Modeling flow of the major canals of Sylhet city using the EPA SWMM runoff model

Bangladesh has a low-lying landscape with more than 230 waterways, making it one of the most vulnerable disaster-prone countries in the world. As the lowest riparian in a huge transboundary river basin, Bangladesh faces an increasing threat of massive flood exposure. Though flood is a natural event and cannot be avoided, the damages can be reduced by responding appropriately and providing necessary drainage channels. However, urbanization and changing demographic features of urban cities have increased the exposure to urban flood hazards (Zhou 2014). Developing countries like it are facing the wrath of this kind of disaster, as the vulnerability level is very high across these regions due to poor socio-economic conditions and haphazard settlements (Khadka & Bhaukajee 2018). The nature of urban flooding is different from riverine flooding due to its hazardous characteristics, including environmental, social, and technical aspects. The flood peaks in urban areas increased from 1.8 to 8 times due to faster flow time (Rangari et al. 2016). So, the increasing trend of urban flooding is now drawing the attention of urban planners worldwide.

Due to the climatic and geological conditions, Sylhet city faces a large amount of rainfall each year (Hasnat et al. 2019). This heavy rainfall generates a large amount of runoff flow, which is evacuated through the channels (locally termed ‘charas’) linked to the Surma River (Munna et al. 2018). These natural channels act as the main drainage network system of the city. However, due to rapid urbanization, improper management, and siltation, these drainage canals within the catchment no longer could convey as much water as they formerly did within their active domain (Ghosh & Mistri 2015). Since the subsidiary natural channels are the main media to carry out the runoff to the river, it is necessary to know whether the existing cross-sections of these channels can carry out heavy runoff and mitigating the flood problem at the time of a heavy shower. This results in the inefficiency of drainage systems and canals and increases the risk of flooding and waterlogging (Ten Veldhuis 2010). In addition, the sewage is disposed of directly to these charas, which reduces the depth and deteriorates the water quality. Such inadequacy in the drainage systems spreads several types of diseases and increases the mortality rate. A defective stormwater management system contaminates the limited freshwater resources and pollutes the environment, whereas a sound management system can minimize the losses if the priorities and the choices are set up.

Watershed development and increased impervious surface development led to decreased infiltration capacity and increased runoff velocity. These circumstances effectively improve the efficiency of water transport to rivers. As rainfall and stormwater runoff from urban areas cause river pollution and watershed impairment, it is essential to integrate drainage system routing into these single and multiple storm event models (Tsihrintzis & Hamid 1998). Many hydrological models are used, such as Storage Treatment Overflow and Runoff Model (STORM), Technical Release Model (TR - 20/55), Hydrologic Engineering Centers River Analysis System (HEC-RAS), Storm Water Management Model (SWMM) and many more. However, studies on urban flood modeling in Bangladesh are limited. Rahman et al. (2014) assessed the drainage system of Chalna city, considering the city's response to hydrology analytically. They included a digital elevation map, satellite picture, and land use for comprehending the city's drainage network (Rahman et al. 2014). Khan (2015) developed a guideline for urban flooding and stormwater drainage for Mymensingh. Afrin et al. (2021) applied Hydrologic Engineering Centers River Analysis System (HEC-RAS) to find the peak runoff considering catchment delineation and the future land use of Dhaka. Kumar & Bhagavanulu (2007) studied an inundation map for a city on the Adyar River's bank. Hossain et al. (2016) performed a study that focused on designing a stormwater drainage system specifically for Sylhet Agricultural University, considering the frequent waterlogging caused by blockages in the existing drainage system, often resulting from sediment and solid waste accumulation (Hossain et al. 2016).

Previous studies have not addressed the development of a rainfall–runoff model specifically for the local canals (charas) within the Sylhet City Corporation (SCC) area. This gap is significant for performing capacity analysis of the major charas, which are crucial in managing most of the stormwater. Furthermore, there is a lack of understanding of the current conditions under various storm scenarios and existing land-use patterns. Notably, no prior research has simultaneously modeled two of the most critical sub-systems: the Malni Chara and Goali Chara sub-systems. These canal sub-systems are pivotal in the urban flooding crisis, often exceeding their capacity in many parts of the City Corporation area, thereby contributing significantly to the havoc caused by urban flooding.

