1 | P a g e

Bark River Watershed Delineation & DEM Accuracy Assessment

Partnership between Save our Lake Foundation and Kennedy Consulting

Created By: Ryan Masek

Introduction: Within the confines of a naturally occurring landscape, hills and valleys act as natural barriers

that separate rivers, creeks, and lakes. This separation of water eventually converges at one point and

can be classified as the end of a watershed, also known as the outlet. To accurately delineate watershed

boundaries a Digital Elevation Model (DEM) must be used. DEMs can have a variety of resolutions

which will affect the accuracy of the watershed that is being delineated. The spatial resolution of the

DEMs refers to the pixel size. Accurate representation of a watershed’s boundaries is crucial to reduce

criticism and also have the most accurate representation, given the available resources. Watersheds are

useful for many reasons, such as the ability to show the specific percentages of land use classifications

in the area. The watershed that is of particular importance for this report was the Bark River Watershed

of Waukesha County.

The Bark River Watershed is located in the northwestern portion of Waukesha County and the

southern portion of Washington County. The primary water body within the watershed is Nagawicka

Lake. Nagawicka Lake is surrounded by a variety of land uses such as agricultural, urban, and forested

lands. Urban and agricultural areas can produce runoff which can contain many nutrients such as

nitrogen and phosphorus which can be detrimental to lake health, chemically and physically. Excess

nutrients entering a water body can cause a lake to turn green. Not only is this green color unappealing

to the residents and visitors of Nagawicka Lake it can also be an indicator of poor of lake health. Under

the right conditions, excess nutrients can cause algal blooms such as blue green algae which is toxic to

aquatic life and if ingested, toxic to humans and animals as well. Delineating the watershed

surrounding Nagawicka Lake allows for the land uses to be determined and know the major source of

pollution.

In this study, the Save our Lakes Foundation hired Kennedy Consulting® to delineate the Bark

River Watershed and compare the accuracy of the delineations with three resolutions of DEMs (5-

Foot, 10 Meter, and 30 Meter). The accuracy of the delineations will be cross referenced with a

credible source (United States Geological Survey) based upon the area, perimeter, surface volume of

the watershed. Through this delineation process, I hope to answer three research questions to assess

the accuracy and consistency of DEMs with different spatial resolutions.

1. What is the area of each watershed delineation?

2. What is the perimeter of each watershed delineation?

3. What is the surface volume of each watershed delineation?

Methods The geodatabase was organized approximately as follows in table 1. The primary data layers

were the three DEMs, (30 Meter, 10 Meter, and 5-foot). A land use and land cover raster data layer

was used to show the percentages of land use types in the watershed. The vector layers were organized

into a feature dataset. The only features in the dataset were a study area layer showing Washington and

Waukesha counties and a hydrology layer, which can either be a polygon or a polyline feature class.

The hydrology allowed for the lake to be located more easily and allowed for the outlet of the

watershed to be located more easily.

2 | P a g e

Table 1: Physical model for watershed delineation of the Bark River Watershed which surrounds Nagawicka Lake.

Results:

When comparing the data to the United States Geological Survey the 30 Meter DEM had the

least amount of difference. The 30 Meter delineation had an increase in 681 acres over the actual size

which is about 2.36% difference, (Table 2). The 10 Meter DEM was similar to the 30 Meter in almost

all values, but they had a higher percentage of difference than the 30 Meter. The 10 Meter DEM

delineation was 991 acres larger than the USGS delineation, which is 3.42% difference, (Table 2). By

comparing the 30 Meter DEM delineation to the 10 Meter delineation, a difference of 309 acres is

determined. This is a 1.06% difference in area.

The 5-foot DEM was not used in comparing any of the parameters since the source data did not

encompass the entire watershed. Because of this, there was a noticeable difference of 13,029 acres

between the 5-foot and the USGS data. This is a 61.43% difference, (Table 2). There is also a large

difference between the 5-foot and the 30 Meter and/or the 10 Meter. This data is not useful because it

shows only about half of the watershed. If the results of this comparison were actually considered,

without taking into account the lack of data for the 5-foot, there would appear to be a very large

difference in the quality of data from the 5-foot and the 30 Meter and 10 Meter. Because of this, the 5-

foot values are essentially excluded from this report. The values are listed in Table 2 for reference, but

they are not influential or at least should not be considered to be until data for Washington County is

located.

An authoritative data source was not located for both the perimeter and surface volume.

