Kimberly C. Mayeski, Faye Krueger
Technological Building Blocks for Deriving Timberland
Suitability Answers
With the advent of GIS and image processing capabilities, the process of
calculating timberland suitability for Forest Planning required through the National
Forest Management Act has changed. This paper compares the traditional way of
determining suitability with innovative, new technology. The format of this paper
begins with a description of what is currently available, explains the process for
determining information needs, and concludes with how we derive the final product,
timberland suitability. The vegetative layer is one of the most critical layers for the
suitability model. The process of creating this layer using satellite imagery is described.
The vegetative layer is used in several spatial analyses that can be vital in making good
management decisions. Various ways of using image processing for incorporating
resource data from remote sensing systems into GIS, are discussed.
Introduction
Forest plans provide broad, programmatic direction to manage the resources of the
National Forest in a coordinated and integrated manner. This direction conveys the
intent of governing laws, regulations, policies, and Regional guidance through Forest
Plan goals, objectives, standards, and guidelines. Collectively, this guidance is
developed for resource professionals to help them attain the described desired
conditions anticipated on the Forest.
Direction in Forest Plans is reviewed at the District level. Districts compare the
existing condition on the land with the desired conditions defined in the Plan. District
personnel find opportunities to move an area toward the desired conditions and design
projects to achieve the outcome. The Forest Plan becomes a commitment to the public
based on their participation. Money is allocated based on projections in the Plan for
goods and services. Forest Plans also describe the basis for monitoring resource
conditions. Monitoring information often results in changes in management.
The planning process addresses all resources. One emphasis area in the National Forest
Management Act (NFMA) relates to timber resources. Specific direction is given for
determining suitable acres of timber. Suitability is defined in NFMA as "the
appropriateness of applying certain resource management practices to a particular area
of land, as determined by an analysis of the economic and environmental consequences
and the alternative uses foregone. A unit of land may be suitable for a variety of
individual or combined management practices."
Identifying Land Suitable for Timber Production
The National Forest Management Act (NFMA) directs that during the forest planning
process, lands that are not suitable for timber production will be identified. NFMA
also requires that lands not suitable for timber production be reviewed at least every 10
years. NFMA regulations outline this process in Section 219.14. The process is
carried out in three stages:
Stage 1 National Forest land is tested against criteria for biologic capability,
availability, or physical suitability. Lands failing to pass these tests
are set aside from further consideration for timber production and
classed as "not suitable." (See Land Classification List in Table 1).
This paper identifies the steps taken for this stage and compares it to
the process used in the initial round of planning.
During Stage 1, lands not suitable for timber production are identified
by the following criteria:
- The land is not forested.
- Technology is not available to ensure timber production from
the land without irreversible resource damage to soils
productivity or watershed conditions.
- There is no reasonable assurance that such lands can be
adequately restocked.
- The land has been withdrawn from timber production by an
Act of Congress, the Secretary of Agriculture or the Chief of
the Forest Service.
Stage 2 Lands passing the tests in Stage 1 are assessed to determine the cost
and benefits for a range of timber management intensities or regimes.
No land is discarded at this point.
Stage 3 The lands from Stage 2 are tested against criteria representing Forest
objectives, silvicultural requirements, and cost efficiency. Lands failing
to pass these tests are classed as "not suitable" also.
How Land Not Suitable for Timber was Determined in the First Round of Forest
Planning
During the Caribou National Forest's initial planning process in 1985, non-forested
land was identified by taking the vegetation types from the rangeland analysis and
timber types from the 10-year timber inventory. Landtype associations were also
mapped. Using clear mylar overlays, vegetation polygons were mapped. Those stands
that were forested were identified by cover type and structure. Landtype capabilities
were mapped on mylar and landtype groups were identified that could not sustain
timber management activities. Also areas that were withdrawn from timber production
were mapped on mylar, all at 1:24000 scale.
Why GIS Today?
The process used in the first round of planning to identify vegetation and landtypes and
to hand-draw polygons on mylar overlays was labor intensive. In many cases, different
people gathered the same data across the forest, introducing their own bias into the data
collection. With the advent of GIS (Geographic Information Systems) and remote
sensing, data collection becomes more cost effective and efficient. GIS is an organized
collection of computer hardware, software, geographic data, and personnel designed to
efficiently capture, store, update, manipulate, analyze, and display all forms of
geographically referenced information. Remote sensing is the art of collecting and
interpreting information about an area from a remote vantage point.
