Spatiotemporal dynamics of population change in flood risk zones of the Pearl River Delta (1985–2024)

Abstract

The increasing likelihood of extreme adverse impacts driven by climate change has significantly impeded regional sustainable development, with flood-induced losses emerging as a critical challenge. This study selects six flood return periods, namely the 10-year, 50-year, 100-year, 200-year, 500-year, and 1000-year ones, to establish a complete gradient of low, medium–high, and extreme flood risks, breaking the limitation that existing studies mostly focus on the 100-year return period flood risk. By matching dynamic factors such as population size changes and flood control project upgrades in the Pearl River Delta (PRD), it quantifies the chain reaction law of "elevation of return period levels → shrinkage of safe zone scope → expansion of population size in flood risk zones" and reveals the dynamic evolution characteristics of flood risks, thereby providing precise scientific support for regional differentiated flood prevention and control. This study explores the spatiotemporal characteristics and regional disparities of population changes in flood risk zones and identifies the commonalities and specificities of such changes across different cities based on the scope of flood risk zones and exposed population, by adopting population distribution data from 1985 to 2024 and flood maps under six return period levels, to provide more robust and precise theoretical support for disaster prevention and mitigation in the Pearl River Delta in the future. As indicated by the cumulative growth rate, except for the scenario of a 10-year flood return period, the population growth rate within the extreme high-risk zone is significantly higher than that within the low-risk, moderate-risk and high-risk zones. the population growth rate within flood risk zones has exceeded that of the entire region, and in 8 cities except Shenzhen, the population growth rates within the high-risk and extreme high-risk zones have generally been higher than those within the medium-risk and low-risk zones; external conditions have influenced human activities during the change of return period grades, leading to distinct regional characteristics.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-56972-x.

Introduction

A significant portion of the population has been exposed to flooding events globally, as the frequency of extreme weather events and the severity, duration, and frequency of destructive floods have been increased in recent years¹. However, the space for human activities is influenced by economic, social, and political factors, and areas of higher exposure are chosen to be settled in by a significant portion of the population. According to the Emergency Events Database (EM-DAT), maintained by the National Center for Research on the Epidemiology of Disasters (CRED) at the University of Leuven, Belgium, and based on criteria including the number of deaths, people affected, and declarations of international emergency assistance, a total of 56 flooding events has been recorded in the PRD and surrounding areas since 1985, with reported economic losses exceeding US$1,100 billion^2,3. Future climate-related flooding is projected to increase with climate extremes^4,5 leading to further economic losses.

The interaction between large populations in urban, driven by rapid urbanization processes, and flooding events has become increasingly pronounced in recent decades. Expanding the scientific understanding of this dynamic relationship and elucidating the mechanisms by which demographic shifts change interacts flood risk have consequently emerged as critical focal points in contemporary disaster risk research. A body of literature that has grown rapidly has emphasized the coupling between population dynamics and flood hazards within international studies. For instance, Tellman et al. , utilizing satellite-derived datasets, reported in Nature that the proportion of the global population exposed to flooding increased by 20–24% between 2000 and 2018⁶. Their projections suggest that approximately 57 countries or regions worldwide are likely to experience heightened flood risk in the future, potentially endangering human settlements and production spaces. Rentschler et al., through a global-scale analysis of population exposure to varying flood intensities, demonstrated that the expansion of high-risk settlements in East Asia outpaces that of safer zones by approximately 60%. Their findings also revealed considerable heterogeneity in both flood risk intensity and population dynamics across different countries and regions⁷. Similarly, Jeremy et al. investigated the nexus between population change and flood exposure under climate change scenarios, providing forward-looking estimates of demographic distributions within flood risk areas⁸. In a case study focusing on Qatar, it was concluded by Ajjur that temperature, wind speed, precipitation and potential evapotranspiration showed strong positive (p < 0.05), strong negative (p < 0.05), moderate positive and moderate negative correlations with flood risk, respectively, via the quantification of flood risk. Scientific evidence was furnished for flood mitigation and climate adaptation planning in rapidly urbanized arid regions against the backdrop of global warming, frequent extreme rainfall events and sea level rise, through an examination of the relationships among urban growth, climate change and flood risk⁹.

There is now a general consensus in the literature that populations residing within flood risk zones have been increasing globally, albeit with substantial geographical variability. Notwithstanding this progress, the focus of current research has remained predominantly on 100-year flood events, with more extreme scenarios seldom being incorporated, thereby constraining its applicability under the climatic extremes projected for the twenty-first century.

Disaster and risk are core concepts in the field of disaster risk management, with an intrinsic distinction of actuality versus potentiality in their essence, and they are interrelated to form a complete risk governance logic. Risk is an aggregation of potentiality and uncertainty, oriented towards the future in terms of temporal attribute, featuring assessability, manageability and reducibility. Disaster arises when the risk threshold is breached and the capacity for response proves inadequate.

Empirical evidence indicates that the severity, duration, and frequency of catastrophic flooding events are likely to escalate as a consequence of synergistic effects from demographic expansion and climate change in flood risk areas. Projections suggest that, economic damages induced by flooding could increase by a factor of up to 20 by the end of this century¹⁰. The PRD is considered to exhibit heightened vulnerability to flooding, attributed to its low-lying geomorphology, dense fluvial networks, high annual precipitation, and pronounced seasonal and inter-annual variability, in comparison to other regions¹¹. Consequently, advancing knowledge of demographic changes within global river flood risk zones, together with elucidating their underlying drivers, has been deemed essential for the formulation of effective strategies in flood disaster prevention and management.