Because of this, the results could not be accurately assessed to a known value. All data and values are

listed in Table 2, but much of the collected data is not fully usable apart from comparing between the

30 Meter and the 10 Meter delineations created during this project. The 5-foot DEM is much different

than the 30 Meter and the 10 Meter DEM because of its lack of data from Washington County. The

surface volume is actually noticeable because it is actually larger than the 30 Meter and the 10 Meter

while encompassing a smaller spatial extent. This may be due to the fineness of detail of the spatial

resolution. Since more specific entities are noticed by the 5-foot DEM, more surfaces may be able to

be detected thus increasing the surface volume.

Comparing the spatial extent of the three DEMs’ watersheds is slightly difficult because the

differences between them are often quite minute. The 30 Meter and the 10 Meter look very similar and

their areas are actually quite similar to support this statement. There are some slight variations at

certain points, but in general the 30 Meter seems to be slightly rounded-off likely due to the lower

resolution imagery. The 30 Meter lumps more into each cell, likely creating a more generalized and

less specific image. This pattern is true for the 5ft image as well. There are more very slight and minor

alterations along the periphery of the watershed. There are some areas that are not included in the 5-

foot DEM. For example, the very southwestern most portion of the 30 Meter and 10 Meter DEMs are

not present in the 5-foot. The three DEM’s are similar, but the 5-foot is the most different of the three.

Geodatabase Feature Dataset Feature Class / Raster

Class

Type Field Name Field Type

Sample

Value

Domain

Values Topology

FID Short Integer 2131

Shape length Double 6,000 Feet

Shape Area Double 4,000 Acres

30 Meter DEM Raster

10 Meter DEM Raster

5 Foot DEM Raster

Land Cover Raster

Watershed_Delini

ation_Features

Watershed_D

eliniation.gdb

Raster_Features

Length DoubleHydrology 12.02 ft Hydro_Clip

Lakes, rivers

and streams

must be within

the study area

Polyline

Study Area (Waukesha

County and Washington

County)

Polygon

3 | P a g e

One of the main

limitations of this project

was the limitation of data

for the 5-foot DEM .

There was no readily

available 5-foot DEM for

Washington County. This

was problematic because

the watershed stretched

from Waukesha County

into , the 5-foot Another

limitation is that the 5-

foot DEM is almost too

accurate and cut off the

watershed due to a few

road crossing on the north

east and south western corners of the watershed (Map 1). This misrepresents the watershed due to the

fact that approximately 47% of the watershed is missing from part of Washington and Waukesha

County.

Conclusions:

According to the United State Geological Survey, the Bark

River watershed is 28,493 acres (Table 2). In this watershed

delineation, 30 Meter DEM came back with a watershed size of

29,174 acres (Table 2). This difference in watershed size is 2.36%

which means the 30 Meter DEM is the most accurate representation

of the Bark River Watershed (Table 2). Although, the 10 Meter DEM

was not far off; it had an area of 29,483 acres (Table 2). This was a

percent difference of 3.42% (Table 2). The area of the 5-foot DEM

came out to 15,464 acres (Table 2). The 5-foot DEM is almost too

accurate and cut off the watershed due to a few road crossing on the

north east and south western corners of the watershed (Map 1). Due

to a few road crossings and the lack of data for Washington County,

almost half of the watershed is not being represented. If an individual

was to delineate the sub-watersheds within the bark river watershed,

a 5-foot DEM would not be a bad idea. This smaller pixel size could

find some of the smaller watersheds that are cut off by roads. But if

an individual was going to just delineate a watershed, a 30 Meter or a

10 Meter would produce the best results.

Table 3 shows the land use percentages for the Bark River Watershed. Agriculture is the

dominant land use of the watershed (38.71%) which could be a cause for concern for the Save the

Lakes Foundation due to possible runoff which could lead to lake pollution. Washington County and

Waukesha County may want to develop some best management practices for the future health and

stability of Nagawicka Lake. This high percentage of agriculture could cause an increase in pollution

due to fertilizer runoff which could possibly decrease the health of Nagawicka Lake.

Area

(Acres)

Perimeter

(Meters)

Surface Volume

(Meters Cubed)

30 Meter 29,174.18 103,799.54 3,590,356,950

10 Meter 29,483.63 106,329.82 3,747,404,009

5 Foot 15,464.00 90,357.71 4,277,621,216.95

Actual Size (USGS) 28,493.00 Data not available Data not available

Difference (10m - 30m) -309.45 -2,530.28 -157,047,059

Percent Different (10m - 30m) 1.06 2.41 4.28

Difference (5ft-10m) -14,019.63 -15,972.12 530,217,207.95

Percent Different (5ft-10m) -62.38 -16.24 13.21

Difference (5ft-30m) -13,710.18 -13,441.84 687,264,266.95

Percent Different (5ft-30m) -61.43 -13.85 17.47

Difference (30m-actual) 681.18

Percent difference actual 30m 2.36

Difference (10m-actual) 990.63

Percent difference actual 10m 3.42

Difference (5ft-actual) -13,029.00

Percent difference actual 5ft -59.28

Land Use

Count (#

of

pixels)

