Content uploaded by Christopher Stockdale
Author content
All content in this area was uploaded by Christopher Stockdale on May 29, 2016
Content may be subject to copyright.
Katelyn Loukes and Oliver Clovis December 2009
Georeferencing the Mountain Legacy
Project oblique photos to evaluate historical
landscape change
By:
Chris Stockdale
Ol iver Clovis
An d Katelyn Loukes
Report Prepared for: Don Podlubny, Foothills Research Institute, Sept 22, 2010
EXECUTIVE SUMMARY
It has been observed from a variety of sources that the structure of the ecosystems along the Rocky Mountain
East Slopes has changed considerably over the past century. The most significant drivers of this change are a
fundamental disruption of the natural disturbance dynamics of the region caused by modern European
settlement (which includes fire suppression, fire exclusion, railway construction, land clearing and other
factors), and climate change.
The Mountain Legacy Project provides us with an invaluable view of this past. A series of comprehensive
phototopographic surveys were conducted along the length of the Alberta Rocky Mountains from the US
border to the Willmore Wilderness from the late 1800’s to the early 1900’s. These photographs show a
dramatically different landscape than today: total forest cover was less prevalent, alpine treeline was at lower
elevation, glaciers were larger, there were fewer mature forest stands, and grassland and meadow cover were
significantly greater.
The overall ecosystem complexity appears to have declined over the past century, and with it, potential losses
in local level biodiversity, plus significantly elevated Mountain Pine Beetle and wildfire risk, loss of habitat for
many species, and a general shift towards a late-succession vegetation landscape. Development of management
solutions to mitigate and reduce these problems is not a simple task. While the MLP images have been used to
qualitatively show landscape change, they have not been used to quantitatively assess these changes due to the
lack of a user friendly method that allows for spatial interpretation of the images.
There are many applications that can do image processing and converting 2D to 3D images, however, most of
these applications require long training, are very expensive, and require advanced workstations to operate. Our
goal was to identify and test an existing solution, or combine systems, to allow for a relatively low cost, easy to
operate solution to permit users to georeference the MLP images. We chose to use the freeware Corripio
Application, as it has been successfully tested in similar studies. As the MLP images were not originally
collected with spatial referencing in mind, a large portion of the study involved trial and error to correctly
format the necessary data inputs for the spatial referencing process.
The process involves combining the image to be referenced with a digital elevation model (DEM) with two
primary inputs: the point of view of the camera, and the centre point of the image. These must be known
precisely. The application then combines the DEM data with the image data, and creates a spatially referenced
image that can be examined in a GIS program to create spatial data. The spatial accuracy of the georeferencing
method places points within 37.7m (se 4.9m) of their real world locations. Relative distances and spatial
extents within the images themselves are accurate. A detailed methodology document is included to allow the
user to replicate the process. We determined that the best approach is to interpret the images in their non-
referenced state as the referencing process degrades the resolution of the images to the resolution of the DEM
used for the referencing process itself.
The labour involved in preparing the data can be decreased significantly by cropping the input photos to
remove all sky and distant objects, and centering the image on a readily identifiable landmark that can be
located on corresponding orthophotos with ease. This may involve dividing MLP images into multiple, smaller
images, but the tradeoffs in accuracy and processing are positive. The Corripio Application itself could be
modified to improve the resolution, however this comes with a significant financial investment in software
beyond the scope of this project.
INTRODUCTION
In Alberta, the single most dominant historic agent of disturbance throughout our ecosystems has been fire.
Fire from both lightning and aboriginal sources has been commonplace across the east slopes of the Rockies
for millennia. In recent decades, however, a variety of sources of data suggest that the disturbance regime of
the entire length of the Rocky Mountains is significantly different today than 100 years ago due to
settlement, fire suppression (putting out fires that start), and fire exclusion (preventing the start of aboriginal
fires) on the landscape. This shift in the fire regime of the Rocky Mountain Eastern Slopes has coincided
with the arrival of the Mountain Pine Beetle (Dendroctonus ponderosae) in Alberta, and a significant
increase in wildfire threat across the landscape.
Managing this landscape level change is not without challenges. The majority of MPB and fire experts
advise that the best long term strategy to minimizing losses due to beetle outbreaks or wildfire is to adjust the
landscape level susceptibility to MPB and fire. For forest health managers this requires a sound
understanding of how the age classes and spatial distribution of pine affects the spread dynamics of MPB.
For fire managers, the variations in the spatial arrangement of fuel on the landscape are critical. These
problems are interconnected: the landscape level pine patterns are influenced by the fire regime, which in
turn is driven by the vegetation structure of the landscape.