In response to these challenges, this study presents a comprehensive analysis aimed at characterizing the spatiotemporal evolution of flood risk zone extents and associated demographic changes in the PRD, considering six flood (10, 50, 100, 200, 500, 1000) return periods from 1985 onwards. A scientifically rigorous foundation for future urban planning and the development of protective infrastructure in the PRD is sought in this study through the characterization of spatial and temporal correlations between flood risk exposure and demographic distributions, and the identification of both common and region-specific patterns of population evolution.

Data and methods

Research area profile

The PRD is situated in the south-central part of Guangdong Province which borders the South China Sea, lies adjacent to Hong Kong and Macao, and faces Southeast Asia across the sea (Fig. 1). The PRD represents a composite delta formed through the deposition of sediments and sands transported by the Xijiang, Beijiang, and Dongjiang rivers of the Pearl River system, together with their tributaries—the Tanjiang, Suijiang, and Zengjiang rivers—in the estuarine bay at the mouth of the Pearl River. Administratively, it comprises nine cities: Guangzhou, Shenzhen, Zhuhai, Foshan, Huizhou, Dongguan, Zhongshan, Jiangmen, and Zhaoqing. These nine cities, which occupy less than one-third of Guangdong Province’s total land area, have been identified as accommodating over half of the province’s population and contributing to more than three-quarters of its economic output.

Fig. 1.

Location map. (This map was produced based on multi-year prefecture-level administrative boundary datasets of China released by the Resources and Environmental Science Data Platform; the base map remains unmodified.)

Following the initiation of the reform and opening-up policy, the scale of construction land in the PRD has undergone substantial expansion, and the regional spatial pattern has been observed to stabilize gradually over time ¹². The total area of construction land increased from 2828.75 km² in 1980 to 8348.89 km² in 2020, representing nearly a threefold increase. As urban development encroaches on various land types, the scarcity of available space drives construction activities to spread irregularly into multiple regions, including flood‑risk zones. This process has further contributed to the concentration of population within flood risk zones.

Data sources and processing

Multi-period land use remote sensing data since 1980, along with prefecture-level administrative boundaries, were obtained from the Resource and Environment Science Data Center of the Chinese Academy of Sciences (https://resdc.cn)¹⁴. Flood simulation data were sourced from the Global Resource Information Database (GRID-Geneva) (https://unepgrid.ch/en). Population data were acquired from the Global Human Settlement Layer (GHSL) (https://ghsl.jrc.ec.europa.eu¹³, whereas GDP and related socioeconomic data were extracted from the statistical yearbooks of the respective cities (Table 1).

Table 1.

List of data involved in the study.

Data classification	Data name	Data sources	Data type	Illustration
Land use data	Administrative division data	Resource and Environment Science Data Platform	Vector data	boundary,
	Remote sensing monitoring data (1980–2020)		Raster data (1km)	Topographic analysis
Flood data	Flood simulation data	Global Resource Information Database (GRID-Geneva)	Raster data (100m)	Flood risk area analysis
Socioeconomic data	Population (1985–2050)	Global Human Settlement Layer	Raster data (100m)	Analysis of population density change in the study area
	Permanent population (1985–2024)	Statistics departments of Guangdong Province and cities concerned	Numerical statistics	Population change analysis in the study area
	Gross regional product (1985–2024)			Analysis of current situation

Demographic data

Raster data acquired are collated and summarized, and population data for the nine cities in the Pearl River Delta are collected and compiled at five‑year intervals in this study.Upon acquisition of the raster data, calculations are performed with the population data documented in the statistical yearbooks of the respective cities, and population correction coefficients are derived.The corrected population rasters thus obtained are employed for subsequent research on population data.

Classification of water risk areas

The flood risk map adopted for the study was obtained from the Global Resource Information Database (GRID-Geneva), jointly established by the Swiss Federal Office for the Environment, the United Nations Environment Programme and the University of Geneva.

In the process of flood risk zone classification, the flood risk classification method proposed by the Standing Committee on Agriculture and Resource Management (SCARM) of Australia in Floodplain management in Australia: best practice principles and guidelines was selected for this study.

Based on the coupled dynamic composite curve of flood water depth (m) and flow velocity (m/s), the classification method has been developed by comprehensively considering the differences in physiological characteristics between adults and children as well as the functional attributes of ordinary buildings and flood control-specialized structures, with differentiated risk adaptation criteria formulated to achieve the synergistic balance between safety and economic efficiency. Key influencing factors, including frequent extreme rainfall and sea level rise under the background of climate change, were incorporated into the classification process, breaking through the limitations of current climatic conditions and reserving management resilience for future flood risk prevention and control. In addition, this method was highly compatible with the PPRR (Prevention-Preparedness-Response-Recovery) flood management framework, with long-term planning effectively linked to short-term emergency response practices. On the basis of this method, flood hazard was divided into four grades, namely extremely high, high, medium and low, in this study¹⁵ (Table 2).

Table 2.

Flood risk classification table.

Flood rating	Divide the criteria
Very high-risk areas	Depth > 1.2 m and Flow rate ≥ 1.6
High-risk areas	0.6 m ≤ depth < 1.2m and 0.8 ≤ Flow rate < 1.6
Moderate risk area	0.3 m ≤ depth < 0.6m and 0.4 ≤ Flow rate < 0.8
Low risk area	Depth < 0.3 m and 0 ≤ Flow rate < 0.4

Flood water depths and ground slopes were acquired using ArcGIS. On this basis, flood flow velocities at different grid points were determined using roughness coefficients, ground slopes and water depth data, in accordance with Manning’s coefficient and the wide-shallow approximation theory in hydrodynamics. Flood risk zones corresponding to six return periods were classified into four grades respectively in line with Table 2. These risk zones were then overlaid with revised population data for different years and statistically analyzed, from which population data within risk zones required for subsequent analysis was obtained.

