Post on 28-Jan-2022
Accessibility and Cost Surfaces in the Boundary Waters Canoe Area, Minnesota
Caroline Rose, Chandler Sterling, Chase Christopherson, Matthew Smith
Geography 578
May 13, 2011
1
Table of Contents
Capstone …………………………………………………… 2
Introduction ………………………………………….………… 2
Conceptualization …………………………………………… 3
Implementation ………………………………………………… 5
Results …..………………………………………………… 8
Discussion ……………………………………………..……… 9
Conclusion ………………………………………………...…… 12
References ……………………………………………...……… 13
Figures ……………………………………………………… 13
Appendix A: Metadata …………………………………………… 18
Appendix B: Programming ………………….…………………… 22
2
Capstone Statement
Analyze the accessibility for canoeing in the Boundary Waters Canoe Area Wilderness.
This assessment could aid canoe trip leaders in planning for travel and emergency evacuation.
Introduction
The Boundary Waters Canoe Area
(Figure 1)
The Boundary Waters Canoe Area Wilderness (BWCAW) is located in northeastern
Minnesota on the Canadian border (Figure 1). The BWCAW is nearly 1.3 million acres in size
with over 1,000 lakes and streams, over 1,500 miles of navigable canoe routes, and over 2,000
designated campsites for canoeists. To traverse the BWCAW one must navigate its elaborate
network of lakes, streams, and portages. Dan Pauly, an expert on the BWCAW, defines
portaging “as an overland trail connecting two bodies of water” (56). To get from one lake to
another one must cross a portage, carrying all gear and canoes across the length of the trail. All
portages are not equal; they vary in length, width, slope, maintenance regime, and trail and
3
landing condition. The accessibility of areas within the BWCAW depends on the distance from
the entry point and difficulty of each traversable portage.
Why Accessibility within the BWCAW?
Every year thousands of individuals set off to explore remote areas within our country‟s
National Parks and Wilderness systems. The BWCAW sees over 250,000 people annually
navigate through more than one million acres of remote wilderness, making it the most used area
within the National Wilderness Preservation System (NWPS). The high volume of visitors
contains people with various degrees of wilderness experience, and it is important that people
choose routes and trips that appropriately reflect their ability. Through this project we visualized
relative accessibility by canoe throughout the BWCAW from entry points to remote locations.
Conceptualization
The goal of this project was to visualize accessibility from entry points throughout the
BWCAW. The key concept employed in our project is accessibility from entry points. The
project utilized 59 entry points, where non-motorized craft can officially enter the BWCAW. The
extent of the project was the network of streams, lakes, and portages accessed by canoe and on
foot. A portage is an over-land trail that connects two bodies of water; in the BWCAW, portages
connect to both lakes and streams. We used a cost distance analysis to derive accessibility. A
cost distance analysis “uses the cost or impedance to traverse each cell as a distance unit”
(Chang, 234). We defined high accessibility as a low cost distance, where cost distance is a
total cost accumulated, beginning at a source point, through travel over bodies of water (lakes
and streams) and portages.
4
Variables
The variables used to analyze accessibility were cost and distance. Cost values were
assigned to each cell in a grid, representing the difficulty of traversing that cell. A cost distance
surface was created by examination of the network of most efficient possible routes within the
BWCAW beginning at entry points, traversing lakes and portages, and avoiding barriers to
travel. Utilizing the variables cost [difficulty] and distance, we were able to visualize
accessibility from entry points to all locations within the BWCAW.
Operationalized Variables
A cost distance analysis was used to determine accessibility across the Boundary Waters
Canoe Area Wilderness (BWCAW) based on the least cost path from an entry point. The cost
distance analysis took into account a cost surface, consisting of the cost of paddled surfaces,
which are crossed by canoe, as well as portaged surfaces, which must be crossed on foot. All grid
cells (lakes, streams, and portages) in the analysis received a base value of 1. Open water, which
contained both streams and lakes, was considered the least difficult to traverse and was given no
additional cost. Portaging was more costly than paddling as it included the unloading and
loading of gear, transport of all gear by foot, and even multiple trips over the portage trail. For
this reason, all portaging areas were assigned a greater difficulty than paddling areas. To reflect
the greater difficulty of portaging, many portages were assigned a difficulty rating based on
expert opinion; this system was extended to all portages based on length. Portage ratings had a
range of 10-100. This rating was divided by portage length in grid cells to assign each grid cell
within a portage an equal share of the overall difficulty for that portage. Certain areas, including
5
shallow streams, rapids and waterfalls, posed significant barriers to travel and were assigned
„NoData‟ values, excluding them from the network of travel surfaces. For the purposes of this
analysis, land without portages could not be crossed and all water within the lakes and streams
layers was considered paddleable. All land that was not considered a portage was assigned a No-
Data value. These values were combined into a cost surface layer to guide the GIS in identifying
the cost distance from the entry points to any part of the network. For a diagram of the
conceptualization, see figure 2.
