Impact of Extreme Rainfall Event over Swedish Urban Catchments
A study on catchment characterization in the context of Aerial Reduction Factor and storm movement
Md Didarul Alam Tusher
Supervisor Joakim Riml
Associate professor in River Engineering, Department of Sustainable development, Environmental science and Engineering,
KTH Royal Institute of Technology Examiner
Anders Wörman
Professor, Head of Division in River Engineering, Department of Sustainable development, Environmental science and Engineering, KTH Royal Institute of Technology
Supervisor at Tyréns AB Johan Kjellin, Vattenspecialist
Degree Project in Environmental Engineering and Sustainable Infrastructure KTH Royal Institute of Technology
School of Architecture and Built Environment
Department of Sustainable Development, Environmental Science and Engineering SE-100 44 Stockholm, Sweden
ii
Sammanfattning
Konstruktion för dagvattenstruktur är påverkade på overallurladdning och toppurladdning av en nederbördhändelse som beror på nederbörds belopp. Aerial Reduction Factor (ARF) är nödvändigt för att beräkna den verkliga mängden nederbörd. Användning av ARF vid nederbördsmätning kan minska den totala förväntade nederbörden från en händelse och därför minska urladdningsvolymen för konstruktionsuppskattning. Den här praktiken ska sänka sannolikheten för överdimensionering av en struktur. Å andra sidan kan molnrörelse öka avrinningsuttags toppurladdning, känd som resonanseffekten när avrinningsflöde och molnhastighet blir samma i samma riktning. Uppskattning till molnrörelse hjälper att öka modell exakthet och garanterar säkerheten i dagvattenstruktur.
I den här studie analyserades användbarheten av ARF och betydelsen av molnrörelse över svenskt urbana avrinningsområden. Det observerades att 90 % av avrinningsområdena är mycket mindre i storlek (<5 km2). Tidigare forskning rapporterade att ARF kan användas för avrinningsområden som är stora än 5 km2. Därför har ARF mycket begränsad användbarhet för enstaka avrinningsområden i svenskt urbant området. Molnrörelse har också begränsad betydelse på grund av stor hastighetsskillnad mellan avrinningsflöde och genomsnittlig molnhastighet under storm. Men vid strukturell design föreslås djupanalys för ett avrinningsområde baserat på lokal kondition.
iii
Abstract
The design practices of urban hydraulic structures are required to be based on enough safety measures in addition to cost efficiency for sustainable development. Overestimation of catchment runoff generated from a storm event can increase the project cost unnecessarily. For urban pluvial studies, use of Aerial Reduction Factor (ARF) helps to estimate the probable precipitation intensity inside a catchment. By understanding the necessity of using ARF from a national context and applying it in present design practices can pave the way towards sustainable design practices. In this study the applicability of ARF from Swedish urban context was analyzed. The urban catchments for fifty biggest Swedish cities were delineated and the catchment parameters were analyzed. Application of ARF depends on the size of catchment.
To experience a significant reduction in catchment outlet discharge, the minimum catchment area for application of ARF was reported as 5 km2. According to the analyzed parameters, ninety percent of catchment sizes in Swedish urban area were found less than 5 km2, which are quite small in respect of ARF applicability. It was realized from the analysis that application of ARF within a single catchment is not much necessary for pluvial studies in Swedish urban catchments due to catchment properties.
In addition to rainfall intensity, rainfall movement also changes the runoff behavior from a catchment. When catchment’s flow velocity and direction through main channel and storm moving velocity and direction over the channel coincides with each other, then the outlet peak discharge magnifies in comparison to a stationery storm, known as resonance effect. The impact of storm movement over catchment was analyzed using HEC-HMS modelling with varying storm movement velocity over catchment. The analysis was performed on 12 catchments of different size and flow velocity. It was found that the peak discharge can increase up to 46 percent depending upon catchment characteristics. The flow velocity through all the delineated catchments of fifty cities were calculated using USDA’s NRCS TR-55 method and then compared with usual storm moving velocities in Sweden. It was found that, due to flat nature of Swedish urban areas, the flow velocities are very low (<2 m/s) in compare to average storm velocity (8 m/s), portraying the fact that there is limited probability that these velocities will coincide. But for any area where storm velocity is normally low, resonance effect can happen. Precise analysis based on local conditions are suggested while modelling a particular area, since impact of resonance effect can overrun the design considerations.
iv
Acknowledgement
This thesis was submitted as the final part of Master degree requirement for the Environmental Engineering and Sustainable Infrastructure program at KTH Royal Institute of Technology, Stockholm, Sweden. It was done in a project jointly collaborated by Tyréns AB, KTH Royal Institute of Technology and Swedish Meteorological and Hydrological Institute (SMHI). The work took place between February to November 2019 under the supervision of Johan Kjellin, Vattenspecialist, Tyréns AB and Joakim Riml, Associate professor in River Engineering, KTH.
I would like to thank both of my supervisors, Johan Kjellin and Joakim Riml for their support, inspiration, knowledge and guidance throughout the work. I would also like to thank Jimmy Olsson from Tyréns AB, Michael Henrich from KTH and Jonas Olsson from SMHI for their valuable views and suggestions. I enjoyed my time working in this project. This thesis was produced during my scholarship period at KTH, thanks to Swedish Institute for providing me the Swedish Institute Study Scholarship 2017 for conducting master study in Sweden. Finally, I express my gratitude to Firoza Akhter and other members of my family, who have encouraged and supported me throughout my studies.
