Determination of subcatchment and watershed boundaries in a complex

and highly urbanized landscape

Aimé Kayembe1 and Carl P.J. Mitchell1*

1. Department of Physical and Environmental Sciences, University of Toronto

Scarborough, 1265 Military Trail, Toronto, ON, Canada, M1C 1A4

*corresponding author. Email: carl.mitchell@utoronto.ca; tel: 416-208-2744

Keywords: urbanization, watershed delineation, GIS, stormwater, DEM, remote

sensing.

This article has been accepted for publication and undergone full peer review but has not been
through the copyediting, typesetting, pagination and proofreading process which may lead to
differences between this version and the Version of Record. Please cite this article as doi:
10.1002/hyp.13229

This article is protected by copyright. All rights reserved.

 Abstract

Urban development significantly alters the landscape by introducing widespread impervious

surfaces, which quickly convey surface runoff to streams via stormwater sewer networks,

resulting in “flashy” hydrological responses. Here, we present the inadequacies of using

raster-based digital elevation models (DEMs) and flow-direction algorithms to delineate large

and highly urbanized watersheds and propose an alternative approach that accounts for the

influence of anthropogenically modified land cover. We use a semi-automated approach that

incorporates conventional drainage networks into overland flow paths and define the

maximal runoff contributing area. In this approach, stormwater pipes are clustered according

to their slope attributes which define flow direction. Land areas drained by each cluster and

contributing (or exporting) flow to a topographically delineated catchment were determined.

These land masses were subsequently added or removed from the catchment, modifying both

the shape and the size. Our results in a highly urbanized Toronto, Canada area watershed

indicate a moderate net increase in the directly connected watershed area by 3% relative to a

topographically forced method; however, differences across three smaller scale

subcatchments are greater. Compared to topographic delineation, the directly connected

watershed areas of both the upper and middle subcatchments decrease by 5% and 8%,

respectively whereas the lower subcatchment area increases by 15%. This is directly related

to subsurface storm sewer pipes that cross topographic boundaries. When directly connected

subcatchment area is plotted against total streamflow and flashiness indices using this

method, the coefficients of variation are greater (0.93 to 0.97) compared to the use of DEM-

derived subcatchment areas (0.78 to 0.85). The accurate identification of watershed and

subcatchment boundaries should incorporate ancillary data such as stormwater sewer

networks and retention basin drainage areas to reduce water budget errors in urban systems.

This article is protected by copyright. All rights reserved.

1. Introduction
Urbanization causes profound changes to natural watershed conditions by modifying

vegetation and soil characteristics, introducing drainage and flood control infrastructure, and

sealing much of the surface with impervious and low-permeability surfaces such as parking

lots, roadways, sidewalks and rooftops. These surfaces impede infiltration, surface storage

and evapotranspiration, and increase the volume and velocity of surface runoff, altering the

hydrological cycle (Ahiablame and Shakya, 2016; LaFontaine et al. 2015). Numerous studies

have documented the impacts of urban runoff on stream water quality. Regional case studies

highlight increased loading to receiving streams of pollutants such as nutrients (Newcomer

Johnson et al. 2016), suspended solids (MacAvoy et al. 2016), oil and grease (Sood et al.

2012), fecal coliform bacteria (Paule-Mercado et al. 2016), trace metals (Ruchter and Sures,

2015), thermal energy (Sun et al. 2015) and salts (Corsi et al. 2015). These studies indicate

elevated but highly variable concentrations of solutes in streams with several constituents

surpassing public health standards, stressing aquatic biota and potentially impairing any

beneficial use of water resources in these landscapes.

In urban watersheds, pollutant loads in runoff are consistently correlated with

watershed impervious area (Arnold and Gibbons, 1996). More specifically, directly

connected impervious area (DCIA), the portion of the total impervious area (TIA) which is

directly connected to the storm drainage system, very strongly controls runoff generation in

urban and urbanizing watersheds (Hwang et al. 2017). DCIA differs from other measures of

imperviousness such as effective impervious area (EIA) in that it is solely derived from

surveys and geospatial data analysis. EIA represents the subset of TIA that generates a rapid

response in smaller storms and is based on rainfall-runoff response derived through

modelling (Ebrahimian et al. 2016). One main limitation with EIA in particular is the greater

uncertainty it may be subjected to because the model should be calibrated for parameters such

This article is protected by copyright. All rights reserved.

as infiltration and connectivity of rooftops to the drainage system, which are difficult to

derive at a larger scale (Ebrahimian et al. 2016). Many studies suggest that impervious areas

display a threshold effect, with evidence of stream quality alteration detected at

approximately 7% imperviousness (Schueler et al. 2009), while major stream degradation

takes place where as little as 10% of the surface is impervious (Carle et al. 2005). Although

the direct estimation of impervious surfaces has potential for increasing the predictive power

of urban models of hydrology and water quality, a fundamental problem exists in quantifying

the spatial extent of these surfaces in highly urban areas. Unlike in natural landscapes, urban

areas are unique because they often comprise a mosaic of land use types and include features

such as storm sewer networks and numerous different impervious surfaces. In particular,

storm sewer networks facilitate the quick delivery of surface runoff to streams and increase

hydrologic connectivity within a watershed (Kaushal and Belt, 2012). A correct

understanding of hydrologic connectivity, specifically the areas contributing flow to urban

streams, is critical for managing both the quantity and the quality of stormwater (Jankowfsky

et al., 2013). However, the determination of drainage networks and watershed boundaries in

highly urbanized watersheds remains a complex endeavour and as a result, has often taken an

overly simplified approach.

In natural landscapes, remotely sensed data have been successfully employed to

generate gridded digital elevation models (DEMs), which are used in geographical

information systems (GIS) for drainage network extraction, determination of flow direction

and accumulation, and watershed delineation (Luo et al. 2016; Wechsler, 2007). These

systems use algorithms such as the d8 flow direction (O’Callaghan and Mark, 1984), which

assumes that flow direction always originates at the centre of each DEM pixel and travels

downslope, following the steepest gradient towards one of the eight nearest pixel neighbours

until the watershed outlet is reached (Freeman, 1991). In this algorithm, flow is modelled as a

This article is protected by copyright. All rights reserved.

vector, therefore computing a more precise estimation of the distance to the outlet, surface

drainage and average flow-path length. Over the years several add-ons to the d8 algorithms

have resulted in satisfactory determination of catchment boundaries in natural landscapes

(Wilson et al., 2008). These add-ons include routines such as stream burning, which assigns

all stream cells the initial elevation values (as calculated by the DEM); while each off-stream

cell is “raised” by adding a constant value to the corresponding original DEM elevation

value. This ensures that flow is constrained to only those cells identified as stream cells

(Jones, 2002). Other methods such as surface reconditioning have been incorporated to

eliminate flat areas and spurious pits from the elevation grid values (Hutchinson, 1989). Flow

direction forcing is used to derive hydrologic characteristics of surfaces and to improve

catchment boundaries identification.

