Introduction

Flash floods remain one of the most destructive natural disasters worldwide, causing substantial casualties, property losses, and long-term impacts on ecosystems and human living environments1,2. Recognizing their significance, the United Nations included rainstorm-induced floods in the priority list during the International Decade for Natural Disaster Reduction3. Between 1970 and 2019, floods accounted for approximately 44% of all natural disasters and contributed to 31% of global economic losses4. The occurrence of such floods is rarely due to a single factor, instead resulting from the complex interplay of environmental and societal drivers, including climate patterns, regional geography, population density, and urban expansion5,6,7,8. With ongoing climate change and rapid urbanization, both the scale and unpredictability of rainstorm floods have increased significantly in recent years9. Notable incidents, such as the 2021 Zhengzhou floods in China, which claimed over 300 lives, and the 2022 flash floods in the United Arab Emirates, highlight the urgent need for systematic flood risk assessment to mitigate impacts and enhance resilience10,11. In this context, systematic approaches and adaptive strategies are essential to improve flood resilience and reduce future risks12,13.

The CPEC spans diverse terrains, supporting major infrastructure, energy projects, and urban development. It cuts through densely populated, economically active lowland regions, which act as key hubs for growth, as well as ecologically sensitive and climate-vulnerable areas, exposing both communities and ecosystems to flood hazards14,15,16. In the high-altitude Himalayas, steep slopes, glacial-fed rivers, and complex weather patterns make floods from extreme rainfall and glacial melt particularly probable17,18,19,20. Before 2000, relatively stable monsoon patterns facilitated water management and economic development21. However, since the early twenty-first century, climate variability has intensified, leading to more frequent extreme rainfall events, fluctuating precipitation, and increased flood risk22,23. The impacts are uneven: some regions face frequent flooding, while others suffer from drought, reflecting the spatial heterogeneity of rainfall24. Combined with the socio-economic fragility of downstream populations and infrastructure, these natural vulnerabilities amplify overall flood risk along the corridor25.

Recent research on CPEC flood risk has focused on local hydrodynamic modeling and indicator-based assessments that integrate socio-economic and environmental factors26,27,28. Other studies have emphasized the rising frequency of extreme rainfall and its consequences for downstream communities29. Nevertheless, most analyses remain region-specific or cross-sectional, lacking comprehensive, multi-indicator, and temporally dynamic evaluations of flood risk30. This limitation makes it difficult to identify flood hotspots accurately and hinders the development of targeted disaster management strategies. To address these gaps, this study develops a dynamic, multi-indicator framework that combines hazard, sensitivity, vulnerability, and coping capacity (HSVC) with an AHP–Entropy weighting method. The framework evaluates flood risk across multiple historical periods, capturing changes over time while providing spatially explicit insights for disaster management and infrastructure planning. This approach allows a comprehensive assessment of rainfall-induced flood risk along the CPEC corridor and clarifies how risk evolves across space and time, offering a practical tool for informed decision-making.

Materials and methods

Study area

The CPEC region is in the central part of the Eurasian continent and is one of the main corridors of the Belt and Road Initiative31. The corridor extends from the Kashgar region in China, running southward to the Gwadar Port in Pakistan, covering a total length of approximately 3,000 km and an area of about 104.4 × 104 km2. The region has a complex topography, with a general trend of higher elevations in the west and north, and lower elevations in the east and south, leading to significant topographic differences between the northern and southern areas (Fig. 1). In the northwest of the study area, three world-renowned mountain ranges—the Himalayas, the Karakoram, and the Hindu Kush converge. The central region is dominated by the Indian River Plain, Pakistan’s most important agricultural area, while the southern part borders the Arabian Sea32. The northern mountainous areas of the CPEC are characterized by a cold mountain climate, whereas the central and southern regions are primarily influenced by tropical monsoon and tropical desert climates33.The average annual minimum and maximum temperatures are – 7 °C and 28 °C, respectively34. Annual precipitation in the study area ranges from 100 to 3000 mm, with uneven spatial distribution and significant interannual variation23. Spatially, precipitation tends to increase and then decrease from north to south35. Rainfall follows a clear seasonal pattern, with the majority concentrated between July and September. Flood events are most frequent during the monsoon season.