Md Didarul Alam Tusher Stockholm, November 2019
v
Contents
1 Introduction ... 1
1.1 Background ... 1
1.2 Research objective... 3
2 Study setup ... 4
2.1 City selection ... 4
2.2 Types of floods ... 6
2.3 Watershed delineation criterion ... 6
3 Methodology ... 11
3.1 Catchment delineation ... 11
3.1.1 Data file processing... 11
3.1.2 Point delineation method for catchment ... 15
3.1.3 Catchment parameterization ... 15
3.2 Influence of storm movement ... 17
3.2.1 Flow velocity distribution ... 17
3.2.2 HEC-HMS modelling ... 19
4 Result ... 26
4.1 Catchment parameter analysis ... 26
4.2 Strom movement analysis ... 33
4.2.1 Velocity parameter analysis ... 33
4.2.2 HEC-HMS analysis outcome ... 35
4.2.3 Angular difference and shape factor analysis ... 45
5 Discussion ... 49
5.1 Applicability of ARF ... 49
5.2 Effect of storm movement ... 51
6 Conclusion ... 53
7 References ... 54
8 Appendix A ... 57
9 Appendix B ... 59
10 Appendix C ... 61
11 Appendix D ... 73
11.1 Catchment shapes selected for HEC-HMS analysis ... 73 11.2 Catchment shapes and related Hydrographs generated using HEC-HMS model . 74
1
1 Introduction
Background
The historical and design storms used in analytical purpose for engineering structures are based on point measurement of rainfall data. The reading from rainfall station is applied to surrounding area to calculate the related runoff from the area, though the rainfall intensity is not same throughout the area (Svensson & Jones, 2010). It is expected that the rainfall intensity differs in spatial content over the whole catchment as the cloud density is not same everywhere.
The actual amount of precipitation fall over a catchment varies from the amount calculated based on the nearby rain station data. The core area of the cloud has the maximum potential rainfall intensity and the cloud density reduces over distance from core (Bengtsson et al, 1986).
This reduced amount of rainfall intensity based on spatial extent over a catchment can differ the design context of urban drainage structures. This cost-efficient design can save overestimation of structural extent. In addition to rainfall intensity, the rainfall movement also plays an important role in the outlet discharge of a catchment (Bengtsson et al, 1986). The peak discharge at the outlet of the catchment differs based on the relative movement direction and velocity of the cloud in relation to the catchment flow direction and velocity.
The Aerial Reduction Factor (ARF) is the ratio of the aerial average rainfall over an area to the point rainfall. ARF is necessary since application of point rainfall data over the surrounding area can lead to overestimation of rainfall (Bengtsson et al, 1986). ARF can be calculated using empirical method or analytical method based on ground station data. Modern radar data can also be used as an alternative to ground station data. Empirical methods are based on two approaches: fixed area method and storm centric method (Omolayo, 1993). In fixed area approach, ARF is the ratio of the average rainfall intensity inside a fixed size area over point rainfall data from rainfall station. This is geographically fixed ARF. The storm centric approach considers the ratio of the average rainfall inside isohyet over rainfall intensity of the storm center, which is based on individual storm event. In storm centric approach, the value is less than one and reduces with the distance from the core which depicts the reduction of rainfall intensity from the highest intensity core area of the cloud (Bengtsson et al, 1986), whereas in fixed area method the ARF can be greater than one. ARF values has dependency on geographical location, catchment size, storm duration and return period (Svensson & Jones, 2010). ARF values decrease for storm with high return period. Reduction of ARF is much noticeable in convective rain event compared to frontal events (Skaugen, 1997). The ARF
2 curve difference between these two rain categories is more pronounced when rain with higher return period is used.
In context of rainfall movement, a downstream moving storm generates higher peak discharge than an upstream moving storm (Sargent, 1981). In compare to a stationery storm, the downstream moving storm gives higher peak discharge for same amount of rainfall (Seo et al, 2012) and can be upto 19% higher in peak discharge for an individual event (Andersen H.S.
1991). According to Niemczynowicz 1984, the peak discharge due to directional bias can be increased upto 20-30% in compare to stationery storm for short duration rainfalls (5-20 min);
but for long duration storm the value seldom exceeds 10%. An empirical study conducted by Veldhuis et al. (2018) states that storm with slow velocity produces higher peak flow than the faster moving storms. The hydrograph created by upstream moving storm has early rise, low peak discharge, not so steep rising limb, and long base time (De Lima et al. 2002) and vice versa for downstream moving storm. When downstream moving storms direction is same as the flood response direction with same velocity, then there creates the highest peak discharge which is known as resonance effect (Ngirane-Katashaya et al., 1985). This resonance condition peak discharge can be surpassed by storm duration effect with additional precipitation if the storm velocity is lower than and, in the vicinity (0.60-0.98) of the flow velocity of the flood response channel (Seo et al, 2013). But the movement sensitivity of a storm depends on its size in relative to the catchment. Change in outlet discharge due to movement of rainstorm depends on spatial distribution of rainstorm. When storm size is bigger than the catchment size and already covers the whole area, then the movement of storm does not have any effect on the peak discharge and the storm starts to act as a stationery storm (Seo et al, 2012; Surkan A.J., 1974). But when the storm or cloud size is smaller, then the movement is sensitive in relation to peak discharge. In this case the catchment is partially active for the rain which influences the peak discharge in relation to movement. This phenomenon is more obvious for rapid and short duration rainfall. When storm event is short and rapid, the cloud movement has more influence on the runoff hydrograph. It was reported that when storm movement is considered in rainfall-runoff model, the difference between simulated and observed hydrograph becomes less (Sigaroodi S.K. 2016). In addition, the downstream movement is more sensitive to peak discharge than the upstream movement for different rainfall patterns (De Lima et al. 2002). If the runoff water contains pollutants from the discharge area, then the peak pollutant concentration is also affected by storm movement (Chang, C. L.2007), which exposes another perspective of considering storm movement in addition to discharge volume.
