Flooding is one of the most common and destructive natural hazards, endangering lives and the economy (Khan et al., 2011). Floods are becoming more intense due to human activities that lead to land use change and climate change (Khan et al., 2011). Floods are natural hazards that are inevitable and are expected to be more severe in the future (Allafta & Opp, 2021). Therefore, current pattern and future flood hazard scenarios require accurate spatial and temporal inform-ation on the potential flood hazards (Ouma & Tateishi, 2014). Flood is defined as a flow of water that inun-dates higher ground under abnormal conditions or at a level above the typical water surface (Rahman, 2006). It is a common natural disaster brought on by exces-sive rain that destroys property, claims lives, and destroys a large area of agricultural and plants. In general, disaster management can be divided into three phases: preparatory, which involves identifying threat zones before a disaster; mitigation, which involves conducting emergency evacuation, tracking, and exe-cuting contingency plans beforehand or during a disaster; and response, which involves assessing damage and implementing recovery measures soon afterwards (Jeyaseelan, 2003). Recent advancements in space technology allow researchers and agencies to use satellite images. Flooding period and extent may also be roughly determined by these images (Veljano-vski et al., 2011), and mapping flooded areas is a crucial step in understanding the altered land use and land cover (DAddabbo et al., 2018).  Thus, evaluating flood risks and adopting appropriate management and mitigation strategies can significantly reduce related risks (Allafta & Opp, 2021; Le Bihan et al., 2017). Determining flood risk areas and applying appropriate mitigation measures can significantly reduce flood damage (Le Bihan et al., 2017; Naulin et al., 2013). In addition, flood risk mapping plays a significant role in land use management, early warning systems, emer-gency response design and flood risk mitigation estimations (Allafta & Opp, 2021; Zhang & Chen, 2019). Central Asian countries, including Afghanistan, Kazakhstan, Uzbekistan, Turkmenistan, Tajikistan, and Kyrgyzstan, are considered to have extremely continental climates (Gerlitz et al., 2018; Sahak et al., 2023). They are typically counted as arid and semi-arid regions due to less rainfall during the summer season compared to North and South Asian countries (Gerlitz et al., 2018). 

