FLOOD HAZARD MODELLING USING HYDRAULIC SIMULATION MODEL AND SATELLITE IMAGES: A CASE STUDY OF CHAJ DOAB, PUNJAB, PAKISTAN

: The main objective of the study is to determine the highly risked areas and to estimate flood extend through different return periods. HecGeo-RAS is used along with Hydrologic Engineering Center River Analysis System (HEC-RAS) to build a model which have analyzed the flow of water. The river and floodplain geometry were obtained by SRTM-DEM and flood mapping is finalized by the vector layer analysis in ARC GIS. Frequency analysis for the period 1986-2014 (29 years), instantaneous discharge data performed by Gumbel's, and LP3. Results depicts that District Mandibahudin always flooded on both sides by Jehlum and Chenab rivers. An interpretation of findings explicitly revealed that the 50-year return cycle flood at Trimmu (5299 m 3 s-1 ) would inundate 500 percent more area than the natural surge. Urban areas are also vulnerable to flooding such as Mandibahudin, which were also evident in the July / August 2014 flooding.


INTRODUCTION
Floods are one of the crucial hazardous pure catastrophes, which might affect localities, livestock and nearly all the humanitarian activities.Amongst the south Asian countries, Pakistan is the country which is extremely affected by Monsoon induced flooding (Khalil & Khan, 2017).The major supply of water provide is Indus River along with its tributaries, and to make the most of this source in better approach dams and barrages have been established on them (Ogras & Onen, 2020;Raza et al., 2019) .Data that is collected through remote sensing techniques is used to build digital elevation models, but land surveying may also be used to build DEM.In geographic information systems, digital elevation models are frequently used and are the most popular source for maps that are created digitally (Das et al., 2016).
Remote sensing is an instrument or technique just like mathematics.By designing the danger risk chart, we can constantly track the potential dangers by Remote Sensing, and these maps can aid with coping with emergency conditions following any disaster (E. et al., 2009;Pradhan, 2009).The primary source of knowledge for many researchers is remote sensing for threat or catastrophe evaluation.Several research on flood hazard zoning and flood risk mapping, using GIS software and remote sensing data, have been reported.Data from optical remote sensing has been commonly used around the globe to track the flooding (Elkhrachy, 2015;Ullah & Zhang, 2020).
The RS satellite will immediately see a view of the broad flooded city (Burns et al., 2014).According to (Klemas, 2015;Samboko et al., 2020) assess that by implementing remote sensing techniques, river inundation zones, river level and river discharge are quickly defined.
The use of GIS began at the beginning of 1990 with hydrological modelling (Devantier & Feldman, 1993;Miller et al., 2007).DTM digital terrain models have been used to simulate surface water flow and assess drainage basins and, under varying environmental and climatic conditions, forecast runoff and flooding.When handling natural threats, GIS plays a significant role (Deborah Thomas, 2005).