Research method

In this study, vector data, including population and flood risk maps of the PRD from 1985 to 2024, were selected. Relevant spatial data were extracted using the spatial classification techniques implemented in ArcGIS to obtain the spatial distribution and growth patterns of the population, as well as the spatial distribution and extent of flood risk zones¹⁵, thereby enabling the visualization of regional disparities (Table 3).

Table 3.

Statistics on the change of flood risk area in return periods.

Changes in the scope of flood risk areas in the PRD
Variation of range of flood risk areas within return periods intervals of 10 years and 1000 years (km2)					Risk range of 1000 years return period
City	Low	Moderate	High	Very high	Area (km2)	Proportion of the city area(%)
Foshan	−205.83	−324.97	−186.89	1024.77	3044.23	80.2
Zhongshan	−46.06	−39.81	−7.5	287.65	1341.99	76.94
Zhuhai	−24.45	−14.66	126.51	85.28	883.61	55.72
Dongguan	−13.95	−71.17	151.84		1029.98	42.05
Guangzhou	−80.01	−41.1	487.63	56.92	1978.78	27.43
Jiangmen	−36.34	70.34	−26.08	236.08	1559.91	16.65
Zhaoqing	−17.15	−44.99	−98.84	613.44	2363.1	15.86
Huizhou	34.37	39.84	48.28	6.08	1135.02	10.03
Shenzhen	40.53	2.58	0.03		55.9	2.87

Significant common characteristics were identified between population agglomeration and the spatial distribution of flood risk zones in the PRD during 1985–2024, with their core spatial scopes exhibiting a high degree of agglomeration along the banks of the Pearl River’s main and tributary rivers and coastal areas.The temporal and spatial evolution laws both presented a tendency of radial diffusion from high-core areas to surrounding peripheral regions with gradual grade decline²², and the spatial patterns all formed a concentric zonal distribution feature characterized by high centrality and low peripherality ²³.

The analysis commenced with an examination of population agglomeration in the PRD since 1985. Subsequently, population changes within flood-safe and flood-risk zones were analyzed both qualitatively and quantitatively. This approach facilitated the exploration of population dynamics within flood risk zones, including their characteristics and regional disparities, and supported the discussion of potential underlying drivers of these spatial variations.

Lorenz curve and Gini coefficient

In this study, prefecture-level cities within the study area were ranked according to population density. Using cumulative percentages of population density and area, Lorenz curves representing the population density distribution of each prefecture-level city were constructed¹⁶. The convergent characteristics of population distribution in the PRD were identified through these Lorenz curve diagrams, enabling interpretation of the association between regional population concentration and changes within flood-safe and flood-risk zones.

The mathematical expression of the Gini coefficient is given as follows:

1

(i = 1,2,…,n) denotes the rank of each prefecture-level city, ordered by resident population density in descending order, where n represents the total number of cities. denotes the proportion of the population density, represents the proportion of land area, and refers to the cumulative percentage of land area.

Spearman correlation coefficient

Flood risk zone levels may exert a measurable influence on variations in population density. To evaluate this relationship, Spearman’s rank correlation coefficient was employed to comprehensively assess the association between flood risk zone levels and population density across different return period scenarios during the study period. A model was subsequently established to investigate the regional characteristics of population changes under flood risk.

The primary formula is given as follows:

2

ρ denotes the Spearman’s rank correlation coefficient, ranging from –1 to 1. A value of ρ=1 indicates a perfect positive correlation, ρ=–1 indicates a perfect negative correlation, and ρ=0 indicates the absence of correlation.

Results and analysis

Spatiotemporal convergence of flood risk area and population dynamics

Spatiotemporal characteristics of rapid population agglomeration and flood risk zones in the Pearl river

The Global Human Settlement Layer (GHSL) population vector data were data-corrected, and population density data for the PRD from 1985 to 2024 were compiled. Population density was subsequently rasterized¹⁷, and spatiotemporal evolution maps of population distribution across the study period were generated (Fig. 2).Results demonstrate a rapid population increase within the study area between 1985 and 2024. The total population increased from 17,394,200 in 1985 to 80,226,000 in 2024, representing an approximate 4.61-fold increase over 39 years. The peak population per grid cell increased from 459,600 in 1985 to 1,205,743 in 2024.

Fig. 2.

Temporal and spatial evolution of population from 1985 to 2024. (This map was produced based on multi-year prefecture-level administrative boundary datasets of China released by the Resources and Environmental Science Data Platform; the base map remains unmodified.)

Overall, the study area exhibited a continuous expansion of high-density population zones alongside a gradual contraction of low-density zones, reflecting a generalized and sustained population growth trend within the PRD.