Afghanistan experiences hot summers and cold win-ters, with the lowest annual precipitation of about 30 mm in the southwest and the highest precipitation exceeding 100 mm in the northeast (Sahak et al., 2023; Nur et al., 2021). According to officials, intense rainfall from August 20th to 23rd, 2022, led to flash floods in Logar, an eastern province of Afghanistan, resulting in the deaths of over twenty individuals and the destruction of over 3000 residences. Additionally, numerous canals were ruined, approximately 5000 acres of agricultural land, primarily orchards, were devastated, and around 2000 livestock perished (ARAB, 2022). Logar Province, due to its location and the fact that most of its residential settlements and agricultural lands are close to the river and its low ele-vation, is prone to flooding. Among the most affected areas was Pol-e-Alam District of this province, which was completely submerged. Flood risk evaluation using numerical models is a common method for flood hazard estimation (Vu et al., 2015). Hydrological and hydrodynamic models are widely used to evaluate floods according to their magnitude, extent, and frequency (Aribisala et al., 2022). The runoff effici-ency model, another hydraulic method, mainly exa-mines flood routing issues in waterways (Dilley, 2005; Khan et al., 2011). These quantitative models can evaluate various datasets and offer important insights on the potential for flooding (Wang et al., 2011). However, the most prevalent and challenging issue with such a system is the lack of hydro-meteorological data (Cabrera & Lee, 2019). Many researches have utilized GIS-based multi-criteria evaluation analysis (MCEA) to assess flood risk by investigating the role of factors that control floods (Allafta & Opp, 2021; Desalegn & Mulu, 2021; Hasanloo et al., 2019; Saha & Agrawal, 2020; Tavus et al., 2022). The GIS-MCEA approach utilizes the advantage of GIS for spatial data processing and the adaptability of MCDA to integrate factual data, such as rainfall, land use, slope, soil, and drainage density, with weights-based data (Adesina et al.; Stefanidis & Stathis, 2013; Yahaya et al., 2010). GIS-based MCEA explores complicated decision-making problems by hierarchi-cally stacking control factors (Chen et al., 2011). The model is experts knowledge-based and was first introduced by Saaty (Razandi et al., 2015a; Saaty, 1980). AHP is an essential way to compute the weights of each parameter for achieving the goal in the decision-making of complex problems (Ahmadi et al., 2020). Comparing the influencing parameters based on their relative importance to the target decision by a pairwise comparison matrix is considered the early stage of the AHP model (Ghosh et al., 2020; Şener et al., 2018a). In numerous studies focused on natural hazard assessment, researchers have shown that inte-grating GIS and AHP within an MCEA framework has proven to be effective, particularly in the context of flood hazard mapping (Feizizadeh, 2013), ground-water potential zoning mapping (Arshad et al., 2020), and soil erosion susceptibility mapping (Kachouri et al., 2014). The efficiency of such a method (i.e. combination of GIS with AHP in the MCEA para-digm) in hazard mapping is significantly due to its capacity to deal with data scarcity (Cabrera & Lee, 2019). In this study, the most common factors used in flood risk mapping were slope, digital elevation model (DEM), drainage density, LU/LC, NDVI, and distance from rivers (Table 2). Variables are often chosen according to a comprehensive review of the literature, and their weights are assigned based on expert knowledge using the AHP approach (Adesina et al.,; Allafta & Opp, 2021; Cabrera & Lee, 2019; Hasanloo et al., 2019; Rahman, 2006; Stefanidis & Stathis, 2013; Yahaya et al., 2010). This research attempted to evaluate the areas affected by flooding pre- and post-flood, and also applied the Analytical Hierarchy Process (AHP) algorithm to evaluate the flood risk assessment zone in Logar, Pol-e-Alam district, Afghanistan. This is the first study to be conducted in the Pol-e-Alam district, and is based on a spatial analysis carried out using the AHP model and taking the most relevant factors influencing natural hazards into account. The novelty of the methodology and the outcomes of this study will scientifically assist the managers and policy makers of the Office of State Minister for Disaster Management of Afghanistan, and other involved national and inter-national organiza-tions with a more comprehensive analysis and clear instructions for creating early warning systems, emer-gency response processes, flood risk mitigation estimations, and suggesting where future development should be avoided or restricted. Consequently, the objectives of this study are as follows: 

a) Estimating land use and land cover using sen-tinel-2 satellite imagery. 

b) Evaluation of flood affected map using land use and land cover map. 

c) Utilizing the analytical hierarchy process (AHP) to analyze and produce a flood risk assessment map. 

Study area Pol-e-Alam district is the capital of Logar province in Afghanistan (Fig. 1). Logar is generally described as a relatively flat river valley in the north and central zones. The east, south and southwest of Logar province is surrounded by rugged mountains. The Logar province is located at an elevation of 2186 meters above sea level; it has a humid, continental, and warm summer climate. The population of Logar was 121,935 in 2021. The average annual temperature is 11.4°C (52.52°F), which is 4.3% lower than Afgha-nistans norms. The climate in the area is extremely conducive to agriculture and varies according to ele-vation. The average annual rainfall in Logar is roughly 32.07 millimeters, and there are 81 wet days per year (Nasimi et al., 2020).

Fig. 1: Location of study area.

Dataset

In this study, Sentinel-2 imagery for the study area is acquired from the Copernic website. Images are chosen for two distinct times, pre-flood and post-flood, with dates of August 12, 2022, and August 28, 2022, respectively. To map the flood-affected areas, Sentinel-2 images were chosen based on these dates. Two images tiles were used to cover the study area. To obtain the image of the Pol-e-Alam district, both image tiles are mosaicked and stacked for each date. Digital Elevation Model was downloaded from the USGS Earth Explorer website (https://earthexplorer. usgs.gov/). Two tiles of elevation images that are compatible with the research region were downloaded and mosaicked. In order to calculate flood risk assess-ment map using AHP algorithm, various thematic maps (slope, drainage density, lulc, elevation, distance from river, and NDVI) are created. 