Multiple-neighbour flow algorithms have also been developed. In particular, these

methods incorporate divergent flow which spreads from one cell to several downhill cells

(Park, 2013). The most common methods include D-infinity, DEM Network (DEMON), and

TOPMODEL. These methods are adequate for handling geomorphologically complex

terrains such as karsts, and landforms created by running water on hillslopes (e.g. gullies) (Su

et al. 2014). However, these algorithms suffer several limitations including their tendency to

generate over-dispersion. These methods can cause flow to spread too much, with some

fraction nearly flowing along the contours (Gruber and Peckham, 2009), leading to

inconsistent flow geometries on peaks and ridges. The widespread use of the d8 algorithm on

the other hand stems from its simplicity and realistic depiction of convergent flow conditions

(Freeman, 1991), as well as in its ability to estimate reasonable contributing areas between

flow patterns (Ariza-Villaverde, 2015).

The use of DEMs for watershed delineation in urbanized landscapes is dependent on

the accuracy of the DEM. It is essential that DEMs in urbanized watersheds accurately

This article is protected by copyright. All rights reserved.

represent topographic features at a scale where drainage features are controlled (Jenson and

Domingue, 1988). Inaccurate DEM representation may generate oversimplified or distorted

drainage patterns in grid-based models (Duke, 2006). For instance, in their floodplain

delineation study Charrier and Li (2012) established that fine scale DEMs provided a

complex and detailed representation of floodplain boundaries corresponding to field-mapped

exercises, but they also produced erroneous watershed boundaries. This was attributed to

minor topographic features that increased simulation uncertainties. Fine scale DEMs are, in

some cases, unable to discriminate between surfaces. For instance, a bridge across a stream

can be viewed as a topographic divide for water flow, while in reality stream water flows

beneath the bridge. In another study Yang et al. (2014) examined the effect of raster cell size

on hydrographic feature extraction and determined that finer resolution DEMs do not

necessarily improve stream network and watershed boundary determination. Furthermore,

despite their greater accuracy in defining terrain features, fine resolution DEMs (e.g. < 1m),

primarily derived from LiDAR, cannot also reveal buried stormwater infrastructure (e.g.

pipes and culverts) (Murphy et al. 2008; Meesuk et al. 2015). The impact of water

conveyance infrastructure on catchment natural (DEM-derived) boundaries has been depicted

in different landscapes. For instance, Duke et al. (2005) delineated a rural watershed,

reporting differences of up to 49% between the area derived using DEM and the area

estimated by incorporating roads, ditches, irrigation canals, culverts, siphons and flumes data.

In recent years, peri-urban areas have been the focus of studies seeking to provide a

better understanding of the hydrological cycle within mixed and artificial landscapes

(Branger et al. 2013). Catchments in peri-urban landscapes comprise relatively larger and

distinct areas of natural cover (e.g. forest), agricultural land and built regions which lead to a

mix of fast and slow hydrologic responses and variable flow paths. The spatial arrangement

of impervious and pervious surfaces and their connectivity to the drainage network are the

This article is protected by copyright. All rights reserved.

main factors controlling the magnitude of runoff responses in these landscapes (Fletcher et al.

2013) . Peri-urban environments do however have a relatively greater stormwater retention

capacity compared to highly urbanized landscapes (Miller et al. 2014). Studies have

demonstrated that the built environment in these landscapes can significantly impact the

natural (pre–urbanization) catchment boundaries. For instance, by incorporating sewer

network, sewer overflow, ditches and retention ponds in the delineation process of a peri-

urban catchment, Jankowfsky et al. (2013) found a 25% increase in the new catchment area

relative to the DEM-based area.

In urban areas surface runoff can be redirected across topographic boundaries mainly

via storm sewer systems (Schmitt et al. 2004) with highly urbanized catchments typically

having a much greater DCIA percentage relative to peri-urban or primarily natural

catchments. Hence, in such landscapes, where the nature of hydrologic connectivity is

significantly altered, the traditional approach for watershed delineation based solely on

terrain topography may fail to capture all features that influence drainage patterns

(Jankowfsky et al., 2014). In most cases, the lack of data on storm sewer infrastructure (or

their completeness), as well as the availability of adequate geospatial software tools capable

of describing urban water systems, are the major impediments limiting the proper analysis of

highly urbanized watersheds (Parece and Campbell, 2015). These factors are suggestive as to

why a relatively large number of studies in urban watersheds still use data obtained from

traditional delineation methods in hydrologic studies.

In this study we propose and evaluate an approach to define the boundaries of a large

and highly urbanized watershed and its subcatchments by incorporating into the delineation

process built environments such as storm sewer systems, stormwater management

infrastructure (e.g. retention ponds), terrain and land cover (impervious and pervious). This

This article is protected by copyright. All rights reserved.

study highlights the inadequacy of relying only on topographic data to determine directly

connected drainage area in highly developed urban landscapes.

2. Methods

2.1 Study Site

The Mimico Creek watershed (Figure 1) is located in the Greater Toronto Area

(GTA), Canada’s largest metropolitan area. The watershed crosses three municipalities, with

its headwaters in the City of Brampton, crossing into Mississauga, then Toronto, where it

empties into Lake Ontario. The Mimico Creek watershed is long and narrow, with an

elevation ranging from 80 to 250 m above sea level (TRCA, 2010). The hydrology of the

stream is characterized by very low summer base flows and high, but short peak flows,

especially during spring and fall related to melt and storm events. The area receives an

average annual precipitation of approximately 760 mm, with the highest average monthly

precipitation of about 80 mm in August and the lowest of 50 mm in February (TRCA, 2014).