3 The significance of rainfall intensity over the hydrological response (e.g., prediction of peak runoff discharge, time to peak as well as the shape of the overland flow hydrograph) is inversely related to storm velocity. When the storm speed is higher (>2 m/s), the rainfall intensity does not have any significant influence (De Lima et al. 2002). Speed of storm is the most important parameter than other parameters like direction, rainfall intensity in respect of hydrograph (Ogden et al. 1995). The basin shape, orientation, antecedent condition, degree of urbanization, channel distribution also influence the peak discharge (Nikolopoulos et al., 2014; Pechlivanidis et al. 2016; Miller et al.2017). Longer, narrow and less branched catchments are much influenced by storm speed and direction (Andersen H.S. 1991, Woods and Sivapalan 1999, Smith et al. 2000). It was reported by Sturdevant-Rees et al., 2001 that storm motion plays important role when catchment size is between 500 and 1000 km2 (Nikolopoulos et al., 2014).
The area of a catchment can be compared with the area of extreme rainfall cloud area to investigate whether the catchment is entitled to highest precipitation all over it. The extreme rainfall clouds are more likely to form elliptical isohyetal pattern with a ratio of major to minor axes of 2.5 to 1 (Hansen, E. M. et al, 1982). The area of ellipses bounding the catchments can be used to identify the spatial extent of catchments in respect of ARF. The applicability of ARF in Swedish urban context can be seen by comparing the areas. To find the significance of storm movement over catchment, rainfall-runoff model can be generated. Then by considering the storm moving velocity and direction, the prospect of happening resonance affect can be identified for the urban catchments.
Research objective
The aim of this research is to analyze the applicability of Aerial Reduction Factor (ARF) and impact of storm movement over Swedish urban catchments to evaluate the effect of extreme rainfall event. The specific research objectives are:
1. To define the catchment parameters for the 50 biggest Swedish cities which includes area, longest flow path, catchment flow velocity, smallest ellipse to cover the whole catchment and hence find the applicability of ARF.
2. To find the impact of ARF over different return period storms.
3. To perform rainfall-runoff analysis using HEC-HMS model for selected catchments to evaluate the impact of storm movement over Swedish urban catchments.
4
2 Study setup
The application of Aerial Reduction Factor (ARF) can make a project more cost efficient. On the other hand, the consideration of rainfall movement in design structures will ensure more safety for design calculations. These factors are more obligatory for urban areas, where impact of a single event is much evident on densely organized population and structures. The cost- efficient designs with reasonable safety measures can ensure better utilization of resources.
Therefore, a study on application of ARF and storm movement over fifty major cities will direct the importance of consideration of these factors from Swedish context.
City selection
The fifty biggest Swedish cities were selected based on the demographic data of 2018 (SCB).
The list of the cities is provided in appendix A. The cities are distributed all over Sweden as seen in Figure 1.
5 Figure 1: Location of fifty selected cities
6 Types of floods
There are three common types of floods which can cause severe damage to infrastructures.
Fluvial flood: When heavy rainfall happens over the catchment area of a river for an extended period, then the river’s discharge capacity gets overrun by the extra water.
This extra water then overflows the banks of the river and cause heavy damage.
Pluvial flood: When heavy rainfall overrun the drainage capacity of an area, then waterlogging happens in the area, causing flood for limited period. This type of flood happens mostly in urban area.
Surcharge flood: This is coastal flood, causes by tidal effect or storm surge. The affected areas are the low-lying coastal areas.
This study is oriented to analyze the pluvial flooding extent for Swedish city catchments.
Fluvial and surcharge flooding consequences are ignored in the analysis.
Watershed delineation criterion
The watersheds or catchments are selected based on four types of catchment selection criteria.
The main purpose is to cover the whole city area with multiple catchments as necessary. Outlets are selected along the borderline of the city boundary (mostly along the boundary of adjacent waterbody) to create the catchments. Waterbody means rivers or lakes which are defined by polygon feature in the Swedish University of Agricultural Sciences (SLU) database. The rivers are more than 6m wide. This catchment delineation was continued until enough catchments are delineated to cover the whole city. The main purpose of this study is pluvial flooding. When any catchment is found containing rural area with size greater than 10 km2, that area is excluded from the catchment considering that the accumulated flow from that region will generate fluvial flow which will pass through the catchment in a defined order through stream and hence will not influence the generated pluvial flow due to extreme rainfall event. The four watershed delineation criteria are:
1a. Outlet point is selected at the point where catchment stream is leaving the city area or joining a water body (river or lake) within the city area. Here the watershed may contain rural areas (<10 km2) which are adding pluvial flow to the watershed (Figure 2).
7 Figure 2: Watershed delineation criterion 1a. The hatched areas are several catchments
covering the city
8 1b. Outlet point is selected at the point where stream is leaving the city area or joining a water body (river or lake) within the city area. Here all the rural parts (<10 km2) are excluded from the watershed to find the urban only watershed. The rural parts are excluded based on the assumption that the runoff response from the rural parts are slower than the urban part.
