Reconstructing a georeferenced inventory of flooding hazards in Raoping, Guangdong province, China, from 1492 to 1985

Flooding, a common natural hazard, has a severe impact on human settlements (Kreibich et al. 2022). Flood risk management of rivers and coasts has been a millennia-old practice of human civilization (Hallegatte et al. 2013; Winsemius et al. 2016). The impact of flooding is increasing in tropical and subtropical regions, due to the intensification of climate change and the acceleration of population settlement and growth in coastal areas in recent decades (Kallis 2010; Formetta & Feyen 2019).

With long coastlines of 32,600 km and around one-third of the land in the tropical and subtropical region, China has long suffered from flooding hazards which put human communities at risk (Dottori et al. 2018). Those cities in China's subtropical region with large populations, such as Guangzhou, Shantou, and Chaozhou, have rich records of flood events with significant destruction in terms of economic and human security (Swain et al. 2020), drawing people's attention to accurate flooding-forecast systems (Cammalleri et al. 2020).

The fundamental of a comprehensive flooding-forecast system requires a detailed analysis of flooding (Turner et al. 2003). Conventionally, the analysis of flooding hazards depends on hydrological modeling approaches and datasets of flow discharge and rainfall gauges (Jongman et al. 2015). China's rainfall and streamflow measurement system started in the 1960s, following the development of hydrological science (Chacon-Hurtado et al. 2017). However, due to the lack of measurement data over a centennial scale, it is difficult to complete a systematic analysis of flooding risks, relying only on the fragmented records by local bureaucracies or landowners (Kreibich et al. 2014, 2022).

The flood events records are a very important reference for researchers to understand the characteristics and vulnerability of floods in the region, and thus help improve the predicting accuracy of the hydrological models in flood-modeling studies (Douben 2006). There are several existing global datasets of flood risks, including the International Disaster Database (EM-DAT) (Donatti et al. 2024), the International Flood Network (IFNET) (Alfieri et al. 2018), and the Global Active Archive of Large Flood Events (Brakenridge 2016). Most of the database of flooding hazards was established depending on hydrological modeling approaches and datasets of flow discharge and rainfall gauges (Jongman et al. 2015). However, many low-income countries have not established systematic measuring and monitoring networks for precipitation and flow discharge. Due to the lack of measurement data in these data-scant regions, it is difficult to complete a systematic analysis of flooding risks (Kreibich et al. 2014, 2022). For example, China's rainfall and streamflow measurement system started in the 1960s, following the development of its hydrological science (Chacon-Hurtado et al. 2017). Analyzing the historical flooding events before the 1960s relies only on the fragmented records of local bureaucracies or land-owners (Pelling 1999; Reale & Handmer 2011).

The reconstruction of databases of historical flooding hazards has been developed in many regions in recent decades. During this process, an advanced suggestion was proposed that a comprehensive analysis of flood risks could be conducted by combining historical documents obtained from local authorities and measured (or modeled) flow discharge data (Llasat et al. 2009, 2010a, b). Hundreds of historical records and works were then carefully studied and evaluated to be involved in database construction, such as those of Europe (Gaume et al. 2009; Llasat et al. 2010a, b, 2014; Mouratidis & Sarti 2013; Taboni et al. 2022; Pappalardo & La Rosa 2023), Africa (Rasmussen & Houze Jr 2012; Hoedjes et al. 2014; Fofana et al. 2022), and the United States (Barrera et al. 2006; Špitalar et al. 2014; Smith et al. 2019). The flood database constructed based on the historical press archives has also been used in flood risk management (Seidel et al. 2009). There are numerous ancient records about flash floods, as well as county chronicles compiled by local bureaucracies and experts (Guzzetti & Tonelli 2004). However, few works have utilized them for study or flooding broadcasting purposes through modeling. China has abundant documents that describe the hazard events recorded in the local chronicles. However, until now, nearly all of China's flooding hazards have been based on flow discharge (Zhang et al. 2002; Wang et al. 2021; Yang et al. 2021). The spatial information in the literature is also missing in its current form. Following the great demand for building a geographic information system for natural resource management at the national level. It is urgent to investigate how to build a comprehensive database combining different data sources, that can better understand the history, evolution, and mechanism of flooding hazards for current and future periods.

