ABSTRACT
This study focuses on South Bandung, an area where flooding is a recurring problem, causing significant damage to infrastructure and displacing local residents. The region's vulnerability stems from its geographical location, surrounded by rivers and low-lying areas, coupled with challenges such as the heavily polluted and sedimented Citarum River. This research examines the role of retention ponds in Andir and Cieunteung as a strategic response to these flood management challenges. The study employs a mixed-methods approach, integrating remote sensing and Geographic Information System (GIS) analysis with deep learning-based semantic segmentation techniques to evaluate the hydraulic performance of the retention ponds. Normalized difference water index and semantic segmentation approaches using multispectral images (red, green, and blue and near-infrared) from the PlanetScope constellation satellite are utilized to assess water bodies and flooding patterns, while field observations and stakeholder interviews provide qualitative insights into the operational effectiveness of these infrastructures. While this study highlights that the construction of retention ponds can help mitigate flood events, it also emphasizes the importance of an integrated approach combining urban planning, community engagement, and multi-scale infrastructure development in addressing the complex challenges of flood resilience and urban sustainability.
HIGHLIGHTS
Evaluated flood mitigation efficacy of retention ponds in South Bandung using mixed methods.
Combined normalized difference water index and deep learning with the PlanetScope multispectral imagery to map floods.
Demonstrated reduced flood impacts post-construction of Andir and Cieunteung retention ponds.
Highlighted urban resilience and sustainability through integrated flood policies.
Advocated for holistic flood management strategies.
INTRODUCTION
Urban flood management has become increasingly critical due to the rapid pace of urbanization and the intensifying effects of climate change. As of 2018, 55% of the global population resides in cities, a figure projected to rise to 68% by 2050, with significant urban growth occurring in the Global South, particularly in Asia and Africa (United Nation 2019). Southeast Asia, including Indonesia, is experiencing substantial urban expansion, which exacerbates flood risks due to increased impervious surfaces and reduced vegetation, leading to higher surface water runoff and decreased infiltration. Between 2013 and 2020, dryland and irrigated agriculture in the study area declined by 3 and 4%, respectively, while forest, bush, and plantation cover decreased by less than 1%, and industrial and settlement areas expanded by 11% (KLHS Bandung Regency 2023–2043).
The Bandung Metropolitan Area, including the neighbouring five districts of Sumedang Regency, exemplifies the challenges faced by Indonesian cities. The conglomeration had a total of 8.2 million residents in 2014, with Bandung City, as the development hub, being one of the five largest cities in Indonesia (Tarigan et al. 2016). It is surrounded by mountains and 2,400 m-high volcanic terrain, and this seismically active region with its tropical monsoon climate, characterized by wet and dry seasons and heavy rainfall during the wet season, making it highly susceptible to natural hazards such as floods, earthquakes and landslides (Gumilar et al. 2015; OECD 2018). Dayeuhkolot, as well as other sub-districts in Bandung Regency, including Baleendah, Sapan, Bojongsoang, and Margarahayu, has a history of flooding incidents (Al Farisi 2021; BNPB 2021; Putra 2024). Heavy rainfall, sedimented rivers and inadequate drainage systems in the area are some of the main drivers of flooding. The region's vulnerability is exacerbated by its geographical location, surrounded by rivers and low-lying areas, and the heavily polluted and sedimented Citarum River. Despite efforts to enhance drainage systems and construct flood barriers, frequent and intense flooding continues to damage infrastructure, disrupt livelihoods, and impact health due to contaminated water sources (Gumilar et al. 2015; Tarigan et al. 2016; Ritonga & Sinaga 2019).
Flooding has resulted in significant infrastructure damage and disruption to residents' lives in Bandung. Frequent flooding has had a negative impact on a range of rural activities, such as agriculture, causing crop damage, horticulture, and small-scale businesses. These activities are vital for the local economy and provide livelihoods for many residents. Additionally, the flooding has also led to impacts on health due to contaminated water sources and increased mosquito breeding grounds. It is therefore crucial for authorities to address these issues promptly to ensure the well-being of the community and prevent further economic setbacks (Ritonga & Sinaga 2019).