Population distribution inequality in the PRD was analyzed by constructing Lorenz curves for population distribution¹⁸. As shown in (Fig. 3), significant deviations of the Lorenz curves from the line of perfect equality have been observed since 1985. Throughout the study period, the Gini coefficient was consistently maintained above 0.5, indicating a highly uneven relationship between population and land area and a pronounced concentration of the population in specific regions.In addition, a rapid increase of 0.051 in the Gini coefficient was recorded over the 39-year period. In 1985, 80% of the population in the Pearl River Delta was concentrated on about 40% of the total land area of the region. By 2024, the level of spatial agglomeration was further enhanced, the scope for population distribution was reduced from 40 to 33.4%, and the imbalance within the region was increasingly intensified. These findings highlight a growing concentration of the population within a relatively small proportion of the PRD, reflecting a substantial imbalance in spatial population distribution.

Fig. 3.

Lorenz curve of population distribution from 1985 to 2024.

The PRD, recognized as one of China’s flood risk and high-frequency flood regions¹⁹, is likely to experience varying levels of flooding across approximately one-quarter of its total area. Due to the flat topography of the central PRD, the presence of the Pearl River main stream and numerous tributaries flowing through it, and the high density of the river network, which collectively provide conditions conducive to flood formation²⁰, the region has been deeply affected by floods of different return periods. Floodwater depths under various return periods were collected and organized in this study, and the spatial distribution of flood risk in the PRD was subsequently delineated (Fig. 4).

Fig. 4.

Spatial distribution of flood risk under different return periods (This map was produced based on multi-year prefecture-level administrative boundary datasets of China released by the Resources and Environmental Science Data Platform; the base map remains unmodified.)

The flood risk area in the PRD has been observed to expand with increasing flood return periods²¹, with the additional areas mainly distributed in Zhongshan, Guangzhou, eastern Foshan, northern Shenzhen, as well as in low-lying zones and regions along both sides of the river. From an overall perspective, flood risk zones in the PRD have been found to extend outward from the Pearl River coastline in Zhaoqing, Foshan, Zhongshan, and Zhuhai, with risk levels gradually diminishing toward peripheral areas, forming a spatial pattern characterized by a high-risk center and a low-risk periphery. Flood risk levels have been classified into two distinct patterns: a dense, high-risk concentration in the central region centered on the Pearl River and its tributaries, and a scattered distribution of low-density, low-risk zones in outlying areas. Flood risk levels are characterized by significant spatial variability, with their extent and trends differing across regions, as influenced by both topography and the hydrological network. (Fig. 5).

Fig. 5.

Area of flood risk zone in the Greater Bay Area under different return periods.

The sizes of flood risk areas in the PRD have been observed to consistently follow the order Foshan > Zhaoqing > Guangzhou > Jiangmen > Zhongshan > Huizhou > Dongguan > Zhuhai > Shenzhen, under six return period scenarios.The extent of flood risk zones in each city has increased with higher return periods, with very high-risk zones in Foshan and Zhongshan expanding notably beyond the regional average. Conversely, low-, moderate-, and high-risk zones in other cities, except Shenzhen and Huizhou, have contracted to varying degrees.

Population size–flood risk zones

The spatiotemporal coupling characteristics between population distribution and flood risk zones are further clarified. Data deviations caused by variations in the scope of flood risk zones are avoided through analysis of the overlapping rate between high-density population areas and flood risk zones under different return periods. Accordingly, high-density population areas with a population density of ≥ 5000 persons per km² in the PRD are extracted in this study.Their distribution areas both inside and outside the flood risk zones were calculated separately. Based on the above basic data, further calculations were performed to determine the proportion of high-density population areas within the flood risk zones, as well as the spatial overlapping area between high-density population areas and flood risk zones. Finally, the proportion of high-density population areas within the flood risk zones was multiplied by the ratio of this overlapping area to the total area of the flood risk zones, yielding the area overlapping rate between the two (Fig. 6).

Fig. 6.

Overlap rate of population and risk zone extent.

The main calculation formula is presented as follows:

3

wherein denotes the proportion of high-population-density areas within flood risk zones; represents the overlapping area between high-population-density areas and flood risk zones; refers to the total area of flood risk zones.

During the study period, overlap rates were observed to increase with higher return period levels, with the maximum and minimum differences between the 1-in-10-year and 1,000-year return period reaching 9.39% (in 1985) and 6.11% (in 2010), respectively. The elevated overlap rates observed in 2000 and 2024 for the same return periods suggest that these years were characterized by rapid population growth. Overall, the patterns of population growth and flood risk zone expansion in the PRD during the study period exhibited strong similarities, with the spatial extents of high-density population areas and flood risk zones largely coinciding during periods of rapid expansion.

Spatiotemporal differentiation of population exposure

Population growth intensifies population exposure

Taking the 1-in-10-year flood event as an example, approximately 20% of the PRD population in 2024 was exposed to flood risk, with about 40.78% of this population residing in high- or very high-risk zones²⁴. Notably, even under this return period, areas with floodwater depths exceeding 60 cm were inhabited by over 6.41 million people. Between 1985 and 2024, the total population of the PRD increased by 345.56%, whereas the population within flood risk zones grew by 356.62%, indicating a higher growth rate in flood risk areas relative to safer zones.

As return period levels increased, both the spatial extent and severity of flood risks expanded, leading to progressive growth in the affected population (Fig. 7). Under the 1-in-10-year flood scenario, the majority of the affected population was concentrated in low-risk zones, while moderate- and high-risk zones exhibited comparable population sizes. However, driven by rapid urbanization, populations in moderate- and high-risk zones experienced accelerated growth over the 39-year period, reaching levels similar to those in low-risk zones by 2024. The population sizes of the other three risk grades have been surpassed by the population size in the high-risk zone in most years, starting from the flood scenario with a 50-year return period. Under the flood scenario with a 200-year return period, the population sizes of the other three risk grades have been completely surpassed by the population size in the high-risk zone. It is indicated hereby that certain proportion of the total population within flood risk zones has been consistently accounted for by the population in the high-risk zone.