Mimico Creek consists of a main channel fed by an upper west branch and upper east

branch. The stream has a total channel length of 58 km, an average slope of 3.4% and a mean

annual flow rate of 0.8 m3/s. The surficial geology in this watershed is characterized

primarily by glaciolacustrine geologic units deposited ̴ 12,500 years ago. These units vary

from near shore sand and gravel beach deposits of the Lake Iroquois shoreline situated within

the lower section of the watershed, to the fine sands, silts and clays of glaciolacustrine

pondings commonly found north of the Lake Iroquois shoreline (TRCA, 2010). Three distinct

physiographic units, associated with flat to low slope topography, dominate the physiography

of the watershed. These units include the South Slope, Peel Plain and the Iroquois Plain

(Baker et al. 1998; TRCA, 2010). The South Slope occupies parts of the lower half and a

small section of the upper region of the watershed. The clay till at the surface of the South

This article is protected by copyright. All rights reserved.

Slope limits groundwater recharge in these areas. The Peel Plain physiographic unit is located

in the upper and central sections of the watershed. Similar to the South Slope, these areas are

characterized by low hydraulic conductivity. Historically, this region included wetlands

which played a major role in flow attenuation (Chapman and Putman, 1973; Snell, 1988).

Finally, the Lake Iroquois Sand Plain unit splits into two the region covered by the South

Slope Plain in the lower half of the watershed. This physiographic unit straddles the current

Dundas Street and comprises soil silt, clay, sand and gravel deposits (Chapman and Putnam,

1973). Historically, this region facilitated groundwater recharge and provided baseflow to

Mimico Creek (TRCA, 1998). However, these functions have been significantly diminished

because of intensive urban development.

Urbanization in the Mimico Creek watershed increased from approximately 55%

cover in 1978 to 77% by 1998 and currently, approximately 98% of the watershed is

considered urbanized, making it the most developed watershed in Ontario (TRCA, 2010).

The majority of the urban land use is industrial (in the northeast and central sections) and

residential (northwest and southern sections), together covering approximately ¾ of the

drainage area. Transportation lands include the eastern reaches of Toronto’s Lester B.

Pearson International Airport, which is situated in the mid-west section of the watershed. The

watershed also features some of the busiest transport corridors in the country such as major

highways (i.e. highways 401, 403, 407, 410, 427 and the Queen Elizabeth Way) and several

important regional roads.

The Mimico Creek watershed has substantial drainage and flooding problems as a

result of its rapid and relatively early urbanization, during which there was little regard to

stormwater management (TRCA, 2010). Most existing stormwater management ponds (17

out of 23) are located in the upper reaches of the watershed. Most of these ponds fail to meet

required standards as they were developed before the implementation of water quality and

This article is protected by copyright. All rights reserved.

erosion control standards (TRCA, 2010 ). Only approximately 30% of the developed areas

have stormwater management plans, thus leaving the vast majority of the watershed without

viable stormwater management controls (TRCA, 2010).

2.2 Spatial Data

The GIS shapefiles of the storm sewer networks and impervious surface (roads, sidewalks,

and parking lots) used in this study were based on data generated in 2014 and later.

Shapefiles were obtained directly from municipal offices or downloaded through the

University of Toronto Map and Data Library (Table 1). Road shapefiles were based on DMTI

Spatial’s 2014 data, which provide a detailed road and highway network for the entire

province; these data were accessed through the University of Toronto. The parking lot

database comprises data current to 2015 and was accessed through municipalities’ websites.

Sidewalk shapefiles were based on data posted by the three municipalities as open source

data (e.g. the City of Toronto) or shared with the University of Toronto (e.g. Municipalities

of Mississauga and Brampton). Additionally, a stream network file from the hydrology

dataset generated by the Ontario Ministry of Natural Resources and Forestry was also used to

aid with the delineation process. This set uses a common map projection and incorporates

raster elevation data, stream networks and a flow direction grid based on the standard d8

algorithm. Other data included a shapefile of the streamflow monitoring stations along the

Mimico Creek based on their coordinates, a 1 metre contour map of the study area generated

by J.D Barnes Limited (2002) which was accessed through the University of Toronto Map

and Data library, and aerial photographs of the study site as found in the ArcGIS software. In

particular, the Esri World Imagery map service used in this study contains high-resolution

(0.08 m) satellite and aerial imagery generated in 2015 and embedded in ArcGIS as a

basemap. All maps were georeferenced to the coordinate system NAD83 UTM Zone 17N and

This article is protected by copyright. All rights reserved.

processed in ArcGIS 10.3. In particular, flow accumulation and direction was modelled using

the Spatial Analyst Hydrology toolset found in ArcGIS.

2.3 Watershed and Subcatchment Delineation

A preliminary topographic delineation of 3 discrete subcatchments and the full Mimico Creek

watershed was carried out using the spatial analysis tools in GIS and the flow accumulation

network from the 1-meter elevation model. In this exercise we considered the flow

monitoring stations in each subcatchment as outlets. Subcatchment delineation was

performed automatically using standard terrain analysis methods based on the d8 direction

algorithm and enhanced flow direction forcing. The latter ensures that flow is assigned more

accurate directionality (Kenny and Matthews, 2005) (Figure 1).

Subsequent to this analysis, directly connected subcatchment areas were derived by

superimposing the storm drain infrastructure and impervious surfaces layers on DEM maps to

assess how man-made infrastructure alters each subcatchment’s DEM-derived boundaries

through flow control. Digital maps of impervious surfaces such as roads and sidewalks were

available in line format. The lines spatially represent the centerlines of these impervious

surfaces. We defined roads and sidewalks by buffering centerlines using specified width

tolerances relative to each type of surface, based on the road classification criteria (City of

Toronto, 2013a), and by using the dissolve and merge geoprocessing tools on the buffered

geometries. To ascertain accuracy, we overlaid the polyline layer on aerial photos and

examined the match visually in multiple areas of the watershed, selected randomly. This

consisted of verifying that road features such as horizontal bends, road location and road

width (including lane and shoulder widths) corresponded in both layers. In this study we

assumed that roof runoff from residential areas is rerouted onto pervious surfaces. The

premise of this assumption is based on a mandatory downspout disconnection bylaw in the

This article is protected by copyright. All rights reserved.

City of Toronto (City of Toronto, 2013b). In the Region of Peel, subscription to the

downspout disconnection program is voluntary, but financial incentives are provided to

eligible homeowners (Region of Peel, 2016). While the number of residents in Region of Peel

that had subscribed to this rebate program was not available, program uptake may be as high

as 80% (Region of Peel, personal communication). Hence, we treated rooftops as isolated

impervious surfaces that were not directly connected to the drainage network.