These hybrid green-blue-grey retention pond infrastructures function as Nature-based Solutions, delivering a range of co-benefits beyond flood control, including the enhancement of urban ecosystem services (Lechner et al. 2020; Lourdes et al. 2021). Co-benefits from these types of infrastructure developments include opportunities for recreational activities such as fishing, jogging, and picnicking, which are particularly popular during weekends and afternoons (Lechner et al. 2021). These activities not only promote physical health and relaxation but also strengthen social bonds as families and friends gather to enjoy leisure time together (Nath et al. 2018; Lourdes et al. 2021). Integrating green and blue spaces and accessible recreational areas into the infrastructure contributes to a higher quality of life, encouraging a healthier, more active, and socially connected community. These benefits are integral to Nature-based Solutions, which aim to provide multi-functional infrastructure addressing both hydrological and socio-environmental needs. In addition to retention ponds, other types of Nature-based Solutions, such as bioswales, rain gardens, and constructed wetlands, can play a crucial role in addressing hydrometeorological hazards in urban landscapes (Debele et al. 2019) and have been trialed in Bandung (Sagala et al. 2022).
This study evaluates the effectiveness of these retention ponds within the context of current flood management policies and planning practices. To achieve this, we undertook a holistic analysis using a mixed-methods approach, including a review of management strategies, field observations, remote sensing and GIS analysis. The paper begins by outlining the background of South Bandung's flood issues and the role of the retention ponds. We then conduct a review of management strategies supported by field observations. In the final part of the paper, we apply flood mapping methodologies: application of the normalized difference water index (NDWI) spectral index and deep learning-based segmentation approaches to map flooding in the study area before and after the construction of the retention ponds. Finally, we discuss the broader implications of these findings for flood management strategies in urban areas prone to hydrometeorological disasters, particularly in the context of climate change. Our research provides insights into the potential of infrastructural interventions to enhance flood resilience in rapidly urbanizing regions. The findings have implications for policymakers and urban planners seeking to develop comprehensive flood management strategies that align with sustainable urban development goals.
BACKGROUND
Based on the 2015 population data for West Java Province, the Citarum River Basin is home to 18,640,000 people, or 35.8% of the province's total population (BBWS Citarum 2023). Bandung Regency is the most populous administrative region in this basin, accommodating 21% of the basin's total population. Situated at an elevation of 600 m above sea level, Andir comprises 11.07% of the total area of Baleendah, or 378.30 hectares. With 36,038 inhabitants, the population density is 95 individuals per hectare (BBWS Citarum 2023). Situated at an elevation of 700 m above sea level, Cienteung comprises 16.98% of the total area of Baleendah and encompasses 590.20 hectares. The population density of this 60,299-person village is 104 inhabitants per hectare (BBWS Citarum 2023). Both Andir and Cieunteung are situated in close proximity to estates and industrial facilities, in addition to agricultural regions that have suffered the negative consequences of urbanisation. These villages are predominantly populated by individuals who are engaged in labour, employment, or entrepreneurship. The educational attainment of households resides within the range of primary school to senior high school, with only a minority having completed tertiary education (Nurdini et al. 2021).
The region serves as a natural reservoir for water from various river systems; this area transforms into a ‘bowl’ during intense precipitation events, increasing the likelihood of flooding. Additionally, the area experiences land subsidence of up to 5 cm/ year, primarily caused by excessive groundwater extraction. This issue is further compounded by the accumulation of massive amounts of solid waste – up to 500,000 m3 – clogging the river (BBWS Citarum 2023). Before the construction of the Andir and Cieunteung Retention Ponds, flooding affected approximately 370.8 hectares of land in this area. These retention ponds, located in Dayeuhkolot, were designed to mitigate flooding by storing up to 80% of excess water, significantly reducing the impact of heavy rainfall, which is common in this region (BBWS Citarum 2023).
The construction of the Andir and Cienteung Retention Ponds in Dayeuhkolot is meant to significantly store excess water and prevent flooding during heavy rainfall, as heavy precipitation is common in this area. Due to heavy precipitation, the region is especially susceptible to flooding during the rainy seasons, specifically from January to March and November to December (BBWS Citarum 2023). Additionally, the area experiences intense flooding every 5 years, known as the 5-year flood pattern.
The Bandung basin, including the area near the Citarum River, such as Dayeuhkolot in which Andir and Cienteung retention ponds are located, has experienced land subsidence in the past couple of decades with an average subsidence rate of −8 cm/year (Gumilar et al. 2015). These land subsidences that happened near Andir and Cienteung Retention Ponds have been linked to various factors (Solihuddin et al. 2021). These include excessive groundwater extraction, the build-up of sediment, the specific geological conditions in the area, and the potential impact of nearby industrial structures. In particular, previous studies (Gumilar et al. 2015) suggested that there is a strong correlation between the land subsidence in Cimahi, Dayeuhkolot, Majalaya, and Rancaekek with excessive groundwater extraction. Nevertheless, the intensive construction of industrial buildings and complexes in those areas may have contributed to the high land subsidence, which can reach −16.9 cm/year in a specific location and time. These factors may contribute to the worsening of land subsidence near the retention ponds, increasing the risk of flooding. Additionally, the lack of proper drainage systems exacerbates the situation.