When precipitation happens, the runoff from the urban parts will flow and reach quickly to the outlet of the watershed causing the immediate discharge volume. The surrounding rural parts runoff will reach later causing another peak discharge. Therefore, to distinguish the effect of urban only watershed, this catchment delineation criterion was applied (Figure 3).
Figure 3:Watershed delineation criterion 1b
9 2a. Outlet point is selected at the point where a river which is flowing through the city area leaving the city. Here the watershed may contain rural areas (<10 km2) which are adding pluvial flow to the watershed (Figure 4).
Figure 4:Watershed delineation criterion 2a
10 2b. Outlet point is selected at the point where a river flowing through the city area leaving the city. Here all the rural parts (<10 km2) are excluded from the watershed to find the urban only watershed (Figure 5).
Figure 5:Watershed delineation criterion 2b
11 All the cities under consideration have catchments from criterion 1a and 1b which are subjected to pluvial studies. Only 25 cities have catchments defined by 2a and 2b criteria as other cities do not have any nearby flowing river.
3 Methodology
The analysis was divided into two parts based on the objectives of the study. In the first part, the Swedish city catchments were delineated and analyzed to define the catchment parameters using ArcGIS software to investigate the ARF applicability. Then in the second part, both ArcGIS and HEC-HMS software were used to analyze the significance of storm movement over selected catchments.
Catchment delineation 3.1.1 Data file processing
The hydrological behavior of a catchment can be presumed by several parameters which include area, longest flow path, time of concentration etc. To define the catchment parameters for the selected cities the following files (Table 1) were downloaded from the Swedish University of Agricultural Sciences (SLU) database for every city.
Table 1: List of files downloaded from SLU for analysis Folder name
(SLU database)
File name File type
description Reference system
Höjddata 2m hojd2m raster Digital Elevation Model (DEM) of the selected areas with 2m resolution
SWEREF99_TM
Fastighetskartan hydrografi vector
hl_get, mv_get vector 1. Polylines defining existing streams less than 6m width (hl_get)
2. Polygons defining rivers, lakes and sea (mv_get)
SWEREF99_TM
Tätorter vektor To2010_SR99 TM_region
vector Polygon defining city area.
An area is considered as city area when there are at least 200 inhabitants with buildings in less than 200 m distance.
SWEREF99_TM
12 Folder name
(SLU database)
File name File type
description Reference system
Fastighetskartan markdata vektor
my_get vector Polygon defining land uses SWEREF99_TM
Jordarter 1:25 000 - 1:100 000 vektor
jordart vector Polygon defining soil types SWEREF99_TM
The DEM (hojd2m) and stream file (hl_get) were first loaded in the ArcMap. The Archydro tool was mostly used for the processing of catchments. At first from Archydro tool, DEM reconditioning process was run using the loaded files. As the streams represented by stream file are less than 6m wide, they were hardly visible or represented in the DEM file. This DEM reconditioning process created a stream cut section in the DEM file by burning along the line of stream line. The burning parameters were selected as to create a 2m wide and 3m deep rectangular channel along the stream line (hl_get). This DEM reconditioning feature ensured the existence of the naturally available streams in the DEM. The rivers and lakes were easily distinguishable in the DEM file; thus no burning was applied on those features. The output file from this step (DEM reconditioning) was ‘AgreeDEM’ which was then processed using ‘fill sinks’ tool. The sinks were the areas from where water cannot flow and get trapped. This tool filled the sinks in the DEM to eliminate the water trapping problem and kept the water flow running towards outlet. Then ‘flow direction’ tool was used which created a flow direction grid from the output of fill sinks tool. This flow direction grid represented the direction of steepest cell around a cell to where the water will flow. After flow direction, flow accumulation tool was used which counts the accumulated number of flowing cells upstream of a single cell, and then gave a value to that cell. The stream definition tool was then used to define the streams based on the flow accumulation grid. A stream definition value defined the minimum area for runoff which will initiate a stream. Here a stream definition of 0.1 km2 was inserted to initiate a stream based on professional practices. A value lower than 0.1 km2 would create more streams with unnecessarily long processing time. On the other hand, higher values greater that 0.1 km2 would create very few streams which would make it difficult to find outlet points for defining catchments. The areas close to water bodies (river, lake) with less than 0.1 km2 drainage area, were omitted from the analysis assuming that these areas have less significance in the study objectives. The output from stream definition file was used for ‘stream segmentation’ step. This step assigned a unique value for every segment of streams whether it is a head stream or intermittent stream between two stream junctions. Then came ‘catchment
13 grid delineation’ step where every cell got a value according to the catchment they belong.
Here the input file was flow direction output file and stream segmentation file. The assigned value was same as the adjacent stream segment value which drains the water from the catchment. The next ‘catchment polygon processing’ step created vector polygons based on the catchment grids. Then the output from stream segmentation raster file was converted to drainage line polyline shape using the ‘drainage line processing’ tool with the help of flow direction file. These drainage lines illustrated the flowing streams in the area based on the previously selected stream definition (0.1 km2). Then the ‘adjoint catchment processing’
function was used. This tool created a polygon for the upstream flow contributing area for every catchment. The output file from this step speeded up the point catchment delineation process used later.