This study presents the reconstruction and analysis of historical flood events in the Huanggang River basin of Raoping county. A 500-year flood database with explicit spatial and temporal coverage was developed based on ancient records and maps. In addition the seasonality and evolution of flooding hazards and the temporal evolution of the study area were investigated. The preliminary findings from the construction of the new database were discussed. The novelty of this study is not only creating a database for the study region but also providing a protocol for application in China's other cities with similar documents. The application of the protocol developed in this study can help in understanding human adaption to climate change, which will inform future developments for flooding prevention at the national scale.

The county has a subtropical marine monsoon climate with abundant but unevenly distributed rainfall. Disasters of typhoons and floods frequently occur in summer, accounting for 81% of the total annual rainfall, with southerly winds prevailing. In winter and spring, the prevailing wind is northerly and rainfall accounts for only 19% of the total annual rainfall. As a result, there are more floods and waterlogging in summer and autumn, and more droughts in winter and spring. According to historical records, the most abnormal weather in the past 500 years occurred on the second day of December in the third year of the Yongzheng period of the Qing Dynasty (AD 1725), followed by a destructive flood and forest damage.

Before the foundation of the People's Republic of China (PRC), this county suffered multiple huge death-totals and economic losses from floods, mainly due to the lack of flood-control strategies and management policies provided by local governments and landlords. After the foundation of the PRC, governments at all levels actively led the carrying-out of large-scale water conservancy projects. Specifically, the policy includes early preparation for flood control, waterlogging before the flood season, and water storage and preparation for agriculture at the end of the flood season, to greatly reduce losses caused by disasters.

In the Huanggang River basin, two hydrologic gauge stations were established for measuring flow discharge and water-quality data (Figure 1(b)). The Tangxi Station (longitude 116°53′14″E, latitude 23°52′46″N), which is located adjacent to the city center, was built in October 1959. This station is the basic outflow flow-control station of Tangxi Reservoir, a large reservoir on the Huanggang River, and is an important national hydrological station, with a water collection area of 667 km². The station was established for the country to collect basic hydrological information on the upper reaches of the Huanggang River for a long time, and to analyze hydrological characteristics and the evolution of river geomorphology. It provides services such as flood control, water-supply safety, comprehensive dispatching of water conservancy projects, water ecological restoration and protection, water environment management, and water resource allocation in Raoping county. In addition, there is one new station (the Raoping Water Level Station) located at the new city center (longitude 116°59′21.0″E, latitude 23°40′35.2″N), established in June 2012. Since our flooding hazard records do not include the ones after 1985, the hydrological analysis in this study would not use the data of the Raoping Water Level Station.

In order to organize the records in the GIS formatted database, a flood event had to be compiled in a standardized format with spatial location, date, number of victims, information on economic losses, administrative units, and name of the river section. The catalog of flood events is given in Table 2.

It is of note that biases and uncertainties may arise from ancient toponymic matching, which requires significant efforts in quality control to ensure the accuracy of determining the river segment that flooded. Therefore, cross-checks were carried out in terms of sources, reported data, damage, and fatalities with multiple groups of local experts to alleviate biases and uncertainties. After compiling the records of flood events, we perform mathematical calculations based on the GIS platform to analyze the spatial and temporal evaluation of river flooding in the Raoping region, which shows significant changes over time.

As shown in Table 2, this geospatial flood-event database of historical floods in the Raoping region had compiled a total of 44 flood events with their dates, locations, fatalities or injuries, and notes on economic losses. Twenty-six records among them were documented before the foundation of the PRC, a few of which had detailed notes, such as the ones that occurred in August 1719, August 1891, and June 1960 with hundreds of deaths, which can be considered as among the largest disasters in a long time across the study region. The other 18 records, which were recorded after the foundation of the PRC, were noted by local authorities on the basis of a standard criterion. It is of note that before the establishment of the PRC, the boundary definition of the administrative unit in the records was blurry due to the limitation of technology, but was made clear thanks to the development of survey technology and hydrological engineering standards of China's government after the establishment.

