Landslide Forecasting at the HJ Andrews
Experimental Forest Using GIS
Development of HJ Andrews Digital Watershed using ArcView GIS Tools
Creating Calibration Region Theme
This term project has been prepared to showcase one
tool that may be used to aid in educated and controlled development and related
population growth in areas of the United States susceptible to terrain
instability. The Stability Index
Approach to Terrain Stability Hazards Mapping (SINMAP) model is a program
designed to assess terrain stability conditions in the geographical information
system (GIS) framework. As presented in
this report, this tool was used to identify and map potentially unstable
regions of the HJ Andrews Experimental Forest (Figure 1.1), located
approximately 50 miles east of Eugene, Oregon.
Figure 1.1 Location Map of HJ Andrews Experimental
Forest
Terrain stability has plagued the western United
States ever since adventure-seekers were tempted by promises of gold, land, and
freedom in the late 1800s. For the
people from the East and Midwest, the route west traversed rugged and difficult
land, but, inevitably, this “untamed” land began to be settled. Along with the influx of population,
however, came the haphazard use of the abundant natural resources of the
west. The necessity of transportation
yielded roads in inhospitable country; houses needed to be built, and
homesteaders turned to the forests for answers. Development continues to this day, but for the past several
decades Americans have begun to see the effects of uncontrolled growth.
Too often, this development has had adverse effects
on the natural landscape – one such example is landslides. Although they occur in every state and U.S.
territory, some areas are more vulnerable than others. The Rocky Mountains, the
Appalachian Mountains, the Pacific Coastal ranges, and parts of Alaska and
Hawaii all have areas of very weak or stressed material resting on steep slopes.
Together with the construction of homes and other structures in these areas,
heavy logging and the associated roads constructed to access harvestable
timber, and increased groundwater flows and surface water runoff due to
development, landslides have become prevalent in these areas, especially during
seasons of heavy rain and snowfall.
Mudslides have plagued California, among other states, for many years,
and Colorado had several fatal avalanches in 1999.
Determining the locations of areas potentially susceptible to these natural phenomena has become critical as the population in the west continues to strain the limits of the surrounding resources. In particular, GIS has come to light in the last decade as an effective tool to map dangerous areas throughout the United States.
The purpose of this report is to determine the
effectiveness of mapping potentially unstable areas using GIS and digital
terrain data that is readily available over the Internet. The HJ Andrews Experimental Forest was
selected because data specific to this application has been developed in the
past several years, including:
§
Digital Elevation
Models (DEMs) with a resolution fine enough to accurately represent the
terrain;
§
Groundwater recharge
data necessary to populate the model;
§
Soils data specific to
the sub-basin of interest; and
§
Maps of the locations
and types of landslides that have occurred in the sub-basin necessary to
calibrate the model.
In particular, this report will present the methodology used to define a digital representation of the terrain. The infinite slope stability model will be discussed, and the SINMAP model will be applied to the HJ Andrews site, generating a stability index for the site, as well as a map of the areas most susceptible to landslides, which will be calibrated with actual landslide data.
Development of HJ Andrews Digital
Watershed Using ArcView GIS Tools
Established in 1948 by the US Forest Service, the HJ
Andrews Experimental Forest has been the root of the HJ Andrews Long Term
Ecological Research (LTER) program - a major center for analysis of forest and
stream ecosystems in the Pacific Northwest.
Since its inception, predominantly Forest Service research has been
conducted on the management of watersheds, soils, and vegetation in the
sub-basin. LTER work has developed a backbone of long-term field experiments as
well as long-term measurement programs focused on climate, stream flow and
water quality, and vegetation succession.
During LTER3 (1990-1996) increasing emphasis was placed on developing
the concepts and tools needed to predict effects of natural disturbance, land
use, and climate change on ecosystem structure, function, and species
composition. Portions of this data have
been used for the application of the SINMAP tool to the HJ Andrews sub-basin.
With only basic terrain data, such as DEMs, GIS tools are able to provide environmentalists, natural resource planners, and developers alike with a general idea of areas most susceptible to terrain instability. ArcInfo is a powerful GIS tool used to create GIS data, while ArcView is a particular GIS tool that is capable of manipulating existing GIS data.
The first step taken to determine the potential
susceptibility of terrain to landslides in the HJ Andrews sub-basin was to
develop an accurate representation of the terrain. DEMs, readily available at
several different resolutions over the Internet (http://edcwww.cr.usgs.gov/doc/edchome/ndcdb/ndcdb.html),
are packets of data encompassing a prescribed area that provide
three-dimensional data, much like an electronic topographic map of the area. A
DEM consists of a sampled array of elevations for ground positions that are
normally at regularly spaced intervals.
DEMs can be imported into ArcView using the Spatial Analyst extension.