In contrast to the rainy season, the dry season brings drought to the region surrounding Andir and Cienteung Retention Pond, an area primarily characterised by agricultural land and villages. During this period, water resources can become scarce, impacting both agricultural activities and the local communities. The scarcity of water resources during the dry season not only affects agricultural activities but also poses challenges to the daily lives of local communities. Access to clean drinking water becomes a concern, and people may have to travel long distances to fetch water for their households.
FLOOD MANAGEMENT PRACTICES AND MITIGATION STRATEGIES IN THE CITARUM RIVER BASIN
In recent years, major Indonesian cities have grappled with challenges related to rapid urbanisation, such as escalating population growth and environmental threats, exemplified by recurrent monsoon season floods. Bandung faces a heightened susceptibility to floods, with 577 high-intensity flood events recorded from 2018 to 2019 (BNPB 2021). The Citarum River's overflow contributes significantly to the elevated frequency of flood events in Bandung, with villages along the river, particularly in the Baleendah and Majalaya sub-districts of Bandung Regency, experiencing approximately ten flood events annually.
A series of studies have been carried out concerning flood management in urban and peri-urban areas; especially for low-lying lands, extra flood prevention and mitigation efforts are imperative (Tanuwidjaja et al. 2010). However, the social and economic development of the area is critically influenced by the sustainability of management (Asian Development Bank 2019). Flood management initially requires a time investment, and over the intermediate and long terms, it potentially offers more efficient decision-making and implementation for land cover and water resource management (Asian Development Bank 2019). Politics does, nevertheless, play a role in flood management. Local government policies and regulations can impact the success of flood prevention and mitigation efforts, as well as the overall sustainability of management practices (Simanjuntak et al. 2012; Houdret et al. 2014).
Neighbouring Jakarta serves as a model for how Indonesian governments have worked to prevent flood damage, addressing increasing severity and frequency over the decades. The primary mechanism for flood mitigation consists of the east and west canals, which redirect the excess water away from the 13 rivers running through the city (Octavianti & Charles 2019). However, as the city experienced increasing population density and regular inundation caused by the decline in its river, flooding emerged as a consistent concern. In response, the urban polder concept was formulated where flooding is managed through an integrated drainage system consisting of pumping stations, dykes, drains, retention ponds, and outfall structures (Tanuwidjaja et al. 2010). Nevertheless, inundation zones still occur within the polder systems during intense flood events of certain heights and durations as a result of inadequate polder maintenance (Noviadriana et al. 2019). With proper management and operation, there is potential for an increase in the number of polders and a reduction in flood risk (Djalante et al. 2017).
Another example, from Indonesia, is the Bengawan Solo River region, where authorities have implemented a series of polders as a preventive measure against flooding. This area experiences near-annual flooding as a result of the river's expansive and winding course and currents (Mawandha et al. 2018). Apart from protecting agricultural regions and settlements, the polders were also constructed to regulate irrigation. Various scenarios of pumping system capacities and crest elevation in flood control gates are presented in the study, which employs mathematical models to simulate hydraulic processes in flood-prone regions. The integration of lateral gates into the design of polders as flood control structures has the potential to enhance their efficacy in alleviating the problem of inundation in agricultural and settlement areas.
Outside of Indonesia, the implementation of controlled flooding in retention areas along the Elbe River in Germany has demonstrated the efficacy of the retention areas in mitigating flood peaks. This approach was particularly successful in reducing the damage in the town of Wittenberge, 30 km downstream (Förster et al. 2005). A conceptual model and simulation of an economic evaluation indicate that, from an economic perspective, the use of retention areas for flood protection is extraordinarily cost-effective. Retention ponds solely for flood prevention without sedimentation calculation may result in unexpected costs and reduced effectiveness over time (Verstraeten & Poesen 1999). Therefore, regular monitoring and maintenance of retention ponds are crucial to ensure long-term success in flood protection efforts.
