Using GIS to Determine Horizon Belts[2]

USING GIS TO DELINEATE HORIZON BELTS
By Kirk D. Sinclair, PhD
GIS Manager
Housatonic Valley Association
Introduction
The Planning and Zoning Commission of Kent, Connecticut approached the
Housatonic Valley Association (HVA) to assist them with scenic ridge zoning. The
purpose of the zoning was to preserve the scenic integrity of the town by restricting
development that could impair prominent ridges. The challenge was to establish an
empirical, defensible method for targeting the zones where development should be
restricted.
Earlier attempts to target these zones had some basis in empiricism—well-defined
data such as elevation or ridgelines were being considered—but none of the available
data correlate directly with the commission’s objective. Higher elevations are thought to
be more of a scenic resource than lower elevations, but in reality the surrounding
topography may obscure higher elevation and expose lower elevation areas. A similar
problem confronts the use of ridgeline buffers. While conceptually the ridgeline buffer is
appealing for preserving a scenic ridge resource—analogous to the use of stream buffers
to preserve a water resource—the surrounding topography complicates the determination
of what this buffer should be. Indeed, the ridgeline itself may not be a scenic resource in
places. While elevation and ridgelines are empirical data, they are not defensible as
indicators of a scenic resource zone.

One advantage of a Geographic Information System (GIS) is the ability to model
new information from existing spatial data. Could GIS be used to model zones important
to the town’s scenic integrity from existing elevation and ridgeline data? The intuitive
approach was to conduct a viewshed analysis from digital elevation models, determining
the area that can be viewed from a viewpoint or collection of viewpoints. Draping
ridgelines over the resulting viewsheds would resolve the issue of which stretches of
ridgelines actually serve as a scenic resource.
Ongoing discussions with Kent P&Z indicated that their conservative intent for
the methodology was to target only the areas of the landscape where development would
break the horizon; this means areas where sky, rather than land, forms the backdrop for
some portion of a manmade structure. A house whose backdrop was land might be
camouflaged and in other ways made less obtrusive than a house whose backdrop was
naked sky. This clarification of the objective revealed two shortcomings in the
methodology that had been adopted up to that point. First, vertical delineation of scenic
resource zones on ridges was more the challenge than horizontal delineation. Second, the
focus for this delineation needed to be the horizon line rather than the ridgeline.
The horizon line is the place on the landscape where ground level of the earth and
sky meet from the perspective of a particular viewpoint. Along this line, objects of any
height would break the horizon. Tangents to the horizon line form sightlines. With
increased distance beyond the horizon line objects would fall below the sightlines; unless
they exceeded a certain height they would be out of sight. With increased distance before
the horizon line objects also would fall below the sightlines; in this case they fall below
the height needed to have a backdrop of sky instead of land.

The challenge to be met by the new methodology was to determine the zones,
both beyond and before horizon lines, where developed structures would rise above the
sightlines. The maximum heights to be considered for these structures would be
determined by current zoning regulations. Since these zones had width derived from the
location of horizon lines, the areas delineated by this new methodology were called
horizon belts.
Since expansive horizon belts exist in areas of gentle slopes—everything lies
within a horizon belt in flat desert—two additional criteria needed to be met to preserve
the original intent of scenic ridge zoning. Horizon belts were intersected with ridgeline
buffers to select those horizon belts that corresponded with actual ridges. Since a
ridgeline can occur in the middle of a plateau, while the actual horizon line for that ridge
occurs at the edge where the ridge starts to drop, the horizon belts were intersected with
steep slope gradients as well.
Methodology
The source data used for this horizon belt methodology were roads, lakes,
streams, digital elevation models (DEMs), digital topographic maps, and subbasins. The
roads, lakes, stream, and subbasins data were obtained from the Connecticut Department
of Environmental Protection’s (DEP) Environmental GIS Data CD and based on DLG
files from USGS topographic maps. The digital topographic maps also were obtained
from the DEP’s Environmental GIS Data and based on the DRG files from USGS
topographic maps. The DEMs for this project were 10 meter resolution, 1/3 arc second,
obtained from USGS at their National Elevation Dataset website.