When the Caribou National Forest embarked on revising their Forest Plan, different
methods for developing a vegetation layer were discussed. In the past, a timber
inventory was taken every 10 years and used to determine suitability and other
management activities. For this new planning effort, forest managers agreed to use
digital imagery to derive a consistent vegetation layer. By using GIS technology, spatial
depictions are highly accurate.
Step 1 - Information Needs Assessment
An information needs assessment was the first step. When assessing needs for a GIS
project of this size, it is important to know what questions need to be answered through
the analysis. By knowing which questions need to be answered, appropriate spatial
layers can be determined and assembled. From monitoring results and other available
information, the Forest needed to ascertain where suitable timber exists today and
whether a change in management direction is needed. The National Forest
Management Act establishes the criteria for determining suitability. One criteria is to
identify non-forested lands. With GIS technology, roads, utilities, administrative sites,
streams, and standing bodies of water can be buffered and eliminated from the suitable
base. During the initial Forest Plan efforts, these features were not captured and were
included in suitable acres.
The information needs assessment identified the following thematic layers the Caribou
National Forest needed to complete a suitable timber model:
1) Vegetation Layer (forested vs non-forested)
2) Administrative Sites
3) Improved Roads
4) Utilities or Utilities Corridors
5) Private land
6) Mines
7) Standing Bodies of Water
8) Streams
9) Slopes from 0-45% and slopes from 0-65%
10) Land types that are unstable, unstockable, and have low productivity.
11) Research Natural Areas
Step 2 Automate Data for Non-Forested Lands
Each layer or theme had to be automated for input into the GIS. Automation is
transferring hard copy data to a digital format that can be used in a GIS environment.
There are several different ways of automating layers so they are in a usable format.
Two common ways are "digitizing" and "deriving". Digitizing is probably the most
common way of automating spatial data. This process converts graphic information
scribed by interpreters into digital form subject to manipulation by the computer.
Deriving layers from other sources is another way of obtaining necessary GIS data for
an analysis. A vegetation layer was derived from Landsat Satellite Imagery, a remote
sensing system that contributes data at low cost and in a form that is compatible with
GIS requirements.
Lands which fall into one of the categories on the Land Classification List (Table 1)
generally should not be included in the suitable timber base. Although these lands may
have suitable timber, special management designation or the inability of the land to pass
criteria under NFMA regulations explained earlier in this paper, eliminate any future
timber harvest opportunities. The order in which these layers are discussed below
follows the order of the thematic layers in the land classification list.
Vegetation Layer: Satellite imagery from July, 1991 was purchased, and a vegetative
layer was derived from the imagery, using Erdas image processing software. This layer
contained the complete forest vegetation, both forested and non-forest lands. The
process and guidelines for deriving a vegetative layer for use in managing national
forests, published by the Nationwide Forestry Application Program, was used. The
raw imagery was clipped to the forest boundary, and an unsupervised classification was
completed. Landtypes that were easily recognized, like bodies of water and non-
forested land, were labeled and verified. Training sites were established for classes
that could not be readily identified and required further investigation and refinement.
Training sites and the associated field forms were entered and used to classify areas
with similar spectral signatures. Labeling of the classes was based on a visual and
statistical analysis with vital assistance from Forest Service personnel who were
familiar with on-the-ground conditions. For a more detailed description of deriving a
vegetative layer from satellite imagery, please refer to the above mentioned publication.
Initially, "Past Harvest Units" were not identified as a required layer. However, this
layer was identified as necessary during the process. If the units were cut before the
date of the satellite imagery, even though saplings may have been planted, these
saplings were too young to be identified as coniferous stands. They would likely have
the spectral signature of a mountain brush or sagebrush type. The Forest decided to
include all sales that were sold before fiscal year 1995 in the past harvest unit layer.
Units were identified through fiscal year 1995 to aid in yield calculations at a later date.
For the purposes of the suitable timber model, these units were digitized and added into
the vegetation layer.
Administrative Sites: Administrative sites include campgrounds, guard stations,
ranger districts, warehouses or other Forest Service buildings. Location and
identification of these sites were obtained from Cartographic Feature Files (CFFs).
Cartographic Feature Files are digitized from Primary Base Series Maps at the
Geometronics Service Center, a Forest Service mapping facility in Salt Lake City,
Utah. These 1:24000 scale maps contain information on roads, trails, waterways, and
administrative boundaries. Administrative sites in the CFFs are represented as point
features. These points represent the site location on-the-ground, but not the actual size
of the facility. It was necessary to identify the size of the site, because timber within
these sites is withdrawn from harvest. From land status documentation for the Caribou
National Forest, the exact acreage of each site was entered into GIS. A "look up" table
was created in ArcInfo and the points were buffered to the actual acres that comprise
each site.
