Climate-resilient urban drainage planning: An approach using a GIS-based SCS-CN model

Urbanization as a global trend keeps growing mostly at the cost of the environment. Any urbanization alters the existing land-use and land-cover (LULC) and impacts the natural water cycle. As world population and urbanization phenomena are on the rise, this precious resource is now sometimes abundant, scarce, and polluted (Kundzewicz & Kowalczak 2009; UNESCO 2021). The concept of sustainable urban drainage management is looming large across the world due to its abundance and climate change (CC) impacts with increased frequency and intensity (Soto-Montes-de-Oca et al. 2023). Urban flooding is not constrained by any geographic boundary, rather it is now a global concern faced by megacities and small towns of both developed and developing countries (Browne et al. 2021; Soto-Montes-de-Oca et al. 2023).

The causal relationship between urban flooding and the driving factors is indeed complex. Many studies are available to identify the causes and driving factors of urban flooding. One of the recent studies investigated the correlation between urban flooding and flood hazard due to edge expansion patterns in built-up land and infilling in 12 megacities (Guangzhou, Tokyo, Jakarta, Seoul, Mexico City, Sao Paulo, Lagos, Moscow, Cairo, Los Angeles, Buenos Aires, and London) (Idowu & Zhou 2023). Besides, the physical geomorphic features, i.e., 2D and 3D landscapes of the urban growth have been identified as the key driving factor for urban flooding in nine megacities (Beijing, Tianjin, Shanghai, Xi'an, Nanjing, Wuhan, Guangzhou, Shenzhen and Shenyang) of China (Li et al. 2022; Wang et al. 2022). Intense short duration or incessant rainfall on reduced blue-green space and increased impervious surface due to spatial–temporal variability of LULC change and inadequate drainage capacity and system are the main causes of urban flooding, (Guptha et al. 2021; Li et al. 2022). Many megacities are also facing the adverse impacts of extreme rainfall events due to rapid urbanization, population density, topography and inadequate drainage infrastructures (WWAP 2012; Ahmed et al. 2020).

Among the 44 megacities in today's half-urban world, 19 megacities out of the first 25 are from Asia which are prone to recurrent urban flooding (Idowu & Zhou 2023). Dhaka, the capital and megacity of Bangladesh with the highest population density in the world, is also susceptible to frequent flooding due to CC impacts, intense rainfall during and beyond the monsoon season and inadequate urban drainage facility. The recorded rainfall of 255 mm by the Bangladesh Metrological Department (BMD) on 24 October 2022 and 125 mm on 6 October 2023 in 1 day flooded many parts of Dhaka city for hours and days. IPCC (2022) identified and predicted the high-risk evidence of urban floods and sea level rise for Dhaka. Sustainable Development Goals (SDGs), especially SDG-11 (sustainable cities) and SDG-13 (CC adaptation) and goals of the National Adaptation Plan (NAP) of Bangladesh (2023–2050), articulate the need for developing sustainable urbanization with climate resilience (Iftekhar & Islam 2022). Urban growth in Dhaka continues through the reduction of healthy blue-green space (Nawar et al. 2022; Nithila et al. 2022). Arguably, urban drainage is planned in silos without much consideration for climate resilience, urban hydrology and ecology in Dhaka.

Any new urbanization should consider climate resilience for urban drainage before adopting and finalizing the plans for development. Many studies have explored the context of CC and/or urbanization impacts on the design of combined drainage systems. The combined effect of CC and urbanization could create a dreadful situation in a coastal city in terms of flooding, by making an 80% increase in the maximum inundation depth of a previous rainfall event (Huong & Pathirana 2013). There may be an increase in the volume of combined sewer overflow by 200% over the design flow due to urbanization alone and by 450% due to the combined impact of urbanization and CC (Semadeni-Davies et al. 2008). Hollis (1975) and several other researchers reported that it was not uncommon for an event with a return period of one in 10 years to become one in two years with increased urbanization. Another important feature of urbanized watersheds is that they generate runoff even for relatively small rainfall events. For a duration of 2 h, which would be a representative period for storms that are most critical in urban areas, the results of the study suggest a range of increase from 13 to 21% (Moujahid et al. 2017). For CC effects, the current guidance in the UK is to make an allowance of an increase of 20% in peak flows or volumes over the next 50 years. Other developing countries like Malaysia that receive rainfall similar to India are now following higher return periods (one in five years for residential medium density and one in 10-year for residential high density) for urban drainage design (Mailhot & Duchesne 2010). Although there is no universal model for IUWM/IWRM to minimize the adverse impacts of development on the environment, sustainable drainage/water management concepts such as ‘Sustainable Urban Drainage System (SUDS)’ in Europe, ‘Low-Impact Development (LID)’ and ‘Best Management Practices (BMPs)’ in the United States, and ‘Water-Sensitive Urban Design (WSUD)’ in Australia have emerged during the last five decades (UNESCO & NIUA 2021).