Fig. 1
figure 1

Location of the CPEC.

Full size image

Indicator selection and data sources

The selection of appropriate assessment indicators is a crucial step in flood risk evaluation, as it directly influences the results36,37. Considering the extreme meteorological events, topography, and socio-economic conditions of the CPEC region, this study selected 13 relevant indicators across four dimensions: hazard, sensitivity, vulnerability, and coping capacity. Table 1 summarizes the attributes and data sources for each indicator.

Table 1 Data sources and details.
Full size table

Hazard (H) quantifies the likelihood of flooding, primarily driven by extreme rainfall events38,39. In this study, three precipitation-based indicators were derived from the ERA5-Land dataset (1979–2024) to capture different facets of flood-triggering rainfall: maximum rainfall over three consecutive days (Rx3day) for short-term intense events, maximum daily rainfall (Rx1day) for single-day extremes, and the number of consecutive wet days (CWD) for prolonged precipitation periods40,41. These indicators collectively reflect rainfall intensity, duration, and accumulation, providing a robust basis for assessing flood hazard potential along the CPEC corridor. In addition, historical flood events triggered by extreme rainfall in the CPEC region from 1985 to 2024 were plotted as reference points, providing a visual validation of the spatial distribution of flood hazards and thereby enhancing the reliability and contextual interpretation of the flood hazard assessment.

Sensitivity (S) reflects a region’s susceptibility to flood impacts due to its physical and infrastructural characteristics42. This study selected four sensitivity indicators: elevation, slope, drainage density (DD), and road density (RD). These indicators represent geographic and physical conditions that influence flood propagation and water accumulation. They remain relatively stable over time, making them suitable as static variables. Elevation and slope strongly affect the flow and accumulation of floodwater43. River network density represents potential water pathways, while road density indicates infrastructure sensitivity that can affect flood dynamics44,45,46,47. Together, these indicators provide a comprehensive measure of inherent environmental susceptibility to flooding, independent of social or demographic factors.

Vulnerability (V) assesses the susceptibility of communities and individuals to flood threats48,49. Four indicators were selected to capture both demographic and infrastructural aspects of social vulnerability. Population density (PD) and the ratio of vulnerable populations (VPR) were included because areas with higher concentrations of people, especially socially vulnerable groups, are more likely to experience severe flood impacts50. VPR is defined as the percentage of children under 14 and elderly over 65 in the total population. It was calculated for 2015 and estimated for other years using a linear function derived from historical population data in the GPWv4 dataset, enabling reconstruction of temporal changes. In addition, the urbanization index (UI) and urban built-up area (UB) were selected to reflect the concentration of infrastructure and assets, which influences the potential scale of flood damage51,52.

Coping capacity (C) measures a community’s ability to respond, adapt, and recover when exposed to flooding. Key indicators include regional GDP levels and the major crop production (MCP), which together reflect economic resources and food security53,54. Regions with higher GDP generally have better infrastructure, emergency services, and financial resources to support timely disaster response and recovery. In addition, stable agricultural output ensures food availability, reduces social stress after floods, and helps communities maintain resilience and stability33,55,56. By integrating these economic and agricultural indicators, coping capacity captures a society’s ability to withstand flood impacts and accelerate post-disaster recovery.

Data processing

Data preprocessing

Before data standardization, all datasets were resampled to a spatial resolution of 0.1° raster data using the bilinear interpolation method. After resampling, the indicators were projected into a consistent coordinate system. Normalization scales each indicator to a range between 0 and 1, where higher values represent a stronger influence on flood risk. This approach allows indicators with different units and magnitudes to be analyzed together. The normalization is calculated as Eq. (1)

$$X_{i}{\prime} = (X_{i} - X_{\min } )/(X_{\max } - X_{\min } )$$
(1)

where \(X_{i}{\prime}\) is the normalized indicator, \(X_{i}\) is the original value, \(X_{\min }\) and \(X_{\max }\) are the minimum and maximum values of the indicator.