Methodology 

The general methodology followed for this study is represented in (Fig. 2). Sentinel-2 imagery is utilized to identify the areas that have been flooded during 2022. A flood affected map is created based on this information. Then the area of each LULC class that was affected by the flood is determined. Finally, the AHP technique is utilized to map the flood risk assess-ment for the entire research area. The next sections provide a thorough description, step by step, of each part of these methodologies.

Fig. 2: Methodological scheme.

Assessment of flood-affected land cover classes 

In order to calculate and evaluate the flood-affected land cover types, Sentinel-2 images from August 12, 2022, pre-flood, and August 28, 2022, post-flood, are used. The study area is covered by two tiles of image, so the images are mosaicked, the study area is clipped based on the shape file of the study area, the prepro-cessing of the image is done in SNAP, and then supervised classification of Sentinel-2 images of the before and after flood is done to produce the land use 

and land cover map. As a supervised classification technique, the support vector machine algorithm has been utilized. The LULC map is classified into four categories: agriculture, built-up areas, barren areas, and water (Fig. 3). To identify the water and non-water areas, each classified image is reclassified, and finally, it is determined how much of each LULC class is affected by the flood as demonstrated in (Fig. 4) and (Table 1).

Table 1: Analysis of Land Use and Land Cover Change for pre-flood and post-flood.

Fig. 4: Change detection of land use/land cover in the study area pre - flood and post - flood.

The land use change pattern in the study area is con-sistent with the classification result of Sentinel-2 image. The pre-flood image makes agriculture and built-up areas along the river very evident (Fig. 5a). On the other hand, the post-flood image (Fig. 5b) exhibits an abnormal change in water area that exp-anded sharply and resulted in flooding of all nearby built-up and agricultural areas.

Fig. 5: (a) - Sentinel-2 image Pre-flood, 12 August, 2022, and (b) - post-flood, 28 August, 2022, are shown some of the validation points.

AHP Technique for Evaluating the Flood Risk Assessment 

AHP is a critical method for calculating the weights of each factor in the decision of complex problems (Ahmadi et al., 2022). The approach relies on expert knowledge and was initially developed by Saaty (Razandi et al., 2015b; Saaty, 1990). The initial step in the AHP model involves creating a matrix for pairwise comparisons of influencing factors and their respective significance in the decision-making process (Şener et al., 2018b). The normalized weights calculation, con-sistency ratio calculation, and ultimate decision of making steps are the other phases of this model. In this study, six factors (elevation, drainage density, slope, land use and land cover, distance from the river, and NDVI) associated with hydrological processes are evaluated using the Saaty, (1977) scale ranging from 1 (equal importance) to 9 (very high importance). These factors are compared in pairs to create a matrix, as illustrated in (Table 2). Elevation: is considered as one of the key causes of flooding (Seejata et al., 2018). Flood risk is inversely related to elevation since lower elevations are more susceptible to floods than higher elevations (Tang et al., 2018). The depth, direction, and extent of the flood are significantly influenced by elevation. Digital Elevation Model (DEM) serves as the elevation layer. In this study, the study area is divided into five categories: very low, low, medium, high, and very high, with respective ranges of 1868 - 1929, 1929 - 1981, 1981 - 2041 - 2041 - 2113 and 2113 - 2264 meters. Drainage density: is still another crucial factor of flood risk, that the chances of higher flood occurrence are associated with higher runoff, which is directly related to the higher value of drain-age density (Mahmoud & Gan, 2018). The greater drainage density is a favorable indicator that the basin has a higher flow accumulation channel. A higher streamline density suggests a greater amount of excess runoff and, consequently, a higher risk of floods. The stream polyline feature is created for the drainage density layer. After the drainage density layer is formed, it is divided into five classes: 0 - 128, 128 - 276, 276 - 423, 423 - 619, and 619 - 1018. The higher chance of water accumulated is related to the higher range of drainage density. Slope: Floods are more likely to occur in flat or low-sloped locations (Seejata et al., 2018). Slope is directly associated with runoff velocity and vertical percolation. Slope and stream power are directly correlated in the downstream. The slope is measured in degrees and divided into five groups, ranging from 0° - 2°, 2° - 4°, 8° - 16°, and 16° - 34°. LULC and NDVI: The infiltration rate is directly affected by LULC and NDVI. In comparison to urban areas, the vegetated areas can accommodate more infiltration (Seejata et al., 2018). Flooding frequency can be significantly influenced by land use patterns (Das, 2019). Urbanized and developed sur-faces produce greater runoff, which is much more stubborn to subside with time. As a result, urbanized areas and developed shorelines are more flood-prone than bare soil and vegetated land covers. As previ-ously stated, the identified parameters were grouped into six categories and assigned rankings based on their relative significance within each subcategory, as demonstrated in (Table 2). The significance of each category and subcategory was derived from existing literature and understood in terms of their implications for flood risk assessment. In order to generate the flood risk map, six important hydrological phenomena parameter layers are created. Then, the criteria classes are reclassified in order to provide weight. Each parameters weight is shown in (Table 5), in a hier-archical order. Multi-criteria decision-making is used to create the flood risk map. As a tool for multi-criteria decision-making, the AHP technique has been emp-loyed.