The modern construction of army corps, arc GIS is a useful tool in the simulation of the flood, 2000 (Cameron, director of flood plain) related to assess the impact of used to provide RAS products were introduced in GEO.The function of the GIS pre and post-process from the HEC RAS -geo spatial data (RAS), to use the extension and extent inundation depth of the results of the model in the flooding of conditions to provide effective and efficient it.GIS environments and flood hazard level to indicate the flood risk of decision-making authority to a very efficient way to communicate (Demir & Kisi, 2016;Yerramilli, 2012).
Pakistan is among the five South Asian states with the highest regular number of folks bodily subjected to floods, that come about normally resulting from storm systems that begin from Bay of Bengal in the course of the monsoon from Jul-Sep.Floods significantly strike Punjab and Sindh whereas hill torrents tend to prompt the hilly areas of North Western Frontier Province, Baluchistan and the northern federally administered areas.Pakistan is a rustic with humid climate situations which ends up in excessive precipitation within the region.Pakistan has suffered many floods with different severity ranges and plenty of different natural disasters in 68 years after its independence.Severe floods occur almost yearly at some parts of Pakistan mostly in areas of KPK and in south Punjab on account of which tremendous loss of life, massive scale damage of property is faced.Floods largely occur throughout monsoon season from July to October (Shahabi et al., 2020).
Over the past decade, GIS has attained a fundamental position in flood risk mapping.GIS performs a necessary function in flood risk management (Isma & Saanyol, 2013;Kobayashi & Porter, 2012).One of many difficulties with Mangla, Marala and Trimmu barrages is the rising riverbed owing to very large deposit disposal around the upstream regions.Mangla Dam, Marala and Trimmu barrages traps big sediments surplus from the upstream storage and digression structures.Moreover, the pond area is also fed yearly along with massive quantities of sediment battered from the greatly dissipated catchment areas of the Suleiman Range.These tedious silts are carried away figuratively all the way through western tributaries of the Indus River.
The primary goal of this venture is to assemble a river flood mannequin in GIS atmosphere using the satellite imagery.This analysis work will likely be useful to make a way to plan and improve, in an effort to reduce the flood losses in the examiner area and a gateway in the direction of flood risk management utilizing RS and GIS techniques.I.
To GIS based mapping of highly risked areas II.
To identify the safer places during flooding situation III.
Estimate flood extend for return periods The outcomes will be furnishing support for evaluating the present circumstances of the realm, in addition to predict future risk zones for the upcoming potential natural flood events.Floods hazards maps and evaluation for many susceptible areas for flooding can be utilized for future planning purposes.
Study area: River flooding is a major problem in Pakistan because of the monsoon rains and melting snow that cause Pakistan's rivers to overflow their banks.There are three significant rivers that generate flooding in Pakistan the River Jhelum, River Chenab, and the lower portion of both rivers (Khan et al., 2011).Jhelum River and Chenab Rivers form the Chaj doab, which encompasses the land between them.From 73°30'-74°28' East, it is between 32°7' and 33° 0' North.It is located in the southernmost part of Kashmir.Gujrat, Mandi Bahauddin, and Sargodha are three of the most important districts in this region (Figure 1).