A concern of forest managers is that solutions created for one problem may result in new, unforeseen
problems, such as increases in wildfire rate of spread or intensity, increases in susceptibility to other forest
pests, loss of critical habitat for keystone species or reductions in biodiversity. This is where using historical
fire regimes to guide overall management actions can be of great benefit: by knowing what the historical
landscape looked like, and using that as a target to work towards, managers can have more confidence that
the majority of the ecosystem’s key functions will be maintained and the requisite parts should be present.
An ideal management solution would be one that provides a landscape structure that:
Is resistant to MPB epidemics
Is within the Natural Range of Variation (and the ancillary benefits associated with this relating to
ecosystem function and processes)
Accounts for climate change uncertainty
Supports a viable timber industry
Is more fire resistant than today’s landscape
Provides for the habitat needsof known endangered or at-risk species
Is valued for recreation
While it is difficult to know with certainty what the landscape has looked like over the millennia since the
glacial retreat, detailed photos provided by the Mountain Legacy Project (Higgs et al, 2009) reveal a
dramatically different landscape of a century ago, compared to today, along much of the East Slopes. These
photographs show far less total tree cover, and more landscape level diversity. While the Mountain Legacy
Project (MLP) photos represent only a single snapshot in time, they provide us with the opportunity to
understand what the landscape looked like at the turn of the 20th century, which was a time of considerably
less human induced disturbance than it is today.
The questions these photographs can answer regarding changes between c. 1900 and 2010 are numerous:
How much has mature lodgepole pine cover increased?
How much has landscape level connectivity of maturelodgepole pine increased?
How much has overall tree cover increased?
How much higher is the alpinetreeline?
How much smaller are meadows?
How much has grassland andshrubland decreased?
What is the average size of mature forest patches?
The landscape level changes observable within the MLP photos are striking, but they have not been
quantified except in the cases of Rhemtulla (1999) and Watt-Gremm (2004), both of which were highly
labour intensive approaches focusing on small areas in one or a few photographs. What is needed is a robust
method by which the changes can be measured across landscapes in multiple photos. In September 2009, we
submitted a proposal to The Foothills Research Institute Mountain Pine Beetle Ecology Program (FRI
MPBEP) titled Using Oblique Historical Photos to Determine Past Mountain Pine Beetle Susceptibility
(attached as Appendix A). The following were the objectives of the full proposal:
1. Identify the most pragmatic method of rendering the oblique Mountain Legacy Photos into a spatial
format.
2. Select a test landscape for this study.
3. Interpret vegetation into classes of MPB Susceptibility
4. Describe past MPB susceptibility of the landscape, and to compare it to the current situation.
5. Using modeling, discover whether landscape structure was one of the primary limiting factors
preventing previous MPB outbreaks.
The FRI MPBEP funded objectives 1 and 2, and chose to defer a decision on funding objectives 3-5 pending
the success of achieving Objectives 1 and 2. That is the goal of this work, and this report describes a method
for achieving objective 1, and the data sources and landscape (Objective 2) that was used to do so.
BACKGROUND/PROJECT RATIONALE
Mountain Pine Beetle (MPB)(Dendroctonus ponderosae) is one of the most widespread and destructive
forest insect pests in North America, with a geographic range from north-central British Columbia in Canada
all the way to Mexico. However, with numerous known historical infestations of MPB throughout its range,
it is only known to have been present in Alberta on two occasions: in the 1970’s-80’s in the Bow Valley and
Crowsnest Pass, and in the Bow Valley in the 1940’s (D. Lux, personal communication). In both cases, these
events were “spillovers” from large infestations in British Columbia: as infestations waned in BC, the
populations in Alberta disappeared. It is suspected by leading scientists that these past Alberta MPB
infestations were only possible due to immigration from BC, and that beetle populations were unable to
successfully breed and maintain themselves on the east side of the Rocky Mountains.
Historically, the climate has not been favourable to the survival and spread of MPB in Alberta (Safranyik
and Carroll, 2006). MPB has never been proven to exist north of Banff before. No records of MPB presence
exist in the Canadian Forest Service’s Forest Insect Disease Survey (FIDS), which has been maintained since
the 1930s, nor is there any known historical reference to this type of disturbance. Mountain Pine Beetle is
present in Alberta today from Waterton Lakes National Park in the south, beyond the Chinchaga Hills in the
north, and past Slave Lake to the east (Alberta Sustainable Resource Development, 2009). The situation is
unlikely to improve as climate change is also creating more favourable conditions for the establishment,
entrenchment and spread of MPB throughout Alberta (Carroll et al, 2004), which in turn could provide
access to the jack pine forests of the entire eastern boreal ecosystem.