We combined three different types of boundary identification to determine the

directly connected subcatchment areas. The choice of a delineation approach was based on

the extent of urbanization within a specific location. Firstly, to identify boundaries in highly

developed zones, we determined built structures (stormwater pipes and impervious surfaces)

that either contribute flow to a subcatchment or reroute it away (Parece and Campbell, 2015).

We expanded the DEM derived boundaries by appending sewershed that contributes outside

flow to the subcatchment, or removed such areas when flow was exported to adjacent

subcatchments. Flow direction within a pipe was based on its gradient attribute, calculated

using pipe length and rise. We treated the point at which the DEM-derived boundary

intersects with a stormwater pipe as an outlet and subsequently quantified the upstream

runoff-contributing land area. This was achieved by using the intersect tool of ArcGIS, which

takes in a polygon and a line as inputs and generates an output type parameter set to point.

We then summarized the clipped boundary with a count operation to count the number of

stormwater pipes crossing the delineated boundary. We then used a recursive Visual Basic

for Applications function that compared slope attribute values to identify pipes belonging to

the same “sub-sewershed”. The function takes the slope of a pipe crossing from outside into

the watershed (or subcatchment), recursively compares it to slopes of adjacent pipes moving

upstream (opposite to the direction of the flow). The limiting condition to the function is

found once a discontinuity in the series (e.g. first pipe of a series) or a “ridge” position

This article is protected by copyright. All rights reserved.

(change in slope sign) in a pipe series is encountered. Since flow is gravity-driven in these

systems (Marsalek et al. 1993), a change in slope indicates a discontinuity in a given

direction of flow. Finally, the function returns an array (attribute ID) of pipes belonging to

the same sub-sewershed. In each case, the sub-sewershed area defined by a cluster of pipes

was determined by altering the initial topographic data to delineate runoff contributing

impervious surfaces.

Secondly, in developed regions adjacent to open areas (e.g. parks or golf courses), we

combined DEM based methods and an object-oriented approach. This is because such areas

(e.g., parcel units adjacent to large open areas) comprise a relatively higher percentage of

pervious surfaces compared to highly urbanized areas. By “object-oriented”, we refer to built-

up regions adjacent to pervious landscapes and connected to a sewer system (Jankowfsky et

al. 2014 ). In this approach, a single parcel of land or a block is connected to the closest and

lowest sewer pipe (Mitchell and Diaper, 2006). Lastly, to identify catchment boundaries in

open areas (e.g. parks) we relied on DEM data to model the convergence of flow. In all three

scenarios we combined aerial basemap photographs with vector data and visual analysis to

determine flow direction. Prior to determining the directly connected subcatchment area, we

also identified retention and detention basins in subcatchments. Stormwater detention basins

control runoff through temporary reservoirs, eventually releasing water at a controlled rate to

a receiving water body. Retention basins, in contrast, are stormwater management systems

that constantly maintain standing water. Although both retention and detention basins are

designed to attenuate runoff volume in order to limit peak flow rates at downstream sites,

retention basins in particular decrease the net surface water budget of a catchment by

recharging aquifers (Salvadore et al. 2015). In Southern Ontario, retention ponds are lined

with low permeability materials to maintain a permanent pool of water promoting the

sedimentation of particulates, organic matter and metals (Drake and Guo, 2008). We treated

This article is protected by copyright. All rights reserved.

these ponds as local pour points, and the contributing upstream lands were delineated as

separate drainage basins. These areas were subsequently removed from the maximal

theoretical subcatchment area. We then generated a multi - polygon layer that defines

hydrologically accurate drainage areas of how stormwater is conveyed through the three

subcatchments and eventually discharged to Mimico Creek.

2.4 Hydrologic analysis

We used stream stage records obtained from average hourly measurements at three

monitoring stations along Mimico Creek to calculate streamflow records. These stations,

managed by the University of Toronto, are fitted with unattended, telemetric water level

sensors. Stream stage data between December 2015 and May 2016 were converted to

discharge using established rating curves. In addition to comparisons between watershed area

and streamflow, we calculated streamflow flashiness using both the Richards Pathlength (LR;

Gustafson et al. 2004) and the Richards-Baker flashiness index (RBI; Baker et al. 2004). The

LR calculates the length of a hydrograph line over a period of time and divides it by the

median daily flow during the study period and the duration of the study period, according to:

∑𝑛𝑖=1 √(𝑞𝑖 − 𝑞𝑖−1 )2 + (𝑡𝑖 − 𝑡𝑖−1 )2

𝐿𝑅 =
(𝑡𝑛 − 𝑡0 ) ∗ 𝑄

where q is the mean daily flow, t is time and Q is the median flow during the study period.

The RBI is a dimensionless and relative estimation of stream flashiness that measures the

magnitude of oscillations in flow relative to the total streamflow over a given period of time,

according to:

∑𝑛𝑖=1|𝑞𝑖 − 𝑞𝑖−1 |
𝑅𝐵𝐼 =
∑𝑛𝑖=1 𝑞𝑖

This article is protected by copyright. All rights reserved.

3. Results

3.1 Maximal theoretical subcatchment areas and boundaries

The DEM-derived watershed and three subcatchment areas differed more

substantially from the maximal theoretical area of the directly connected watershed at the

subcatchment scale than at the overall watershed scale (Table 2; Figure 2). The directly

connected watershed overall increased in size by 2.8% increase compared to the DEM

derived watershed. The directly connected upper and mid subcatchments lost approximately

5.0% and 7.9% of their land areas, respectively; whereas the lower subcatchment gained

14.6% relative to the DEM derived area. The directly connected watershed perimeter

increased by approximately 8.0% relative to the DEM-derived perimeter, with much of this

increase as the result of increased surface area in the lower subcatchment. Additionally, we

measured a 12.1% decrease in the common boundary between the upper and the mid

subcatchments.

The stormwater sewer pipe network can have a strong influence on watershed

boundaries. This is particularly apparent when one focuses in on the lower subcatchment

(Figure 3). In delineating the directly connected drainage areas for each subcatchment we

identified 338 locations where stormwater sewer pipes traversed the DEM derived watershed

boundary. A total of 180 stormwater inlets contributed flow to the Mimico Creek DEM-based

watershed and the remaining pipes (158) diverted flow outside of the watershed.

