision-makers plan actions before, during, and after fires. Effective firefighting strategies can be implemented by considering vulnerability levels, including reviewing fire action plans, organizing fire response teams, and strategically positioning fire-watch towers [48]. The model also aids in fire prevention efforts by informing the deployment of firefighting teams, assessing road network efficiency, and identifying suitable locations for water resources. Additionally, they can guide the creation of buffer zones between forests and neighboring residential and agricultural areas in high-vulnerability zones. Satellite data and GIS integration offer effective tools for identifying and categorizing forest areas based on factors like topography, vegetation type, average temperature (ERA-5 data [143]), and proximity to roads and settlements. This integration facilitates the identification of high-vulnerability areas and aids in forestry management planning post-fire [47,49,50,51,144,145]. According to the study [146], vandalism generally exhibited high values of forest fires in North Macedonia, especially in the eastern part of the country (MKM). More attention is needed to understand the social causes of forest fires and effectively target prevention efforts. Prioritizing communication to specific population segments is crucial. Emphasis should shift from intervention to prevention, with a balanced budget allocation and community engagement to reduce negligent fires.

As is the case with flash floods, the situation concerning forest fires is similarly characterized by insufficient systematic historical observations. Authors recognize that factors beyond geophysical ones (e.g., anthropogenic factors) must be considered when analyzing forest fires. This complexity adds a layer of delicacy to estimating model accuracy associated with ROC/AUC values. Furthermore, the use of MODIS remote sensing products makes it difficult to attribute the given climatological hazard within the existing methodological framework due to the limitations of their high resolution.

While the study provides invaluable insights into the assessment of natural hazard vulnerability, it is imperative to acknowledge certain limitations in the methodologies employed. Firstly, despite exhaustive validation efforts encompassing various calculations, measurements, and field research, inherent uncertainties and limitations persist within the datasets utilized, potentially undermining the accuracy of the findings. Secondly, the absence of local pluviometry stations may have compromised the precision of the flash flood susceptibility assessment, given its heavy reliance on precipitation data. Thirdly, a notable constraint in assessing landslide susceptibility lies in the inadequacy of the inventory database, which may have impinged upon the completeness and accuracy of the model. Additionally, it is crucial to recognize the limitations in assessing forest fire susceptibility, including the reliance on remote sensing data and the inherent uncertainties in fire behavior modeling. Moreover, while the FFPI and RC method offers a comprehensive evaluation of flash flood and forest fire potential, its applicability is subject to variation based on regional characteristics, underscoring the necessity for further validation and refinement tailored to specific geographical areas.

5. Conclusions

The study provides comprehensive insights into sediment transport, erosion rates, landslide susceptibility, flash flood vulnerability, and forest fire susceptibility in the MKM region. The findings emphasize the critical need for proactive measures to mitigate the impact of natural hazards on the local environment and community. Collaborative efforts involving stakeholders at various levels are essential for implementing effective conservation and management strategies. Further research and field studies are recommended to enhance understanding and address the complex interplay of factors influencing vulnerability to natural hazards in the region.

Integrating the findings with field research is crucial for preserving and enhancing geodiversity in this municipality area. According to one study [46], these areas must be managed carefully, so it is necessary to conduct advanced research to identify the kind of measurement or management action to be implemented in the future. Collaborative efforts involving public nature protection agencies, provincial and national environmental institutions, and local environmental groups are essential for implementing conservation measures and monitoring environmental elements to reduce the vulnerability of natural hazards in MKM. Standardizing and implementing multi-hazard methodologies would improve monitoring and identifying natural disasters in North Macedonia on local and regional scales. This underscores the importance of developing southeastern Europe vulnerability assessments and management programs.