Fig. 7.

The population in flood risk areas under different return periods.

Moreover, populations in low- and moderate-risk zones gradually converged in size with increasing return periods, and under the 1-in-500-year and 1,000-year return period, the population in moderate-risk zones surpassed that in low-risk zones. Although very high-risk zones contained the smallest population, they exhibited the fastest population growth rate. Specifically, Compared with that in 1985, the population within the very high-risk zones of the 2024 once-in-a-millennium flood disaster has been found to have increased by approximately 5.85 times.

Notably, we observed a considerable slowdown in population growth within high-risk zones in both 1995 and 2015.In terms of spatial scale: during the study period, the population size of high-risk zones increased steadily as the risk zone expanded with higher return-period levels, consistent with the pattern observed in the study.

In terms of temporal scale: first, population growth in the Pearl River Delta slowed around 1995 and 2015.Around 1995, affected by the decline after the peak of the migrant labor wave and the contraction of inflows, population growth was halved, entering a period of “fluctuating slowdown”²⁵. Around 2015, China reached the Lewis turning point, and its working-age population began to decline²⁶.

In addition, the catastrophic floods in South China in 1994 prompted the optimization of national and local policies and regulations, and the Pearl River Delta launched governance projects for some river channels²⁷. In 2015, national and local authorities implemented policies such as sponge city construction and ecological control of flood-risk zones. These measures further constrained population agglomeration in risk zones and directly contributed to the reduction of their population size.

Finally, 1995 and 2015 were important nodes for industrial transformation and upgrading in the Pearl River Delta²⁸. Population relocation out of risk zones along with industrial upgrading²⁹ and shrinking demand for low-skilled labor led to a continuous slowdown in population growth³⁰.

In summary, the population reduction in risk zones in these two years resulted from the combined effects of population growth dynamics, spatial development orientation, flood control policy constraints, and industrial restructuring at these critical junctures. Essentially, it reflects the dynamic adaptation between population space and flood risk space in the urbanization process of the Pearl River Delta.

As shown in Fig. 8a, the cumulative growth rate is confirmed to have remained positive in flood risk areas under different return periods since 1985, with growth rates in high- and very high-risk zones generally observed to exceed those in low- and moderate-risk zones. Except for the low-risk zone, where population growth has consistently been the slowest, growth in the moderate-, high-, and very high-risk zones has been found to fluctuate annually.

Fig. 8.

The cumulative population growth and the year-on-year population growth in flood risk areas under different return periods.

Under the six return periods, the growth rate in the very high-risk zone is identified as consistently higher than that in the high-risk zone, representing the highest rate among the four flood risk categories. As shown in Fig. 8b, year-on-year growth is confirmed to peak in 1990 and 2000 across all risk classes, indicating periods of rapid population expansion in flood risk zones. These peaks coincide with the overall population growth of the PRD between 1985 and 2024, although growth rates in low- and moderate-risk zones during these peaks have been observed to be significantly lower than those in high- and very high-risk zones.

Population dynamics in flood risk zones exhibit two distinct patterns under stable growth, as influenced by the return period levels: fluctuation amplitudes in low- and moderate-risk zones are found to decrease as return period levels increase, whereas those in high- and very high-risk zones are observed to increase correspondingly.

Population exposure exhibits regional characteristics

Significant intercity differences in population growth and flood exposure were identified across the PRD²⁴. This study takes the average population growth multiple in the risk zones of the nine prefecture-level cities as the demarcation criterion, classifying their population growth status into red and blue categories for presentation (Fig. 9). The population in safe zones was found to have increased by 383.81% between 1985 and 2024, whereas populations in high- and very high-risk zones grew by 422.32%, under the 10 year return period.The fastest population growth within flood risk zones was recorded in Zhongshan, Dongguan, Foshan, and Huizhou, with growth rates exceeding the regional average during the study period.

Fig. 9.

Population growth rate from 1985 to 2024 under different return periods.

Population increases in flood risk zones were found to intensify as return period levels rose, with the most rapid growth occurring in Zhongshan, Huizhou, and Shenzhen when the flood risk level increased from the 10 year return period to the 50 year return period. Growth rates generally decelerated as return period levels continued to rise. Peak population growth in Zhuhai was recorded under the 1-in-200-year to 500 year return period, whereas in Shenzhen and Foshan it was observed under the 50 year return period to 100 year return period scenarios. Shenzhen was identified as the most affected city under escalating return period levels, with its flood risk zone population growth rate increasing by 39.17% when the risk level rose from the 10 year return period to the 1000 year return period. Zhongshan and Dongguan followed, exhibiting increases of 20.56% and 13.45%, respectively. In contrast, Jiangmen was least affected, with only a 1.09% change recorded.

Population growth rates of risk zones in the nine cities of the PRD under different return periods were calculated by multiplying the growth from 1985 to 2024 at each risk level by the average population of each city during the study period (Fig. 10). With the elevation of flood return period levels, the PRD safe zone scope has been continuously narrowed during the study period, leading to the population size reduction within safe zones.The expansion multiples of population size within flood risk zones of different levels are found to range between 0.04 and 42.01, with changes in flood return periods; and larger values are commonly observed in high-risk and extreme high-risk zones.

Fig. 10.

Population growth rate of flood risk areas in different cities from 1985 to 2024.