Implementation
Data Preparation
The general implementation of our analysis followed figure 3. The layers used included
the BWCA boundary, lakes, entry points, portages and barriers. The final output was a
combination of all layers into a cost surface raster. The data layers had to be properly prepared
before they were ready for the cost surface raster operation.
The procedure to prepare the lakes, streams, and portages layers was repetitive. We first
set the spatial reference then used the BWCAW boundary layer to clip the layers to the study
area. The three layers were first clipped within the study area. Some entry points, however,
existed on lakes outside of the BWCAW. The layers were then selected by location, selecting
for all lakes, streams, and portages that intersect with the BWCAW boundary, to extend the
study area to include the entirety of lakes partially inside and partially outside of the BWCAW
boundary.
After selecting for all attributes within the boundary itself, we then applied topology rules
to the streams layer because many of the stream polylines ran through the lake polygons. We
6
erased any section of a stream that intersected a lake. Next, we rasterized the stream and lake
layers to a grid cell size of 10 meters. After rasterization, we reclassified each grid cell in both
the stream and lake layers to a value of 1. Reclassification was the final step in data preparation
for these two layers before adding them to the cost surface.
After we prepared the lakes and streams layers, we began preparation of the portage
layer. We obtained the portage layer in .kml format and converted it into a shapefile using
OGR2OGR (see Appendix B). Once in shapefile format, we investigated the file and noticed that
the polylines of the portages did not precisely match up to the lake polygons. To remedy this
issue, we used spatial adjustment (edge snapping) to match the portage polylines to the lake
polygons. After snapping, errors still occurred in the portage layer because some of the polylines
were within the lake polygons. They were identified using topology rules and manually adjusted
to remove portage polylines that were within the lake polygons.
Once we fixed the portage polylines and each portage was in its correct location, we had
to assign a cost rating to each portage for addition to the cost surface raster layer. We created a
list of each portage‟s difficulty, the lakes that the portage went to and from, and the length of the
portage. We determined the difficulty of each portage based on the book Exploring the Boundary
Waters by Daniel Pauly. He lists many portages within the BWCAW and gives each portage a
difficulty rating from one to ten based on characteristics such as portage length, slope,
muddiness, rockiness, and landing quality. We built the table of portage difficulties using these
ratings and then joined the table to the existing portage attribute table. Because not all portages
within our shapefile were included in the constructed portage table, difficulties of the unmatched
portages were generated based on length class. We determined length class by grouping all
portage lengths into 10 classes using the natural breaks classification scheme. We generated the
7
average difficulty of known portages in each class and then applied that value to the unmatched
portages.
Once each portage had its own difficulty rating, we created two rasters of the portage
layer. One raster layer was rasterized by difficulty rating and the other was rasterized by shape
length. We then performed raster calculation to derive the final value for each grid cell within the
portage layer. The raster based on difficulty rating was divided by the raster based on length to
give each grid cell its appropriate cost rating, distributing the difficulty of each portage among all
of its grid cells. With this final step, the portage layer was ready to be added to the cost surface
raster.
Within the BWCAW there are waterfalls, rapids, and other geographic features which are
unsafe or impassible for canoeing. We did not have a shapefile of the locations of the barriers, so
we investigated the portage layer for suspected barrier locations. For each portage that appeared
to avoid a stream, a square polygon was placed there as a barrier. Barriers were rasterized and
assigned a NoData value, effectively deleting a section of water, to force the cost raster to seek a
different route (i.e., to take the nearest portage instead of paddling). The barriers layer was
created to model the travel surface more realistically under the assumption that portages are
created to bypass some hazard or barrier to paddling.
The entry points layer was also obtained in .kml format. We used OGR2OGR to convert
it to a shapefile (see Appendix B). We discovered that some of the locations of the entry points
were not touching the lakes or streams, which resulted in an erroneous cost surface raster
excluding those points. By moving the entry points to be completely surrounded by water, then
rasterizing the entry points, we successfully generated an acceptable layer of source points to be
used as points of origin in the cost distance analysis.