Analytic hierarchy process

The Analytic Hierarchy Process (AHP) was used to calculate the subjective weights of factors (\(X_{i}\)) and indicators (\(X_{j}\))57. Construct a pairwise comparison matrix \(Y = \left[ {y_{ij} } \right]_{n \times n}\) using a 1–9 scale to reflect the relative importance between elements. Ensure the reciprocal property \(y_{ji} = 1/y_{ij}\). Compute the maximum eigenvalue \(\lambda_{\max }\) of the matrix. Check consistency using the consistency ratio (\(CR\)):

$$CI = \frac{{\lambda_{\max } - n}}{n - 1}\;，CR = \frac{CI}{{RI}}$$
(2)

where \(RI\) is the random consistency index. Accept the matrix if \(CR \le 0.1\); otherwise, adjust the comparisons. The calculated \(CR\) values for each indicator and the overall factor layer are summarized in Table 2, showing all matrices meet the acceptable consistency threshold.

Table 2 CR values for factor and indicator layers in the CPEC rainstorm flood risk assessment.
Full size table

Normalize the eigenvector corresponding to \(\lambda_{\max }\) to obtain the weights of factors and indicators:

$$W_{{X_{i} }}^{AHP} = \frac{{e_{i}^{F} }}{{\sum\nolimits_{k = 1}^{n} {e_{k}^{F} } }},W_{{X_{j} }}^{AHP} = \frac{{e_{j}^{I} }}{{\sum\nolimits_{k = 1}^{n} {e_{k}^{I} } }}$$
(3)

where \(e_{i}^{F}\) and \(e_{j}^{I}\) are the elements of the eigenvectors for the factor and indicator layers, respectively.

Entropy weight method

The entropy weight method (EWM) objectively determines indicator weights based on information entropy58. Indicators with lower entropy provide more information and receive higher weights, complementing the more subjective AHP approach.

Assuming there are \(m\) spatial units and \(n\) indicators, the indicators are arranged into a matrix \(X = \left[ {x_{ij} } \right]_{m \times n}\), where \(x_{ij}\) represents the value of indicator \(j\) in spatial unit \(i\). The normalized feature ratio of indicator \(j\) in unit \(i\) is calculated as:

$$p_{ij} = \frac{{x_{ij} }}{{\sum\nolimits_{i = 1}^{m} {x_{ij} } }}$$
(4)

The information entropy of each indicator \(j\) or factor \(i\) is calculated by:

$$H_{j} = - k\sum\limits_{i = 1}^{n} {p_{ij} } \ln p_{ij} ,k = - \frac{1}{\ln m}$$
(5)

If \(p_{ij} = 0\), set \(p_{ij} \ln p_{ij} = 0\). The entropy weight of indicator \(j\) or factor \(i\) is then:

$$W_{{X_{i} }}^{EWM} = \frac{{H_{i} }}{{\sum\nolimits_{i = 1}^{n} {H_{i} } }},W_{{X_{j} }}^{EWM} = \frac{{H_{j} }}{{\sum\nolimits_{j = 1}^{n} {H_{j} } }}$$
(6)

where \(H_{j}\) and \(H_{i}\) are the information entropies of indicators \(j\) and factors \(i\), respectively.

Combined weights

The distance function is a method used to combine different weighting results by minimizing the discrepancy between them. In this study, it is employed to integrate the subjective AHP weights (\(W^{AHP}\)) and the objective entropy weights (\(W^{EWM}\)), achieving a balanced combined weight5. The combined weight \(W^{comb}\) is determined as:

$$W^{comb} = a \cdot W^{AHP} + b \cdot W^{EWM}$$
(7)

where \(a\) and \(b\) are the weight distribution coefficients, constrained by \(a + b = 1\). The distance function \(d\) between the subjective and objective weights is defined as:

$$d\left( {W^{AHP} ,W^{EWM} } \right) = \left[ {\frac{1}{2}\sum\limits_{i = 1}^{n} {\left( {W^{AHP} - W^{EWM} } \right)^{2} } } \right]^{\frac{1}{2}} = \left( {a - b} \right)^{2}$$
(8)

By solving this equation for \(a\) and \(b\), the distribution coefficients for each weight are obtained. Substituting these coefficients into the equation for \(W^{comb}\) produces the combined weights for the factor and indicator layers, as shown in Table 3.