Additionally, we also identified 7 pipes in the mid subcatchment that contributed flow to the

upper subcatchment and 19 pipes in the upper subcatchment that were adding flow to the mid

subcatchment. We identified 3 pipes in the lower subcatchment that directed flow to the mid

subcatchment and 6 pipes that contributed stormwater flow from the mid to the lower

subcatchment. We also identified and quantified 5 areas that contribute runoff to retention

ponds. Of these, 2 were located in the upper subcatchment and the remaining in the mid

This article is protected by copyright. All rights reserved.

subcatchment. In the upper and mid catchments, this resulted in 0.25% and 1.1% losses in

catchment area, respectively.

3.2. Relationships with streamflow and hydrograph flashiness

Coefficients of variation in linear relationships between directly connected catchment areas

and both streamflow and flashiness indices were greater than when DEM-derived catchment

areas were used (Figure 4). Admittedly, the low sample numbers in these relationships should

be treated with a great deal of caution; however, the consistency of better explained

relationships using the directly connected catchment areas suggests that this method is a more

accurate determination of the actual contributing area.

Discussion

The determination of catchment boundaries in this study was approached using a

multi-step and iterative process that integrates flow direction and flow accumulation methods

with vector datasets of the stormwater conveyance network and with contributing areas to

stormwater retention ponds to derive flow contributing areas. Using this approach, changes in

area of a highly urbanized Toronto-area watershed were significant at the subcatchment scale,

indicating that directly connected urban infrastructure, such as roads and stormwater sewer

network pipes, have a significant control over the size and shape of urban drainage basins.

This approach for delineating large watersheds in highly urbanized landscapes holds strong

potential for application elsewhere, though further study across a larger number of urban

watersheds is needed.

Based on the findings of this study, we infer that the relationship between stream

hydrologic metrics and catchment area is sensitive to the way catchment boundaries are

defined. Accounting for the influence of the built environment in the delineation process

This article is protected by copyright. All rights reserved.

resulted in a more linear relationship between subcatchment area and streamflow metrics

including total flow and flashiness. Total streamflow usually does scale closely with

watershed area even in natural areas (e.g., Gianfagna et al. 2015). Hydrograph flashiness (LR)

also tends to decrease as watershed areas increase because of longer average flow pathways

for water to reach the outlet, however the RBI index is normalized to median flows in a

watershed and thus relationships with watershed area are more subtle than for LR (Baker et al.

2004). Given the high degree of impervious cover across our entire study watershed and thus

smaller influences and spatial heterogeneities in storage, we assume a stronger linear

relationship being related to a more accurate watershed area delineation to be more valid here

than in natural landscapes. It is worth noting that the stronger relationships observed come

primarily from the loss in land masses recorded in the mid subcatchment (8%) and the gain

reported in the lower subcatchment (15%). While the scalability of these relationships is

questionable because of the analysis of only three subcatchments, the relationships were

consistently weaker using topography-based watershed delineation, thus our findings are less

likely simply related to other errors such as in flow determination. While ours is not the first

work to add storm sewer networks or retention basins into the watershed delineation process

(see e.g., Jankowfsky et al. 2013; Parece and Campbell, 2015), previous work has not

included this type of flow comparison to watershed delineation approaches, perhaps because

the focus of previous work has been at a much smaller geographic scale.

A major distinction of this work from previously reported research is the relatively

large scale at which we could accomplish a semi-automated determination of watershed area.

Previous research has focused on relatively small, neighbourhood-scale delineations (~0.2 – 9

km2 in Jankowfsky et al. 2013; Parece and Campbell, 2015; Rodriguez et al., 2003; Sample et

al. 2001) where the inclusion of storm sewer pipes crossing topographic boundaries is based

on individual point identifications and upstream route tracing. Such manual identification in

This article is protected by copyright. All rights reserved.

the relatively large (~74 km2) and heavily urbanized Mimico Creek watershed of this study

would have been extremely time consuming because of the hundreds of sewer pipes in the

watershed. Our approach which automatically clusters storm sewer pipe gradients into

directional clusters at watershed boundaries is a more automated approach, which should

enable its more efficient application in other larger and highly urbanized watersheds.

Specific to this study, we encountered several shortcomings. For instance, we

identified isolated stormwater pipes that were not connected to the main sewer network, the

stream channel, or to stormwater management ponds. It was not therefore clear whether the

land drained by these pipes contributed or removed land area in the subcatchments. This

situation was prevalent in the oldest sections of the upper and mid subcatchments, and is most

likely a function of older, less accurate, and perhaps unmapped sewer main records. Data

sources may not be exhaustive in the sense that municipal authorities may perform alterations

to the drainage infrastructure and such changes may take longer to be updated into spatial

databases (Parece and Campbell, 2015). Additionally, some datasets may contain incomplete

documentation and the missing information may be a source of analytical uncertainty. When

accurate and current storm sewer data is difficult to obtain, further research is needed to go beyond

mapping exercises and possibly use other types of data to infer changes in catchment areas. One

possible avenue may be the use of isotope or hydrochemical tracers. For example, Christian et al.

(2011) have shown that 87Sr/86Sr values in urban rivers correlate strongly with urbanization indices

such as impervious cover. Testing hydrochemical approaches against spatial mapping approaches

would be a logical first step. One main question arising from the use of the object-oriented

approach in this study relates to whether or not a parcel or a block unit, in moderately

developed areas situated on drainage boundaries, constitutes a hydrologic unit; which is a

pure hydrologically defined drainage area. In reality, flow into residential backyards, for

instance, can follow multiple pathways and contribute hydrologically to different areas

This article is protected by copyright. All rights reserved.

compared to the flow collected on rooftops, which is often, but not always, channelled to

storm sewers.

Inaccurate delineation of watersheds may lead to jurisdictional cross-boundary

contentions, possibly exacerbating boundary choice issues in watershed management (Cohen

and Davidson, 2011). This is because regions that contribute flow to the stormwater

management areas may extend well beyond initially estimated land areas, particularly if such

areas have undergone iterative further developments with additions of sewer pipes and

alterations to the surface topography. Oftentimes, spatial records are also not accurately

updated, exacerbating this issue. Studies indicate that across the globe, urban land expansion

rates are higher than or equal to urban population growth rates (d’Amour et al. 2017). It is

therefore highly likely that the current morphology of urban watersheds will also change as

continuing urbanization will increase the hydrologic connectivity between regions across

topographic boundaries, adding possibly poorly otherwise delineated areas to urban

watersheds. Urban watershed and subcatchment boundaries therefore need to be updated as

urban areas expand. The approach in this paper, although it requires further validation across

more watersheds, does provide a more automated means for both catching up with better

characterizing other poorly delineated urban watersheds and to assist in keeping up with

ongoing development, particularly for larger (10’s km2) watersheds. Maintaining an updated

and accurate infrastructure database will enhance management capacity (e.g. flood

forecasting) and help to more efficiently resolve the myriad issues related to urban runoff

(Paul and Meyer, 2001).