A spatial cluster of flooding hazards was found in the northern and southern parts of the Raoping region, primarily due to the high population and frequent exposure to both riverine floods and coastal storm surges. Less than 10% of the river segments have more than 13 flooding hazards. The river across the current Raoping county downtown, which is located at the outlet of the Huanggang River basin, has more than 13 hazards. Furthermore, the former administrative center of the Raoping region prior to the foundation of the PRC, which was named Sanrao City and Xinfeng City, located in the northern part of the study area, also has massive flooding hazards.

Collectively, this study shows that the flood hazard in the Raoping region is highly affected by meteorological and hydrological factors, as well as the population density and migration during the 500 years. In particular, the Raoping region is a representative watershed for the coastal area of south China, where the coastal regions are prone to high-intensity storms and tropical cyclones. This study also presents a caveat for the migration of the population and political center from the highlands to the coastal region for development purposes.

In addition, the three most destructive flooding hazards occurred in August 1719, August 1891, and June 1960. All these three events occurred around the same location: between the Shifan Stream, Xintang Stream, and Huanggang River. The landform near this location is also extraordinary, mostly a plain area surrounded by multiple mountains. Such terrain made this place suitable both for humans to settle and for water to overflow, finally leading to those tragic results. Such terrain would also highly accelerate the formation of cumulonimbus clouds for mitigating the drainage of the hydrosphere from local human activity. The location south of the study area (near Zhangxi) exhibited a higher number of flooding hazards for the time being. For most of the locations where flooding hazards were recorded, the affected areas were obviously spreading.

This study provides a spatially explicit database of flooding events obtained from literature spanning from ancient times to contemporary periods in a representative region of China. Most of the other flooding hazards databases around the world were developed based on instrumental data (Guzzetti & Tonelli 2004; Barnolas & Llasat 2007; Gaume et al. 2009). However, peak flow discharge does not absolutely indicate flooding events, because humans can build infrastructure such as levees to prevent inundation in their settlements (Stamataki & Kjeldsen 2021). The advancement of the database constructed in this study is that ours has a much larger time-span. The centered political condition for most of the time in China's history allows the local government to comprehensively record the hazardous events that occurred in their managed region (Woodside 2016). Secondly, due to the imperial examination system in ancient China, the literacy rate in rural regions of China is relatively high (Jowett 1989). This phenomenon allows the local landowner to describe local events (such as strong wind, wildfire, and flooding) in different forms such as poetry, essay, and drama (Jao 1949).

The flooding events recorded in our database can be used to implicate the impact of climate change and the interaction with human activities. Flooding events can also be indicators to evaluate the impact of climate change. Here, we can observe the widespread distribution of flooding hazards after 1960, while most of the hazards in the ancient period were observed in limited river segments. Many factors can result in this great difference, such as the lower population density, the scant information collection ability, as well as smaller climate variability in ancient times. Although directly comparing the ancient period with the contemporary period may suffer large uncertainties as discussed, the phenomena we observed can still indicate the meteorological and hydrological characteristics of the study region influenced by human behavior and climate change.

The uncertainty of this work should be acknowledged, which is mainly due to the accuracy of the spatial location when doing the georeferencing of the location matching of flood events (Unnikrishnan et al. 2023). Due to the ancient period in our investigation, the changes in names and movement of the city center could significantly affect the spatial matching process. However, this study provides the most reliable resource for this study region from the perspective of documentary and personal negotiation. The locations of flood events were confirmed by at least three experts in our workshop. The locations with vague descriptions were excluded from this study because they do not meet the criteria of our study. Also, due to most of the records being from the ancient period, the climate data used in this study may suffer uncertainty because the national-scale systematic observation network for China's meteorological records has been established only in recent decades. It is assumed that at least the monthly variation of rainfall and precipitation across this region does not have significant changes (Figure 5(c)).