Table 2: Flood Susceptibility Criteria and sub - criteria Ranges for Flood Susceptibility Assessment.

Pairwise comparison the factors 

The Saaty, (1980) comparative scale is one of the most common methods for comparison. Considering this method, a comparative scale is made up of integers from 1 to 9. As a result, the number one represents the least important factor, while the number nine repre-sents the most important factor. The comparison process was done for all six factors, and the relative weight of each factor was evaluated (Table 3). Fur-thermore, the normalized matrix and weight for each parameter were calculated, as shown in (Table 4, 5). To examine the discrepancy between pairwise com-parisons and the reliability of the obtained weights, the consistency ratio (CR) should be calculated.

Table 3: Pairwise comparison matrixes.

Table 4: Normalized Pairwise matrix calculated.

Table 5: Weighted of each parameter.

Flood Factors Criteria weighted Criteria Normalized weighted Influence (%)

In AHP, consistency is used to construct a matrix and is expressed by a consistency ratio that must be < 0.1 to be accepted. Otherwise, the subjective judgments (Saaty & Vargas, 2001) need to be revised and recal-culated. The following formula is used to calculate the consistency ratio (CR):

  ………………………………………….(1) 

Where IR is the random inconsistency, which was standard using Saaty, (1980) and the value depends on the number of aspects (n); in this study, there are six factors denoted as n = 6, with a consistency index (CI) calculated using equation 2 and a random index (RI) value of 1.24. 

  ……………………………………….(2) 

Where n is the number of elements and λ is the consis-tency vectors average value. Furthermore, a weighted overlay analysis was performed to create a flood risk assessment map. The Flood Risk Index (FRI) layer is calculated using equation 3.

  ……………………………………(3) 

Where W is the weight of factors i and P is the rating of an individual parameter, moreover, the values that were derived from the FRI index were grouped into four hazard classes according to the probability of flood occurrence. This classification was done using multi-criteria decision analysis with the help of the spatial analyst tool in ArcGIS. 

Assessment of flood affected area 

The land use and land cover map was developed to assess the impact of flooding on the areas land use and cover. In this research, the support vector machine supervised classification algorithm was employed to categorize the study region into four main types: agri-culture, urban areas, barren land, and water bodies. By reclassifying the LULC map into "water" and "non-water," classes, the flood-affected map was generated, the outputs, indicate that 11.88 km2 were affected by the flood. Additionally, the area of each class of land use that is affected by flood is calculated, and the results are presented as a statistical graph and tabula-tion (Fig. 6, 7, 8). The majority of the flood-affected area is located close to the Logar main river. It is evident from the map of the flood-affected land use and land cover that the agricultural area is severely affected by flooding; 10.36 km2 of agricultural land has been affected, resulting in loss of peoples lands and agricultural products. The majority of the affected agricultural area is situated on both sides of the Logar River. While 0.67 km2 of built-up area has been affected by flooding, resulting in the loss of both property and lives, 0.31 km2 of barren area has also been impacted by the floods. (Table 6) shows the area of each land use and land cover class that was impac-ted by the flood. As it was shown, flooding is a frequent occurrence in the floodplains close to the Logar River. As there are no significant water bodies or river channels close to the affected areas, the north-eastern section of the research area has only experi-enced minor flooding as a result of rainfall. The sec-tions on either side of the river that were inundated were primarily caused by the accumulation of excess water during the summer rainfall.