5. Conclusions

This study presents an alternative approach to watershed delineation in complex and

highly urbanized areas by demonstrating that landscape transformation resulting from

This article is protected by copyright. All rights reserved.

urbanization can significantly alter the perceived drainage area that is derived using

topography. The application of terrain methods for boundary identification to our study site

underestimated the total watershed area and its lower subcatchment whereas the mid and

upper subcatchment areas were inflated. From these results we conclude that conventional

flow-direction algorithms such as the D8 are not designed to predict watershed structure

accurately in highly urbanized landscapes. Accounting for land use features such as

impervious surfaces and sewer network can significantly improve urban watershed

identification. Additionally, incorporating aspects of urban stream flow metrics such as

flashiness and total flow are useful additions to this process in providing lines of evidence

toward inferring a more accurate watershed area calculation. While there are previous studies

that have delineated catchments in urbanized landscapes, these studies have comparably more

limited scope (small catchment area and mostly homogeneous land use). A large-scale

perspective is important not only for informing jurisdiction, management and conservation

decisions but also for effectively assessing the cumulative effects of hydrologic connectivity

in highly urbanized landscapes on flow regime in local streams.

The accurate identification of watershed boundaries is a major building block towards

gaining insights into stream hydrological and geochemical responses in urban landscapes. It

will lead to a more precise determination of key parameters (e.g., runoff ratio and percentage

of effective or directly connected impervious catchment area) that can provide not only a

better understanding on the hydrological functioning of urban landscapes but also aid in

defining useful indicators for assessing the impacts of urbanization on catchment ecosystems.

Although this study was conducted on only one heavily urbanized watershed, it forms a basis

for comparison in future work, particularly at the extreme end of impervious cover and

urbanization.

This article is protected by copyright. All rights reserved.

Acknowledgments

Funding support from the NSERC Discovery program, as well as from FedDEV

Ontario and the Ontario Ministry of Research and Innovation via the Southern Ontario Water

Consortium, is gratefully acknowledged. The authors are especially indebted to the

University of Toronto Map and Data Library, the municipalities of Brampton, Mississauga

and Toronto and the Region of Peel for their collaboration and provision of data. Finally, a

special thank you to the two anonymous reviewers and Dr. Claire Oswald for their

constructive input to this manuscript.

References
Ahiablame, L., and Shakya, R. (2016). Modeling flood reduction effects of low impact

development at a watershed scale. Journal of Environmental Management, 171, 81-91.

Ariza-Villaverde, A. B., Jiménez-Hornero, F. J., and de Ravé, E. G. (2015). Influence of

DEM resolution on drainage network extraction: A multifractal

analysis. Geomorphology, 241, 243-254.

Arnold Jr, C. L., and Gibbons, C. J. (1996). Impervious surface coverage: the emergence of a

key environmental indicator. Journal of the American Planning Association, 62(2), 243-

258.

Baker, C. L., Lahti, L. R., and Roumbanis, D. C. (1998). Urban geology of Toronto and

surrounding area. In Urban Geology of Canadian Cities, Karrow P.F., and White, O.L.

(Eds). Geological Association of Canada, GAC Special Paper, 42, 323-352.

Baker, D. B., Richards, R. P., Loftus, T. T., and Kramer, J. W. (2004). A new flashiness

index: characteristics and applications to Midwestern rivers and streams. JAWRA Journal

of the American Water Resources Association, 40(2), 503-522.

This article is protected by copyright. All rights reserved.

Branger, F., Kermadi, S., Jacqueminet, C., Michel, K., Labbas, M., Krause, P., Kralisch, S.

and Braud, I. (2013). Assessment of the influence of land use data on the water balance

components of a peri-urban catchment using a distributed modelling approach. Journal of

Hydrology, 505, 312-325.

Carle, M. V., Halpin, P. N., and Stow, C. A. (2005). Patterns of watershed urbanization and

impacts on water quality. JAWRA Journal of the American Water Resources

Association, 41(3), 693-708.

Chapman, L. J., and Putnam, D. F. (1973). Physiography of southern Ontario. Published for

the Ontario Research Foundation, University of Toronto, Canada.

Charrier, R., and Li, Y. (2012). Assessing resolution and source effects of digital elevation

models on automated floodplain delineation: a case study from the Camp Creek

Watershed, Missouri. Applied Geography, 34, 38-46.

Christian, L. N., Banner, J. L., and Mack, L. E. (2011). Sr isotopes as tracers of

anthropogenic influences on stream water in the Austin, Texas, area. Chemical Geology,

282(3-4), 84-97.

City of Toronto. (2013a). City of Toronto Road Classification System: Summary Document.

Retrieved on May 28, 2017 from

https://www1.toronto.ca/City%20Of%20Toronto/Transportation%20Services/Road%20Cl

assification%20System/Files/pdf/2012/rc_document.pdf

City of Toronto. (2013b). Mandatory Downspout Disconnection. Retrieved on October 23,

2017 from https://www.toronto.ca/311/knowledgebase/kb/docs/articles/toronto-

building/customer-service/zoning-front-yard-back-yard-landscaping-requirements-

driveway-widths.html

Cohen, A. and Davidson S. (2011). The watershed approach: Challenges, antecedents, and

the transition from technical tool to governance unit. Water Alternatives, 4(1), 1-14.

This article is protected by copyright. All rights reserved.

Corsi, S. R., De Cicco, L. A., Lutz, M. A., and Hirsch, R. M. (2015). River chloride trends in

snow-affected urban watersheds: increasing concentrations outpace urban growth rate and

are common among all seasons. Science of the Total Environment, 508, 488-497.

d’Amour, C.B., Reitsma, F., Baiocchi, G., Barthel, S., Güneralp, B., Erb, K.H., Haberl, H.,

Creutzig, F. and Seto, K.C. (2017). Future urban land expansion and implications for

global croplands. Proceedings of the National Academy of Sciences, 114(34), 8939-8944.

Drake, J., and Guo, Y. (2008). Maintenance of wet stormwater ponds in Ontario. Canadian

Water Resources Journal, 33(4), 351-368.