The main tasks of the methodology are listed and annotated as follows:
Create a triangulated irregular network (TIN) model and viewshed ridgeline data
DEMs are a type of raster data representing terrain as a series of discrete steps, 10
meters squared. TIN models can be derived from DEMs, and are a more realistic
representation of terrain with a surface of continuous elevation change. The ArcView 3D
Analyst Extension was used to generate a TIN for the town of Kent. Lakes are features
of constant elevation and were used as breaklines for refining the TIN model, assigning
those areas of the model a constant elevation. Viewshed ridgelines were derived from the
boundaries delineated in the subbasins data, after eliminating subbasin boundaries that do
not correspond with topographically distinct ridges. Some topographically distinct ridges
not corresponding to a subbasin boundary were manually added.
Create the road viewpoints file
The road viewpoints were obtained from public roads in the town of Kent.
Viewpoints were chosen that corresponded to intersections of these road segments with
other roads, the town boundary, subbasin boundaries (an indicator of high points along
the road), or known vistas (Figure 1). The viewpoints were converted to a 3D file, with
the elevation derived from the TIN model, and a target offset (OffsetB) value of 35 feet
was assigned as an attribute to each viewpoint.

Figure 1: Viewshed analysis for Carter Road, Kent, Connecticut.
A. Carter Road viewpoints were chose from intersections with other roads, basin
boundaries, or known vistas. B. Ridge viewpoints were chosen from along the edge of
the viewshed analysis, and corresponded with high points along the edge, or at viewshed
junctions. A digital elevation model (DEM) is used for the display.
Run a road viewshed analysis
The target offset value of 35 feet corresponds to current zoning regulations for the
town of Kent. The viewshed analysis used the TIN model to calculate the areas of the
landscape where a 35 foot structure would be visible to at least one of the selected
viewpoints. The output of a viewshed analysis is a raster grid, with each discrete cell of
the grid given the value of how many viewpoints are visible to that cell. The grid was
reclassified to indicate whether a cell was visible to any (value = 1) or none (value = 0) of
the viewpoints (Figure 1).

Create the ridge viewpoints file
The ridgeline data was used to indicate what portions of the road viewshed
corresponded to ridges. A ridge viewpoint file was created that corresponded to the back
edge of the viewshed traversing a ridge (Figure 1). The ridgeline was used only as a
horizontal guide for the location of the ridge. The back edge of the viewshed, and the
corresponding “ridge” viewpoints could have extended beyond or fell short of the actual
ridgeline. At least three ridge viewpoints were selected, two from near the ends of the
ridge and one from a highpoint along the ridge. Additional viewpoints were added for
divides, as indicated by ridgelines. They were also added at viewshed junctions where
the back edge of the viewshed changed direction sharply. These viewshed junctions
represent locations viewed by more than one road viewpoint. The ridge viewpoints were
converted to a 3D file, with elevations derived from the TIN model. Both target offsets
and observer offsets (OffsetA) of 35 feet were assigned as attributes to all viewpoints.
Create an analysis mask
Analysis masks were used to constrain the viewshed analysis from the ridge
viewpoints to a limited area (Figure 2). This served two practical purposes. Areas
“behind” the ridge viewpoints were not of concern and eliminating them from
consideration speeds up processing time. Analysis masks also were used to segment
concave ridges and create separate horizon belts for each ridge segment.

Figure 2: Horizon belt northwest of Route 341.
An analysis mask was used to segment a viewshed analysis from the northwest ridge
points. If the analysis had included the west ridge points, then the area labeled “A”
would have been included in the horizon belt, being visible from both the west ridge and
Route 341 viewpoints. However, structures in this area would not break the sightline
between the northwest ridge points and Route 341, and are not along the sightline from
the road to the west ridge points. A separate analysis was done for the west ridge
segment.
Run a viewshed analysis from the ridge viewpoints
This step is the crux of the horizon belt methodology. The ridge viewpoints are
located at the back edge of a belt where a 35 foot structure rises above the sightline from
the original road viewpoints (Figure 2). Giving these ridge viewpoints an observer offset
of 35 feet locates the viewpoint at the height of this original sightline, rather than ground
level. The ridge viewshed analysis used the TIN model and the 35 feet target offset to
calculate the areas of the landscape where a 35 foot structure rises above the original
sightline. Beyond this viewshed edge a 35 foot structure would be out of view from the