Table 3 Combined weights of factor and indicator layers.
Full size table

Rainfall flood risk assessment framework

Flood disaster is regarded as a complex system covering multiple aspects such as natural, social, and economic. Disaster risk assessment is a systematic process to calculate the level of risk of a disaster to a community by quantitatively analyzing the hazard, potential damage, and coping capacity of the disaster. Based on the frameworks proposed by IPCC and the INFORM Index60,61. Integrating additional relevant factors, this study defines rainstorm flood risk as a composite function of hazard, sensitivity, vulnerability, and coping capacity61,62. It is expressed as Eq. (9) and (10):

$$R = H \cdot d_{H} \cdot W_{H} + S \cdot d_{S} \cdot W_{S} + V \cdot d_{V} \cdot W_{V} + C \cdot d_{C} \cdot W_{C}$$
(9)
$$\begin{gathered} d_{i} = \left\{ {\begin{array}{*{20}c} {1,\;if\; the\; indicator\; increases \;flood \;risk\; \left( {positive\; indicator} \right)} \\ { - 1,if \;the\; indicator\; decreases\; flood\; risk \left( {negative \;indicator} \right)} \\ \end{array} } \right. \hfill \\ i = H,S,V,C \hfill \\ \end{gathered}$$
(10)

where \(R\) represents the composite flood risk index; \(H\),\(S\),\(V\) and \(C\) denote the hazard, sensitivity, vulnerability, and coping capacity indices, respectively. \(W_{H}\) ,\(W_{S}\),\(W_{V}\) and \(W_{C}\) are the corresponding factor weights. The direction coefficient \(d_{i}\) is introduced to standardize all indicators in a unified negative-direction framework. In this formulation, positive indicators increase the flood risk index, while negative indicators, such as DEM, slope, and other mitigating factors, decrease it.

This study divides the timeline into two stages, 1979–1999 (P1) and 2000–2024 (P2), to investigate the temporal evolution of flood risk. Sensitivity indicators, based on relatively stable geographic and geological data, remain unchanged, while hazard, vulnerability, and coping capacity factors are reconstructed for each period using long-term averages or data from specific years, depending on availability. All indicators are classified into five levels: low, middle-low, middle, middle-high, and high, using the natural breaks (Jenks) method. The workflow of the rainfall flood risk assessment is illustrated in Fig. 2, and the spatial distribution of the indicators except hazard is shown in Fig. 3. Comparing risk levels between P1 and P2 reveals temporal changes and shifting spatial patterns, offering a quantitative foundation for managing and mitigating flood risk in the CPEC region. The maps presented in Figures were generated using ArcGIS software (version 10.8.2) through spatial overlay analysis, which was applied to integrate hazard, sensitivity, vulnerability, and coping capacity layers to ensure consistent spatial resolution and accurate representation of flood risk across the corridor.

Fig. 2
figure 2

Framework for rainstorm flood disaster risk assessment based on the AHP–Entropy method.

Full size image
Fig. 3
figure 3

Spatial distribution of non-precipitation indicators. Panels (a–d) show sensitivity, (e–h) vulnerability, and (i–j) coping capacity across different periods.

Full size image

Results

Spatial and temporal analysis of rainfall events in CPEC

Analysis of long-term data from 1979 to 2024 reveals significant spatial and temporal variations in extreme rainfall along the CPEC corridor. Overall, central plains and southeastern lowlands experienced pronounced increases, whereas some northern and northwestern regions showed declining trends (Fig. 4). High-risk zones are mainly concentrated in central and southern sections, forming a continuous belt, while low- and mid-low risk areas are largely situated in northern mountainous and transitional regions. Comparing the P1 and P2, low- and mid-low risk areas remained relatively stable, whereas middle- and high-risk zones expanded, indicating a gradual increase in overall flood risk along the corridor.

Fig. 4
figure 4

Spatial distribution of trends and interannual change rates of hazard indicators Rx3day, Rx1day, and CWD during 1979–2024.