Figure 1 Study area map
Trimmu barrage was built to control water flow in the river Jehlum and Chenab for and flood control purposes across different areas of Mandibahudin District.The barrage is premeditated for a low flood as well as a super flood ranging from 1000000-1200000 cusecs (Sajjad et al., 2019).Mandibahudin is one of Pakistan's most historic districts.Additionally, it is Punjab's most populous and biggest district.Total area of the district is 8249 km 2 .The population of the district is approximately 3826000 (Mahmood et al., 2019).Soil condition of Trimmu barrage is dry, barren land.There are two major rivers which flow through the District Mandibahudin (Jhelum and Chenab).Trimmu barrage is built on the River Indus while a tributary is joining it.Any fluctuations in water level of Trimmu barrage has a direct impact on the localities (Change et al., 2014;Stewart et al., 2018).Trimmu barrage is situated in Tehsil Kot Addu of District Mandibahudin.It feeds four canals of the Tehsil and has a capacity of 1 lac cusecs (Mahmood et al., 2019).

MATERIALS AND METHODS
Datasets and software:DEM is the fundamental input of the hydraulic modeling software, as it contains all the information about the water profile like depth, slope, aspects etc.We used DEM with resolution of 30 m.We used SRTM DEM and our study area was covered within a single tile, so we extracted the area in Arc Map.To make it appropriate to be able to use it in HEC-RAS, it was re-projected in Arc Map 10.8 from GCS to Projected Coordinates System.After that contours of interval were made of the AOI and then TIN was generated using these contours to create the geometric data.
The CSI (Consortium for Spatial Information) website offers interactive elevation models (DEM) for the entire globe, covering all countries in the world.These data were obtained using the international initiative of the Shuttle Radar Topography Mission (SRTM), which created the fullest high-resolution digital topographic database on Earth (Rabus et al., 2003).Several studies have used SRTM data (Gorokhovich & Voustianiouk, 2006;Sanders, 2007).Watershed delineation was performed using the hydrology methods in Arc Map 10.8.The river and floodplain geometry were obtained by means of the SRTM-DEM.Satellite data, included Landsat 8 images were downloaded from the website of USGS https://earthexplorer.usgs.gov.Field survey was done by the researchers herself.The data collected from the field consisted of latitude, longitude, elevation, discharge, cross section.To define land uses, color images from Landsat images were used, and then these classes are used to estimate the n-values of Manning needed by HEC-RAS for hydraulic computations.Two gauging stations are available with discharge data for an appropriate number of years: the Marala Barrage and the lower Trimmu Barrage.From 1986 to 2014, the full instantaneous discharge data at the Marala Barrage and the lower Trimmu Barrage was available.The regular mean discharge was possible at the Trimmu Barrage, without any holes, for the duration 1986-2020.Regression models were used to fill in missing data points from the Marala and lower Trimmu Barrages' maximum instantaneous discharge data.For the linear regression model, twelve pairs of daily and instantaneous discharge data were used at Marala Barrage and lower Trimmu Barrage.Using Eq. 1, the linear connection between the maximum instantaneous discharge and the maximum daily discharge may be found (1).It was observed that the correlation coefficient (r) was 0.94 between the actual instantaneous and daily median flow.
= 1.238  − 305.47 (1) where For the period 1986-2014 (29 years), the overall instantaneous discharge data as well as the mean regular discharge are available at Chinote.The values of maximum and minimum instantaneous discharges produced Trimmu barrages.The lower-reach start node refers to the upper-reach end node, while the lower-reach end node was expected to be at Marala Barrage and the lower Trimmu Barrage.The flow data at the lower Trimmu was used for the entire upper reach, assuming stable state conditions.It was believed that the discharge at Chinote was flowing over the entire lower reach under steady state conditions.

METHODOLOGY
Floodplain maps for Pakistani territory along the Chaj Doab were developed as part of the current investigation.
Figure 1 depicts the stages involved in the development of floodplain maps.In order to obtain floods that correspond to various return periods, we must first conduct an analysis of the frequency of available observed discharge data, prepare DEMs using this data, delineate watersheds and drainage networks using HEC-GeoRAS, prepare geometric data using HEC-GeoHMS, implement HEC-GeoHMS, and create floodplain maps using GIS.
Flood intensity and water surface profiles may be estimated using peak flows for varying return periods and return times.Flood frequency analysis (Chow et al., 1998) was conducted to achieve flood peaks at the Trimmu for various return times.For frequency analysis, three widely used frequency distribution functions have been used to predict extreme floods, namely the log-Pearson form III distribution (Vb, 1982), the Gumbel or extreme value distribution (Ang, A.H.-S.and Tang, 1975;Ullmann, 1961) and the log-normal distribution (Hoshi et al., 1984;Stedinger et al., 1993).Chow (Chow et al. 1998) provides a detailed overview of each of the processes.The Kolmogorov-Smirnov (KS) test with a confidence interval of 95% was used to establish the best match distribution for the calculation of flood peaks.The KS test protocol is stated in (Khattak et al., 2016).For flood frequency study, the maximum and minimum instantaneous measured discharge data of the Trimmu are used (Tables 1).Using LP3, the flood peaks for various return times were collected and used as an input to the HEC-RAS model.For flow inputs, two stations were selected; the first station was upper and downstream of Trimmu Barrage at some distance.In order to locate the predicted flood levels along river stretches running across inhabited parts of the basin, the peak flows are introduced into HEC-RAS.For 5, 10, 50, 100, and 150 year floods, flow inputs were supplied along with the full instantaneous discharge value of the Jhelum and Chenab River 2014 flood.The HECRAS model was performed upon completion of data inputs to conduct a comprehensive flow analysis.The model produced a comprehensive report of the study showing the flow depth, discharge at each cross section, and other data.