There is another factor that cannot be overlooked to explain the suspected historical failure of MPB to
establish firmly in Alberta, and that is the historic landscape structure. For MPB to reach epidemic
proportions, numerous thresholds must be overcome at the tree, stand and landscape level (Raffa et al. 2008).
If the landscape does not have large amounts of contiguous mature pine, an outbreak cannot be sustained, no
matter whether the other barriers have been overcome, including favourable climate conditions.
Evidence from a variety of sources suggests that the landscape condition of the east slopes is well beyond its
historical, natural range (Hawkes, 1979; Arno, 1980; Barrett, 1996; Feller et al, 2004; Jevons and Donelon,
2008; Rogeau, 1999; Rogeau, 2004; Rogeau, 2005). A comprehensive systematic repeat photography survey,
the Mountain Legacy Project (Higgs et al, 2009), shows dramatic changes in the landscape between roughly
1910 and today. At the beginning of the 20th century, forested areas were less prevalent, less dense, and less
homogenous in structure and composition than they are today at the start of the 21st century. Not
surprisingly, coinciding, and quite possibly the result of this past-century shift in landscape condition is the
epidemic scale arrival of the MPB in Alberta.
The Mountain Legacy Project holdings provide us with a collection of photographs that can be used to
evaluate changes in historic landscape structure. We propose using the historical oblique photos to allow us
to quantitatively measure the changes in MPB susceptibility between the landscape of the early 1900’s and
the present day. This would also provide land managers with a more strategic overview of how to potentially
alter today’s landscape to reduce the potential of epidemic spread of Mountain Pine Beetle. The current
targets call for a 75% reduction over 20 years in susceptible pine along the East Slopes (Alberta SRD, 2010).
A reduction of that scale over that timescale is operationally difficult to achieve. We may demonstrate that
much smaller, well targeted reductions may be highly effective.
Despite a significant financial investment in these historical photos, their potential for addressing ecological
studies in a quantitative manner remains largely untapped. For example, this proposal is designed to test and
identify the most promising method of using these photos to allow for spatial analysis. By fully exploring the
potential of these photos using innovative techniques, we may be opening up an entirely new area of
decision-support for land managers.
GEOREFERENCING OBLIQUE PHOTOGRAPHS
When the historical photos were first discovered, their value in studying landscape change was immediately
noted, however not necessarily for the purpose of spatial analysis in a GIS. While the initial photographs of a
century ago were taken primarily for use in the creation of topographic maps, once the maps were created,
the data associated with the creation of the maps (ie precise location of the photopoint, azimuth of
photograph etc) was not necessarily retained. In addition to this, the retaken photos were not initially taken
specifically with spatial analysis in mind, and therefore many attributes that would be useful in such analyses
were not recorded.
Orthorectification of aerial photographs is commonplace and used widely in innumerable applications
(Petrie, 1977), however, orthorectification and georeferencing of oblique angle photos is distinctly less
common. While georeferencing oblique photos is not technically any different than aerial photography (both
are based upon the principles of photogrammetry), the rarity of these oblique data sources being used has
resulted in far fewer computer applications being available to tackle the problem. In comparison, the number
of computer applications available for the spatial interpretation, georeferencing and manipulation of aerial
photos is massive.
Previous Attempts to Georeference Mountain Legacy Project Images
This is not the first study to look at spatially referencing the MLP images. Numerous graduate students have
worked on the Mountain Legacy Project (and its predecessor the Rocky Mountain Repeat Photography
Project). Two students in particular had projects related to developing methods to georeference small
samples of paired photos, Jeanine Rhemtulla and Graham Watt-Gremm.
Rhemtulla (1999) made the first attempt to conduct a spatial analysis of the MLP images. She picked out
control points on 20 paired images. The original and repeat photos were interpreted using a vegetation
classification scheme which included physiognomic classes, and crown closure. Each image was overlaid
with an acetate copy of the image and control points. These were laid onto a Cartesian grid to obtain a rough
coordinate system, and subsequently digitized for analysis in a GIS.
Watt-Gremm (2004) initially attempted to rectify the images to a digital elevation model using ground
control points obtained from orthophotos. He found that the results were too distorted for analysis and that
obtaining ground control points directly from the field was not feasible. He eventually settled on a modified
version of the photogrammetry methods used in the historical phototopographic surveys.