Under the influence of risk-level boundary shifts driven by return period changes, low- and moderate-risk zones were transformed into high- and very high-risk zones, causing population growth to decelerate. For instance, declines in population growth rates were observed in low-risk zones of Jiangmen, moderate-risk zones of Dongguan, Zhuhai, and Zhongshan, and low-, moderate-, and high-risk zones of Zhaoqing.

Strong regional heterogeneity was exhibited in population distribution within flood risk zones. In Shenzhen, population in all risk zones was found to increase progressively with the elevation of return period levels. Population growth within safety zones, driven by rapid urbanization and expansion of flood risk zones, was observed to exceed that of other PRD cities. Conversely, in most other cities, populations in low- and moderate-risk zones were found to contract with increasing return period levels, while populations in high- and very high-risk zones expanded. Population growth within flood risk zones, owing to the larger spatial extent of flood risk areas, surpassed that within safety zones in Foshan and Zhongshan.In contrast, in Zhaoqing and Jiangmen, where population bases were relatively small, population growth rates in safety and risk zones were similar.

Reemergence level-population size

According to the statistical results of the study, an increase of 11,660,500 people was observed in the exposed population of flood risk areas in 2024, compared with 4,061,700 people in 1985, even under the 10 year return period. This increase was accounted for approximately one-quarter of the total population growth in the PRD region during the study period. Along with the rise in the return period level, a continuous expansion was observed in both the extent of the flood risk areas and their affected populations.

To facilitate a more detailed analysis of populations in flood risk and safety zones, population densities of flood risk zones at different grades in nine PRD cities were calculated under varying return periods (Fig. 11)³¹. It was shown by the chart that, in all studied cities except Shenzhen and Zhuhai, a slight decrease was exhibited by the population density within safety zones as the flood return period increased. In contrast, a general increase was displayed by the population density within flood risk zones across all levels.

Fig. 11.

Population density of flood safety zone and risk zone.

Based on these observations, the nine cities were classified into four categories by the study:

Shenzhen – As the most densely populated city in the PRD, a generally higher population size was recorded in both safety and risk zones compared to other cities. Due to the relatively small extent of flood risk areas within the city, a population growth pattern similar to Zhuhai was exhibited. A steady increase in the safety zone population was observed with the rise in the return period class, while in flood risk zones, an initial increase followed by a gradual decrease in population density was noted at higher risk levels. These trends were distinguished clearly from those in the other seven cities.

Dongguan, Guangzhou, and Zhongshan – In these cities, a decrease in population density within safety zones was exhibited with increasing flood levels, whereas a steady increase was shown within flood risk zones. Notably, at higher return periods, a population density within low-risk zones surpassing that of safety zones was identified.

Foshan and Huizhou – A consistently higher population density was observed in low-risk zones than in safety zones. In Foshan, a rapid increase in population density within moderate-risk zones was exhibited with rising flood levels. By contrast, no significant change in population distribution was identified in Huizhou under varying flood levels.

Jiangmen and Zhaoqing – As two cities characterized by relatively low population densities and GDPS in the PRD, a higher population density was observed in flood risk zones than in safety zones. Moreover, a faster increase in population density within higher-risk zones compared to lower-risk zones was exhibited as the return period increased.

The proportion of population within risk zones to the total city population from 1985 to 2024 was averaged to derive population proportion data for six return period scenarios, for statistical purposes Table ( 4). Among the nine cities in the PRD, a population share in risk zones exceeding half of the total city population was consistently recorded in Dongguan and Jiangmen, and this share was further increased as the return period level rose. Jiangmen, identified as the city with the highest proportion of population in risk zones, exhibited the largest sensitivity to changes in return period level, with the proportion increasing by 12.71% as the level was raised from 10 year return period to 1000 year return period. In contrast, Zhongshan and Zhuhai, identified as the two cities with the smallest population proportions in risk zones, were the least affected by changes in return period level, with increases of only 1.46% and 2.43%, respectively, from a 10 year return period to 1000 year return period. Small decreases in population share during the return period were also observed in Zhongshan and Huizhou, where Zhongshan’s share declined by 0.32% from a 500 year return period to a 1,000 year return period, and Huizhou’s share fell by 0.97% from a 50 year return period to a 100 year return period, with this decline even offsetting the overall increase across the full return period range.

Table 4.

The proportion of the population in the risk area in different return periods.

Proportion of population in risk areas (%)
City	Return period
	10	50	100	200	500	1000
Dongguan	84.19	85.35	85.52	85.59	86.42	86.64
Foshan	32.54	33.47	34.53	35.09	35.51	32.45
Guangzhou	13.32	14.57	15.05	14.98	14.04	16.41
Huizhou	43.30	47.36	48.90	48.67	45.63	53.32
Jiangmen	54.92	60.07	62.02	61.73	57.87	67.63
Shenzhen	14.72	16.10	16.63	16.55	15.52	18.13
Zhaoqing	15.90	17.91	20.00	23.19	25.71	26.21
Zhongshan	5.89	6.44	6.65	6.62	6.21	7.26
Zhuhai	10.51	11.50	11.87	11.82	11.08	12.95

In addition to changes in regional population density being examined, the Spearman correlation coefficient was used by the study to measure the strength of the relationship between flood risk class and population density (Fig. 12)³². Strong regional differences in this relationship were observed among the nine cities studied³³. From the correlation results, coefficients of 0.85 and 0.87 (significant at p < 0.05) were recorded for Jiangmen and Zhaoqing, respectively, indicating a positive correlation between flood risk class and urban population density, where population density within flood risk zones increased with higher risk zone levels. In Zhuhai and Shenzhen, a negative correlation was identified between population density and flood risk class (significant at p < 0.05), suggesting that flood risk influenced residents’ migration activities in these cities, with some residents choosing to relocate out of flood risk zones under its influence. Positive correlations (p < 0.05) between risk level and population density were also found in Foshan, Huizhou, and Guangzhou, with correlation coefficients decreasing in that order. In contrast, no significant correlation between flood risk level and population density was observed in Dongguan and Zhongshan, as indicated by Spearman’s correlation coefficient. On this basis, it can be further illustrated that the elevation of flood risk levels is correlated with population size changes to a certain extent, namely that variations in flood return period levels induce the transformation of population size within the flood risk zones and safe zones of the PRD³⁴.