idge viewpoints, and would have a backdrop of only land from the perspective of the
road viewpoints. Thus, the purpose of the ridge viewshed analysis is to subtract the areas
from the road viewshed analysis that would have a backdrop of land rather than sky. The
output of the ridge viewshed analysis was reclassified to indicate whether a cell was
visible to any (value = 1) or none (value = 0) of the ridge viewpoints.
Create horizon belts
The horizon belt corresponds to the intersection of the road viewshed and ridge
viewshed analyses. This was obtained by multiplying the two grids. Any area common
to both grids receives a calculated value of one, while an area just common to one grid
receives a calculated value of zero. These belts delineate the area where at least some
portion of the backdrop to a 35 foot structure would be open sky, as viewed from the
public roads.
Implement horizon belt constraints
Horizon belts are a subset of a viewshed, the area within a viewshed where
structures have a backdrop of sky rather than land. Further constraints were added to this
subset for the town of Kent. A 1000’ buffer from the viewshed ridgelines was created
and given a grid value of 1. A slope gradient map was generated from the DEMs and
reclassified to indicate slope gradients >= 15% (grid value = 1), or slope gradients < 15%
(grid value = 0). A dataset of steep slopes OR ridgeline buffer was combined. A subset
of all horizon belts was derived from an intersection with this grid of constraints.

Results
All the public roads in the town of Kent were used for the analysis, a total of
twenty-one. From these roads a total of twenty-five ridges were determined as scenic and
targeted for horizon belt delineation. Most of these ridges were small, a few hundred
yards in length.
The town of Kent covers 31,800 acres. Due in large part to the presence of the
Appalachian Trail corridor, more than one third of the town, 11,919 acres, are protected
by conservation ownership or easement. There were 8416 acres of horizon belts
generated by the methodology. After the ridge and slope constraints were applied this
was pared down to 6304 acres. Out of this total, 2903 acres occurred on the protected
lands, leaving 3401 acres as the subset of horizon belts to be affected by new zoning
regulations (Figure 3).
Discussion
Sources of Potential Error
The original 10 meter resolution of the DEMs used for creating the TIN model
causes some error, but this source of error is unbiased, potentially leading to either
overestimation or underestimation of horizon belt boundaries. Due to the sensitive nature
of creating new areas to be zoned, sources of error creating a conservative bias
(underestimation) were considered a desired attribute of the methodology. A subset of
road viewpoints would not generate the entire viewshed that might be seen from the
entire road. While these points were selected to maximize the road viewshed, they were

still only a subset. Similarly, a subset of ridge viewpoints underestimates the ridge
viewshed created.
Figure 3: Horizon belts for the town of Kent, Connecticut.

Concave ridges pose a problem for the methodology that was circumvented by
using analysis masks to segment the ridges. A false horizon belt along a concave ridge
would result from the viewsheds created from the ridge viewpoints on the opposing arms
of the ridge. By using analysis masks that constrained viewsheds to the ridge segment
that fell directly below the ridge viewpoints, the false horizon belts that would be created
on opposite ridge segments were avoided. Straight or convex ridges curving away from
the road do not create this problem and were not segmented by masks.
For the town of Kent the horizon belts were created for ridges and their adjacent
roads. Horizon belts could be created for a ridge and viewpoints from additional roads,
or even other ridges, but multiple analyses would need to be run. The part of a ridge that
forms the horizon line differs between an adjacent road below the ridge and observation
points that occur further away or higher up, due to the change in viewing angle. If you do
an analysis with viewpoints from both an adjacent road and an opposing ridge lumped
together, the result will be biased towards the opposing ridge. The ridge viewpoints
would be pushed farther back by the viewshed from an opposing ridge, placing them out
of sight from the closer road viewpoints. The horizon belt for the opposing ridge thus
may omit areas that might have been horizon belts for the road, yet would include new
areas that could not be seen from the road.
Horizon Belt Criteria
Areas that are scenic are subjective by definition. Beauty is in the eye of the
beholder. This does not mean that scenic resources can neither be empirical nor