Duke, G. D., Kienzle, S. W., Johnson, D. L., and Byrne, J. M. (2006). Incorporating ancillary

data to refine anthropogenically modified overland flow paths. Hydrological

Processes, 20(8), 1827-1843.

Ebrahimian, A., Gulliver, J. S., and Wilson, B. N. (2016). Effective impervious area for

runoff in urban watersheds. Hydrological Processes, 30(20), 3717-3729.

Fletcher, T. D., Andrieu, H., and Hamel, P. (2013). Understanding, management and

modelling of urban hydrology and its consequences for receiving waters: A state of the art.

Advances in Water Resources, 51, 261-279.

Freeman, T. G. (1991). Calculating catchment area with divergent flow based on a regular

grid. Computers and Geosciences, 17(3), 413-422.

Gianfagna, C.C., Johnson, C.E., Chandler, D.G. and Hofmann, C. (2015). Watershed area

ratio accurately predicts daily streamflow in nested catchments in the Catskills, New York.

Journal of Hydrology: Regional Studies, 4, 583-594.

Gruber, S., and Peckham, S. (2009). Land-surface parameters and objects in hydrology.

Developments in Soil Science, 33, 171-194.

This article is protected by copyright. All rights reserved.

Gustafson, D.I., Carr, K.H., Green, T.R., Gustin, C., Jones, R.L. and Richards, R.P. (2004).

Fractal-based scaling and scale-invariant dispersion of peak concentrations of crop

protection chemicals in rivers. Environmental Science & Technology, 38(11), 2995-3003.

Hutchinson, M. F. (1989). A new procedure for gridding elevation and stream line data with

automatic removal of spurious pits. Journal of Hydrology, 106(3-4), 211-232.

Hwang, J., Rhee, D. S., and Seo, Y. (2017). Implication of directly connected impervious

areas to the mitigation of peak flows in urban catchments. Water, 9(9), 696.

Jankowfsky, S., Branger, F., Braud, I., Gironás, J., and Rodriguez, F. (2013). Comparison of

catchment and network delineation approaches in complex suburban environments:

application to the Chaudanne catchment, France. Hydrological Processes, 27(25), 3747-

3761.

Jankowfsky, S., Branger, F., Braud, I., Rodriguez, F., Debionne, S., and Viallet, P. (2014).

Assessing anthropogenic influence on the hydrology of small peri-urban catchments:

Development of the object-oriented PUMMA model by integrating urban and rural

hydrological models. Journal of Hydrology, 517, 1056-1071.

J.D Barnes Limited. Toronto Contours 1 metre intervals. (2012). University of Toronto.

Retrieved April 3, 2017 from

http://maps.library.utoronto.ca/datapub/toronto/contours_2002/index.html

Jenson, S. K., and Domingue, J. O. (1988). Extracting topographic structure from digital

elevation data for geographic information system analysis. Photogrammetric Engineering

and Remote Sensing, 54(11), 1593-1600.

Jones, R. (2002). Algorithms for using a DEM for mapping catchment areas of stream

sediment samples. Computers and Geosciences, 28(9), 1051-1060.

Kaushal, S. S., and Belt, K. T. (2012). The urban watershed continuum: evolving spatial and

temporal dimensions. Urban Ecosystems, 15(2), 409-435.

This article is protected by copyright. All rights reserved.

Kenny, F., and Matthews, B. (2005). A methodology for aligning raster flow direction data

with photogrammetrically mapped hydrology. Computers and Geosciences, 31(6), 768-

779.

LaFontaine, J. H., Hay, L. E., Viger, R. J., Regan, R. S., and Markstrom, S. L. (2015). Effects

of climate and land cover on hydrology in the southeastern US: Potential impacts on

watershed planning. JAWRA Journal of the American Water Resources

Association, 51(5), 1235-1261.

Luo, G. J., Wang, S. J., Bai, X. Y., Liu, X. M., and Cheng, A. Y. (2016). Delineating small

karst watersheds based on digital elevation model and eco-hydrogeological

principles. Solid Earth, 7(2), 457-468.

MacAvoy, S. E., Plank, K., Mucha, S., and Williamson, G. (2016). Effectiveness of foam-

based green surfaces in reducing nitrogen and suspended solids in an urban

installation. Ecological Engineering, 91, 257-264.

Marsalek, J., Barnwell, T.O., Geiger, W., Grottker, M., Huber, W.C., Saul, A.J., Schilling, W.

and Torno, H.C. (1993). Urban drainage systems: design and operation. Water Science and

Technology, 27(12), 31-70.

Meesuk, V., Vojinovic, Z., Mynett, A. E., and Abdullah, A. F. (2015). Urban flood modelling

combining top-view LiDAR data with ground-view SfM observations. Advances in Water

Resources, 75, 105-117.

Miller, J. D., Kim, H., Kjeldsen, T. R., Packman, J., Grebby, S., and Dearden, R. (2014).

Assessing the impact of urbanization on storm runoff in a peri-urban catchment using

historical change in impervious cover. Journal of Hydrology, 515, 59-70.

Mitchell, V. G., and Diaper, C. (2006). Simulating the urban water and contaminant

cycle. Environmental Modelling and Software, 21(1), 129-134.

This article is protected by copyright. All rights reserved.

Murphy, P. N., Ogilvie, J., Meng, F. R., and Arp, P. (2008). Stream network modelling using

lidar and photogrammetric digital elevation models: a comparison and field

verification. Hydrological Processes, 22(12), 1747-1754.

Newcomer Johnson, T. A., Kaushal, S. S., Mayer, P. M., Smith, R. M., and Sivirichi, G. M.

(2016). Nutrient retention in restored streams and rivers: A global review and

synthesis. Water, 8(4), 116.

O'Callaghan, J. F., and Mark, D. M. (1984). The extraction of drainage networks from digital

elevation data. Computer Vision, Graphics, and Image Processing, 28(3), 323-344.

Parece, T. E., and Campbell, J. B. (2015). Identifying Urban Watershed Boundaries and Area,

Fairfax County, Virginia. Photogrammetric Engineering and Remote Sensing, 81(5), 365-

372.

Park, J. (2013). Review on remote sensing data-based stream mapping algorithms. Forest

Science and Technology, 9(2), 87-96.

Paul, M. J., & Meyer, J. L. (2001). Streams in the urban landscape. Annual Review of

Ecology and Systematics, 32(1), 333-365.

Paule-Mercado, M. A., Ventura, J. S., Memon, S. A., Jahng, D., Kang, J. H., and Lee, C. H.