Fig. 12.

The correlation between flood risk level and population density.

Changing trends of population exposure

To further enhance the practical significance of the research for flood risk prevention and control in the Pearl River Delta region and promote the transformation from traditional static risk assessment to a dynamic risk cognition system, population projection data for 2030, 2040 and 2050 provided by the Global Human Settlement Layer are adopted in this study. Population projection data for 2030, 2040 and 2050 from the Global Human Settlement Layer are adopted in this study to further forecast the population change trends in flood-risk zones of the Pearl River Delta. To reduce the impacts induced by uncertainties in the simulated population data, quantitative comparisons of risk exposure levels across different years and return periods are conducted by calculating the population exposure rate (Fig.  13).

Fig. 13.

Population exposure ratio to flood risk 2030–2050.

The main formula is given as follows:

where Ti is the exposed population under return period T, and Pi is the total regional population in the current year.

As the flood recurrence period extends from a 10-year to a 1000-year event, the exposure ratios in both low-risk and very-high-risk zones have been continuously rising. In 2030, the low-risk zone expanded from 10.15 to 14.74%, while the very-high-risk zone increased from 2.41 to 6.16%; in 2050, the low-risk zone rose from 10.75 to 14.97%, and the very-high-risk zone climbed from 2.08 to 6.26. In contrast, the exposure ratio in high-risk zones decreased significantly, and that in moderate-risk zones remained stable within the range of 9.3–9.9%, showing weak sensitivity to changes in the recurrence period (Table 4).

Between 2030 and 2050, the low-risk zone maintained steady growth under all recurrence periods; the very-high-risk zone saw continuous increases in exposure ratios under long recurrence periods (200-year, 500-year, and 1000-year events); the moderate-risk zone exhibited a mild upward trend in exposure ratios; and the high-risk zone experienced an overall downward or slightly declining trend. As a result, population exposure pressure has been gradually shifted toward low-risk, moderate-risk, and very-high-risk zones.

From 2030 to 2050, the population exposure to flood risk zones exhibited a core divergent trend of expansion at both ends and stability in the middle, accompanied by contraction of high-risk areas.

In the future, extreme flood events will lead to a more concentrated population exposure at the two ends of low-risk and very-high-risk zones, further reducing the concentration of population exposure in moderate- and high-risk zones. This pattern not only reflects the continuously strengthened effect of population agglomeration in low-risk areas but also highlights the intensifying long-term threat posed by extremely high-risk floods to the population, thus providing a key basis for the formulation of future flood risk management and adaptation strategies.

Conclusions and discussion

Conclusions

The population size of the Pearl River Delta and the extent of flood-prone zones are demonstrated to present a certain degree of synchrony in their spatiotemporal evolution. Overall, driven by the rapid process of urbanization, the regional population has been continuously increased; the period of population growth is found to be roughly consistent with the period during which the flood-prone zones were expanded.Spatially, the scope of population distribution, the extent of urban construction land expansion, and the spatial distribution of flood-prone zones are observed to overlap and align to a certain extent at the macro scale.It should be noted, however, that such spatial consistency is characterized mainly as a descriptive feature, which reflects the geographical superposition between urban expansion and population growth.It does not indicate that population growth or urban expansion can be directly regarded as the cause of changes in the extent of flood-prone zones, nor is there a simple causal relationship between the two.

The population growth rate within flood risk zones in the PRD exceeded that of the entire region. During the study period, population within flood risk zones increased rapidly, with growth rates on average 93.91% higher than those in flood safety zones under the six return period scenarios. This rapid population growth, driven by urbanization and the expansion of flood risk zones, further intensified population exposure across the PRD. In addition to the expanding scope of extreme weather events and urbanization-induced population concentration from rural to urban areas, the presence of a long coastline and a well-developed Pearl River water system facilitated the development of economically robust ports leveraging geographic advantages and economic fundamentals^35,36. This infrastructure enabled the realization of intermodal transport across the Pearl River Basin. After weighing the benefits and risks, most people chose to settle within flood risk zones. Consequently, human activities advanced into these areas, and population growth within risk zones outpaced that of the overall region during the 35-year period.