Roads: Arcs representing the existing transportation system were pulled from the
CFFs and buffered 30 feet, 15 feet on each side of the road. Overall, road width
averaged 30 feet forest-wide.
Utilities: Utilities include powerlines, gaslines, slurry lines, and water lines. These
arcs were pulled from the CFFs. The Forest determined that the average width of these
disturbances was sixty feet, 30 feet on each side of the utility corridor.
Private Land: Areas within the Forest Boundary that are privately owned were also
identified. This thematic layer was pulled from the CFF's. These areas did not need to
be buffered, because the areas represent the exact location and acreage of the private
land.
Mining: Open-pit mining of phosphate occurs on the Caribou National Forest. Areas
were identified where mining impacts further preclude timber production. The actual
lease boundary was not used, because much of the area within the boundary is not
disturbed. The areas within the lease boundaries that have already been mined or
disturbed were digitized. These areas represent where actual phosphate mineral is
located and include the associated pit and borrow areas. Geologists identified these
areas and digitized this layer for the GIS environment.
Standing Bodies of Water: Standing bodies of water were also pulled from the CFFs.
These standing bodies of water include ponds and lakes. These areas were not
buffered.
Streams: The streams, however, were buffered 30 feet for perennial streams and 15
feet for intermittent streams. Since the width of a stream on the Forest varies from
streambank to streambank, 30 feet for perennial and 15 feet for intermittent streams
were considered average based on specialists' recommendations. These buffered areas
do not represent riparian areas. Riparian areas will be addressed in stages 2 and 3 in
the Forest Planning process. (See Page 1, stages 2 and 3)
These eight layers represent the total non-forested land. (Refer to Table 1, Section 1 of
the Land Classification list.) This process was more refined and more accurate than the
one used during the initial round of Forest Planning. The non-forested areas can now
be computed and removed from the land base; the resulting layer would be forested
land.
Step 3 Automate Data for Lands Withdrawn, Not Capable, or Not Suitable for
Timber Harvest
Land Withdrawn from Timber Production: Other considerations were made to
remove certain lands from the suitable base through legislation regarding the
management of National Forest System lands. (Refer to Section 2 on the Land
Classification list Table 1.) Forested land withdrawn from timber production includes
wilderness areas and Research Natural Areas. Although the Caribou National Forest has
two proposed wilderness areas, neither of these areas have received legislative action.
Therefore, these areas are included in the suitable timber base. Research Natural
Areas, on the other hand, are special management areas and are excluded from the
suitable timber base. RNA administrative boundaries were pulled from the CFFs.
Land Not Capable of Producing Crops of Industrial Wood: Areas that are rocky,
talus, and have low productivity are identified as not capable of producing crops of
industrial wood. (Refer to Section 3 on the Land Classification list Table 1.) The
Forest's soil inventory was digitized, and contained land type attributes. Using this
inventory, the land types were evaluated and a determination was made as to what land
types are suitable for harvesting timber and what land types are unsuitable. Polygons
that fit the description of rocky, talus, and low productive areas were excluded from the
suitable base in the model.
Forested Land Physically Unsuitable: Lands that fall into this category are describe
as unstable and not restockable. These land types are unsuitable for harvesting timber.
In addition, forested land physically unsuitable was also defined as land with slopes
greater than 65% or greater that 45% depending on which harvest method was
proposed. The slope layer was derived using Digital Elevation Models (DEMs).
DEMs are created by the U.S. Geological Survey. Using the ArcInfo module,
GRID, these DEM's were manipulated to create a layer called "Slope". Through
mitigation measures and unconventional logging methods, harvest may occur on some
slopes within the 45-65% range.
Step 4 Building the Model
Several different ways of running the suitability model were discussed. Consideration
was given to the accuracy needed for the resulting suitable timber layer. The vegetative
layer was derived using satellite imagery with 25 meter pixels in a raster format. For
convenience, a decision was made to design the model in a grid environment.
To prepare the individual layers discussed above for use in the suitability model, an
additional field of information was added in the vector environment in ArcInfo. A
field or an item (ArcInfo term) was added to the vector coverages and then values of
zero or 1 were entered into the field: an attribute of 1 indicates the land is suitable for
timber harvest; an attribute of zero indicates the land is not suitable for timber harvest.
For example, on the Research Natural Area layer, RNA polygons were attributed with
a value of zero and areas outside the Research Natural Area was attributed with a value
of 1. The Vector layers were converted into 5 meter grids.