8
Cost Raster
Paddling surfaces, consisting of the lakes and streams, were considered the easiest to
travel. Each paddling surface grid cell was given a base travel cost of 1. Portages were modeled
as more difficult than paddling and received the base cost of 1 plus an additional cost based on
rating. The portage ratings from Exploring the Boundary Waters, ranging from 1 to 10, were
multiplied by ten, producing ratings of 10-100. The new rating for each portage was divided by
the total number of grid cells in that portage to find a cost value for each grid cell in each portage
(See Figure 4). The length of the portage in grid cells was found by dividing the length of the
portage by 10 meters, the grid cell size. The barriers layer was created over streams with known
or suspected impedance to travel such as shallow water, rapids, or waterfalls. We wanted to
include only travelable areas. Barriers were given a rating of No-Data to make them impassable
and force the analysis to take portages and avoid non-navigable streams. Land cover that was not
distinguished as a portage was given a value of No-Data. Portaging over land that covered by
vegetation is very difficult; we considered this impassable and out of the scope of our project.
All layers were added together using raster calculation to make a cost surface at a spatial
resolution of 10 meters. Cost distance analysis was used to create the final output map. This
operation calculated the least accumulative cost distance from each cell to the nearest entry
point.
Results
Our output is a static map of cost distance throughout the study area (see fig. 5). The low
value is green in color. This indicates relatively easy access. The high value is red in color and
indicates relatively difficult access. The numeric values of low and high are not important; the
project can only meaningfully show relative cost distance and accessibility. Cost distance values
9
were produced by accumulation of travel cost over each grid cell on the cost surface. The inset
depicts a good example of accessibility within the network of lakes, stream, and portages. The
area closest to the entry point in green as it is close and there are not many portages to cross. The
output becomes green to yellow to orange moving further from the entry point. There are,
however, streams that are close to the entry point and yet have a red color. This is because there
are no portages connecting to these streams. The intervening land has a No Data value. To get to
these areas one must travel through the network of streams, lakes and portages, accumulating
travel cost over more grid cells and making it relatively more difficult to get to.
The same analysis was run after increasing the portage ratings per grid cell by an order of
magnitude from the original design (see fig. 6). The output produced a similar map to our initial
map, highlighting the same general areas of relatively harder accessibility. When portages were
reassigned to just the base value of 1, the output produced a near identical map as our original
design (see fig. 7).
Discussion
Assumptions
We must acknowledge that our model is a drastic simplification of the world. Some
assumptions were inherent in our model of the study area as a raster grid: the grid takes a
complex and detailed physical world and simplifies it to 10 meter blocks of homogeneous area
belonging to one of a finite number of classes. The canoeist‟s true travel surface of marshes and
shallows and of rocky, muddy, slick or steep ground has been reduced to only squares of „water‟
and „portage‟ in a vast landscape of NoData values. The model assumes that all „water‟ cells are
equally easy to traverse, that each cell in a particular portage is equally difficult, and that each
portage‟s difficulty can be reduced to a static, general rating. While many portages were given a
10
rating from reference material, ratings for many unlisted portages were assumed to be the
average for that portage length. This led to another major assumption: that the length of each
digital „portage‟ line in our model was representative of true portage length on the ground. We
recognized that these lengths were very approximate, but we had no better alternative; no
existing map accounts for every twist and turn of a portage trail.
„Water‟ cells were assumed to be passable except where a portage provides an
alternative. Many of these areas may be too shallow to travel either seasonally or in general. The
barriers layer was based on the assumption that all portages exist for a good reason. A barrier
was positioned to block any paddling route that would bypass a portage. While some barriers to
paddling -- rapids, waterfalls and shallow, rocky streams-- are known through a group member‟s
field experience, most were simply suspected to exist.
Variable conditions were impossible to capture in our assessment. The wilderness
changes hour to hour, month to month, and year to year. Forest fires come and go, blocking
access while they burn and leaving a changed landscape behind. Wet and dry years mean
variable water levels. Inclusion of these conditions, however, would make the analysis too case-
specific. We had to assume fair weather conditions, having no way to account for three foot
waves and headwinds blasting across the vast expanse of Saganaga. High winds
disproportionately affect large bodies of water, where waves build up, while small or narrow
areas may be sheltered; this might affect the geographic distribution of cost distance. Although
difficult to measure and constantly changing, variable conditions have a great impact on
wilderness travel; their exclusion is a limitation of our project.
Our approach is limited; the travel surface is simplified, portage lengths (which affect
weighting scheme) are very approximate, we assume the presence of barriers, and the model
11
offers no way to account for variable conditions. We recognize these limitations, but feel that we
have done our best with the available resources.