defensible, as long as the criteria for what is scenic are clearly defined and accepted. For
using horizon belts as scenic resource zones these criteria correspond to:
• Vistas,
• Ridges,
• Obtrusiveness,
• Constraints.
The town of Kent used a comprehensive dataset of all public roads as the criteria
for vistas. More restrictive datasets could be formed by the use of publicly determined
scenic roads or landmarks. Recreational groups or watershed organizations may have
different criteria for vistas, such as trails or rivers. All distinguishable ridges within the
town were considered for Kent. As a parallel to publicly determined scenic roads, a town
may have publicly determined scenic ridges as a more restrictive dataset. A town
commission or other organization such as a Land Trust could determine the horizon belt
for just one ridge of scenic importance.
Using what breaks the horizon as the criterion for obtrusiveness depends on the
height of the structure that would obtrude from within a horizon belt. The zoning
regulations for the town of Kent impose a height limit of 35’ for houses and this became
the empirical basis for obtrusiveness. Regulations determining structures allowed and
height limits will vary by town.
The criterion used for constraining horizon belts around ridgelines was
determined by consensus of the Kent Planning and Zoning Commission for this particular
regulation. The constraint for steep slope gradients was based on preexisting town
regulations for building roads. Other constraints might be based on soil type definitions

or slope gradient requirements for other built structures. A constraint could also be used
to ensure that horizon belts occur a specified distance away from roads.
While the land use attorney for the town of Kent found the methodology for
producing horizon belts defensible, almost one-third of the town lies within all the
horizon belts that were determined. This prompted the use of constraints to remove the
portions of horizon belts that occur on gentle slopes beyond 1000’ from ridgelines. In
general, constraints of a political nature could be the driving force behind the applied
physical constraints that determine a subset of horizon belts.
Horizon Belt Regulations
The constrained horizon belts formed the Horizonline Conservation District, as
referenced in the zoning regulations created for this purpose by Kent’s Planning and
Zoning Commission (Appendix A). As stated in the opening section of the regulations:
The purpose of this Horizonline Conservation District is to conserve and
protect the hill summits and ridges which form the high horizon visible
from the Town’s system of roads while allowing reasonable, appropriate
and compatible uses of the land. The specific goals of the District include
the preservation of scenic views and vistas that are critically important to
the rural landscape and character of the Town, and the minimization of
erosion and sedimentation hazards caused by the development and use of
steep hillsides and ridges.
The development restrictions in the Horizonline Conservation District pertain to
building construction and tree cutting. Site plans must be submitted for review if these
proposed activities fall within the District. If a portion of a lot lies within the District, but
the proposed activities are for the portion outside the District, the restrictions do not
apply. To the extent that building and/or cutting is allowed within the district, minimized
impacts on scenic vistas and steep slope gradients must be demonstrated.

Site plans for building or cutting activities in the District must indicated specific
trees to be removed; architectural drawings that include wall elevations, roof lines, and
façade materials; and 10 foot topographic contours for the entire lot. Other components
of a site plan that might be required include 2 foot contours, the flying of a balloon to
mark visibility, and field markings of the District by a surveyor. Other information might
be required as well depending on the proposed activity and location.
Future Direction
In August of 2005 the Horizonline Conservation District was officially approved
by the town of Kent. Other Appalachian Trail towns in Connecticut have been waiting to
see what would happen with this new approach to protecting scenic ridges, and may
follow suit. Some towns outside the Appalachian Trail corridor are implementing or
considering this methodology as well. Other areas of the country could benefit from this
approach to protecting scenic ridges. Western states now feature some of the highest
growth rates in the nation, with ridges that are often exposed with sparse tree cover.
Indeed, the horizon belt methodology could prove beneficial for anywhere that scenic
ridges and development patterns seem destined to collide.
Developing scenic ridges threatens more than local town character. The
recreational experience on trails and watercourses that come in close proximity to private
lands, such as is the case with the Appalachian Trail, can be negatively impacted by the
sight of a “castle on a hill.” Delineating the most obtrusive zone of a viewshed could
have regional applications that serve both land and water recreationists.
By focusing on the most obtrusive part of a viewshed, horizon belts allow for
compromise. A homeowner could locate a house on a portion of a ridge that affords a

ewarding view, while hidden or camouflaged from important vistas. The potential for
such compromise may mean the difference between a Zoning Commission willing to take
a needed step towards protecting scenic ridges that shape a town’s character, or one
unwilling to face the onslaught of legal challenges that a less discriminating approach to
protection may bring.