FLOOD VULNERABILITY IN PUNJAB, PAKISTAN: A GEOSPATIAL ANALYSIS AND CARTOGRAPHIC APPROACH

Introduction

Flooding is one of the greatest dangers globally, which not only destroys infrastructure and damages the economy, but also claims human lives (Bates et al., 2008; Kundzewicz et al., 2014; Rentschler, Salhab and Jafino, 2022; Chen et al., 2024). Pakistan, especially the Punjab region, is affected frequently by floods, due to its high population density, agricultural activities, and the presence of five major rivers (Rahman et al., 2017; Waseem and Rana, 2023; Chen et al., 2024). Yearly flood events are exacerbated by climate change (Aldous et al., 2011; Arnell and Gosling, 2013; Kundzewicz et al., 2014; Youssef et al., 2021) thus demanding for effective flood assessments and effective mitigation strategies.

To address flooding globally, researchers have turned to flooding models in combining remote sensing data, Geographical Information System (GIS) techniques, and various flood drivers, to assess the flood challenge in a study area (Gigović et al., 2017; Hoque et al., 2019; Burayu, Karuppannan and Shuniye, 2023; Hossain and Mumu, 2024; Ibrahim et al., 2024; Roy and Dhar, 2024; Ullah et al., 2024). While remote sensing provides important data about topographic conditions, the GIS environment gives the opportunity to store and process this data. Depending on the study and the area, different parameters are chosen and their influence on the study’s model is determined.

The flood danger and the large number of people affected in the Punjab province was the motivation for this study. Although global research on flooding exists, only a few studies concentrate on Pakistan, and fewer on the highly affected Punjab region. Studies in Punjab mainly focus on the assessment of areas that are prone to floods and where flooding occurs, but the population is not taken into account. As flood events not only affect the landscape, but also communities, it is important to investigate flooding from a multi-dimensional perspective. Such a perspective is used in this master's thesis by developing a flood vulnerability framework. This framework consists of three components which are used to highlight areas most vulnerable to flooding. While one component considers physical and environmental factors, the others reflect the affected population. Another motivation lies in the mapping of the results. Recognizing that good visualizations, which effectively communicate the results, are missing in existing flood research, this study also focuses on creating visually appealing maps. Different mapping techniques are developed and evaluated in user testing to come to a visualization which works for different users.

All results are compiled into a printed atlas, making the outcomes tangible and accessible not only for informing a decision but also for the general or local population.

Objectives

The aims of the master’s thesis is to analyze the flood vulnerability in Punjab, Pakistan by developing a flood vulnerability assessment by integrating environmental, social, and coping capacity indicators, and using Copernicus satellite- and geospatial data and tools. The main focus of the thesis lies in the identification of the flood-prone and the most vulnerable areas. The study aims to enhance flood risk mapping and visualization to support decision-making and disaster risk management. To achieve this, a review is done on floods, risk, and relevant vulnerability indicators.

The results of the thesis will be presented through spatial overlay analyses and a cartographic product. In spatial analysis, the focus lies on identifying the flood-prone areas and determining the main factors of vulnerability. Through user testing, the usability of the different cartographic visualization techniques is evaluated to improve the communication of flood risk information. The final output of the thesis will be a printed Atlas, featuring maps of hazard and vulnerability areas, demographic impacts of floods, infographics, and textual commentary. A story map will serve as a digital supplement to the printed atlas, embedded with an electronic version of the printed atlas.

The objectives of this master’s thesis can be structured and divided as follows:

1. Geospatial Analysis and Mapping of Flood Vulnerability
2. Cartographic Design and User Testing
3. Cartographic Project of the Atlas

Study Area

The study area – Punjab, Pakistan – is one of the five provinces of Pakistan (Figure 1). According to the Pakistan Bureau of Statistic (PBS), almost 130 million people live in the region. Although it is not the largest region, it is the most populated one in Pakistan (PBS, 2023). The region spans an area of 205,345 km² and borders India on the eastern side. The PBS divides the province in their census into 36 districts, as well as 146 Tehsils; these are the administrative regions below the district level. The presence of the five major rivers – Indus, Jhelum, Chenab, Ravi as well as Sutlej – has made Punjab an agricultural center of the country, but prone to catastrophic flooding (Rahman et al., 2017).