Subjectivity
The greatest subjective decision in the development of our project was the cost of portage
travel relative to paddling. As our results showed, this relative weighting makes little difference
in the overall pattern of accessibility across the Boundary Waters. Other decisions were
reinforced by our reference material. This analysis relied heavily on Daniel Pauly‟s book
Exploring the Boundary Waters for subjective information regarding the study area. The
publication supported several important decisions, including:
- The exclusion of land without portages, which could, in theory, be accessed by
bushwhacking; “It is all but impossible to bushwhack significant distances
through the dense forests of the Boundary Waters, particularly with a canoe and
gear” (60).
- The travel cost of portage surfaces; Pauly gives many BWCA portages a difficulty
ranking from 1 to 10, based on personal experience (56).
- The assignment of average rating by length for unlisted portages; “Some portages do
not have rankings, typically because they are of average difficulty for their
length” (56).
While we recognize that these choices might be made differently and may affect the outcome of
the project, we are confident that we have made informed decisions.
12
Uncertainty
Every layer in the project included some level of uncertainty. Portage length is uncertain
as is the exact location of each portage‟s connection to a lake or stream. Portage ratings are
uncertain as difficulty may vary with season, and portaging cost relative to paddling depends on
the gear, efficiency, organization, and abilities of the specific group. The layer of water,
including lakes and streams, is accurate enough for our purposes in terms of location, but its
classification as „paddleable‟ is uncertain. The precise locations and often the existence of
barriers are uncertain. It would take an extensive survey to quantify these uncertainties. Although
they limit our analysis, we consider these ambiguities an innate characteristic of our study area.
It is the nature of wilderness to be wild and uncertain; part of the adventure of wilderness travel
is to face the unknown.
Conclusion
The goal of this project was to visualize accessibility from entry points to all areas
throughout BWCAW. The map output was able to demonstrate areas of harder accessibility
relative to areas of easier accessibility. Portage difficulty did not weigh into the analysis as
originally predicted; it was the distance from entry points that largely drove the cost distance
analysis. The output map will help in planning canoe trips for groups of all skills and abilities.
Beginners will want to stay in areas of easy accessibility that are closer to entry points. This will
hopefully make their first experiences in the Boundary Waters more enjoyable. It will also allow
for a faster exit in case of an emergency situation. The more experienced paddlers who seek a
challenge may want to push themselves to explore areas that are less accessible and relatively
more difficult to get to.
13
References
Chang, Kang-tsung. Introduction to Geographic Information Systems. New York, NY:
McGraw-Hill Higher Education. 2004.
Pauly, Daniel. Exploring the Boundary Waters. Minneapolis, MN: University of Minnesota
Press. 2005.
United States Forest Service (USFS). “The BWCAW Act”. Special Places, Superior National
Forest
http://www.fs.usda.gov/wps/portal/fsinternet/!ut/p/c5/04_SB8K8xLLM9MSSzPy8xBz9C
P0os3gjAwhwtDDw9_AI8zPyhQoYAOUjMeXDfODy-HWHg-zDrx8kb4ADOBro-
3nk56bqF-
RGGGSZOCoCAPi8eX8!/dl3/d3/L2dJQSEvUUt3QS9ZQnZ3LzZfMjAwMDAwMDBB
ODBPSEhWTjJNMDAwMDAwMDA!/?navtype=BROWSEBYSUBJECT&cid=stelprd
b5203434&navid=100000000000000&pnavid=null&ss=110909&position=Not%20Yet%
20Determined.Html&ttype=detail&pname=Superior%20National%20Forest-
%20Special%20Places (accessed 4 March 2011).
Figures
Figure 1 (embedded in text): http://www.bwcaoutfitter.com/images/boundary%20waters.gif
14
Figure 2:
15
Figure 3:
16
Figure 4:
17
Figure 5:
Figure 6:
18
Figure 7:
Appendix A: Metadata
Identification Information:
Citation:
Citation information: Originators: University of Wisconsin Madison, Department of Geography, Geog 578
*Title: barriers_new *File or table name: barriers_new
Publication date: May 13, 2011 *Geospatial data presentation form: vector digital data
Description: Abstract: Known and suspected impedance on water travel. Barriers were placed over streams where there are impassable obstacles such as rapids, shallow water, waterfalls, etc.
On streams that ran parallel to portages, but did not have known barriers, it was assumed the portage was there for a reason and a barrier was created over the stream to force the cost distance analysis over the portage layer. Purpose: Added to cost surface layer. Excludes cost distance analysis over impassable streams.