Table 6: Flood-affected areas within each LULC.

Fig. 6: Statistically analysis flood affected classes of land use and land cover.

Fig. 7: LULC of the study area into water and non-water classes.

Fig. 8: Flood affected classes of land use and land cover.

Flood Risk Assessment AHP 

The AHP technique was used for the weighted classi-fied flood-generating components, and a pair-wise comparison was performed on a 9-point significance scale for all six factors. Based on Saatys, (1980) sugg-estion, weighting methods were utilized to prioritize the relative importance of individual factors in a weighted overlay compared to other factors. The six parameters that are considered to have the most impact on the flood risk is elevation, slope, drainage density, NDVI, distance from a river, and land use and land cover. These criteria have already been explained in detail. We generated weighted maps for these parame-ters using ENVI, SNAP, and ArcGIS software, and the maps for six parameters are shown in (Fig. 9). Based on the results of the flood risk assessment map, the study area was divided into four zones of flood occur-rence: high risk, moderate risk, low risk, and very low risk (Fig. 10, 11). The possibility of flooding on both sides of the Logar River can be well seen. Flooding is more likely in areas near the river and at low eleva-tions. A significant portion of the studied area is under high and moderate risk, due to its proximity to the Logar River, so that the entire study area from Niazi Khel village to Shahghashi village on both sides of the river is under high risk of flooding; respectively, 15.10 km2 are under high risk and 179.50 km2 are under moderate flood risk. Furthermore, based on the Fig. 9, the outputs reveal, that the majority of the villages, which are closer to either side of the river are in a moderate flood zone.

Fig. 9: Flood vulnerability factors maps.

Fig. 10: Flood Risk Assessment Map.

Fig. 11: The graph shows the flood risk zones.

Consequently, based on the out-comes of this study, it will scientifically assist the managers and policy makers of the Office of State Minister for Disaster Management of Afghanistan, and other involved nati-onal and international organizations with a more com-prehensive analysis and clear instructions for creating early warning systems, emergency response processes, flood risk mitigation estimations, and suggesting where future development should be avoided or restricted.

Remote sensing has the benefit of synchronized and cost-effective data for monitoring flood impacts and risks, as well as their environmental effects on local, regional, and global scales. This research evaluated the impacts and risks of the flood that occurred on August 20, 2022, in the Logar province of Afghanistan. For this, the flood-affected area was calculated through the Sentinel-2 satellite images. The study area was classi-fied into four classes using support vector machine supervised classification algorithm pre- and post-flooding (built-up area, agriculture, barren area, and water). Then, the classified maps were reclassified into "water" and "non-water," classes, the flood-affected area was determined. Based on the results, 11.88 km2 of the study area was inundated. And also, according to land use and land cover maps, the outputs showed, that the flood-submerged areas of the different land use and land cover classes (agriculture, built-up areas, and barren) are respectively as follows: 10.37 km2, 0.67 km2, and 0.3 km2. The AHP approach was used to prepare the flood risk map. Therefore, the flood risk parameters or criteria were first identified, and then the AHP was performed. Then, through the weighted overlay analysis, a flood risk map was generated. From the generated flood risk map the study area was divided into four zones: high, moderate, low, and very low. Based on the outputs of the flood risk map, it was revealed that the Niazi Khel and Shahghashi villages on both sides of the river are at high risk of flooding. The outputs of risk map generated based on AHP algorithm also confirmed that, the highly flood affect-ted areas are generally located near the Logar river. Consequently, based on the above outputs, policy-makers in the Office of State Minister for Disaster Management of Afghanistan, and other involved nat-ional and international organizations should be con-cerned with a more comprehensive analysis and clear instructions for creating early warning systems, emer-gency response processes, flood risk mitigation esti-mations, and suggesting where future development should be avoided or restricted. 

I would like to thank our colleagues for their assis-tance and guidance in helping us successfully comp-lete this work. 

The authors declare that there is no conflict of interest to publish it.