14 Figure 6:Archydro tool terrain preprocessing steps
15 3.1.2 Point delineation method for catchment
When all these steps were done, then the area was prepared for catchment delineation. The catchments were defined using ‘point delineation’ method. Here the outlet was defined first according to the catchment delineation criteria and then the contributing area or catchment referred to that point was identified and saved as catchment. From the catchment definitions described in section 2.3, the 1a type catchments were defined first. The city boundary was defined by ‘To2010_SR99TM_region’ shape file. The catchments were delineated to cover the city region, though some catchments include both urban and rural parts. As the study was emphasized on pluvial flood study, therefore the fluvial flow generating area was excluded from all the watersheds. The fluvial flow streams were defined by those streams which generate a flow from rural area with accumulated area size greater than 10 km2 (considering 100 liter/s flow in stream with 10 liter/s. km2 runoff from area). To exclude these areas from watershed, an outlet point was selected where the potential fluvial stream enters the city area. Then this outlet defined the contributing area for that point. When the area was greater than 10 km2, then it was excluded from the watershed. There are on average 40 catchments found per city to cover the whole city. To define the catchments of 1b description, the only urban part form 1a catchments were extracted using ‘clip’ tool. The 2a watersheds were defined where there exists river flowing through the city. The point was taken in the river where the river was leaving the city and thereby the whole city catchment was delineated. The 2b watersheds were defined same as 1b watersheds.
3.1.3 Catchment parameterization
When all the watersheds were defined, then the following parameters were calculated for all the watersheds. The parameters include Area, longest flow path length and slope, straight line path length and slope, straight line bearing (0º-360º, 0º means north), concentration time, water flow velocity, longest straight-line distance, smallest ellipse to cover the whole catchment.
Area means the total area of a watershed. The longest straight-line distance is the length of the longest straight line which can be fitted inside a watershed irrespective of the direction. As storms are formed with the core or eye as an elliptical shape, this longest straight line was used as the major axis to create an ellipse which covers the whole watershed. This was the ellipse with the minimum area to cover the whole catchment.
16 Figure 7: Catchments and covering ellipses (left) with definition of ellipse parameters (right)
The calculated watershed area and its surrounding ellipses area were compared to find the applicability of Aerial Reduction Factor (ARF) for Swedish urban catchments. When the watershed size was less than 5km2, then the whole watershed was subjected to extreme rainfall generated from the core area of cloud (Bengtsson et al, 1986). This comparison was also performed using ellipse areas. When the ellipse areas were greater than 5 km2, then ARF were applicable to balance the spatial change of rainfall intensity.
The impact of considering ARF over total discharge and peak discharge from a catchment with size greater than 5 km2 was analyzed using ARF values from Bengtsson et al, 1986. Rainfall- runoff analysis using Chicago Design Storm (CDS) rain with different return periods or precipitation values were tested to understand the rainfall-runoff relationship for a catchment.
Also, three catchments with size greater than 5 km2 were selected to simulate the change in peak discharge due to application of ARF for different return period of storms. The ARF values reported in the paper were for one- year return period storm, but these values were used for higher return period storms analysis due to data limitation.
17 Influence of storm movement
3.2.1 Flow velocity distribution
For the resonance effect analysis, it was assumed that the longest flow path velocity for catchment was the representative main-stream flow velocity for the catchment. To identify the catchment flow velocities in comparison to storm velocities, the longest flow path velocity for catchments were analyzed.
The ‘longest flow path’ tool was used to calculate the longest flow path for all the catchments using batch delineation process. Longest flow path is the path used by a water drop to travel from most remote point of watershed to the outlet. The slope of longest flow path was calculated using ‘flow path parameter from 2D line’ tool which gave 10-85 slope of the path as follows,
Slope for LFP = % %
LFP= Longest flow path
Water flow velocity through the longest flow path was calculated using TR-55 method (USDA 1986). There are three flow mechanisms from the upstream to downstream section; sheet flow, shallow concentrated flow and channel flow. When rain starts to fall over catchment, sheet flow occurs for a maximum of 90 meter, after that shallow concentrated flow begins. To calculate the sheet flow velocity, site specific 2 years 24 hours rainfall data is necessary. From the length distribution of longest flow paths for all catchments, it was seen that most of the lengths were around 1-2 km. Therefore, for the ease of calculation for all the catchments distributed all over Sweden, the comparatively small segment (90 meter) of sheet flow part was ignored. The shallow concentrated flow and channel flow mechanism was considered for longest flow path velocity analysis. The shallow concentrated flow mechanism was considered when streams flowed through paved or unpaved surfaces as per the land use map. But when streams flowed through water channels according to the land use map, then channel flow mechanism (Manning’s equation) was considered.
The flow length through different land covers were calculated using ‘intersection tool’ in ArcGIS where the input file was longest flow path file and land use file (mv_get). Then the attribute table was analyzed to find the total flow length for paved and unpaved surface for every flow path. From the land use file the paved areas included BEBYGG, BEBLÅG, BEBHÅG, BEBSLUT, BEBIND, ÖPKFJÄLL, ÖPGLAC, ÖPTORG categories whereas unpaved areas included ODLÅKER, ODLFRUKT, ODLEYÅK, ÖPMARK, SKOGBARR,
18 SKOGLÖV, SKOGFBJ categories and water channel with VATTEN. These land use classifications were based on SLU database. For a single longest flow path, the calculations were,
Paved path length= total flow length through BEBYGG, BEBLÅG, BEBHÅG, BEBSLUT, BEBIND, ÖPKFJÄLL, ÖPGLAC, ÖPTORG land use surfaces
Unpaved path length= total flow length through ODLÅKER, ODLFRUKT, ODLEYÅK, ÖPMARK, SKOGBARR, SKOGLÖV, SKOGFBJ land use surfaces
Water channel path length = total flow length through VATTEN land use Velocity for paved area= 20.3282 × (LFP slope) .