Full size image

Flood risk along the CPEC corridor is primarily shaped by extreme rainfall, which has the strongest impact on both the intensity and spatial distribution of floods. Sensitivity, vulnerability, and coping capacity also contribute, modulating the risk pattern. Sensitivity reflects the influence of terrain, river network density, and transport infrastructure on flood occurrence, while vulnerability amplifies potential losses in areas with high population density, intensive land use, or limited infrastructure. Coping capacity, summarizing regional economic strength and emergency response potential, partially mitigates the impacts of floods. Simulation results closely match historical flood records, confirming the reliability of the assessment. These findings provide a solid scientific basis for targeted flood mitigation, infrastructure planning, and adaptive strategies in middle- and high-risk regions along the CPEC corridor.

Analysis of rainstorm flooding factors

Hazard

The spatial distribution of flood hazard along the CPEC during P1 and P2 is shown in Fig. 5. In P1, low-risk areas accounted for 50.45%, mid-low risk for 26.22%, middle risk for 12.36%, mid-high risk for 7.68%, and high risk for 3.29%. By P2, low-risk areas slightly decreased to 43.11%, mid-low risk increased to 30.89%, middle risk rose to 14.76%, mid-high risk slightly declined to 7.31%, and high-risk zones increased modestly to 3.93%. High-risk areas were mainly concentrated in northern Punjab, central KP (Khyber Pakhtunkhwa Province), southern Azad Jammu and Kashmir (AJK), and the Federal Capital Territory(FCT), with several mid-high-risk patches in southeastern Sindh. In contrast, Kashgar and parts of Balochistan, which receive relatively little rainfall, largely remained low risk. Overall, central CPEC formed the main cluster of extreme precipitation hazard. The spatial variation of extreme rainfall further intensified risk differentiation, with central and eastern regions showing an expansion of high-risk zones, while western and northern areas remained relatively stable. Comparing the two periods, high-risk zones persisted around the FCT, though the most extreme areas slightly contracted as middle-risk zones expanded. Hazard levels in northern CPEC decreased further, while southeastern Sindh experienced a slight reduction. Overall, the coverage of middle-to-high risk areas increased from 23.33 to 25.99%, representing a rise of about 2.7 percentage points, reflecting the expansion of regions exposed to stronger extreme rainfall.

Fig. 5
figure 5

Hazard maps of the CPEC region for P1 and P2.

Full size image

Sensitivity

Figure 6 illustrates a clear north–south gradient in flood sensitivity across the CPEC region, with overall sensitivity increasing toward the southern lowlands. Approximately 63% of the study area falls into the mid–high to high sensitivity categories, which are mainly distributed in basins and alluvial plains such as the Kashgar Basin, the Punjab Plain, and the river valleys of Sindh. In contrast, low-sensitivity zones are primarily located in the northern mountainous areas, particularly in the high-altitude terrains around Kashgar. The spatial pattern of sensitivity is strongly shaped by topography and human activities: in high-relief areas with steep slopes, rainfall-induced runoff can be discharged rapidly, resulting in lower sensitivity, whereas in low-lying plains with high drainage density and extensive transportation and agricultural infrastructure, extreme precipitation often leads to runoff retention and waterlogging, thereby amplifying flood risk. Overall, the sensitivity distribution corresponds closely to the underlying geographical setting: higher elevations and steeper slopes generally reduce sensitivity, while flatter landscapes with high drainage density and concentrated road networks experience significantly greater impacts under heavy rainfall events.

Fig. 6
figure 6

Sensitivity map of the CPEC region.

Full size image

Vulnerability

Figure 7 illustrates the spatiotemporal patterns of vulnerability across the CPEC region. During P1, middle vulnerability dominated the landscape, accounting for 67.47% of the total area, followed by mid-low levels at 21.52%. Mid-high and high vulnerability zones were comparatively limited, representing about 7.04% and 1.14% respectively, and were mainly concentrated along the Indus River, with pronounced clusters in southeastern Punjab and the FCT, while the Kashgar region remained at consistently low levels. In P2, however, a structural shift was evident: the share of middle vulnerability declined to approximately 60.32%, whereas mid-high levels increased to 20.44%, suggesting a clear trend toward greater overall vulnerability. Spatially, dispersed hotspots in the middle–lower Indus Plain during P1 evolved into broader and more continuous clusters in P2.