Figure 2 Framework of methodology
After removing all mistakes, the data was transferred to ArcGIS using the RAS GIS Export File.It was then loaded into ArcGIS and a raw floodplain map was created after creating the water surface and floodplain demarcation.The source map only showed the flood depth and a few islands or reservoirs, so these were removed and replaced with a smooth sea surface.On July 31, 2014, MODIS satellites took photos of the 2014 Jhelum and Chenab River flood that were evaluated by the UNITAR Operational Satellite Applications Network (UNOSAT).This image was used to measure the scale of the flood and the damage to a large region of South Punjab.A contrast was made between our model's simulated flood extent and that provided by the satellite image of the MODIS.
Satellite images were processed for the visualization based analysis.Satellite images of Landsat 7 and 8 with resolution of 30m are used.The very first thing to be done with these images was to geo-reference them in Arc Map.These images were separately processed for land use and Land cover mapping.Also SRTM DEM (30m) was used for the watershed delineation and the flood prone areas mapping.The satellite images were primarily stacked in the Envi 5.4.The subset of the study area was extracted in the Arc Map before further processing of the data.Details of all types of mapping with corresponding software are given below.
Land use mapping: Land use analysis of the area showed the different types of man-made settlements.All of this was done in ArcGIS software; which created a separate layer for each of the land use like river, canals, and roads.The Landsat images were in raster format, firstly they were converted to the vector format.The digitization of all the layers was done in Arc Map.

Land Cover Mapping:
Land cover patterns of the study area were obtained by applying the object based classification technique of digital image processing.Different types of land covers may assist us during the time of flood by their obstructive nature.So the Land cover mapping of the study area is necessarily done.For this purpose eCognition software was used to perform Object Based Image Classification (OBIA).
Flood Modeling: DEM is the fundamental input of the hydraulic modeling software, as it contains all the information about the water profile like depth, slope, aspects etc.We have used DEM with resolution of 30 m.We used SRTM DEM and our study area was covered within a single tile, so we extracted the area in Arc Map.To make it appropriate to be able to use it in HEC-RAS, it was re-projected in Arc Map 10.8 from GCS to Projected Coordinates System.After that contours of interval were made of the AOI and then TIN was generated using these contours to create the geometric data.
In order to extract watershed features (such as river centerline) from DEM, HEC-Geo HMS was used.Using HEC-GeoRAS, the flow and geometric data e.g., cross-sections, bank stations, and lines of flow direction were prepared and which were then used as HEC-RAS input.DEM and geo-referenced natural color images of the study region were the files needed for preparing the geometric results.In order to obtain the natural color images, Google Earth was used and the visual inspection tool was used to measure the n values of Manning.The whole region was divided into four categories on the basis of the form of land cover, as seen in Table 2.  (Chow et al., 1998).For single-story and double-story houses, Phillips (Sañudo et al., 2020) proposed Manning's n value of 0.32.'N' 0.32 was included in this analysis.The RAS GIS Import File was imported into the HEC-RAS once all geometric data specifications had been implemented and 'n' values assigned to each land category.This method makes it easier to convert geometric data from ArcGIS to HEC-RAS.The steady flow editor needed the steady flow data and boundary conditions, which were provided in the following phase.
For flow inputs, three stations were selected; the first station was Mangla dam, second was Marala and downstream of Jhelum and Chenab river at Trimmu.In order to locate the predicted flood levels along river stretches running across inhabited parts of the basin, the peak flows are introduced into HEC-RAS.For 5, 10, 50, 100, and 150-year floods, flow inputs were supplied along with the full instantaneous discharge value of the Indus River 2014 flood.The HEC-RAS model was performed upon completion of data inputs to conduct a comprehensive flow analysis.Reports were generated that included the depth of flow and discharge at crosssections, as well as other relevant information.Errors were deleted before being exported into a RAS GIS Export File for ArcGIS.A contrast was made between our model's simulated flood extent and that provided by the satellite image of the MODIS.