Velocity for unpaved area= 16.1345 × (LFP slope) . Velocity for water channel = . ( ) / ( ) / ;
where r = hydraulic radius (ft) = ( )
( ) = 2.36 ft (0.72 m), s = LFP slope, n = 0.03 (natural channels)
the average channel section considered to calculate hydraulic radius is sketched in Figure 8.
Width 4m
Height 1.5 m
bottom 1 m 1.5 m Figure 8: Channel section
Time of concentration = + +
Average flow velocity for LFP =
19 Straight line is the line created by joining start and end point of LFP. The purpose of this line assumed that the clouds move in a straight line compared to the catchment size. Thus, the rainfall and flow were assumed to follow a straight line through the watershed.
Average flow velocity for Straight Line =
The average straight-line flow velocity for all the catchments were plotted to visualize the frequency distribution of velocities for Swedish city catchments. Then this distribution was compared with the usual storm velocities for Sweden as extracted from Niemczynowicz, 1991 (Figure 9).
Figure 9:Average storm velocity measured in Sweden (Niemczynowicz, 1991). The columns represent the percetage of stroms with the velocity
The common region between the two distributions (flow velocity for straight line and average storm velocity) in X axis represents the catchments with flow velocity in the vicinity of storm velocity which creates the likelihood of happening resonance effect, if the storm direction is considered towards the flow direction.
3.2.2 HEC-HMS modelling
The existence and magnitude of resonance effect over catchments were investigated using rainfall-runoff model. The modelling tool used for the creation of the model was Hydrologic Modeling System (HEC-HMS), which is an open source software developed by US Army Corps of Engineers. This software was designed to simulate rainfall-runoff process in a watershed. According to US Army Corps of Engineers, the Soil Conservation Service (SCS)
0 20 40 60 80 100 120
0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Cumulative %
Storm frequency %
Velocity (m/s)
Storm velocity graph
Storm velocity frequency % Cumulative storm frequency
20 curve number method is a well opted method for the analysis of small urban watersheds. To calculate the Curve Number (CN) for different parts of catchment, the soil type and land use pattern were analyzed. Then the CN values were calculated based on the provided table in SCS method. The curve number value ranges from 1-100, where higher numbers indicate hard soil or rock with less abstraction and low value indicates soft surface with high initial abstraction.
The model preparation task was divided into two parts, ArcGIS and HEC-HMS. The HEC- GeoHMS tool from ArcGIS was used to prepare the area for HEC-HMS analysis. All the processed files from Archydro tool as defined in section 3.1, were used here for HEC-GeoHMS processing.
3.2.2.1 ArcGIS analysis
The processing for HEC-GeoHMS analysis followed several steps. At first, a new project for HEC-GeoHMS analysis was created using ‘start new project’ tool. Then the outlet of the intended catchment was selected using ‘Add project point’ button. Then the ‘generate project’
tool was used to create a project catchment for the defined outlet. When the project was created then the river length feature was used to calculate the attributes of the catchment rivers or streams. Then came the river slope and subbasin slope calculations based on the DEM file.
During the project creation, the catchment was divided into several subbasins (sub-catchment) based on the drainage lines. Every single drainage line segment had a single subbasin. The longest flow path for these subbasins were calculated in next step. Then the gravitational centroid, centroid elevation and centroidal longest flow path were calculated for further analysis. Then the HMS processes were selected based on the study preference. There are four internal processes which made the rainfall-runoff simulation functional; basin loss, basin transform (overland flow), baseflow and river routing. There are different methods to choose for these processes. Here for small urban catchments, the processes were selected as per the SCS curve number method as follows:
3.2.2.1.1 Model processes
Basin loss: SCS curve number loss model
The precipitation excess after basin loss was calculated here. It is a function of cumulative precipitation, soil type, land use and antecedent moisture condition.
Pe = ( )
21 Where Pe= precipitation excess accumulated over time t in mm, P= total precipitation over time t in mm, Ia=initial abstraction/ loss in mm, S= potential maximum retention in mm
for small urban watershed,
Ia= 0.2S and S = – 254
Basin transform: SCS unit hydrograph
The process of creating direct run-off from a subbasin due to excess rainfall was defined here.
The shape of the runoff hydrograph was decided based on the input hyetograph subtracting the loss. The input parameter necessary here was the lag time which decides how far (time) the direct runoff hydrograph centroid shifts from the precipitation hyetograph. The lag time was calculated using the SCS 1973 equation as stated below,
tL= ( ). ( ) .
×( ) .
Where, tL = lag time [hrs],
L= Longest flow path [ft],
S= Maximum retention= - 10
y= subbasin slope in % Base flow: None
Base flow or subsurface flow contribution for this model was neglected. As the study was related to extreme rainfall event with high precipitation in small span of time, therefore the comparatively slow baseflow contribution to the outlet was ignored as the arrival time of this component was much delayed and has very little contribution to the mostly paved urban catchments.
River routing: lag method
The routing method determines how the downstream hydrograph of a channel/ river will change based on the given upstream hydrograph from subbasins outflow which is the upstream boundary condition. From several available methods, for urban drainage channels the widely used method is lag method (Pilgrim and Cordery, 1993). As in this study, the urban drainage channels were paved with little loss and friction, the downstream hydrograph would not differ
22 much from upstream hydrograph except the happening time. Therefor this lag time method only shifts the hydrograph ordinate based on the given lag time of the specific channel without altering the shape of hydrograph. The lag time for the channel was calculated based on TR-55 method same as discussed in section 3.2.1 for Longest Flow Path time of concentration.