Fig. 7
figure 7

Vulnerability map of the CPEC region for P1 and P2.

Full size image

This upward shift in vulnerability can be closely linked to changes in population distribution, urban expansion, and land use intensification. Specifically, the vulnerable population ratio rose significantly from P1 to P2, indicating that a larger proportion of residents now belong to socially vulnerable groups exposed to flooding. This demographic vulnerability was particularly marked in rapidly growing urban centers along the Indus corridor. Similarly, the urbanization index increased substantially, reflecting both the outward expansion of existing cities and the emergence of new peri-urban settlements, which tend to concentrate socially and economically fragile populations in flood-prone plains. Moreover, the built-up area expanded considerably during P2, especially in southeastern Punjab and Sindh, where agricultural land was converted into residential and industrial use. The increase in impervious surfaces has not only amplified runoff generation under extreme rainfall but also heightened the vulnerability of local communities and assets to potential flood impacts. Overall, the transition from P1 to P2 highlights that while natural sensitivity factors remain relatively stable, socio-economic processes, including population growth, accelerated urbanization, and expansion of built-up land, are the dominant drivers behind the rising vulnerability of the CPEC region.

Coping capacity

Figure 8 presents spatial and temporal variations in flood coping capacity across the study area. In the P1, areas with low, mid-low, middle, mid-high, and high coping capacity accounted for 53.71%, 15.66%, 15.97%, 8.22%, and 6.43%, respectively. By P2, the proportion of low-capacity areas declined to 48.82%, while mid-low and middle categories rose to 18.43% and 16.81%, and high-capacity areas expanded to 7.25%, indicating an overall contraction of well-equipped regions alongside a gradual increase in less-prepared zones.

Fig. 8
figure 8

Coping capacity maps of the CPEC region for P1 and P2.

Full size image

Spatially, regions exhibiting stronger coping capacity tend to coincide with economically developed areas and zones with concentrated major crop production, particularly the Punjab Plain and northeastern KP, which together account for over 70% of the total high-capacity area. During the second period, these regions maintained their relative advantage, although the distribution became more spatially focused. Meanwhile, low-capacity zones extended into remote mountainous areas and less-developed parts of the plains, producing a distinct pattern of spatial disparity intertwined with temporal change. The distribution of coping capacity clearly reflects the influence of regional GDP and major crop production, highlighting the persistent imbalance between well-resourced and vulnerable areas and emphasizing the challenge for future flood management and resilience building.

Rainfall flood disaster risk assessment

During P1, the flood risk along the CPEC corridor was classified into five categories: low, mid-low, middle, mid-high, and high, accounting for 14.57%, 30.36%, 39.12%, 12.25%, and 3.70% of the total area, respectively. By P2, the proportion of low-risk areas slightly increased to 15.11%, mid-low and middle risk areas decreased to 29.44% and 38.18%, mid-high risk decreased to 11.65%, while high-risk zones rose to 5.62%, indicating a notable expansion of the most vulnerable regions and relative stability of low-risk areas. Provincial-level analysis revealed a clear spatial redistribution of flood risk between P1 and P2. High-risk zones formed a continuous belt across KP, Punjab, the plains, and AJK, characterized by high hazard sensitivity, dense populations, intensive agricultural activities, and limited coping capacity. Southeastern Sindh also showed an upward trend in high-risk classification, primarily associated with inadequate drainage infrastructure and concentrated demographic and agricultural pressures. Historical flood records corroborated the modeled patterns, showing 17 major flood events along the corridor during P1, which increased sharply to 45 by P2, mostly concentrated in high-risk provinces such as KP, Sindh, and Punjab, supporting the observed expansion and intensification of flood hazards (Fig. 9).

Fig. 9
figure 9

Rainfall flood risk maps of the CPEC region for P1 and P2, with historical rainstorm flood events.