Calendar date: unknown Currentness reference: publication date
Status: Progress: Complete
19
Maintenance and update frequency: None planned Spatial domain: Bounding coordinates:
*West bounding coordinate: -92.423368 *East bounding coordinate: -90.035127 *North bounding coordinate: 48.335151 *South bounding coordinate: 47.749112 Local bounding coordinates: *Left bounding coordinate: 543192.389600
*Right bounding coordinate: 719876.617300 *Top bounding coordinate: 5353716.012800 *Bottom bounding coordinate: 5292583.935200 Theme keywords: Barrier Theme keyword thesaurus: Impediment
Access constraints: None Use constraints: None Contact organization: University of Wisconsin Madison, Department of Geography, Geog 578 *Native dataset format: File Geodatabase Feature Class *Native data set environment: Microsoft Windows XP Version 5.1 (Build 2600) Service Pack 3; ESRI ArcCatalog 9.3.1.3000
Data Quality Information:
Process description: Created polygon layer over stream layer. Process date: 20110427 Process time: 20234600
: Process description: Dataset copied. Process date: 20110427 Process time: 20261200
Source used citation abbreviation: E:\Geog578\barriers\barriers_2
Process description: Dataset copied.
Process date: 20110428
20
Process time: 12433100
Spatial Reference Information:
Horizontal coordinate system definition: Coordinate system name: *Projected coordinate system name: NAD_1983_UTM_Zone_15N *Geographic coordinate system name: GCS_North_American_1983
Entity and Attribute Information:
Detailed description: *Name: barriers_new
Entity type:
*Entity type label: barriers_new *Entity type type: Feature Class *Entity type count: 857 Attribute: *Attribute label: OBJECTID_1 *Attribute alias: OBJECTID_1
*Attribute definition: Internal feature number. *Attribute definition source: ESRI *Attribute type: OID *Attribute width: 4
*Attribute precision: 0 *Attribute scale: 0
Attribute domain values: *Unrepresentable domain: Sequential unique whole numbers that are automatically generated.
Attribute: *Attribute label: OBJECTID *Attribute alias: OBJECTID *Attribute definition: Internal feature number. *Attribute definition source:
ESRI *Attribute type: Integer *Attribute width: 4 *Attribute precision: 0 *Attribute scale: 0
Attribute domain values: *Unrepresentable domain:
21
Sequential unique whole numbers that are automatically generated.
Attribute: *Attribute label: Shape
*Attribute alias: Shape *Attribute definition: Feature geometry. *Attribute definition source: ESRI *Attribute type: Geometry *Attribute width: 0 *Attribute precision: 0 *Attribute scale: 0
Attribute domain values: *Unrepresentable domain:
Coordinates defining the features.
Attribute: *Attribute label: Id *Attribute alias: Id
*Attribute type: Integer
*Attribute width: 4 *Attribute precision: 0 *Attribute scale: 0
Attribute: *Attribute label: Shape_Leng *Attribute alias: Shape_Leng
*Attribute type: Double *Attribute width: 8 *Attribute precision: 0 *Attribute scale: 0
Attribute:
*Attribute label: Shape_Length *Attribute alias: Shape_Length *Attribute definition: Length of feature in internal units. *Attribute definition source: ESRI *Attribute type: Double *Attribute width: 8 *Attribute precision: 0 *Attribute scale: 0
Attribute domain values: *Unrepresentable domain: Positive real numbers that are automatically generated.
22
Attribute: *Attribute label: Shape_Area *Attribute alias: Shape_Area *Attribute definition:
Area of feature in internal units squared. *Attribute definition source: ESRI *Attribute type: Double *Attribute width: 8 *Attribute precision: 0
*Attribute scale: 0
Attribute domain values: *Unrepresentable domain: Positive real numbers that are automatically generated.
Metadata Reference Information:
*Metadata date: 20110510
*Language of metadata: en
Metadata contact: Contact information: Contact organization primary: Contact person: REQUIRED: The person responsible for the metadata information. Contact organization: University of Wisconsin Madison, Department of Geography, Geog 578
Appendix B: Programming
Two layers we used to produce our final output, portages and entry points, were originally in
.kml format. To convert them to shapefile format, we issued an OGR2OGR command in the
FWTools shell. FWTools is an open source GIS/RS binary kit for Microsoft Windows and Linux
(more information on FWTools can be found here: http://fwtools.maptools.org/) The command
issued was
ogr2ogr -f “output format” OutputDataSource InputDataSource
where “output format” was ESRI Shapefile, the OutputDataSource was the name of the new file,
and InputDataSource was the .kml file.