The principles governing an integrated stormwater management approach represent a suitable framework for the adaptation of urban drainage systems to CC (Mailhot & Duchesne 2010). Bruce (2002), Watt et al. (2003) and Moujahid et al. (2017) suggested accommodating the effects of CC and achieving resilience in urban drainage systems. Numerous methods have been developed to estimate the runoff volume and peak discharge and mostly consist of empirical peak runoff methods, hydrologic simulation models, or statistical methods based on the analysis of hydrometric records. Both empirical peak runoff methods and hydrologic simulation models are based on parameters that describe the land-use aspects and the local historical rainfall conditions. However, the Soil Conservation Service – Curve Number (SCS-CN) model, since its conception in 1950, has been the subject of numerous analyses on both practical and theoretical grounds with several modifications, adaptation, and improvement attempts (Soulis 2021). The SCS-CN model is now used for computing peak discharge and volumes for urban hydrology and also for the analysis of land-use changes, i.e., urbanization or LID of urban areas (USDA 1986; Banasik et al. 2014). The CN parameter represents the LULC changes and their impact on surface runoff (Sumarauw & Ohgushi 2012). The SCS-CN model provides better details in describing the physical characteristics of the watershed, which led to its use for larger watersheds compared to the rational method (Yazdanfar & Sharma 2015). Recent application of GIS and RS (remote sensing) technology to the SCS model can help identify the precise soil characteristics and LULC types and so the CN values of different soil types. GIS-based integrated approach is also an effective tool for the investigation of groundwater resources with higher precision (Sathiamoorthy et al. 2023).

To the best of our knowledge, there is no study on urban drainage through the inclusion of CC impacts for assessing and mapping the peak discharge of urban catchments using a DEM and GIS-based SCS-CN model. In doing so, this study first attempts to explore the localized climate anomalies of rainfall using the statistical downscaling of General Circulation Models (GCMs) and apply the statistical tools to quantify the frequency of cumulative rainfall occurrence. Having the DEM-based catchment delineation, the results of rainfall and frequency of occurrence have been included to assess the peak runoff rate or discharge due to LULC change with and without CC applying the GIS-based SCS-CN model. The study area has been selected from greater Dhaka where new urbanization is growing. Two scenarios of LULC have been assessed and mapped in order to easily project the differences in discharge due to base (historical) rainfall and rainfall with CC effects for existing and future LULC conditions. In fact, the output of the model can be used for better catchment management and conservation purposes. More so, it would help the urban planners/local authority to get an insight and certainty on climate resilient urban drainage for future urban growth and make sustainable cities.

The CN method implies a non-linear relation between rainfall and runoff. For a given CN value, the runoff coefficient is not constant but increases with increasing rainfall. In addition to this, the CN value itself is not a time-independent constant just depending on the local type of the soil and vegetation. Instead, it is affected by soil moisture resulting from previous rainfall. In addition, there is a strong seasonal effect. The moisture status of the soil is called ‘Antecedent Moisture Condition’ (AMC). Traditionally, three distinct AMCs are distinguished: AMC-I (dry), AMC-II (average) and AMC-III (wet) (USDA 1986). The cumulative rainfall of the preceding 5 days determines which AMC applies. These thresholds (of cumulative rainfall) are different for the ‘growing’ and the ‘non-growing’ seasons. The value of AMC depends on the 5-day cumulative rainfall of that specific catchment. In this study, AMC-II (average condition) was used for the estimation of runoff.