The method used for running the first suitability model was to multiply the layers
together. If one layer is unsuitable in a 5 meter grid cell, then the final suitability layer
is unsuitable in that same 5 meter grid cell. If the unsuitable areas have a value of 0
and the suitable areas have a value of 1, 0 times any number will result in a
determination of unsuitable. The modeling was done with Erdas Image Processing
Software, using the model maker module.
The suitable timber model was tested on the Cache Range, an area on the Montpelier
Ranger District. This area was chosen through a forest-wide assessment of which area
might best represent all the features that are considered when determining timberland
suitability. Using ArcInfo, a hard copy map of this test area was plotted with two
different classes: suitable and unsuitable for harvesting timber. Each class was
represented by a different color. Locational features, like roads and the Forest
Boundary, were used to help the user identify real world locations.
After viewing the hard copy map of timber suitability on the Cache Range, managers
were interested in finding out which layer was responsible for a particular area being
classified as unsuitable. By multiplying all layers together at one time, this question
was unanswered. Instead, multiplying the layers together in sequence as shown on the
Land Classification List (Table 1) provided information on where the greatest change
occurred between layers. (See Table 2 for results of this process.) The results in Table
2 were determined by taking the Rangeland layer times the Administrative Sites layer.
The result of this multiplication was then multiplied by the roads layer and so on
through the list. On Table 2 in the change column, the layer identified as causing the
majority of acres to be classified as unsuitable was the Low Productive layer or Soils
layer.
Things You Want to Consider
Be Customer-Focused
The key to having a product that is usable by the customer or user is to involve
them in its design and production. When the vegetation layer was derived, a
representative from each Ranger District, who was familiar with on-the-ground
conditions, helped classify the vegetative types. It also was vital that District
representatives review suitability mapping results. Technical people can
perform the automation of the layers and the modeling tasks, but if they are
unfamiliar with real world conditions, they lack a key component. Employees
who are familiar with real world ground conditions can fill the missing
component.
Tune in to Problem-Solving
When the suitability for the test area was being reviewed, several questions
arose. The first issue involved isolated stands of suitable timber averaging
around ten acres in size. These isolated areas would not be economical to
harvest, because of road-building costs to access the stands. Different filtering
methods were tested. The best results were attained using the "eliminate"
command in ArcInfo. Polygons, less than 10 acres, that were labeled suitable
were eliminated. This method only eliminated polygons less than 10 acres,
and did not change the shape of the other polygons.
Harvesting practices may leave narrow strips of timber between harvesting
units. These narrow strips did not show up as suitable in the final layer, even
though they are suitable. They did not show up as a pure signature of timber,
on the raw imagery, because of their narrow width, and therefore, are
classified as non-forested. The process that was used to fix this problem
involved changing the existing Past Harvest Unit coverage, so it would include
the stringers. A copy was made of the Past Harvest unit coverage and then the
stringers added between units, using aerial photos for accuracy.
Limber Pine is a coniferous timber type that is not marketable and not suitable
for timber harvesting. From the satellite imagery, limber pines have a similar
spectral signature of other coniferous types and were classified suitable in the
model. Limber pine only exists on the southern part of the Cache Range
within the Caribou National Forest. Looking at the soils layer, limber pine
showed up in areas that were identified on the soils layer as land type 101.
Land type 101 is describe as being areas with high elevations, glaciated and
cyroplanated slopes. Typically, this land type has steep slopes with stony soils
that have low revegetation potential. During the initial identification of land
types, this land type was missed as a type that is not restockable. This error
was fixed by calculating land type 101 as unsuitable for harvesting timber.
Summary
The technological world of GIS will make land management planning more
understandable, consistent, trackable, more efficient, and more exact. It offers the
ability to analyze and make decisions about how land resources can be managed. It
meets needs for both site-specific and programmatic level analyses. It can serve as a
tracking facility for monitoring management activities and land conditions. It can
produce future scenarios for land management options. GIS links real world problems
with powerful analytical tools. For these reasons, and many more, GIS is taking
rudimentary planning, used in the early 1980s, into the 21st century.
GIS will be used to develop alternative management options during the Caribou
National Forest's update to their current Land Resource Management Plan. It will also
provide real time spatial displays and associated attributes that link to other relevant
data bases for on-the-ground management activities. Change detection using GIS for
monitoring will greatly improved the ability of resource professionals to adapt
management practices with changing environments and ground conditions.
Kimberly C. Mayeski, GIS/Remote Sensing Specialist
Faye Krueger, Assistant Forest Planner, GIS Coordinator
USDA Forest Service, Caribou National Forest
250 South Fourth Ave Federal Bldg. Suite 172
Pocatello, Id 83201
Telephone: (208) 236-7539
Fax: (208) 236-7503