A DEM with a 10-meter by 10-meter cell coverage for
the HJ Andrews site was made available and imported into ArcView. The following steps were implemented to
successfully load the DEM into GIS:
1.
The file dem.asc
was imported into ArcView using the File/Import menu.
2.
The file was opened in
a text editor to ensure that the data was correct and in the proper format.
3.
The first six lines of
the file were edited in a text editor to match other ASCII DEMs. The original file read:
north: 4903755
south: 4893465
east: 572175
west: 558465
rows: 343
cols: 457
The first six lines of the file were modified to read:
ncols 457
nrows 343
xllcorner 558465
yllcorner 4893465
cellsize 10
NODATA_value -9999
This was necessary because
ArcView is designed to read this data in a particular order and for unknown
reasons, the ASCII DEM did not supply this data in the correct order.
A representation of the DEM as imported into ArcView is presented in Figure 2.1.
Figure 2.1
HJ Andrews DEM
The projection of the DEM was determined to be in
the Universal Transverse Mercator (UTM) projection, which in this case used the
North American Datum (NAD) of 1927. The
HJ Andrews site falls in Zone 10 of the UTM projection. This was determined by opening the dem.prj
text file associated with the DEM.
Determining the projection of the DEM was critical due to the fact that all of the other data layers that were to be used with the DEM had to be in the same projection, with the same cell size defined for each subsequent data layer as that of the original DEM.
Grid Clipping using CRWR-Raster
The ArcView extension CRWR-Raster was used to clip
the DEM grid to fall entirely within the limits of the Boundary
theme. With the DEM grid active, the
steps followed included:
1.
Select
CRWR-Raster/Clip Grid by Polygon
2.
Select Yes to clip
active theme by polygon to be chosen
3.
Select Boundary as
Clipping Theme
A new DEM grid was generated that had all the elevation values completely within the Boundary theme.
Landslides
Landslide data specific to the HJ Andrews site was a
critical piece of information for this site, eventually used to calibrate the
SINMAP model. The landslide data was
downloaded in ArcInfo format (the file had a .e00 extension) and was
imported into ArcView using the Import 71 feature.
The landslide data was added as a point coverage
theme, with each point having attributes of area, perimeter, slide id, slide #,
id, northing, and easting.
The projection for the landslide data was also
determined to be in the UTM – NAD 27 Zone 10 projection, so no additional
projection was necessary.
In order to successfully calibrate the SINMAP model,
a field called Type had to be defined for each point. Initially, there was no Type field
associated with each landslide. This
data was gathered from personnel at the HJ Andrews site and was added manually
to the attribute table of the point coverage.
With the Attributes to Slides theme active, the Type field
was added using the following steps:
1.
Select Table/Start
Editing
2.
Select Edit/Add Field
3.
At the Name
prompt, input Type
4.
Manually enter the
landslide type for each point
5.
Select Table/Save
Edits
6.
Select Table/Stop
Editing
The landslide type was differentiated between those
that started as a result of road construction (2) and those unrelated to road
construction (1). This allowed the
calibration of the SINMAP model with only the naturally occurring landslides
and gave a more true representation of the areas of the sub-basin susceptible
to landslides due to terrain and moisture conditions in the soil.
Figure 2.2
Type Field in Landslide Attribute Table
Soils
Soils data for the HJ Andrews sub-basin was also
downloaded. This included soils data
specific to the HJ Andrews site available on the project web site in addition
to State Soil Geographic (STASGO) database data downloaded from the USGS (http://www.ftw.nrcs.usda.gov/statsgo2_ftp.html)
The soils data specific to the HJ Andrews sub-basin
was downloaded from the project web site in ArcInfo format and imported into
ArcView using the Import 71 feature.
The attributes associated with this data included:
§
Area
§
Perimeter
§
Soil Survey #
§
Soil Survey ID
§
Soil Unit
§
Unit Name
§
Description
§
Slope Class
§
Depth Class
§
Land Type
§
Map Symbol
§
Map Unit
This site-specific data can be related to the STATSGO soils data (presented below) via the Map Unit and Soil Unit fields.
This data set is a digital general soil association
map developed by the National Cooperative Soil Survey and consists of a broad
based inventory of soils and non-soil areas that occur in a repeatable pattern
on the landscape. The STATSGO soil maps are compiled by generalizing more
detailed soil survey maps. The soil map units are linked to attributes in the
Map Unit Interpretations Record relational database, which gives the
proportional extent of the component soils and their properties in each map
unit.
STATGSO was designed primarily for regional,
multi-county, river basin, state, and multi-state resource planning and, as
such, the large size of the polygons makes it suitable only for large areas.