Figure 1: Study area; (a) the province’s location in Pakistan; (b) study area Punjab.

Flood risk and vulnerability are a central concern in the province. According to Rentschler et al. (2022), Pakistan is among the top ten countries where the population is exposed to high flood risk. In particular, the Punjab region is in third place among the subnational administrative areas with the highest absolute number exposed to floods: 38% of the population lives in high-risk flood zones. 2022 was the severest flood since the 2010 flooding (Waseem and Rana, 2023), affecting 33 million people (WFP, 2024). Almost every three years Pakistan is hit by flood events; between 1950 and 2021 around 21 extreme flood events occurred in the country (Waseem and Rana, 2023). The monsoon season is from June to September and brings severe rainfall (Latif and He, 2025). In the last three years, the average monsoon rainfall was above average (PMD, 2024). This is also reflected in the flood severity, the last year’s floods, in 2024, caused 94 lives in Punjab, among them 46 children, while 238 got injured, including 86 children (Islamic Relief, 2024).

Methodolgy

The workflow of this study followed a structured approach, integrating literature review, expert consulting, spatial data analysis, and user testing to develop and visualize a flood vulnerability assessment (Figure 2). Based on literature review, freely available data, and meetings with experts in disaster risk, climatology, and meteorology, 13 parameters are defined for this study. These indicators are derived from various data sources. Each parameter is mapped, classified, and ranked into classes, ranging from 1, very low, to 5, very high (Hoque et al., 2019; Allafta and Opp, 2021; Ullah et al., 2024).

Figure 2: Workflow of the study.

The component maps are generated by a weighted overlay analysis based on multi-criteria decision analysis, using AHP and applying normalized weights to the indicators. The importance of each parameter is derived using a survey. Similar to other studies (Hoque et al., 2019; Allafta and Opp, 2021; Roy and Dhar, 2024; Ullah et al., 2024), a pairwise comparison matrix is created, relative scores are calculated, and the consistency of the matrix is determined.

After each map is generated, the flood vulnerability is assessed with different approaches, including formula-based results, as well as different cartographic ways. As the study region is very large, it is important to still be able to show regions that have a high vulnerability and to make this message not only available to researchers but also understandable to people. Therefore, the creation of accessible and legible information is important. To achieve that, data has to be generalized to present it in a way that shows high vulnerability areas in a static format. The vulnerability cartographic mapping approach is rated and evaluated with two user groups: the general public and climate risk analysts, identifying which mapping techniques are preferred.

The analysis, the parameter maps, the cartographic maps, as well as the results of the analysis, are compiled into an atlas. Furthermore, a digital product is created.

Datasets and Sources

Different datasets were used for the different parameters used in this study (Table 1). PERSIANN-CSS (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks - Cloud Classification System) data was used for the Annual Rainfall, downloaded via the CHRS data portal (CHRS, no date). Drainage Density, Elevation, Slope, and Topographic Wetness Index were obtained from the FABDEM (Forest And Buildings removed Copernicus 30m DEM). FABDEM is a product from the Copernicus GLO 30 Digital Elevation Model (DEM), delivering a resolution of 1 arc-second grid spacing (approximately 30m at the equator), whereas errors of buildings and vegetation were removed (Hawker et al., 2022). The WorldCover V2 2021 was used for the Land Use Land Cover (Zanaga et al., 2022). Health facilities were downloaded provided by the Humanitarian OpenStreetMap Team via the Humanitarian Data Exchange portal; the same applies to the river stream data. The census data of 2023 was accessed at the Pakistan Bureau of Statistics website (PBS, 2023), and downloaded as PDFs. Sentinel-1 SAR data was used for validating part of the model (Copernicus Sentinel data, 2021-2024), and retrieved via Google Earth Engine (https://earthengine.google.com/).

Table 1: Description of datasets, their sources and outputs.