Percent

Land Use

Agriculture 34,954 38.71

Barren 3,796 4.20

Forests 12,805 14.18

Forested

Wetland 2,848 3.15

Grassland 17,298 19.16

Open Water 4,895 5.42

Shrubland 603 0.67

Urban 7,254 8.03

Wetland 5,845 6.47

Total 90,298 100

Table 3: This table summarizes the

percentages of land use displayed on

Map 1 for the Bark River Watershed

Table 2: This table summarizes the

Area, Perimeter, surface volume as

well as the percent difference of the

30 Meter, 10 Meter, and 5 Foot the

Bark River Watershed

4 | P a g e

Map 1: This map includes the 30 Meter, 10 Meter, and 5 Foot watershed delineations as well as the

landuse surrounding Nagawicka Lake. This watershed is known as the Bark River Watershed. See table 2

for landuse percentage values

5 | P a g e

Workflow Scripts:

The following methodology will primarily cover the steps taken to delineate the Nagawicka

Watershed and the steps needed to compile all data required to answer the research questions.

To delineate the Nagawicka Watershed, the DEMs were initially sized in the computer display

to the area surrounding the watershed but with some extra space on the northern size of the lake to

allow for the watershed to be delineated up into Washington County. Once the study area was sized

into the display, a fill function was used to fill all extra holes or low spots in the landscape that are

represented in the DEM, (figure 1). The fill was limited to the extent of the screen.

Figure 1: Fill Function

After the fill was completed, a flow direction process found under the hydrology tools in

ArcGIS had to be performed, (figure 2). The flow direction was limited to the extent of the screen and

created a new layer based on the slope of the landscape. It created a raster based on the flow direction

from one cell to its steepest downslope neighboring cell. This was useful because it accounted for

natural water drainage based on the elevation changes and allowed for the boundaries of the watershed

to be created later on.

Figure 2: Flow Direction Process

After the flow direction, a flow accumulation, (Hydrology Tools – ArcGIS), had to be

developed, (figure 3). The flow accumulation was limited to the extent of the screen and created a new

layer showing the areas of the study area where the most water would collect. It highlighted cells that

have the most other cells that would likely flow into it. It created a black and white image showing the

areas where water collected as white and the areas where water flowed from as black.

Figure 3: Flow Accumulation Process

6 | P a g e

Once the flow accumulation was created, the pour point had to be placed. The pour point of the

watershed indicated where the outlet of the watershed is located. This pour point helped to limit the

extent of the watershed to stop at a certain point. A new feature class had to be made to allow for a

point to be placed, (figure 4). The point was placed along a high accumulation pixel at the south-

western border of the Nagawicka Lake near a dam.

Figure 4: Pour Point Process

Next came the actual watershed creation (Hydrology Tools – ArcGIS). The watershed was

created using the pour point and the flow accumulation (Figure 5), which was based off the flow

direction. A watershed process was performed thus delineating the watershed around Nagawicka. This

process was performed for each resolution of DEM. At this point, the watershed had been created. The

next step was to perform some alterations to and with the watershed to answer the mentioned research

questions.

Figure 5: Watershed Process

To determine the area and perimeter of the watershed, the watershed had to be converted to a

polygon instead of a raster layer. This was accomplished using a “Raster to Polygon” function (figure

6). Once the new polygon was created, two new fields were added to the attribute table, “Area” and

“Perimeter”. These two fields were set as a double and their geometry was calculated to determine

their quantitative values.

Figure 6: Raster to Polygon Process

To determine the surface volume, an “Extract by Mask” had to be performed on the DEM-Fill

layer, (figure 7). The raster mask was limited to the extent of the watershed polygon. The raster mask

acted as a “Clip” and made a new layer of DEM that showed only the area that was included in the

watershed. Once this new, limited DEM was created, a “Surface Volume” process was performed on

the masked DEM, (figure 8). No boundary height was established because the volume for the entire

7 | P a g e

layer was desired. The surface volume was determined in a table in ArcMap. The values for area,

perimeter, and surface volume were recorded in Microsoft Excel.

Figure 7: Raster Mask Process

Figure 8: Surface Volume Process

References

Garn, H., Robertson, D., Rose, W., Goddard, G., & Horwatich, J. (2013.). Water Quality, Hydrology,

and Response to Changes in Phosphorus Loading of Nagawicka Lake, a Calcareous Lake in

Waukesha County, Wisconsin.

United States Geological Survey. 2015.