Full size image

Temporal comparison between P1 and P2 indicates that extreme rainfall is the primary driver of floods along the corridor. Spatially, hazard distribution is the most direct indicator of flood occurrence, with high-risk areas concentrated in KP, Punjab, the plains, and AJK, while low-risk areas remained largely stable. Sensitivity significantly influenced the risk pattern: flat terrains with dense river networks exhibited higher sensitivity, facilitating flood concentration, whereas high-altitude and steep-slope regions had lower sensitivity and correspondingly lower risk. Vulnerability further modulated risk, with densely populated, rapidly urbanizing, and agriculturally intensive regions showing elevated flood risk, while sparsely populated areas with limited economic activity exhibited lower vulnerability and relatively stable risk levels. Differences in coping capacity also affected flood outcomes, as high-risk zones often faced inadequate infrastructure and limited defense measures, amplifying flood impacts. Mid-low and midlle risk areas were primarily distributed along two belt-like regions: one covering the highlands of Gilgit-Baltistan and KP extending to the Kashgar area, and another stretching from the Punjab plains to Sindh, where flood risk remained middle and relatively stable over time (Fig. 10).

Fig. 10
figure 10

Changes in flood risk levels in different regions of the CPEC for P1 and P2.

Full size image

Overall, the spatial distribution of flood risk is most directly reflected by hazard patterns, while sensitivity, vulnerability, and coping capacity collectively shape the heterogeneity and temporal evolution of risk. These findings provide critical insights for disaster prevention, mitigation strategies, and infrastructure planning in high-risk areas, while highlighting that stable vulnerability and coping capacity help maintain consistent risk levels in regions with middle hazard exposure.

Discussion

Comparison between historical floods and assessed risk

Pakistan has repeatedly faced severe flood disasters, with the 2010 and 2022 events among the most destructive in recent decades. The 2010 floods submerged large areas of Punjab and Sindh as torrential rains combined with surges from the Indus River, overwhelming drainage systems and urban infrastructure63,6464,65. In 2022, rainfall reached up to three times the climatological average, with local peaks seven to eight times higher, triggering extensive flooding across Sindh, Punjab, and KP65. Prolonged storms, extreme climate events, and inadequate drainage compounded the risks, while rapid population growth and urban expansion further reduced local resilience 29.

The spatial distribution of historical floods closely corresponds with the high-risk zones identified in this study. Flood records from 1985 to 2024 indicate concentrated events in Punjab, Sindh, and KP, aligning with areas persistently classified as high-risk in both P1 and P2. Notably, high-risk zones in Sindh and KP expanded during P2, consistent with the observed increase in flood frequency. Severe floods also occurred in regions categorized as mid-low or middle risk, such as Gilgit-Baltistan and parts of Punjab, indicating that lower-risk areas are not immune to extreme rainfall events, since localized heavy rainfall can still trigger substantial flooding. Importantly, the flood hazard evaluation in this study is based on an integrated approach that considers multiple factors, including extreme rainfall, local vulnerability, topography, and adaptive capacity, rather than relying solely on meteorological data. This comprehensive assessment provides a more realistic representation of flood risk and offers a nuanced context when comparing modeled high-risk zones with the actual occurrence of historical floods, highlighting both consistencies and divergences between predictions and observed events.

Limitations and future directions

Comprehensive flood risk assessments that integrate both spatial and temporal perspectives are still rare in the CPEC region. Our study makes an early attempt to bridge this gap by evaluating storm flood risks over multiple decades with a combined AHP–Entropy approach. While this method is not entirely novel, it provides a balanced way to integrate structured criteria with data-driven evidence. In practice, AHP offers a framework for organizing complex factors, whereas the Entropy method reduces reliance on subjective judgment by drawing on the inherent variation within datasets28,66. Even so, the approach is not without weaknesses. The AHP component may introduce bias if the pairwise comparisons are poorly structured, and the Entropy method sometimes amplifies the influence of highly variable indicators, regardless of their practical importance. Finding a workable balance between these two techniques remains an open challenge, particularly in data-scarce regions such as western Pakistan.