RESULTS AND DISCUSSION
Vector Layer Analysis: Maps of all drainages in the research region are also drawn up, and they are shown separately on this map.Chenab and Jhelum river belts include some of the most densely populated areas in Pakistan.Outcomes of supervised classification results in the research area are discussed as part of the digital image analysis process.The findings of the unsupervised classification did not provide enough information on the study area's land cover since the Landsat OLI image utilized for classification had a 30m resolution (Low resolution).For his high-resolution photography of the research region, object-based classification was shown to be the most effective classification method.An appropriate quantity of information was available in the categorized image output.The three primary landformsvegetation, water, and soil-were used to classify the many types of land cover.
Frequency Analysis: The maximum flows derived from frequency analysis carried out using the highest instantaneous discharge available at Trimmu are powered by the analysis of the depth of the region inundated under various return-period floods.Using the LN, Gumbel, and LP3 distributions, the maximal instantaneous discharges at Trimmu for various intervals were As shown by the Figure 4, at upper mean, or Trimmu, the expected peak flood using the LN distribution is greater than that projected by the distribution of Gumbel and LP3.Using the LP3 distribution, the smallest values were generated.
The K -S test results for Taunsa are shown in Table 4.A nonparametric measure that can act as a goodness-of -fit measure is the K-S test.For Trimmu bridge station, the critical value of the K-S test statistical 'D' was 0.190.Although the values of the test statistics are less than the critical value for any of the three distributions, all the distributions used in this analysis are deemed appropriate for use at all gauging stations.The LP3 distribution was used to predict flood peaks for Trimmu on the basis of the values of K-S statics.In two portions of HEC-RAS, gross instantaneous flows obtained from the LP3 distribution were utilized as steady flow inputs for the return periods of 5, 10, 50, 100, and 150 years and the 2014 floods.Both parts were at the Mangla Dam and Maria River/Trimmu River confluence, which were downstream from lower Trimmu and had relatively clear topography compared to the upstream sections.Figure 5 provides a comparison of Trimmu's water surface profiles under the 10-, 50-, and 150-year return cycle floods and the 2014 floods obtained and are shown in Tables 3,  respectively.
As seen in Figure 5, both the 50-year and 150year return periods result in higher levels of water than that created by the flood of 2014.As predicted, the water surface level under the 2014 flood is greater than that under the 100-year return period flood for all areas in the upper and lower reach.In the upper and lower reach, respectively, under 50 years of return-period flood.Area submerged under natural flow, flood frequency for the specified return date, district borders, names of numerous settlements, towns and roads in near proximity to the floodplain, and gauging stations are included in the information provided by the floodplain maps.The area inundated under natural flow corresponds to the area between the banks that is filled and is delineated using square images.
As a result, the flooded area and flow depth in the upper reach are smaller than in the lower reach.The Trimmu barrage is located in the most severely affected area.With a 100-year return time, it's clear from Table 5 that more than 400 percent of the city will be submerged.
Results of the 50-year return cycle flood simulation show that the most significantly affected villages in the districts of Mandibahudin and Chinote, the eastern side of Wazirabad and Gujrat Tehsils (Figure 6), Figure 7 indicates the degree of inundation displayed by MODIS image and by the model.The floodplain map prepared for the 2014 flood using GIS and HEC-RAS closely followed those reported by Hussain (Hussain et al., 2011) andKwak (Hashim Nisar Hashmi, 2012) using various satellite images.The boundaries of the flooded areas simulated by our model showed near alignment with the July 31, 2014 MODIS map, but some positions had minor deviations.The way the water surface is formed by ArcGIS may be due to these deviations.ArcGIS generated a water surface with a number of islands and reservoirs, presumably because the DEM heights had an interval of 1 m whereas the water surface expected constant elevation values.Overall, given that channel resistance was based on very preliminary values and no calibration changes were made, the success of the HEC-RAS model in producing 2014 flood inundation maps is very good.Channel roughness could theoretically increase the model's accuracy.The analysis of the literature discussed in this paper explicitly revealed that researchers used a wide range of different methods for the simulation of flooding incidents.However, owing to the extremely dynamic existence of meteorological and hydrological systems, significant precipitation events leading to catastrophic flooding will not be predicted.Humans are especially vulnerable to floods in developing countries such as Pakistan due to high population density, lack of effective flood protection mechanisms, and lack of zoning laws and emergency preparedness programs.Industrialized nations are armed with improved measures for flood protection.However, developing nations have still recorded incidents of flood-related deaths because it is not possible to offer full flood protection.Moreover, research aimed at enhancing flood forecasting practices and reducing the harm caused by flooding have given impetus to global climate change.
The results of this case study show that for a given flood, a public domain model of HEC-RAS can easily be used to simulate flood levels.To apply the model to simulate flooding, the expensive acquisition of channel geometry data between cities does not need to be needed.It will allow government agencies in Pakistan, by using the strategy proposed herein, to achieve a substantial reduction in flood damage in the Indus River basin.While hydraulic models of erratic flow are proven to be hard to implement, it is not shocking that water system operators and planners are debating the usefulness of their organizational use.
However, several agencies in Pakistan are now familiar with HEC-RAS models that can be used effectively to reinforce and simplify projections of areas that are likely to flood in a given flood.The planning and processing of input data is important because it is a quasihydrodynamic model.The value of the resistance vector, namely Manning's "n", tends to differ with the water depth.As the productive relative roughness reduces and then increases again as the river flows over channels, while the channel roughness is lower than floodplain roughness, this value decreases with increasing water level.