In addition, urban design and water management, including retention ponds and polders, must be incorporated into the macro-spatial plan (Tanuwidjaja et al. 2010). Without proper planning, retention ponds, which are frequently used to reduce flooding in urban and peri-urban river catchments, could exacerbate the problem (Birkinshaw & Krivtsov 2022). In order to enhance the efficiency of the polder system, operational flood management is also crucial. This includes controlling the timing of gate openings for flood water diversion and monitoring dyke breaching (Huang et al. 2007). In order to preserve the effectiveness of flood control measures over an extended period, retention ponds and polders should be designed with a long-term operational guarantee, taking into account the watershed functional limits and the potential anticipation effect (Miguez et al. 2015). Furthermore, the incorporation of biodiversity and environmental elements into the living spaces of local residents through the integration of urban landscape design may be a key component of sustainable urban development (Keyvanfar et al. 2021).
The Citarum River Basin, which is considered the most strategically significant river basin territory in the country, is currently grappling with the adverse effects of industrial expansion and urbanisation. As a result, groundwater has become scarce in multiple areas, and the environment is deteriorating (Asian Development Bank 2016). Land subsidence of 7 cm annually was directly caused by excessive groundwater extraction (JICA 2010; Asian Development Bank 2013; Gumilar et al. 2015). Land subsidence and groundwater extraction have the potential to modify the water flow trajectory within a drainage system, which could potentially worsen the frequency of flooding events (Ge et al. 2014). Furthermore, its vulnerability has been significantly increased due to climate change (Asian Development Bank 2019).
The Citarum River has inundated the Baleendah area extensively, with 38 out of 45 flood events occurring there between 2013 and 2018. Due to its proximity to the Citarum riverside, roughly 36% of Baleendah, encompassing around 1651.5 hectares, faces potential annual flood impact, affecting areas such as Andir and Cienteung (BBWS Citarum 2023). Residents of Andir and Cienteung have demonstrated adaptability in surviving incremental challenges, employing strategies like adding building floors using conventional structures or implementing gradual structural modifications until the building envelope is completed (Nurdini et al. 2021).
Land use change in Dayeuhkolot, Bojongsoang, and Sapan 2010–2018. Source: Development Planning Agency at sub-national level of west Java modified by authors 2024.
Land use change in Dayeuhkolot, Bojongsoang, and Sapan 2010–2018. Source: Development Planning Agency at sub-national level of west Java modified by authors 2024.
The drainage infrastructure in some urban areas may be inadequate to handle heavy rainfall, leading to urban flooding. In addition, the increase in impervious surfaces such as concrete and asphalt further contributes to the problem by preventing water from infiltrating into the ground. This leads to a higher volume of water flowing overland and overwhelming drainage systems. Moreover, the lack of proper maintenance and regular cleaning of drainage systems can also worsen flooding, as clogged drains are unable to effectively channel water away during heavy rainfall events. Self-erected flood mitigation on houses, such as erecting a wall in front of each house, causes more flooding in the city. Furthermore, climate change has led to increased uncertainty in rainfall patterns and intensities (Asian Development Bank 2021). South Bandung may experience more frequent and severe rainfall events, further heightening the risk of flooding.
Several retention ponds and polders have been built to reduce flood risk in some Citarum River areas. The result showed that the development of retention ponds and polders reduced flood occurrence locations, the flood duration from monthly to hourly, the flood elevation from 100–200 to 50 cm, and flood-impacted areas so the people are not required to move their belongings (BBWS Citarum 2023).
MATERIALS AND METHODS: GIS FLOOD MAPPING AND REMOTE SENSING ANALYSIS
To delineate the inundated zones in South Bandung and thoroughly evaluate the efficacy of existing flood mitigation strategies, we utilized advanced remote sensing techniques. To perform the mapping, we collected high-resolution satellite imagery from PlanetScope. PlanetScope is a constellation of approximately 130 satellites operated by Planet.com, covering the entire Earth's land surface with a collection capacity reaching 200 million km2/day. For the same Earth's surface location, images were captured by the PlanetScope constellation satellites every 24 h with around 3 m/pixel resolution. Despite offering more frequent revisit times and higher spatial resolution compared to other well-known Earth observation satellites, such as Landsat and Sentinel, PlanetScope imagery can suffer from uneven geometric and radiometric quality due to technical and processing limitations (Aati et al. 2022). Additionally, common remote sensing challenges, such as cloud and haze obstruction, still affect image clarity. Therefore, image pre-processing techniques, including radiometric correction and multi-date image mosaicking, remain essential for ensuring data quality and consistency.
Regarding the generated products, PlanetScope supports up to 8-band imagery such as red, green, and blue (RGB), near-infrared (NIR), red edge, yellow, green I, and coastal blue. In our task, where our objective is to delineate flooded areas, we acquired not only RGB but also NIR imagery, particularly emphasizing the significant advantage of NIR bands in water detection. NIR bands are very useful in detecting water due to the pronounced differences in reflectance values between water and non-water land surfaces captured by the NIR electromagnetic signals (Wieland et al. 2023).