Figure 10:Lag time method hydrograph formation (from Technical Reference Manual of HEC-HMS)
3.2.2.1.2 Basin file creation
After selecting the HMS processes, the rivers and subbasins of the catchment were given auto name feature using ‘river auto name’ and ‘basin auto name’ function. Then all the parameters were converted to HMS unit (feet, second). The check data function was used at this stage to search for any error in the processing. One usual error that arise at this stage is that subbasin centroid containment, as the centroid might get located outside the subbasin due to irregular shape. This error was resolved by merging the subbasin with an adjacent subbasin to form a regular shape single subbasin whose centroid was contained inside the subbasin boundary.
When all the data check were finalized successfully then the data were made ready for export using ‘prepare data for model export’ command. There are three output files needed to create a HEC-HMS model from HEC-GeoHMS tool; basin file, meteorological file and gauge file. At this stage only the basin file was ready.
3.2.2.1.3 Meteorological and gauge file creation
To create met file and gauge file, a methodological approach was needed to decide on how the hyetographs would be distributed all over the catchment. Based on the intention of this study as to analyze the significance of storm movement, the Delaunay triangulation method (Gauge Thyssen polygon feature) was selected for hyetograph insertion. To insert the hyetograph in the catchment to replicate the rainfall scenario, gauge points were required to be distributed over the catchment. These gauge points were created along the straight line which connected the
23 upstream and downstream point of the longest flow path. The difference from gauge to gauge was selected as 315m based on the stream definition defined during catchment processing. The stream definition was selected as 0.1 km2, which gave a rectangular block of 315m by 315m.
From the very upstream gauge point, it would take 315m distance for the water flow to run into a defined stream and make some hydrological change or significance in the watershed segment.
Hence, all the gauge points were created by maintaining this 315m difference. Figure 11 shows the location of gauge points in a watershed started from north to south. The horizontal lines are separating the blocks under a gauge point. All the points under a same block got the same rainfall as the hyetograph inserted into the gauge point of that block. For stationery storm, all the gauge points got same hyetograph at same time. But for moving storm, the upstream most gauge point got the hyetograph first which lasted for 2 hours. In the meantime, the second gauge point got the same hyetograph with a time difference depending upon the storm moving velocity. By this method, all the gauge points got the same hyetograph at different times based on their distance and storm moving velocity. The Delaunay triangulated blocks were created using ‘Gauge Thyssen polygon’ feature.
24 Figure 11:Gauge points in a catchment (pink tringles)
The required met and gauge file were created using ‘Met model file->gauge weight’ feature.
Then the model was set to be created for HEC-HMS analysis as the basin file, met file and gauge file were prepared. Then using the ‘create HEC-HMS project’ command, the model was created.
3.2.2.1.4 Curve number and impervious cover
The curve numbers for different subbasins were necessary for HEC-HMS analysis. The curve number for every pixel in the catchment was calculated using the ‘Generate CN grid’ command based on the land use and soil type. The curve number value tables are given in appendix B.
Then the average curve number for every subbasin was calculated using the zonal statistics.
The percent impervious grid for the catchment was created using ‘feature to raster’ tool where the input feature was the land use file with assigned impervious cover for different land use types as shown in appendix B. The average impervious percent for every subbasin was calculated using zonal statistics tool.
25 3.2.2.2 HEC-HMS model analysis
The exported model was then opened in HEC-HMS. For basin loss model; the calculated initial loss, curve number and percent impervious number were inserted in the table for all the subbasins. Then for basin transform model, the calculated lag times for every subbasins were inserted. After that the river routing table was prepared using the lag time values for streams/rivers. Then the hyetographs were inserted into all the gauge points. A Chicago Design Storm (CDS) rain hyetograph of 2 hours duration with 100 years return period was created according to guideline provided by Svenskt Vatten, 2011 (Figure 12). The peak of the hyetograph was chosen at 37% of the duration according to normal practice.
Figure 12:CDS rain hyetograph of 2 hours duration with 100 years return period In stationery storm case, the hyetograph started at 1200 o’clock for all the gauge points and stopped at 1400. Total rainfall for every gauge point was 65.26 mm. For moving storm scenario, the time difference for start time of hyetograph from one station to the next was calculated using the following formula.
Time difference = ,
Thus, the hyetographs were started at different times at different gauges based on their distance from the upstream point and lasted for two hours with the same amount of rainfall (65.26 mm).
when all the gauge points were prepared with designated time-oriented hyetographs, then the model was run with an inserted time span enough to expect all the precipitated water flows to the outlet and the outlet discharge curve came to zero at the end of analysis period. The model
0 0.01 0.02 0.03 0.04 0.05 0.06
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120
rainfall intensity (mm/s)
Time in minute
CDS rain hyetograph
26 was run separately for stationery storm and moving storm scenario. The difference in shape of catchment outlet discharge curve due to storm movement was analyzed to figure out the significance of moving storm in comparison to stationery storm with same amount of rainfall.
There were four different storm velocities simulated over 12 selected catchments to analyze the storm moving significance. The storm velocities were: 0 m/s (stationery storm), resonance velocity (same as flow velocity), 0.5 m/s and 5 m/s. The 0.5 m/s velocity represents the impact of slow-moving storm whereas the 5 m/s velocity shows the impact of fast-moving storm. A four directional storm movement analysis was also performed over a single catchment. This analysis gave the idea about the significance of storm moving direction.