The second uncertainty may be due to river-channel migration, which would be a common problem in reconstructing historical records of ancient floods. Although we have ancient maps of river networks from ancient documents, the maps are manually delineated without accurate geospatial information, making it difficult to incorporate them into our database. Due to the relatively large forest cover in the study area, the sediment load is lower than in many of China's northern rivers, such as the Yellow River. In addition, Raoping is in the mountain area, where river channels are mostly created and concreted by mountain trends (rocks) instead of soils on the plain area. In this case, the possible reason for river-channel geomorphology changing would be an earthquake, which is able to affect change in a mountain area. Due to the location of Raoping, which is near the center of the Eurasian Plate, there exists a limited possibility for the occurrence of earthquakes, and few such events were recorded within the studied time-period in this area (Bristow 1987). Therefore, the geomorphology of the main channel, which is the Huanggang River in this region, would not show significant changes over 500 years, and the river network of the contemporary period can be applied as the reference for the whole study period.

In general, it should be noted that the criteria used in this research are consistent with other regional and global studies, although the reconstruction of ancient flood records may face significant challenges due to the error of human interpretation. As the economic condition of China during the 500 years may have undergone significant changes, the estimates of economic loss were not included in this study due to the lack of standard conversion methods. However, the notes referring to the economic loss have been added to the database.

In this study, a systematic GIS database containing comprehensive flood events is reconstructed with spatially explicit reference in Raoping, Guangdong, China, based on the records of non-instrumental resources from 1492 to 1985. Using this database, we analyze the spatial evolution, seasonality, and temporal variability of historical floods in the study region. We found that 44 floods were identified in the study region, of which the flood that occurred in June 1960 caused 119 deaths, 43 injuries and huge economic losses including the collapse of 5,689 houses. The results suggest that most of the flood hazards occurred in summer and autumn, with the greatest hazard occurring in June. Due to the lack of records, trend analysis for the records would suffer from great uncertainty. It is of note that the records are consistent with the movement of the city center from north to south, with the southern region being closer to the coastal zone. The migration of the population from the inland to the coastal region may increase the potential exposure to flooding. Given the importance of analyzing historical flood events, this study can help inform future management strategies for flood-hazard prevention and mitigation (Petrucci et al. 2017; Pappalardo & La Rosa 2023).

The method of this study, using the Raoping region as an example, also provides a reference for other research conducted on China's cities and river basins. Flooding is still one of the most destructive hazards for human society, which may substantially influence the social economy and human settlements for a long time-period, even though engineering facilities have been significantly improved nowadays. Therefore, the analysis of ancient records can still provide valuable insights for studies trying to understand the relations between natural hazards and human behaviors, when systematic observations were scarce in ancient times.

In the future, such a data-processing approach can be applied to the Chaozhou region and Guangdong province with more in-depth perspectives. In the future, more research will be conducted on the changes in flood patterns in the context of increased global temperature and intensified climate extremes at much larger scales.

The authors would like to thank the local administration agency for helping the consultant with ancient names of the flooding hazards during several workshops.

Z.W. initiated and designed this research. L.Z. contributed to data processing, result analysis, and interpretation. J.X. contributed to the writing and development of the manuscript. R.Z. gave technical support and uncertainty analysis. Y.Y. contributed to data processing and result analysis.

This research is supported by the following: (1) National Natural Science Foundation of China (NSFC) (no. 42371410); (2) Soft Science Research Project of Shanghai (24692113800); (3) Hainan Provincial Key Laboratory of Tropical Maricultural Technologies (HNTMTOF202404); (4) State Key Laboratory of Tropical Oceanography, CAS (LTO2325).

The relevant datasets of this study are archived in the box site: https://drive.google.com/drive/folders/1WxCJ4rnqLiSWiTail47rmt-sPkq0nlnF?usp=sharing.

The authors declare there is no conflict.