Historical flood data in Dayeuhkolot, Bojongsoang, and Sapan. Source: KemenPUPR modified by authors 2024. The maps in 2018 and 2020 were produced by the authors using NDWI combined with the area boundary information.
Historical flood data in Dayeuhkolot, Bojongsoang, and Sapan. Source: KemenPUPR modified by authors 2024. The maps in 2018 and 2020 were produced by the authors using NDWI combined with the area boundary information.
Remote sensing technologies enable us to systematically observe and analyze floods on a large scale, offering valuable insights into their spatial extent, dynamics, and impacts. However, the inherent variability in flood events, characterized by diverse receding times ranging from a couple of hours to a couple of days, poses a challenge in detecting inundated regions solely based on remote sensing images. To tackle these issues, we cross-referenced flood event information obtained from Internet news sources, systematically documenting the dates of occurrences. We subsequently cross-verified this data against available imagery from PlanetScope, focusing particularly on RGB images and ensuring cloud-free coverage of the areas of interest. After confirming image availability and clarity, we proceeded to download the images, employing water segmentation techniques such as the NDWI or deep learning-based semantic segmentation approaches to delineate the flooded areas. The NDWI is a spectral index used to identify and monitor water bodies, calculated using the NIR and green spectral bands from satellite imagery. It enhances water features while suppressing soil and vegetation information, making it useful for detecting and analyzing surface water extent, including flooding events. On the other hand, deep learning is a subset of machine learning that uses neural networks with many layers to automatically learn patterns and features from large datasets, enabling it to perform tasks, such as semantic segmentation, which classifies every single pixel in the images into semantically meaningful categories. This method enables the accurate delineation of flooded and non-flooded regions in images, provided the model is properly trained (Feliren et al. 2024).
We generated flood maps for two primary objectives. First, we aimed to address the gaps in flood data provided by the government spanning from 2010 to 2023, as illustrated in Figure 5. Given the government's original data, which displayed flood events every 2 years at the district level, we employed NDWI with −0.1 as the threshold to complement the missing flood map in the years of 2018 and 2020. The threshold here defines NDWI values (computed using NDWI = (NIR – SWIR)/(NIR + SWIR, where SWIR represents short-wave infrared values)) for distinguishing water bodies from non-water, where a larger value than the threshold is classified as water. Since a highly precise flood mapping was not required for this purpose, we utilized NDWI to delineate flood regions, which were subsequently converted into village/district-level boundaries. It is important to note that our methodology involved applying straightforward criteria to determine whether a village or district experienced flooding in a particular time. Specifically, if a substantial amount of water was detected within the village (after cross-referencing with flood event news), we categorized the village or district as being affected by flooding.
Second, we produced flood maps to provide a more granular view and detailed perspective of the Dayeuh Kolot region. To achieve this, we adopted deep learning-based semantic segmentation approaches, and state-of-the-art machine learning models that leverage neural network architectures that excel in pattern recognition and feature extraction from complex datasets (LeCun et al. 2015). These models, when trained on vast datasets of remote sensing imagery, have demonstrated superior capabilities in automatically learning hierarchical representations of features, enabling them to produce more accurate delineations of flooded areas compared to conventional methods such as the NDWI (Wieland et al. 2023). In particular, we employed our previously developed semantic segmentation models, ProCANet (Feliren et al. 2024), designed based on the U-Net model architecture (Ronneberger et al. 2015). We modified the original U-Net by incorporating another encoder where the first one encodes R-G-B-NIR bands and the second one encodes NIR bands. The model has demonstrated an accuracy of 0.988 and an IoU score of 0.815 evaluated using Sen1Floods11 (Bonafilia et al. 2020). We then added an attention layer to combine each modality from both encoders. This mechanism leverages convolutional layers to construct masks that highlight significant regions within the first encoder and the second encoder data independently, thereby enhancing the salience of pertinent details. The details of the approach can be seen in Feliren et al. (2024). This approach was chosen to obtain a finer-grained understanding of how floods propagated within the Dayeuh Kolot region. The outputs of these models are displayed in Figures 7 and 8, showing the flooded regions as bright colours.
RESULTS
NDWI analysis
According to historical records (Figure 5), the size of floods in the area has varied. In some years, the floods have been relatively small and manageable, while in others they have caused significant damage (BBWS Citarum 2022). This variability highlights the importance of implementing effective flood management strategies that can adapt to changing conditions and mitigate the potential impacts of future floods.