Catchment boundaries helped to determine the discharge of the area of interest. After preparing all the data, DEM-based model setup was done to delineate the catchment and sub-catchment boundaries. For delineating the catchment, automatic flow direction, flow accumulation and stream network were generated using the automatic catchment delineation tool of ArcSWAT. From this, streams and outlets of the streams were generated. As the main objective is to calculate the discharge, outlets near the river and the growing township were selected. A catchment of the study area (Jolshiri) was generated focusing on the position of the end of the urban area near the river Balu. The study area is considered a single catchment and then it is divided into 16 sub-catchments (Figure 2(b)). The flow direction of the catchment is shown in Figure 2(c) and the existing soil type in Figure 2(d).

Almost 90% area of the catchment had clay-type soil having a hydrologic soil group (HSG) ‘D’. Due to infilling of sands for urban development most of the low-lying land of group ‘D’ soil has been converted to group ‘A’ soil. This was confirmed by an onsite investigation as per USDA (1986). The calculation of existing land-use has been considered with CN value for ‘A’ and ‘D’ group soils. In future, the existing land-use is likely to be converted to HSG ‘A’ by filling up the low-lying lands. In that case, the entire catchment will be converted with HSG ‘A’.

The annual rainfall of 42 years in Dhaka is relatively varying with an annual average about 2,041 mm. The highest rainfall at 3,028 mm was recorded in 1984 and the lowest at 1,169 mm in 1992. The daily rainfall data in a particular year is converted to 1, 2, and 5 days of consecutive rainfall by summing up the rainfall of corresponding previous days. The maximum amount of 5 days of consecutive rainfall for each year was taken for analysis. Daily rainfall in Dhaka is remarkably varying when observed during the months of the year. The maximum rainfall events within the identified years are observed during the months of May to October in a year. The maximum daily rainfall between 1980–2021 has the highest at 341 mm in 2004 and the lowest at 62 mm in 2012. Different rainfall events have been analyzed from the daily rainfall data to determine the 1-, 2-, 3-, 5- and 10-day cumulative rainfall events with a view to assessing the extreme rainfall events. The yearly maximum rainfall event for the last 42 years is in Supplementary material, Table S2.

The assessment of future CC has been projected by IPCC through the Couple Model Inter-comparison Project (CMIP). A set of GCM simulates the plausible future climate conditions based on different scenarios. Recently, the IPCC released its 6th Assessment Report (AR) that utilized CMIP6 GCMs. In this report, IPCC has introduced the SSPs which are scenarios of projected socioeconomic global changes up to 2100. The assessment of local CC impact demands downscaling of GCM data that are very coarse in resolution (approximately 100–300 km) to capture local phenomena. Using statistical downscaling (SDSM 4.2.9), localized CC information has been identified using the SSP5-8.5 scenario. The annual rainfall for Dhaka is predicted to increase by 6.70% with an average of 2,380 mm. The monthly plausible changes for SSP5-8.5 in rainfall during the 2080s are predicted during December, January, February (DJF) −1.10%; March, April, May (MAM) +4.52%; June, July, August, September (JJAS) +11%; October, November (ON) +5.66%. A downward trend in rainfall is observed during winter while a significant increase is predicted during monsoon (JJAS) for Dhaka and the reduction of rainfall is observed during winter (DJF).

Rainfall event selection is one of the important aspects of drainage planning and for the application of the SCS-CN model. For calculating the discharge, rainfall events are selected by analyzing the rainfall intensity. In general, for the design purpose 5-day, 10, 25 and 100-year return period events are selected. Hence, for calculating the discharge, three rainfall events are selected considering both base (historic) and CC scenarios in this study as shown in the following:

• i Base (Historic) without CC – 5–day, 10-, 25-, and 100-year return period, and

• ii Future with CC – 5–day, 10-, 25-, and 100-year return period.