Population agglomeration in flood risk zones of PRD cities was influenced by socioeconomic factors, exhibiting variability across different locations. Due to differences in geography, economy, industry, and other conditions, distinctive patterns of population change within flood risk zones were observed among the PRD cities^37,38. Human vulnerability in the face of extreme climates is generated by the combined effects of multiple factors³⁹. Based on relevant literature and the findings of this study, the driving mechanism of population change within flood-risk zones in the Pearl River Delta is summarized and reasonably inferred from four dimensions: urbanization and land development intensity, economic development level, infrastructure and public services, and policy and institutional arrangements. On this basis, the nine cities in the Pearl River Delta are roughly divided into three types of regions: industry-oriented regions, economy- and politics-oriented regions, and flood-risk-oriented regions. First, in Huizhou, Jiangmen, and Zhaoqing, where pillar industries remain rooted in traditional manufacturing with heavy reliance on shipping⁴⁰, flood risk zones offering favorable shipping conditions became the primary settlement areas, supported by port-centric logistics hubs enabling intermodal transportation and sustaining industrial development^41,42. Second, in Dongguan, Foshan, Guangzhou, and Zhongshan, where main urban activity zones are generally exposed to flood risk, residents no longer prioritized safety alone but instead favored low- and moderate-risk zones that were economically and politically dominant. Third, in Shenzhen and Zhuhai, where flood risk zones within municipal boundaries were smaller and more spatially constrained, safer zones with better locations became the primary areas for local population aggregation⁴³.

Discussion

Flood disasters with centennial and even millennial return periods have increasingly been recognized as an inevitable issue in social and economic activities, against the backdrop of the escalating global climatic extremization, which further highlights the urgency of population distribution optimization in flood risk zones of the PRD²³. This study breaks the traditional framework of "population change driving flood risk", focuses on population changes within flood risk zones, explores the reverse impact of flood risk on human agglomeration, and fills the research gap in the "risk-human behavior" nexus. On this basis, it avoids the limitation of a single return period and selects six gradient return periods (10, 50, 100, 200, 500, and 1000-year) to meet the practical needs of climate extremization, enhancing the relevance and cutting-edge nature of risk assessment. Taking the PRD as a typical sensitive and vulnerable region, the study aligns with international core research themes. Relying on the common characteristics of the PRD and global coastal deltas as well as coastal urban agglomerations, the conclusions break geographical boundaries and possess reference value for similar regions worldwide. Additionally, the study adds predictions of future population changes in flood risk zones, filling the gap in dynamic research and providing forward-looking support for the formulation of targeted early warning and differentiated flood control strategies.

However, the discussion on the scope of flood risk zones in this study is limited by the flood risk map, and the flood simulation process is primarily conducted by calculating the uniform flow depth of river channels via runoff estimation⁴⁴. While this method enables the determination of flood risk maps over a large scale, it may result in the absence of partial flood data in small-scale areas or regions without water system distribution.

In addition, only precipitation-induced floods are considered in this map, with storm surges, coastal flooding and other potential flood scenarios yet to be discussed herein. Furthermore, due to data constraints, the existing flood risk map fails to take into account the conditions of flood control facilities⁴⁵, which may lead to a higher risk grade indicated on the flood risk map than the actual risk grade under the same flood return period grade to a certain extent, a circumstance that is prevalent in relevant literatures.

Finally, the core objective of this research, as an exploratory study, is to investigate the fundamental question of whether a correlation exists between population change and flood risk. Its main tasks include systematically sorting out existing data and constructing a logically rigorous analytical framework. Therefore, the driving factors of population change proposed in the research conclusions are based merely on reasonable inferences from socioeconomic factors and relevant studies, without involving the collection and verification of empirical data.

In actual flood disasters, the presence of flood control infrastructure would reduce the extent of risk zones, and the severity and spatial coverage of flooding would likely be lower than those discussed herein. Moreover, rapid urbanization has introduced urban flooding as another critical risk, where riverine factors alone are insufficient for delineating flood risk zones.

Despite these limitations, the current flood protection standard in the study area remains at 1-in-200 years or less, underscoring the relevance of the issues examined in this paper under scenarios of intensifying climate extremes.

Future relevant research can further incorporate storm surges and coastal floods on the existing basis, starting from compound flood events⁴⁶ and coupled with the scope of flood risk zones under actual flood disasters, to explore the spatio-temporal evolution laws of population in regional flood risk zones in a more comprehensive and meticulous manner. In view of the limitation that the driving factors of population change in this study have not been verified with empirical data, further in-depth research can be carried out in the future by means of improving the multi-source data integration system, optimizing the model construction methodology, and deepening the scenario-based analysis of the driving mechanism, so as to more accurately reveal the inherent correlation mechanism between population change and flood risk. This will thereby provide more robust and precise scientific support for the planning of urban areas and construction of protective facilities in the future, as well as a scientific basis for formulating comprehensive flood control strategies and elevating the level of urban flood safety resilience⁴⁷.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Wang, Responsible for the generation, review and revision of research concepts, as well as research supervision and guidance; Wu, Responsible for experimental method design, Material preparation,verification and verification, draft writing, and visualization of research results. All authors contributed to the study conception and design.

Funding

This work was supported by the National Natural Science Foundation of China (Grant No. 52578076). The funding source had no involvement in the study design, data collection, analysis, or publication decision. The authors declare that they have no other competing interests.

Data availability

The datasets supporting this study are from the following public sources. Land use and administrative boundary data: Resource and Environment Science and Data Center. (RESDC), https://www.resdc.cn. Flood simulation data: Global Resource Information Database (GRID) , https://unepgrid.ch/en. Population data: Global Human Settlement Layer (GHSL) , https://ghsl.jrc.ec.europa.eu. Population and GDP data: Statistical Yearbooks of various cities, http://stats.gd.gov.cn/gdtjnj/ . All public datasets are accessible via the provided URLs.All datasets generated and analyzed during this study are included in the published article and its supplementary information files.

Declarations

Competing interests

All authors have read and approved the final manuscript and declare that they have no competing interests.