Infrastructures that were already built by the government around South Bandung. Source: BBWS Citarum modified by authors 2024.
Infrastructures that were already built by the government around South Bandung. Source: BBWS Citarum modified by authors 2024.
Comparison before and after the Cieunteng retention pond was built. Source: Authors 2024
Comparison before and after the Cieunteng retention pond was built. Source: Authors 2024
Comparison before and after Andir pond was built. Source: Authors 2024.
Deep learning flood mapping analysis
In order to analyse how effective Cieuntung and Andir retention ponds are in reducing the flood, the PlanetScope satellite was utilised to generate the flood mapping of the DayeuhKolot region from 2017 to 2023. Information will be gathered both prior to and subsequent to the construction of these two retention floods. The effectiveness of the Cieuntung retention pond is illustrated in Figure 7. These images depict the flooding that occurred in 2017 and 2018 prior to the construction of the Cieunteung Pond in 2018 (left image). The water image depicted in the white patch was obtained using satellite data. The area remained severely flooded in 2019 and 2020, even after the completion of the project (right image). Derived from Figure 7, flood-affected areas (white pixels) increased from 26.72% in 2017 and 43.68% in 2018 to 59.58% in 2019 and 61.33% in 2020. Although the Cieunteung Pond can help mitigate flood events, it is not always effective in reducing flooding incidents due to factors such as heavy rainfall exceeding the pond's capacity, which likely contributed to the increased flooding in 2019 and 2020 (Nugroho 2020; Sinaga 2020).
Figure 8 maps the effectiveness of the Andir retention pond prior to and after its construction in December 2020. Here, the white patch indicates that the water has significantly decreased. Derived from Figure 8, flood coverage (white pixels) decreased from 61.33% in March 2020 to 11.70% in April 2020, and further to 7.89% in 2021 and 1.50% in 2023. According to the flood mapping conducted using deep learning techniques, the presence of the Andir Pond subsequent to the construction of the Cieunteung Pond indicates a decrease in the volume of flooding in the vicinity. The combination of these two retention ponds has proven to be effective in reducing the impact of flooding in the area, as shown by the decrease in water levels post-construction. This highlights the effectiveness of synergistic flood mitigation interventions.
While our results are informative, satellite images have several constraints. One significant issue is the difficulty of acquiring clear images during the wet season due to cloud cover, which coincides with the occurrence of flooding. Other dates were devoid of precipitation or cloud cover. In most cases, flood events were only mapped one to two days after the event, specifically after the rain. Furthermore, changes in flood extent can be attributed to various upstream factors, including land use changes that increase impervious surfaces and other drainage infrastructure altering hydrological flows.
DISCUSSION
The findings of this study underscore the role of retention ponds in enhancing flood resilience in South Bandung while also highlighting significant areas for improvement in urban flood management practices. It is apparent from our mapping that the implementation of flood mitigation measures, particularly the construction of retention ponds, could effectively safeguard against flooding. Sustained monitoring and evaluation of the efficacy of these interventions are necessary to guarantee enduring protection against flooding. Satellite data are crucial in assessing the long-term impact of retention ponds on flood mitigation efforts. However, beyond these two catchments, we were unable to map the area because additional data collection and analysis are necessary to determine the potential impact of flooding in those regions.
The functioning of a retention pond aligns with the concept of temporary storage for water that eventually returns to the river. The idea is that the pond serves as a holding area during times of excessive rainfall or flooding, preventing immediate overflow into the surrounding areas. Our mixed-methods evaluation has demonstrated that the retention ponds in Andir and Cieunteung substantially mitigate flood risks by effectively capturing and storing excess rainfall, thus reducing the volume of water flowing downstream. However, the evaluation also reveals critical gaps that must be addressed to optimize the impact of these infrastructures.
The hydraulic performance of the retention ponds, as analyzed through the NDWI and deep learning-based flood mapping, indicates a marked reduction in flood frequency and intensity in the areas surrounding Andir and Cieunteung from 61.33% in March 2020 to 11.70% in April 2020, and further to 7.89% in 2021 and 1.50% in 2023. This aligns with previous studies that highlight the utility of retention ponds in managing floodwaters and reducing peak flow during heavy rainfall events (Verstraeten & Poesen 1999; Förster et al. 2005). The ponds have functioned effectively as temporary reservoirs, thereby alleviating pressure on existing drainage systems and decreasing the risk of overflow in downstream areas. Moreover, the existence of a pump designed to return water to the Citarum River indicates an endeavour to regulate and oversee the water levels within the pond.