Rainfall intensity of 5-day cumulative rainfall for 10- and 25-year return period corresponds to 356 and 421 mm that already happened or closer to it in 1986 was 401 mm on 30 September in Dhaka. Again, the rainfall intensity of 5-day cumulative rainfall of a 100-year return period is 516 mm and more than that already happened with 600 mm on 14 September 2004. Accordingly, percentage deviation was calculated from the historical rainfall of that year before applying the SCS-CN equation. However, for 10- and 25-year return period for base and CC, maximum rainfall of two months in 1986 of August and September were considered for computing the peak discharge from each sub-catchment for existing and future land-use change. Similarly, for a 100-year return period, maximum rainfall of August and September of 2004 was considered. In all cases, CC impacts accounted for an 11% increase in the rainfall considering the results of climate downscaling for the study area of Dhaka. Table 2 shows the design rainfall, i.e., 5-day cumulative rainfall with and without CC for the application of SCS-CN. The daily rainfall data of 1986 and 2004 were adjusted with the increase and decrease according to identified percentage deviation for 10-, 25- and 100-year return periods in order to apply the SCS-CN equation. Considering this, the drainage management system and the structures could be designed for 356 mm rainfall (one in 10 years), 421 mm (one in 25 years) and 516 mm (one in 100 years) return period.

CN is a crucial factor for runoff estimation in the SCS-CN method. A high CN means high runoff and low infiltration; whereas a low CN means little runoff and high infiltration. CN values range between 0 to 100 and higher CN values are associated with higher runoff potential. Traditionally, an area-weighted average CN is used to study the runoff of a catchment. In this study, CN was identified for each sub-catchment to precisely find the runoff condition of each sub-catchment. The CNII values for AMC-II (average condition) are used in general for the runoff calculation. CN values for AMC-I (dry condition) and AMC-III (wet condition) conditions are calculated using conversion equations as given in USDA (1986). In this study, weighted CNs were calculated with respect to the catchment area with AMC-II condition.

Based on the LULC of each sub-catchment, CNII (average CN) values for both existing and future LULC conditions have been determined for each sub-catchment. In most cases, CN reduced in future conditions except for sub-catchment C5. As the value of S depends on the value of CN, the recharge capacity or maximum potential retention ‘S’ was determined by substituting the values of weighted CN. In fact, the CN value remains the same in each sub-catchment as the AMC-II condition is considered. The results of identified CN values and maximum potential retention ‘S’ in each sub-catchment are given in Table 3.

The runoff depth (Q) was calculated by substituting the value of weighted CN in the SCS-CN equation. After finding the runoff depth, discharge (q) was estimated by dividing by the time. Table 4 shows the estimated peak discharge of each sub-catchment as runoff rate for drainage management. While comparing the results of peak discharge for both existing and future LULC conditions, in most cases peak discharge of future LULC has been reduced except for sub-catchments C4, C5 and C16. It is observed that land-use change for urbanization is not always dependent on excess runoff, rather it is the sustainable management of runoff accounting for the CC effects of rainfall.

Detection of rainfall trends and CC impacts is crucial in predicting runoff or peak discharges for resilient urban drainage planning. The approach illustrated in the paper provides a detailed outline and procedure to assess the climate resilient urban drainage planning using the DEM of the selected urbanization and through the application of GIS-based SCS-CN model using existing and future land-use scenarios due to urbanization. Though the approach is time intensive, it can provide the precise output of rainfall-induced discharge that would help towards achieving resilience in urban drainage planning and management.

This study predicted the localized CC impacts of the study area by climate downscaling techniques with relevant statistical analysis. Prediction of CC effects on rainfall is mostly available from GCMs. Climate anomalies of rainfall at a local level have been ascertained for using it as a percentage of increase or decrease from historical rainfall. To achieve climate resilient urban drainage planning and design, the selection of rainfall events with CC is critically important as opposed to using values of IDF curves or historic design storms which are not based on CC. Hence, the foreseeable effects of CC necessitate a change in the overall approach to resilient urban drainage planning. The importance of assessing the changes in localized rainfall trends as opposed to relying on GCMs with statistical analysis has also been highlighted in a recent study by Chadee et al. (2023).

Precise runoff or discharge estimation is always challenging in hydrology. Recent studies involved advanced novel hybrid intelligence predictive numerical model based on successive variational mode decomposition algorithms and also machine learning methods (MLMs) towards achieving precision in predicting runoff and assessing stage-discharge relationship (Kisi et al. 2024; Parsaie et al. 2024). However, this study used the GIS-based integrated approach to delineate the catchment using ArcSWAT and to obtain the detailed flow direction of rainfall-induced runoff from the high-resolution DEM with the application of CC impacts on local-level rainfall.

