Flood inundations and risk mapping in a tidal river: a case study for the Kelani River basin, Sri Lanka

23 The downstream low-lying regions of the Kelani River, including some areas in the Districts of 24 Colombo and Gampaha, Sri Lanka, frequently face severe inundations due to extreme rainfalls 25 in the upper basin. In the present study, 1-D and 2-D hydrodynamic models in HEC-RAS have 26 been used to examine the flood inundations in the tidal influenced Kelani River with ground 27 observations and remote sensing. The HEC-RAS model has been used to produce a flood 28 hazard map for hazard assessment in the lower Kelani River basin under different return 29 periods. Furthermore, expected discharges for different return periods have been estimated 30 using the hydrological model HEC–HMS with the updated intensity depth frequency curves for 31 the Kelani River basin. Sentinel 1 imagery and field survey results are used to validate the 32 simulated flood inundation extent; hydrodynamic model results validated against observed 33 stage measurements; hydrological model validated against discharge measurements. 34 Further, the validated hydrodynamic model showed the high capability to capture flow 35 processes (Nash-Sutcliffe coefficient = 0.90 and Pearson coefficient of correlation = 0.95) 36 along with inundation extent (Success Index = 0.90) of selected historical extreme events. 37 Then the hydrological model is used to predict the flows of the Kelani River basin with a good 38 agreement (Nash-Sutcliffe coefficient = 0.91 and the Pearson coefficient of correlation = 0.93). 39 Finally, flood risk zoning for different return periods are developed based on the present model 40 which would be a useful benchmark to design and implement flood control and mitigation 41 measures for the river basin. 42


Introduction 46
Each year millions of families from different parts of the world are affected due to floods, 47 causing US$662 billion in damages since 1995 (UNISDR 2015). In South Asian, most 48 townships and urban cities are located at popular flood risk areas (Tariq 2011   The use of hydrodynamic simulation to forecast flood extent in the Kelani River is 70 particularly important for water managers and decision-makers in Sri Lanka. The river basin 71 has significant socio-economic benefit. The commercial and administrative capital of Sri Lanka 72 is located in the lower basin of Kelani River. The upper basin is covered predominantly with vegetation, whereas the lower basin is 115 heavily urbanized. The river basin receives an average annual rainfall of nearly 2,400 mm and 116 carries a peak discharge of 800 − 1,500 m 3 /s during the monsoonal periods (i.e. especially in 117 South-West monsoon period from May to September). Frequently, during the South-West 118 monsoon period, the lower reach of the basin is subjected to flooding as recorded from the 119 flood gauge located at Nagalagam Street. 120

Data 121
The data collected for the present project included precipitation data, stream flow data, land-122 use characteristics data, flood field measurements, satellite images and a Digital Elevation 123 Model (DEM) of the river basin. Table 1 Table 2 shows the rank and hazard levels 225 employed in the present study relative to the Nagalagam Street gauge reading (0 MSL). 226 227

Flood Vulnerability 228
The vulnerability was classified into population and building vulnerability. Both population and 229 building vulnerabilities were analyzed based on the age groups and building materials 230 respectively in each GN (Grama Niladari) division. In the present study, the population's 231 vulnerability to flood was computed using population data obtained from the 2012 census. 232 In population vulnerability, the highest ranks are assigned to age groups less than 5 233 years and above 60 years as they are deemed to be more vulnerable. Low and intermediate 234 ranks are assigned to the age group of 26 to 59 year and 6 to 25 years respectively. At the 235 same time, to assess building vulnerability, the highest ranks are given to buildings that are 236 susceptible to damage due to minimal impact. 237 The vulnerability index for each categorization in GN division was calculated by 238 where: is vulnerability in a GN division (-), is a fraction of age group/building 241 material type out of the total in the GN division, is vulnerability rank of each group (-) and 242 n is number of categories respectively. 243 The vulnerability index calculated using equation (12) was standardized by overlaying with the 244 present model's inundation extents to obtain revised vulnerability index for various return 245 periods in the Kelani River basin (as shown in Table 3): 246

Flood Risk 247
The risk maps for various return period floods are computed by assuming the equation × . The Table 4 shows the risk classification employed in 250 the present study. results showed that the HEC-RAS hydrodynamic model is able to capture the temporal 259 variability of the flood flow process very well. However, the comparisons also revealed that 260 the model underestimates the stage on many occasions (Fig. 3). It is understood that the 261 cause for this underestimation is primarily resulting from the negligence of rainfall-runoff at the 262 lower basin in the model.  Table 6. 268

Model simulated inundation comparison with satellite observed inundation related to 269
May-2018 showed that there is an 83% of hit rate and 94 % of accuracy. This shows that the in both events shows that there is a high accuracy (i.e. 90%) but the hit rate was 61% as well 284 as the false alarm rate and the ratios were 17% and 3% with 0.815 bias score. This suggests 285 that the model underestimates the inundations relative to the surveyed inundations. All prediction. 299 Table 8 shows the 3-day total rainfall and the expected discharges derived from the 300 HEC-HMS model using IDF curves. Based on these results, the Kelani River basin appears 301 to have faced only a maximum flood of 10-year return period; Table 8 is a good assessment 302 of potential damage for higher return periods. 303

Hazard Mapping and Risk Zoning 304
Flood hazard mapping for different return periods were conducted with estimated discharges 305 from the previously validated models. According to Tables 9, 15  During this study we considered vulnerability in terms of population and buildings. The evaluation of risk in terms of buildings shows similar behaviour to that with the 320 population (see Fig. 9) but there is an increase in the building risk areas (Fig. 7

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.