Despite their effectiveness, the occurrence of flooding in the areas under study nonetheless prompts enquiries regarding the efficacy of the retention ponds under extreme conditions and potentially requires further interventions to improve their functionality. These interventions could include increasing the capacity of the retention ponds to accommodate larger volumes of water or implementing additional drainage systems to ensure efficient water flow. Additionally, regular maintenance and monitoring of the pump systems may be necessary to ensure their proper functioning during periods of heavy rainfall or flooding.
Beyond the technical aspects of flood management interventions, this study draws attention to larger questions about the policy and planning imperatives. We find that the retention ponds have not been fully integrated into a broader urban flood management strategy. Our study reveals that their design and operation are not sufficiently aligned with other flood management measures and urban planning practices. This disconnection has led to inefficiencies in the ponds' performance and underutilization of their potential benefits. For example, the retention ponds' operation often does not consider seasonal variations in rainfall or the broader hydrological dynamics of the Citarum River Basin, leading to suboptimal flood risk reduction (Miguez et al. 2015).
The misalignment between technical solutions like retention ponds and urban planning frameworks is a common challenge in flood risk management. This issue is not unique to South Bandung but is also evident in cities such as Jakarta and Manila. In Jakarta, where the government has made strides through the Jakarta Flood Control Strategy, flood mitigation efforts have faced difficulties in fully integrating flood risk management with urban planning due to fragmented governance, rapid urbanization, and insufficient coordination between various agencies, leading to inefficiencies in flood management (UNFCCC 2023). The city's focus has often been on reactive measures rather than long-term, integrated strategies. In contrast, Manila's flood control strategy has aimed at improving integration through the National Flood Control Master Plan, which encourages collaboration among local governments, urban planners, and communities. This centralized coordination, along with proactive infrastructure investments and community engagement, has allowed Manila to better align flood risk management with urban development goals, thereby enhancing its overall flood resilience (World Bank 2017).
Moreover, retention ponds not only mitigate flood risks but also contribute to socio-environmental outcomes, such as increased recreational activities and enhanced local biodiversity. These benefits align with the design principles of Nature-based Solutions, where multi-functionality enhances ecological services and community well-being (Keyvanfar et al. 2021). The socio-environmental benefits of the retention ponds include recreational opportunities such as fishing and picnicking and improved community well-being such as jogging, which are significant but often overlooked while devising flood management strategies. Retention ponds enhance local biodiversity by providing habitats for aquatic and terrestrial species while supporting climate adaptation through stormwater management and reducing extreme weather impacts. Our field observations and stakeholder interviews validate these socio-environmental contributions, and underline the importance of embedding such features in urban resilience strategies. Field observations revealed diverse species thriving around these ponds, such as monitor lizards, various fish species, and frogs, demonstrating their ecological significance in urban settings. Stakeholder interviews further highlighted the socio-environmental benefits, such as improved water quality as the retention pond works as natural filtration, the cooling effect on surrounding microclimates, and the creation of recreational spaces such as fishing and community gathering spots that foster community engagement. These findings emphasize the need to integrate retention ponds into broader urban resilience strategies, positioning them as essential tools in mitigating climate-related risks while promoting sustainable and livable urban environments. For instance, by incorporating green spaces and recreational areas into the design, retention ponds can contribute to urban aesthetics and promote social cohesion (Keyvanfar et al. 2021). However, the study also notes the need for better maintenance practices to ensure long-term efficacy and prevent sedimentation, which can reduce the ponds' storage capacity over time (Verstraeten & Poesen 1999).
Community engagement and education play a crucial role in the successful implementation of retention ponds. By involving local communities in the planning and management processes, we can enhance the acceptance and effective use of these infrastructures. Public awareness campaigns and participatory planning initiatives can help integrate retention ponds into the local fabric, ensuring that they are supported and utilized effectively by the community.
Our study thus highlights the critical need for a more holistic and integrated approach to flood management in South Bandung. To effectively address the region's flood challenges, it is essential to embed retention ponds within a comprehensive flood management framework and an integrated approach that combines urban planning, community engagement, and multi-scale infrastructure development so that localized interventions such as retention ponds align with regional hydrological strategies. Urban planners and policymakers must consider a combination of natural and engineered interventions as well as social and community-based solutions as part of a broader flood mitigation strategy that encompasses improved drainage systems, land use planning, and climate adaptation measures. Such a unified approach would strengthen the region's flood resilience and ensure a more effective response to future flood risks.