The spatial–temporal change of the LULC and the soil profile of the urban catchment provided the basis for precise CN estimation. CN estimation with AMC (II) condition is generally considered for SCS-CN application. With the help of area-weighted CN as per sub-catchment, 5-day cumulative rainfall with and without CC effects have been ascertained for the application of SCS-CN equation and quantifying the peak discharge of each sub-catchment. The return period has been selected from the best practices and principally from the requirement of urban development. The maximum potential retention, S, is estimated first for each sub-catchment and then converted to quantify the peak discharge using SCS-CN equation. For both existing and future LULC scenarios, the peak discharge with and without CC effects can be mapped easily using GIS and help differentiate the impacts of CC. Based on the peak discharges of sub-catchments, drainage planning for the entire catchment can be decided that ensure climate resilience. However, options for various runoff reduction methods can further reduce the peak discharge of the catchment. The suggested assessment method and mapping can be a useful tool for future resilient urban drainage planning, design and management at the local level for the urban planners/authority and avoid the risk of urban flooding or waterlogging. The approach illustrated in this study can be compared with other advanced approaches to identify the variations in precision levels of predicting runoff or peak discharge with CC impacts in future studies.

Climate-resilient urban drainage planning is critical for any urban growth as it impacts human life and resources in many ways. CC is now a global concern. Many megacities and cities that never faced any flood are now experiencing the ominous effect of floods due to CC impacts on rainfall. The most common application of drainage planning is the selection of design storms and the calculation of peak discharge for the design of drainage structures. Typically, drainage is planned from the historical rainfall analysis with a selected return period from IDF curves and considering the economic efficiency, spatio-temporal rainfall variability, and antecedent soil moisture conditions. The computational method is either a rational or urban runoff model. However, drainage networks with a high design return period will have a resilient drainage capacity as opposed to a lower design return period. In order to achieve resilience and minimize the risks of urban flooding, there is a pressing need to incorporate localized CC impacts on historical rainfall and choose an optimum design return period for urban drainage planning. The suggested approach below is for the climate resilient urban drainage assessment and mapping of the peak discharge using DEM and GIS-based SCS-CN model:

• i DEM-based catchment delineation using ArcSWAT,

• ii Identification of LULC change and soil type by applying GIS and RS for both existing and future LULC conditions due to urbanization,

• iii Identification of a 5-day cumulative rainfall event from historical rainfall,

• iv Identification of localized CC impacts of rainfall using SSP5-8.5 scenario applying climate downscaling,

• v Inclusion of CC impacts of rainfall as a percentage in a 5-day cumulative historical rainfall event,

• vi Selection of return period as per urban development plan and structural importance of LULC change,

• vii Return period analysis of 5-day cumulative rainfall events using statistical tools and adjustment of percentage deviation of increase or decrease in 5-day cumulative rainfall for observed year,

• viii Identification of area-weighted CN (average) values of different LULC for both existing and future conditions,

• ix Assessment of maximum potential retention (S) of delineated catchment from the different LULC conditions for varying CN, and

• x Finally, the estimation of runoff (Q)/peak discharge (q) of sub-catchments applying SCS-CN equation and mapping the peak discharge (q) with and without CC effects of rainfall for both existing and future LULC conditions.

Local urban planners/authorities may decide the return period based on the urban development plan. However, a wide range of return periods based on long-term rainfall data with CC effects would help assess and address the CC realities. Based on the above assessment, urban drainage can be mapped easily in order to identify the differences in effects of with and without CC on rainfall, runoff and peak discharge for resilient urban drainage planning. It is evident from the assessment that increased rainfall may not increase the urban drainage requirement, rather it depends on the combined response of rainfall, LULC and soil profile of the urban catchment. The ultimate CN due to LULC change will govern the drainage requirement.

This study deeply acknowledges the data support and cooperation of BMD, SOB, Jolshiri Urban Development Authority, Military Institute of Science & Technology (MIST) and Bangladesh University and Engineering Technology (BUET), Dhaka.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.