A second set of challenges arises from the datasets themselves. Rainfall records, socioeconomic data, and infrastructure statistics often differ in temporal resolution. For example, national census and GDP figures are typically available at multi-year intervals, while daily hydrological observations may exist only for certain river basins. This mismatch complicates the analysis of evolving risks and can obscure short-term dynamics, such as the rapid urban growth observed in Punjab after 2010 or the localized impacts of the 2022 floods in Sindh. Furthermore, several indicators used in this study, such as elevation, road density, or long-term GDP averages, are essentially static, which may underrepresent local variations in vulnerability and coping capacity.

Looking forward, three practical steps could make future flood risk assessments more reliable. First, routine land-use monitoring through Sentinel-2 or Landsat imagery could capture seasonal changes in cropland and urban expansion along the Indus basin. Second, real-time measurements of river levels and drainage performance, when linked with regularly updated socioeconomic data, would better represent the dynamics of hazard and vulnerability. Third, structured community surveys could help reveal infrastructure gaps and coping strategies that are difficult to capture with remote sensing alone. By combining these measures, risk assessments would not only improve in technical accuracy but also in their relevance for disaster preparedness along the CPEC corridor.

Finally, methodological refinements could enhance the weighting framework itself. Sensitivity testing, adaptive weighting schemes, or even machine learning–assisted scoring of indicators could reduce subjectivity and improve robustness67,68,69,70,7168,69,70,71,72. Such improvements, paired with better data and locally grounded evidence, would strengthen the credibility and practical value of future flood risk assessments in this strategically significant region.

Flood disaster prevention measures for rainfall in CPEC

Flood prevention along the CPEC must adapt to local geographic and climatic realities rather than rely on a one-size-fits-all solution. As noted in integrated flood risk management literature, successful strategies should be context-specific and multi-layered, accounting for regional differences in exposure, infrastructure, and resilience goals72.

In the FCT and surrounding urban areas, the biggest issue is stormwater congestion. A more resilient drainage network is required, linking sewers, streets, and open channels into one system. Rather than relying on vague routine checks, drainage facilities should be inspected and cleared specifically before and after the monsoon season to keep them functioning effectively73. Expanding green spaces and introducing rain gardens would also help absorb runoff naturally, easing the burden on pipes56,74 Punjab’s vulnerability is tied to agriculture. Here, flood storage ponds and small-scale dams can regulate water flow in crop-growing region75. At the farm level, simple practices like rotating crops or planting cover crops improve soil’s ability to hold water and reduce erosion76. In Sindh, where river flooding is most severe, embankments need constant reinforcement. Rather than one-off projects, a routine program of strengthening and repairing embankments before heavy rain seasons could make them more reliable77. Beyond infrastructure, community preparedness matters as well. Local governments could organize drills, distribute flood response manuals in local languages, and set up neighborhood-level volunteer teams78. Taken together, these practical, region-specific steps can build a stronger flood defense system for the CPEC, protecting both people’s livelihoods and the long-term sustainability of the corridor.

Conclusions

This study assessed rainstorm flood risk along the CPEC corridor from 1979 to 2024, revealing distinct spatial and temporal patterns. Long-term precipitation trends indicate increasing extreme rainfall in the central plains and southeastern lowlands, while changes in the northern and western regions are comparatively minor. High-risk areas form continuous bands across the central plains and southern lowlands, whereas low- and mid-low-risk zones are mainly concentrated in the northern mountains and transitional regions. Over time, the extent of mid- and high-risk areas has expanded, reflecting an overall increase in flood risk, particularly in the central plains. The rise in flood risk is closely linked to the temporal evolution of key factors. Extreme rainfall remains the primary driver, while sensitivity, vulnerability, and coping capacity modulate the spatial distribution of risk. Although improvements in infrastructure and economic capacity have mitigated disaster impacts to some extent, flood risk continues to rise in rapidly urbanizing areas. Aside from occasional localized extreme rainfall events, model results align closely with historical flood records, providing confidence in the assessment. These findings offer practical guidance for disaster preparedness, infrastructure planning, and adaptive strategies, while also addressing a long-standing gap in multi-indicator, long-term flood risk assessment along the CPEC corridor.