For future studies, it may be beneficial to conduct a more comprehensive analysis of the areas surrounding the retention ponds by developing flood risk maps using multi-criteria decision analysis (MCDA). Our current work focuses on evaluating the effectiveness of retention ponds through remote sensing analysis (combining the NDWI, deep learning approaches, and government data) along with field observations and interviews, which differs from studies that produce flood risk maps using MCDA (e.g., Ouma & Tateishi 2014; Gacu et al. 2022; Gacul et al. 2024). To create flood risk maps, methods, such as the analytical hierarchy process (AHP) (Ouma & Tateishi 2014; Gacu et al. 2022) or fuzzy AHP (Gacul et al. 2024), can be utilized, which are widely recognized in flood risk assessment research for providing a structured framework to prioritize and rank factors contributing to flood hazards. These maps would help identify high-risk areas around the retention ponds that require greater attention for flood mitigation.
CONCLUSION
This study demonstrates that while the development of flood prevention infrastructure projects such as retention ponds is effective in mitigating flood risks, their true potential is realized only when they are part of a broader, integrated flood management framework. This framework combines urban planning, policy support, and community engagement, ensuring that localised solutions (like retention ponds) are coordinated with regional or national-level systems (like river basin management). An integrated strategy can harmonise efforts across scales, addressing potential gaps or redundancies and fostering resilience and sustainability in flood-prone areas. Conversely, a disjointed approach, which treats retention ponds as isolated interventions rather than components of an interconnected policy response, limits their efficacy and fails to harness their socio-environmental benefits.
Our findings suggest that South Bandung could significantly benefit from adopting a more coordinated, multi-sectoral approach to flood risk management. First, integrating flood management strategies into comprehensive urban planning frameworks is crucial. Drawing from the successful example of the National Flood Control Master Plan in Manila, which emphasizes collaboration between local governments, urban planners, and communities, and the lessons learned from Jakarta's Flood Control Strategy, a central coordinating agency that links flood mitigation efforts with broader urban development projects would help to ensure effective coordination and integration. The region could implement a blend of nature-based and engineered solutions, as seen in other global cities such as Kuala Lumpur and Singapore, to enhance ecological resilience while controlling infrastructure costs. Regular stakeholder engagement and community-driven planning are also essential so that flood management initiatives are socially inclusive and locally supported. For example, in Semarang, Indonesia, the Building with Nature (BwN) programme has involved local communities in mangrove restoration and coastal infrastructure planning, which has improved flood resilience as well as provided livelihood opportunities for residents through aquaculture and ecotourism. Finally, aligning retention pond operations with regional hydrological dynamics and climate adaptation strategies is critical to ensure their sustained effectiveness in mitigating flood risks, optimizing water management, and adapting to future climatic uncertainties.
The implications of our findings extend beyond technical flood mitigation measures. They highlight the necessity of a paradigm shift in how urban resilience is conceptualized and implemented. This shift requires urban planners, policymakers, and communities to adopt a more integrated and interdisciplinary approach to flood management – one that recognizes the interconnectedness of physical infrastructure, environmental stewardship, and social empowerment.
To advance this agenda, future research and practice should focus on enhancing the integration of retention ponds with other flood management and urban planning strategies. This includes aligning their operation with regional hydrological dynamics, incorporating climate adaptation measures, and ensuring their design supports multi-functional urban spaces. There is also a need for policy frameworks to evolve to provide stronger support for such integrated approaches and facilitate the coordination of efforts across different sectors and levels of governance.
Ultimately, this study calls for a reconceptualization of urban resilience – one that moves beyond conventional flood management practices to embrace a holistic, integrated, and community-centered approach. Such a vision is essential for building cities that are not only resilient to floods but also adaptable to the need for integrating environmental sustainability perspectives into urban development agendas. This integrated approach, if effectively implemented, has the potential to transform flood-prone areas like South Bandung into models of sustainable urban development, where infrastructure, policy, and communities are inextricably linked in the pursuit of a resilient future for our cities.
ACKNOWLEDGEMENTS
We would like to thank Dr Michaela Prescott and Dr Dwinanti Rika Marthanty for their insightful discussions on the topic, as well as BBWS Citarum for providing essential data related to the Andir and Cieunteng retention ponds in South Bandung.
FUNDING
This work was funded by Monash Indonesia Research Innovation Seed Funding Grant (2022).
ETHICS STATEMENT
This research follows the research ethics procedure as regulated by Monash University.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict of interest.