Although many hydrologic and hydraulic models available can predict the effects of watershed changes and characterize stormwater runoff peaks and volumes (Yan et al. 2013), SWMM has been used in the study. SWMM has the key advantage of running the model using a dynamic routing system and is more user-friendly than the others. Though SWMM is developed to model urban drainage network settings, it can be well suited to model natural watersheds (Jang et al. 2007). In addition, the wide acceptance by the scientific society and engineers, user-friendly interface, and technological advancement make SWMM better suited for use. So, after properly calibrating and validating to provide realistic scenarios of runoff generation, hydrological models may be used to comprehend and assess these channel reactions to future land-use changes and climate changes (Du et al. 2012).

The specific objective of this research has been multi-fold. Firstly, rainfall–runoff models were developed for the significant canals (charas) of Sylhet city. Then, calibration and validation of the formulated hydrologic models were carried out respectively to assess the model performance. However, this study focuses only on the calibration and validation of the wide ranges of charas within different storm conditions and to find the sensitivity of different parameters in developing a rainfall–runoff model for the two significant sub-catchments within the area. Also, the study provides a comprehensive understanding of the stormwater management needs in this rapidly urbanizing region, offering significant insights for urban planners, stakeholders, and policymakers.

The required input data were collected from different public institutions or websites. All the inputted datasets that have been used are summarized in Table 1.

Sylhet and its neighboring areas are geologically located in the Sylhet Trough. Physio-graphically, it falls underneath the category of tertiary hills and comprises a tropical climate having an annual average highest temperature of 23 °C (August to October) and an average lowest temperature of 7 °C (January) and the mean annual rainfall is around 3,963 mm (Roy et al. 2014; Bari et al. 2015).

In SWMM, the user must select the infiltration process as well as the flood routing method. For channel flow routing in SWMM, three different types of routing techniques can be used: steady flow routing, kinematic routing, and dynamic routing (Rossman et al. 2008). By modeling the pressured and backwater effects brought on by downstream flow constraints, the dynamic wave routing approach entirely solves the Saint-Venant equations (Rossman et al. 2008; Greenberg 2015). For this reason, we used the dynamic wave routing method (Greenberg 2015). Also, for the infiltration process, the Green-Ampt method was selected.

EPA SWMM requires three major parameter categories for runoff quantity modeling including the physical catchment characteristics, rainfall, and infiltration data (Niyonkuru et al. 2018). The physical characteristics include sub-catchment area, percentage of impervious area, sub-catchment width, average slope, surface depression storage, and surface roughness (Paterne 2019a). This information has been derived by processing the topographic data using the ArcMAP and drainage data collected from the local authority.

However, the soil map for Sylhet city or the specific soil characteristics for Malni and Goali sub-systems were unavailable. Hence, the default value of Type C soil has been used.

Percentage of the impervious area has been calculated using the existing land-use data provided by the Urban Development Directorate (UDD). The land-use map was georeferenced and imperviousness (%) for every sub-catchment within the study area was calculated. The impervious area was divided by the total area to determine the percentage of imperviousness for each sub-catchment.

For conduit properties, ‘irregular’ shapes for the conduits were inputted using the cross-section editor. The cross-section map collected from the SCC provided the length, station distance, and elevations for the section concerned.

Manning's roughness value (n) was selected from the ASCE manual of practice for gravity sanitary sewer design and construction. As the charas are natural channels with an irregular section, the value 0.07 represents the channel roughness (ASCE 1982; Bizier 1982). Other input parameters are mentioned in Table 2.

The standard range is defined between 0 and 1 for the index of agreement where a value of zero means no correlation at all and a value of 1 means that the dispersion of the prediction is equal to that of the observation (Paterne 2019a).