DISCUSSION
The main objective of this study was to evaluate the suitability of the HEC-RAS model in the simulation of water surface profiles and to determine the extent of flooding in Pakistan's Jhelum and Chenab River during different floods during the return period.By using natural channel geometry in the sub-reaches where it is appropriate to route between these sites and flood levels in estimated geometry, flood routing was involved.Study carried out using the Kolmogorov-Smirnov test found that the best distribution match for the gauging station of the Jhelum and Chenab River Trimmu is LP3.Using the LN, Gumbel, and LP3 distributions, the values of the 5, 10, 50-, 100, and 150-year return cycle floods at Trimmu and LN 4887,5467,6521,6907 and 7179. 4968,5417,6405,6822 and 7066 Gumbel's,and LP3,4906,5480,6344,6755 and 7155.This study was conducted using the HEC-RAS model, in conjunction with ArcGIS, of the magnitude of areas likely to be flooded during various return time floods.An interpretation of findings explicitly revealed that the 50-year return cycle flood at Trimmu (5299 m 3 s- 1 ) would inundate 500 percent more area than the natural surge.Floodplain maps have shown that for agriculture, flood-prone zones are often used.Also vulnerable to flooding are urban areas such as Mandibahudin, which were also evident in the July / August 2014 flooding.The magnitude of the flood displayed by the satellite image of the 2014 flood was replicated very well when the 2014 flood was incorporated into the HEC-RAS model, showing the ability of the model to replicate open water flooding and create water levels at the desired locations with fair precision.It is clearly visualized from the floodplain charts that with one flood every 50-year return period, flood levels are around four times those due to natural flow.
Therefore, it is of utmost significance to cities like Chinote On both banks of the river, Khan and Mandibahudin should be secured mainly by the removal of embankments.The decline in the extent of destruction may be done by de-silting the weak stretches of the channel.During field visits, it was noted that some parts of the marginal embankment need to be restored as the villagers have cut them at many locations and constructed passages for their cattle and tractors.The deterioration of the foundations of embankments has contributed to this.Repair and preservation of current embankments at frequent intervals should be at the top of the list for the government institutions responsible for river preservation.The interference by citizens is another related concern with the protection of the floodplain in the basin.

Flood Mapping :
Flood mapping of the previous flood events involves the peak discharge values of the water bodies.Flood data of the recent years have been mapped and their extents in the vicinity of Trimmu barrage.Following are the maps obtained from different sources and shows the extents of flood in the province Punjab.District Mandibahudin always flooded on the both sides by Jehlum and Chenab rivers.Chenab River is on the other side of the Tehsil.Map of flooded area of Chinote Tehsil of district Mandibahudin is shown in the figure 5.

Figure 3
Figure 3 Depicts the research area's land cover categorization map.

Figure 4 .
Figure 4. Water field elevations of multiple floods at some distance downstream of the jhelum and chenab river

Figure 7
Figure 7 Comparison between modis and simulated of actual flood at chaj doab.

Table 4 : K-S test results for the Trimmu Barrage.
Figure 5. Discharges at trimmu obtained using different distributions.