Results

Geospatial Analysis and Mapping of Vulnerability

The FVI consists of three components: FPC, PSC, and CCC (Figure 3). While the FPC represents the physical flood hazard calculated over the whole area, the PSC shows the social vulnerability of populations, and the CCC, representing their ability to respond, is derived from the settlement location. First, the results of the three component maps will be analyzed and then the FVI. After each FVI parameter was mapped and ranked as previously described, and the importance of each parameter was determined through the AHP, the different component maps were generated with overlay analysis.

Figure 3: Flood Vulnerability Index and its components.

Flood-Prone Component

Using the parameters and the respective weights derived from the AHP, the overlay analysis generated the FPC Map. The output of the FPC ranged from 1 (very low flood-prone) to 5 (very high flood-prone); while 8.22% (16,889.62 km²) in low, 48.74% (100,266.5 km²) in moderate, 42.72% (87,882.86 km²) in high, and 0.32% (654.96 km²) is very high flood-prone areas in the study area in the FPC pixel map (Figure 4a).

Figure 4: (a) Flood-Prone Component Map, and (b) aggregated to Tehsils (The top 10 Tehsils with the highest flood-prone score are labelled).

The results indicate that almost half of the study area falls within the moderate flood-prone category, while a smaller portion of ~43% is classified as high and very high. This confirmed the high flood of catastrophic susceptibilities in the Punjab region. The high classes can especially be found near river basins, characterized by high drainage density, high proximity to rivers, flat terrain, and low slopes. While areas, characterized by high elevation, higher slope, and no river basins are classified as low, as well as moderate. Moderate areas are also especially found in flat terrain and lower slopes, However, the topographical characteristics of riverine areas in particular make them higher flood-prone areas. The FPC aggregated to Tehsils (Figure 4b) confirmed these findings. Especially areas in river basins are highly affected.

Population Susceptibility Component

The PSC Map was created with overlay analyses based on its parameters and the weights. The map ranged from 1 (very low population susceptibility) to 4 (high population susceptibility); while 1.89% (133.86 km²) in very low, 48.78% (3454.19 km²) is low, 47.91% (3392.97 km²) is moderate, and 1.42% (100.63 km²) in high population susceptibility in inhabited places in the PSC pixel map (Figure 5a). The PSC aggregated to Tehsils showed that the Tehsils Jalalpur Pirwala, Kot Radha Kishen, Chowk Sarwar Shaheed, Naushera, Rajanpur (Tribal Area), and Koh-e-Suleman are identified as the highest population susceptibility. These Tehsils are mostly located in rural areas, while areas with less population susceptibility are in urban areas. As the AHP evaluated the population density as the lowest influential factor (8%) it makes sense, that the rural areas achieved higher values, as they are characterized by a higher percentage of dependent and disabled population.

Figure 5: (a) Population Susceptibility Component Map, and (b) aggregated to Tehsils (The top 10 Tehsils with the highest flood-prone score are labelled).

Coping Capacity Component

The CCC map was generated with overlay analysis and the obtained weights The CCC map ranged from 1 (very low coping capacity) to 5 (very high coping capacity); while 26.64% (1886.33 km²) of the study area is very low, 32.14% (2275.78 km²) is low, 9.74% (689.99 km²) is moderate, 15.00% (1061.94 km²) in high, and 16.49% (1167.60 km²) in very high coping capacity in inhabited places (Figure 6). Urban areas are characterized by higher coping capacity, as cities have a higher presence of health facilities, and they were ranked as highest in the AHP (67%). Furthermore, the literacy rate is also higher in urban areas and to the north.

Figure 6: Coping Capacity Component Map, and (b) aggregated to Tehsils (The top 12 Tehsils with the highest flood-prone score are labelled).