4 Result
The results from ArcGIS and HEC-HMS analysis are extracted here. The ArcGIS analysis shows the catchment parameters distributions for four watershed delineation criteria distributed over fifty cities. Results from this analysis are compared with the cloud core size (5 km2) whether Aerial Reduction Factor (ARF) is applicable. Further, the longest flow path velocity in comparison to actual storm velocity shows the possibility of happening resonance effect. The impact of various storm velocity over catchment peak discharge are reported here using HEC- HMS analysis for 12 catchments with various sizes distributed over the catchment size distribution graph. The results represent the catchments behavior for all the expected sizes available in Swedish urban area. The analyses show the catchment behavior in respect of storm movement.
Catchment parameter analysis
The catchment parameters from 1874 delineated catchments distributed in fifty cities according to criterion 1a are plot here with size distribution analysis (Figure 13). It is seen that 95% of all the catchments from watershed delineation criterion 1a, are limited within 5 km2 area.
Therefore, only 5% (94 nos) catchments are bigger than 5 km2 area.
27 Figure 13:Catchment size distribution for criterion 1a
When minimum ellipse area to cover the whole catchment with major axis in longitudinal direction is considered, then 91.39% of catchments are within 5 km2 area.
Figure 14: Ellipse size distribution for criterion 1a
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
0 5 10 15 20 25 30 35
0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3 4.5 4.7 4.9 5.1 More Cumulative frequency
Frequency %
Catchemnt size (km2)
Catchment size distribution
Mean= 1.18 , Median= 0.37, 5th percentile= 0.11 , 95th percentile= 4.78
Frequency % Cumulative %
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
0 10 20 30 40 50 60 70
0-1 1-2 2-3 3-4 4-5 5-10 10-15 15-20 20-55
Cumulative frequency
Frequency %
Ellipse size (km2)
Ellipse size distribution
Mean= 1.71 , Median= 0.53, 5th percentile= 0.15 , 95th percentile= 7.19
Frequency Cumulative %
28 The major and minor axis distribution of the ellipses are shown in Figure 15 and Figure 16. It is seen that 85.2% major axis sizes are limited within 2.5 km, whereas 92.07% of minor axis are limited within 2 km.
Figure 15: Ellipse major axis size distribution for criterion 1a
Figure 16:Ellipse minor axis size distribution for criterion 1a
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
0 5 10 15 20 25 30 35 40 45
0.4-0.6 0.6-0.8 0.8-1 1-2.5 2.5-5 5-12
Cumulative frequency
Frequency %
Major axis (km)
Ellipse major axis size distribution
Mean= 1.53 , Median= 1.09, 5th percentile= 0.54 , 95th percentile= 3.97
Frequency Cumulative %
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
0 5 10 15 20 25 30 35
0.2-0.4 0.4-0.6 0.6-0.8 0.8-1 1-2 2-8
Cumulative frequency
Frequency %
Minor axis (km)
Ellipse minor axis size distribution
Mean= 0.89 , Median= 0.63, 5th percentile= 0.32 , 95th percentile= 2.39
Frequency Cumulative %
29 The longest flow path distributions are shown in Figure 17. The most common range is 1-2 km (44.23%), with a total of 79.46% of all the longest flow paths are limited within 3km length.
Figure 17: Longest flow path length frequency distribution
When considering all the flow paths available in the catchments originated from every pixel (2m×2m) in the catchments and end in the outlet, then the lengths of flow path distributions are shown in Figure 18. There are total 20,844 flow paths found in all 1874 catchments. It is seen that the popular range is 1-2 km, same as longest flow path distribution, but 60.84% are limited within 3 km length. Here a significant percentage (39.16 %) is found over 3 km length since the catchments have more widen area in the upstream part due to leaf shape of catchment. This widen upstream part generates several longer flow paths in comparison to narrow downstream part.
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
80.00%
90.00%
100.00%
0 5 10 15 20 25 30 35 40 45 50
0.5-1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 8-9 9-10 10-1111-1212-1313-1414-1515-16
Cumulative %
%
Length (km)
Longest flow path length frequency distribution
Mean= 2.23 , Median= 1.60, 5th percentile= 0.79 , 95th percentile= 5.75
% Cumulative %
30 Figure 18: Length distribution of flow paths originated from every pixel (2m×2m) of catchments till the outlet point. The distribution contains a total of 20,844 flow paths covering
1874 catchments distributed over 50 cities
The extracted figures from watershed delineation criterion 1b, 2a, 2b are presented in appendix C.
The application of ARF over a catchment’s precipitation volume will reduce the total runoff and peak discharge from the catchment. To understand the relationship between rainfall and total runoff of a catchment, an analysis was performed in a catchment. The relationship was found proportional but non-linear as presented in Figure 19.
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
120.00%
0 5 10 15 20 25 30
0-0.5 0.5-1 1-2 2-3 3-4 4-5 5-10 10-15 15-16
Cumulative %
Frequency %
Length (km)
Flow path length frequency distribution
Mean= 3.011 , Median= 2.34, 5th percentile= 0.46 , 95th percentile= 7.91
Frequency % Cumulative %
31 Figure 19: Precipitation vs total discharge relation for a single catchment
The peak discharge also changes with the change in precipitation volume (Figure 20). This change in peak discharge is more pronounced when storm with high return period is used as shown in Figure 21.
0.1 y 1 y
5 y 10 y
20 y (storm return period)
50 y
100 y
200 y
0 10 20 30 40 50 60 70
0 10 20 30 40 50 60 70 80 90
Total discharge (mm)
total precipitation (mm)
Precipitation vs total discharge curve