Another database, called SSURGO (still under development by the USDA), details
soil polygons on the component level as opposed to the map unit level and would
be more suitable to small coverages like the HJ Andrews sub-basin.
The STATSGO soils coverage for the State of Oregon
was downloaded as a polygon shape file and was imported as a theme into
ArcView. The projection of the data was
originally in the Albers Equal Area projection, so it had to be projected to
the UTM-NAD 27 Zone 10 projection to match the other themes in ArcView. The CRWR-Vector extension was used to project
this theme. With the STATSGO
theme active, the steps followed included:
1.
Select Output Units
of meters
2.
Select UTM-NAD 27
from the Category menu
3.
Select Zone 10
from the Type menu
A new polygon coverage of the STATSGO soils was created in the proper projection, and the polygon theme was then clipped using the ArcView Geoprocessing Wizard extension described below.
Additional data
downloaded from the HJ Andrews web site included:
§
The boundary of the HJ
Andrews watershed, added as a polygon theme;
§
Detailed road maps of
the sub-basin, added as a polygon theme;
§
50 ft. contours of the
sub-basin, added as a polygon theme; and
§
Stream gauge
locations, added as a point theme.
As with the rest of the data downloaded from the HJ Andrews web site, the data had to be imported into ArcView using the Import 71 feature.
Clipped Grids (Geoprocessing Wizard)
The ArcView extension Geoprocessing Wizard was used
to clip the Roads and Contours polygons to match the boundary of
the HJ Andrews sub-basin. With the Boundary
Theme active, the steps included:
1. Select View/Geoprocessing Wizard
2. Select Clip on theme based on another
3. Select Boundary as Input Theme to Clip by
4. Select Roads or Contours as Polygon Overlay Theme
A new polygon theme was then generated that fell entirely within the Boundary theme. A three-dimensional representation of the entire HJ Andrews watershed with select themes activated is presented in Figure 2.3.
Figure 2.3
HJ Andrews Watershed with Select Themes Activated
This view was created with the 3-D Analyst ArcView
extension. The 3-D Analyst used the HJ
Andrews DEM to create a triangular irregular network (TIN) to represent the
topography. A clipped RF3 file from the
EPA Basins web site was overlayed on the TIN, along with the roads theme, to provide
the user with reference points. The red
dots represent landslides that occurred naturally, while the yellow dots
represent landslides that occurred in close proximity to roads.
SINMAP is a grid-based ArcView extension developed
at Utah State University with the support of Forest Renewal British Columbia,
in collaboration with Canadian Forest Products Ltd., Vancouver, B.C. This program implements the computation and
mapping of a slope stability index based upon geographic information, primarily
digital elevation data. As applicable
to the HJ Andrews experimental forest, SINMAP was the logical tool for the
prediction of the areas most susceptible to landslides because it could be
populated with site-specific data (gathered over the last several years) to
quickly identify regions where more detailed terrain stability assessments may
be warranted.
There are many approaches to assessing slope stability
and landslide hazards. Field inspection
has been the most widely used approach historically, but this is an arduous
process that is time-consuming and expensive.
The projection of future instability patterns from documented landslide
types and locations has also been used extensively, but this method may not
account for site-specific environmental conditions in all cases. The purpose of the SINMAP software is to
provide an objective terrain stability mapping tool that can compliment the
subjective terrain stability mapping methods currently being practiced in the
field.
SINMAP relies heavily on the coupling of steady
state topographic hydrologic models with the infinite plane slope stability
model, and has its theoretical basis in the infinite plane slope stability
model with wetness (pore pressures) obtained from a topographically based
steady state model of hydrology. SINMAP
uses site-specific data, combined with past landslide history, to create a
probabilistic model of the terrain most susceptible to instability.
It is important for the reader to note that a large part of the text of this section was taken directly from the SINMAP User’s Guide, 1998. For a more in-depth discussion of the theory of the SINMAP model, the reader is referred to this document.
The SINMAP methodology is based upon the infinite
slope stability model (e.g., Hammond et al., 1992; Montgomery and Dietrich,
1994) that balances the destabilizing components of gravity and the restoring
components of friction and cohesion on a failure plane parallel to the ground
surface with edge effects neglected.
Soil moisture (specifically, pore pressure) is also taken into
consideration in this model because it reduces the effective normal stress on
the failure plane. SINMAP populates the
topographic variables of the slope stability model by automatically extracting
elevation data from the DEM, calculating the specific catchment area (sub-basin)
of each cell, and quantifying the corresponding material properties in the
sub-basin on a cell-by-cell basis, such as soil strength and the effects of
climatological factors. The primary
output of the SINMAP model is a stability index used to classify the terrain
stability in each grid cell within the study area.