(2016). Monitoring and predicting the fecal indicator bacteria concentrations from

agricultural, mixed land use and urban stormwater runoff. Science of the Total

Environment, 550, 1171-1181.

Region of Peel (2016). Public Works: Water and wastewater. Retrieved on December 13,

2017 from https://www.peelregion.ca/pw/water/sewage-trtmt/downsput-disconnection-

rebate.htm

Rodriguez, F., Andrieu, H., and Creutin, J.D. (2003). Surface runoff in urban catchments:

morphological identification of unit hydrographs from urban databanks. Journal of

Hydrology, 283, 146-268.

This article is protected by copyright. All rights reserved.

Ruchter, N., and Sures, B. (2015). Distribution of platinum and other traffic related metals in

sediments and clams (Corbicula sp.). Water Research, 70, 313-324.

Salvadore, E., Bronders, J., and Batelaan, O. (2015). Hydrological modelling of urbanized

catchments: A review and future directions. Journal of Hydrology, 529, 62-81.

Sample, D., Heaney, J., Wright, L., and Koustas, R. (2001). Geographic information systems,

decision support systems, and urban storm-water management. Journal of Water

Resources Planning and Management, 127, 155-161.

Schmitt, T. G., Thomas, M., and Ettrich, N. (2004). Analysis and modeling of flooding in

urban drainage systems. Journal of Hydrology, 299(3), 300-311.

Schneider, B. (2001). Phenomenon‐based Specification of the Digital Representation of

Terrain Surfaces. Transactions in GIS, 5(1), 39-52.

Schueler, T. R., Fraley-McNeal, L., and Cappiella, K. (2009). Is impervious cover still

important? Review of recent research. Journal of Hydrologic Engineering, 14(4), 309-315.

Snell, E. A. (1988). Recent wetland loss trends in southern Ontario. In M.J. Bardecki and N.

Patterson. (eds.), Wetlands: inertia or momentum? (pp. 183-197). Toronto, Ontario.

Sood, S., Sood, V., Bansal, R., and John, S. (2012). Further insights into the role of land use

in urban stormwater pollution. Management of Environmental Quality: An International

Journal, 24(1), 64-81.

Su, Z., Xiong, D., Dong, Y., Li, J., Yang, D., Zhang, J., & He, G. (2014). Simulated

headward erosion of bank gullies in the dry-hot valley region of southwest China.

Geomorphology, 204, 532-541.

Sun, N., Yearsley, J., Voisin, N., and Lettenmaier, D. P. (2015). A spatially distributed model

for the assessment of land use impacts on stream temperature in small urban

watersheds. Hydrological Processes, 29(10), 2331-2345.

This article is protected by copyright. All rights reserved.

Toronto and Region Conservation Authority. (1998). State of the watershed report: Etobicoke

and Mimico Creek watersheds. Retrieved June 3, 2017 from

http://www.trca.on.ca/dotAsset/25986.pdf

Toronto and Region Conservation Authority. (2010). Etobicoke-Mimico watersheds technical

update report. Retrieved June 3, 2017 from

http://www.trca.on.ca/protect/watersheds/etobicoke-mimico-creek/emwtu2011/02-

StudyArea-PhysicalSetting.pdf

Toronto Regional Conservation Authority. (2014). Toronto and Region Source Protection

Area. Toronto: Water Budget and Stress Assessment. Retrieved June 3, 2017 from

https://www.ctcswp.ca/wp-

content/uploads/2014/08/TRSPA_Tier1_2WB_September_V3_0.pdf

Wechsler, S. P. (2007). Uncertainties associated with digital elevation models for hydrologic

applications: a review. Hydrology and Earth System Sciences, 11(4), 1481-1500.

Wilson, J. P., Aggett, G., and Yongxin, D. E. N. G. (2008). Water in the landscape: a review

of contemporary flow routing algorithms. In Advances in Digital Terrain Analysis (pp.

213-236). Springer Berlin Heidelberg.

Yang, P., Ames, D. P., Fonseca, A., Anderson, D., Shrestha, R., Glenn, N. F., and Cao, Y.

(2014). What is the effect of LiDAR-derived DEM resolution on large-scale watershed

model results? Environmental Modelling and Software, 58, 48-57.

This article is protected by copyright. All rights reserved.

Figure 1. Mimico Creek watershed and 3 subcatchments delineated using digital elevation
model. Basemap of the area was obtained from ESRI software.

This article is protected by copyright. All rights reserved.

Figure 2. Directly connected watershed and subcatchment areas delineated from stormwater
infrastructure and imperviousness compared to delineation by topography. Areas
contributing flow to the retention ponds are located in the upper and mid subcatchments.

This article is protected by copyright. All rights reserved.

Figure 3. Left image shows DEM –derived watershed boundaries and stormwater network
pipes for the lower portion of the Mimico Creek watershed. Right image demonstrates land
area addition from areas that are drained by stormwater pipes contributing flow to the
watershed.

This article is protected by copyright. All rights reserved.

Figure 4. Relationships between derived watershed area and hydrologic metrics using the
directly-connected (a, b, c) and DEM (d, e, f) approaches. LR and RBI are the Richards
Pathlength and Richards-Baker Index, respectively, which are hydrograph flashiness indices.

This article is protected by copyright. All rights reserved.

Table 1. Sources of geospatial data.
Source Data Obtained from
City of Toronto Storm sewer network City of Toronto
Parking lots City of Toronto (open source)
Sidewalks City of Toronto (open source)
City of Brampton Storm sewer network City of Brampton
Parking lots City of Brampton
Sidewalks University of Toronto Map and Data Library
City of Mississauga Storm sewer network University of Toronto Map and Data Library
Parking lots City of Mississauga
Sidewalks University of Toronto Map and Data Library
Parking lots University of Toronto Map and Data Library
Region of Peel Storm sewer network Region of Peel
Parking lots Region of Peel
Sidewalks Region of Peel

This article is protected by copyright. All rights reserved.

Table 2. DEM-derived compared to directly connected subcatchment areas and perimeters.
DEM-derived Directly connected
subcatchment subcatchment
Subcatchment Area Perimeter (km) Area Perimeter
(km2) (km2) (km)
Upper 14.1 20.5 13.4 23.2
Mid 25.3 37.5 23.3 37.3
Lower 32.2 41.6 36.9 51.5
Watershed 71.6 79.6 73.7 86.0

This article is protected by copyright. All rights reserved.