Flood Vulnerability Index

The FVI was calculated by multiplying the PFC with the PSC and dividing by CCC. The FVI was classified using the Equal Interval method. The calculation of the FVI revealed that 13.28% (941 km²) of the study area is in very high and high vulnerability areas in inhabited places in the FVI pixel map. The FVI emphasizes that areas characterized by large FPC and large PSC values, as well as low CCC are highly vulnerable. The FVI aggregated to Tehsils revealed that Jalalpur Pirwala is the area with the highest flood vulnerability. This Tehsil has an average FPC of 3.89, an average PSC of 4.00, and an average CCC of 1.00, resulting in a raw FVI of around 15.57, which normalizes to 1.00, making it the most vulnerable. The Tehsils identified as the most vulnerable are Jalalpur Pirwala, Shorkot, Khairpur Tamewali, Bahawalnagar, Sadiqabad, Athara Hazari, Pakpattan, Jatoi, Chowk Sarwar Shaheed, and Liaqatpur. While Tehsils like Koh-e-Suleman (avg. PSC 4.00) have high population susceptibility and a lower coping capacity (avg. CCC 1.65), their PFC is lower (avg. PFC 2.51) than other areas, making it less vulnerable when computed by the formula. >In contrast, Tehsils with higher CCC averages, like Multan City, have a high average of CCC (4.68), but low PSC (1.00) and high PFC (3.63), however, vulnerability is still low due to the high CCC. This emphasizes that even urban areas are more prone to flooding overall, due to distance to rivers, but as the ability to cope is higher and the population is less exposed, they are less affected.

Geographic Design and User Testing

Different mapping approaches were visualized. The FVI is aggregated to Tehsils and interpolated. Furthermore, the FPC was mapped in the background, and PSC and CCC was visualized with pie charts, half-circles, and Wurman dots. The maps can be seen in Chapter 5.4 in the Thesis document. These maps were tested in the user testing.

Rating of the different mapping approaches

The users were asked to rate the different maps based on their level of spatial detail and their ease of understanding (Figure 7). The FVI map across Tehsils (a) clearly shows that users in both groups found this approach easy to understand, although it lacks spatial detail. The Kriging Map of FVI (b) shows a similar result; however, a higher detail was evaluated. This makes sense, as due to the interpolation technique approach the vulnerability does not stop at boundaries and has weird cuts between Tehsils. However, opinions are divided when it comes to the next maps, where the components are visualized separately. Users evaluate the FPC background with pie charts of people exposed across Tehsils (c) with a higher spatial detail as before, justified by many respondents that this gives more information it the three components and their influence on vulnerability. However, this means that simple understanding is lost for several participants, and given more time for understanding, the complexity of the map can be overcome. Similar things were said about the FPC background with half-circles of people exposed across Tehsils (d) map. Interestingly were the responses to the Kriging Map of FPC with Wurman Dots of people exposed (e). The spatial detail was rated high, however, for some analysts the Wurman dots were difficult to understand, as their legend title was misleading.

Figure 7: User responses of climate risk analyzers and general public; a) FVI across Tehsils, b) and interpolated, c) FPC background with pie charts of people exposed across Tehsils, d) FPC background with half-circles of people exposed across Tehsils, e) Kriging Map of FPC with Wurman Dots of people exposed; overlapping points exists.

Considering the average of the users’ responses, although the FVI map across Tehsils (a) was the easiest to understand, it was also the one with the lowest spatial detail (Figure 8). This map received a score of 37.00 for Level of Spatial Detail (LSP) and 83.70 for Ease of Understanding (EoU) by the climate risk analysts user group, 39.33 LSP, and 83.33 EoU by the general public (Table 2). The Kriging Map of FVI (b) was rated as having higher spatial detail but with a slightly less ability to understand: 62.05 LSP and 76.30 EoU by the climate risk analysts, and 59.83 LSP and 83.00 EoU by the general public; this visualization approach was therefore rated the best map in using the FVI formula, also highly rated overall. The FPC background with pie charts of people exposed across Tehsils (c) was rated better by the general public than the climate risk analysts. This makes sense, as analysts might understand the problem of the pie charts better and know that they cannot directly get an understanding of the underlying values. This visualization approach achieved 70.60 LSP and 45.75 EoU by the climate risk analysts, and 78.33 LSP and 55.42 EoU by the general public; indicating that this map provided the most information but was harder to interpret. The FPC background with half-circles of people exposed across Tehsils (d) was rated with 64.60 LSP and 56.95 EoU by the climate risk analysts, and 72.08 LSP and 47.50 EoU by the general public; this map was quite similar to (c) but better understood by analysts. The Kriging Map of FPC with Wurman Dots of people exposed (e) was rated with best values in the maps by visualizing the components separately: 69.50 LSP and 56.00 EoU by the climate risk analysts, and 77.00 LSP and 74.58 EoU by the general public.