The following input parameters are recognized to
vary in each sub-basin and are specified in SINMAP by the user as upper and
lower boundaries on the ranges these values may take:
§
T/R
§
Cohesion
§
Angle of Internal
Friction
§
Lower Wetness Line
Percentage
The first parameter listed above is the ratio of
transmissivity of the soil (m2/hr) to the effective steady-state
lateral recharge rate of the groundwater in the sub-basin (m/hr), with a
default range of 2000 to 3000 (m).
Transmissivity data is collected in the field and is specific to each
soil type in the sub-basin of interest.
As such, it is difficult to quantitatively estimate specific values for
both transmissivity and recharge and, thus, SINMAP uses a range of values to
model the uncertainty of these values.
The second parameter is the cohesive properties of
the soils of the sub-basin, with a default range of 0 to 0.25
(dimensionless). This data is also
collected in the field and, for the same reasons described above, is input as a
range of values.
The third parameter is the angle of internal friction (F), with a default range of 30 to 45 degrees. Although the actual angle of internal friction for specific soils can only be determined in the field, F can be determined in general terms from Table 3.1, extracted from the Standard Handbook of Civil Engineering (McGraw-Hill, 1995).
Table 3.1
Angles of Internal Friction and Unit Weights of Soils
|
Type of Soil |
Density or Consistency |
Angle
of Internal Friction, Phi, degrees |
Unit
Weight (lb/ft3) |
|
Coarse
Sand or Sand
and Gravel |
Compact Loose |
40 35 |
140 90 |
|
Medium
Sand |
Compact Loose |
40 30 |
130 90 |
|
Fine
Silty Sand or Sandy
Silt |
Compact Loose |
30 25 |
130 85 |
|
Uniform
Silt |
Compact Loose |
30 25 |
135 85 |
|
Clay-Silt |
Soft to
Medium |
20 |
90-120 |
|
Silty
Clay |
Soft to
Medium |
15 |
90-120 |
|
Clay |
Soft to
Medium |
0-10 |
90-120 |
The last parameter is the SA plot lower wetness line
percentage. This dimensionless value
represents the boundary wetness between the low moisture and partially wet
zones on the saturation map. It is also
the wetness of the lowest line on the SA plot.
Table 3.2 provides an example of the stability classes that are defined in terms of the Stability Index (SI) for this model. The SI is the factor of safety that gives a measure of the magnitude of destabilizing factors required for terrain instability and is defined as the probability that a location is stable assuming uniform distributions of the parameters over the uncertainty ranges specified above. The SI generally ranges between 0 (most unstable) and 1.0 (least unstable). However, where the most conservative set of parameters still result in stability, the stability index is defined as the factor of safety at this location under the most conservative set of parameters and may yield a value greater than 1.0.
Table 3.2
Stability Class Definitions
|
Condition |
Class |
Predicted
State |
Parameter
Range |
Possible
Influence of Factor Not Modeled |
|
SI >
1.5 |
1 |
Stable
slope Zone |
Range
cannot model instability |
Significant
destabilizing factors required for instability |
|
1.5
> SI > 1.25 |
2 |
Moderately
stable slope zone |
Range
cannot model instability |
Moderate
destabilizing factors required for instability |
|
1.25
> SI > 1.0 |
3 |
Quasi-stable
slope zone |
Range
cannot model instability |
Minor
destabilizing factors could lead to instability |
|
1.0
> SI > 0.5 |
4 |
Lower
threshold slope zone |
Pessimistic
half of range required for instability |
Destabilizing
factors are not required for instability |
|
0.5
> SI > 0.0 |
5 |
Upper
threshold slope zone |
Optimistic
half of range required for instability |
Stabilizing
factors may be responsible for stability |
|
0.0
> SI |
6 |
Defended
slope zone |
Range
cannot model instability |
Stabilizing
factors are required for stability |
The selection of breakpoints (1.5, 1.25, 1.0, 0.5
and 0.0) in Table 3.2 is subjective, requiring user judgment and interpretation
in terms of the class definitions. The
terms ‘stable’, ‘moderately stable, and ‘quasi-stable’ are used to classify
regions that, according to the model, should not fail with the most
conservative parameters in the parameter ranges specified. The terms ‘lower threshold’ and ‘upper
threshold’ are used to characterize regions where, according to the parameter
uncertainty ranges, the probability of instability is less than or greater than
50%, respectively. The term ‘defended
slope’ is used to characterize regions where, according to the model, the slope
should be unstable for any parameters within the parameter ranges
specified. In general, if the SI is
greater than 1.0, there is a probability that the terrain is unstable given the
most conservative parameters specified in the user-defined ranges.
The general infinite slope stability model factor of safety (ratio of stabilizing to destabilizing forces) is given by (simplified for wet and dry density the same, from Hammond et al., 1992):