Figure 4(b) illustrates the ratio of variations for different outputs which represents the sensitivity to changes in various parameters. The roughness of conduit has an average sensitivity of 0.55 for runoff while flow width (K) has an influence with 0.83. Imperviousness, conduit roughness, and conduit length have negative coefficients, which means that lowering this input parameter will lead to higher output values (Niyonkuru et al. 2018). From this figure it is clear that the imperviousness percentage, flow width (K), and roughness are the most vital parameters in the Goali Chara sub-catchments and Malni Chara sub-catchments.

Both figures collectively emphasize the substantial impact of impervious percentage and flow width on peak flow during simulations of tropical urban catchment runoff, both in terms of quantity and quality, as per SWMM modeling (Chow et al. 2012; Paterne 2019b). This is in line with Li et al. (2016), who found that conduit roughness and imperviousness percentage greatly affect runoff components, a conclusion also supported by our analysis. Table 4 shows the correlation and the sensitivity class for different parameters that have been used in this study.

Continuity errors verify the accuracy of the SWMM's calculations. Two different kinds of continuity errors are calculated: one for flow routing and the other for runoff modeling (Greenberg 2015). These errors are calculated for the system by summing the final storage and total outflow and then subtracting it from the sum of the initial storage and total inflow (Rossman et al. 2008; Greenberg 2015). For maintaining the validity of the model and the system, the continuity error should not exceed 5% (Rossman et al. 2008). The runoff and routing continuity error found for the Malni Chara sub-system is −0.13 and 1.18%, whereas for the Goali Chara sub-system it is −0.15 and 1.49%.

To enhance the accuracy of the runoff rainfall model, it is crucial to have access to high-resolution data with shorter intervals. However, the absence of such detailed data in the study area limited our ability to match the peak flow for the catchments precisely. Additionally, the lack of soil data for the canal area also constrained the model's robustness. Access to these data could have significantly improved the precision of runoff and groundwater recharge estimates. Additionally, for calibration, we utilized runoff data from 2016, and for validation, data from 2019 were used. The model's accuracy would have benefited from a broader range of recorded discharge data. We could only collect discharge data corresponding to rainfall for these two years due to data scarcity. Integrating more extensive recorded data would significantly enhance the model's accuracy and predictive capabilities.

This research focused on calibrating and evaluating the EPA SWMM for stormwater runoff modeling in Sylhet city's main drainage canals. A critical aspect of this study was the division of the total drainage system into two sub-systems and further segmentation into sub-catchments, facilitating precise identification and analysis. The EPA SWMM model parameters were meticulously developed, taking into account the unique characteristics of these sub-catchments, along with the area's topography and drainage network data. A significant part of this study was the parameter sensitivity analysis, which underscored the robustness of the chosen parameters. This robustness was further affirmed by the successful calibration and validation of the model within acceptable performance ranges. The indices used in the study uniformly indicated an excellent fit for the modeled data, reinforcing the model's reliability in simulating the runoff quantity of the city's drainage canals.

The practical implications of this study are significant. The calibrated model is a powerful tool for assessing the performance of drainage systems and crafting flood mitigation strategies. It offers a framework for other urban areas facing similar challenges, highlighting the potential of such models in urban planning and decision-making processes. By applying these findings, urban planners and policymakers can better evaluate the effectiveness of existing drainage systems, identify solutions to flood-related issues, and develop comprehensive stormwater management strategies for the watershed.

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

The authors have no relevant financial or non-financial interests to disclose.

Gulam Md Munna prepared the manuscript. Gulam Md Munna received cooperation from Md Mahmudul Hasan, Ahmed Hasan Nury, and Jahir Bin Alam in performing computation, developing the methodology, and manuscript preparation. Md Misbah Uddin, Mohammad Shahidur Rahman, and Shriful Islam assisted in analysis and draft preparation. Furthermore, the authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

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

The authors declare there is no conflict.