Figure 8: Average user responses of all interviewees, climate risk analyzers, and of the general public; (a) FVI across Tehsils, (b) and interpolated, c) FPC background with pie charts of people exposed across Tehsils, (d) FPC background with half-circles of people exposed across Tehsils, (e) Kriging Map of FPC with Wurman Dots of people exposed.

Given the rating, the Kriging Map of FVI (b) achieved the highest score of 140.03 in total (summing both dimensions together) in combining the user groups, with 61.22 LSP and 78.81 EOU. Followed by the Kriging Map of FPC with Wurman Dots of people exposed (e), with a value of 135.28, with 72.31 LSP and 62.97 EoU. As Map (b) was highlighted to understand it easier and that it makes more sense and reflects the vulnerability compared to the aggregation in Tehsils, this map should be given when using the formula. This statement is supported by different participants, who stated that this map would be good to show to the public; also, some people stated, that this map is also good for decision making, to get a quick overview there exactly in the Tehsils help are needed. As some people mentioned the interpolation resolution, different resolutions should be given, to give the user different views for Tehsil, or settlement scale views. Although map (e) did not achieve the most scores, it was still the best-rated map in separately visualizing the vulnerabilities components and receiving feedback on how to improve it. Several participants rated this map as a lack of understanding as the legend title of the Wurman Dots was misleading; would this explanation had been given on the map, or the title better, they might have rated it better. The resolution could also be considered here. Furthermore, the feedback, adding the number of people in the Wurman dots, might draw a better picture to get more insight into vulnerability. A bivariate mapping method could be used: the size shows the risk values (calculated by dividing PSC by CCC), and the color intensity of the dots tells something about how many people live there. This would give a good overview of the final flood vulnerability while being easy to understand as well as giving also information on the people, as well as their level of vulnerability. Also changing the color background slightly, making it not so prominent, to increase the readability. Given that the interpolation makes consumption, for the atlas maps with finer hexagon resolution of should be provided, and the above-mentioned feedback. With that, the user has the possibility to get a broader overview of vulnerability in the province and can look more specifically at different maps capturing a smaller scale, and a higher resolution in a particular area . Overall, general feedback from the risk analyzers came that they really appreciated the different approaches to mapping, as they are usually confronted with GIS Outputs.

Table 2: Mean values of user testing.

In summary, through the user feedback, the following changes were made:

• Changed the color scheme of the FPC.
• Improved the description to enhance understanding of the methodology.
• Corrected legend titles that were previously misleading.
• Adjusted symbology colors to improve readability and clarity.
• Provided different resolutions of interpolation maps.
• Added inset maps to the atlas to zoom into key areas.
• Enhanced the Wurman dots map by including additional information.

Cartographic Project of the Atlas

Atlas

According to the layout structure presented in Chapter 6.1 (see the Thesis document) the atlas was designed. The parameters are visualized covering two pages (Figure 9) This gives the possibility to present the spatial distribution of the classes in one map, while a figure shows the covered area. Another map shows the distribution of the values. The mapping approaches are mainly visualized on two pages (Figure 10). Providing more space for explanation and for inset maps.

Figure 9: Sample image of two pages of the atlas; here the AR parameter.

Figure 10: Sample image of two pages of the atlas; here FVI drawn with FPC in the background and PSC and CCC as half-circles.

The improved maps, based on the feedback from the user testing, can be seen in the atlas at https://gernotnikolaus.github.io/MasterThesis_FloodVulnerabilityPunjab.

Digital Product

The digital product was developed in leaflet and hosted on a GitHub page https://gernotnikolaus.github.io/FVI_Punjab. This interactive dashboard allows the user to explore the layers of FVI and its three different components, aggregated to the Tehsil level (Figure 11). Explanation text on the left side gives a summary about the study analysis. While the different layers can be switched on or off, the user can zoom to the specific Tehsil and investigate the different values of the analysis by hovering over the respective administrative boundary.

Figure 11: Screenshot of the interactive dashboard.

Contact

• Twitter
• Facebook
• Instagram
• GitHub

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';