First, he draped a grid and orthophoto of each image location onto a digital elevation model (DEM) using the
freeware Geographic Resources Analysis Support System (GRASS GIS), and then used the NVIZ extension
to view the combined DEM/image scene from the camera perspective. The resulting scene was then exported
to PhotoShop 7, where the NVIS scene was overlaid onto the oblique photo in question. The combined the
scene and oblique photo were then imported back to GRASS. Here the oblique photo grid cells were
georeferenced using data from the NVIS scene, and subsequently interpreted.
Both the Rhemtulla (1999) and Watt-Gremm (2004) methods were effective in allowing for the spatial
referencing of interpreted data from the MLP images, but both of these methods were labour intensive,
subject to an unknown amount of error and unsuitable for processing a large volume of images.
Georeferencing Mountain Legacy Project Images: Applications and Data
The goal of this project is to take the lead in implementation of spatial analysis of the Mountain Legacy Project
photographs and to quantify historical MPB susceptibility of the landscape.
The objectives of this project were to:
1. Identify the most pragmatic method of rendering the oblique Mountain Legacy Photos into a spatial
format
2. Select a test landscape for this study.
Past attempts have been limited by the technology available to the researchers. While Rhemtulla (1999) and
Watt-Gremm (2004) developed methods specifically for examining the MLP images, they are by no means
the only attempts at georeferencing oblique photos. Our challenge was to look at other approaches and to
determine which of these other solutions could be applied to the MLP images so as to improve the accuracy,
and reduce the labour required to reference the images as per Rhemtulla or Watt-Gremm. We needed to
ensure that the data inputs required for these other methods were readily available to us from the MLP
images, which culled the number of potential applications significantly.
Several promising methodologies have been published which are less reliant on ground control points and
take advantage of advancements in computing power. Corripio (2004) developed a program (hereafter
referred to as the Corripio Application) which georeferenced oblique images using a DEM model, the x,y,z
(latitude, longitude and elevation) coordinates of the camera location, the angular field of view, the x,y,z of
the exact center of the image and extra ground control points for increased accuracy.
Aschenwald et al. (2001) developed a similar methodology to Corripio and a software application called the
JUKE method which requires similar inputs, but also includes an error measurement. While the application
showed considerable promise, there was little documentation available to allow us to use the application
successfully. We contacted the authors of the program for assistance, but were unable to receive a timely
reply, thus this method was not explored further.
Anderson et al (2009) conducted a thorough statistical analysis of the Corripio Application by comparing
oblique photos with orthorectified aerial images in US National Parks. They concluded that based on measures
of geometric accuracy, standard error and spectral accuracy, the application functioned well. The Corripio
Application has been used successfully to monitor snow accumulation in the Rockies (De Beer and Pomeroy
2009). Photos taken repeatedly at identical locations throughout the season were successfully georeferenced
and used to measure how the areas of snow/non-snow changed over time. They used a GPS and prominent
features to acquire ground control points and ended up with a RMS error of ~ 3m. However, the error tended
to be in a consistent direction, so the surface areas calculations were affected very little. Our challenge was to
determine whether or not we had the necessary data inputs required to run the application successfully on the
MLP photos.
CORRIPIO METHODOLOGY
Background
This methodology was originally developed by Javier Corripio to georeference terrestrial oblique photos in
order to quantify changes in snow surface albedo. The quality of the resulting georeferenced image hinges on
the resolution of the DEM, the image quality (ie. presence of lens aberrations/distortions) and the number of
ground control points (Corripio 2008). His method involves projecting a DEM into the camera coordinate
system and rotating and scaling it so that the DEM cells match up with the corresponding pixels in the image
(Corripio 2004). The DEM information is then mapped onto the image pixels, resulting in a georeferenced
image (geotiff format file).
To accomplish this, the internal camera specifications (ie. the focal point and the exact location/rotation of the
camera must be known. In order to avoid mapping photo pixels to DEM locations not viewable from the
camera location, a viewshed analysis must be performed in order to generate a DEM which contains only
locations that correspond to pixels in the photograph. This may be accomplished using ArcGIS 9.x or other
GIS packages which include DEM and viewshed functionality.
Software Used
1. Georeferencing Application (Corripio's Application)
The main Georeferencing application which runs off the IDL virtual machine platform. It is
downloadable from the IDL user code libraryat:
http://www.ittvis.com/UserCommunity/CodeLibrary.aspx
Inputs:
•3 band Geotiff Image File to be georeferenced
•Full DEM in Geotiff format
•.dat file in csv format containing ground control point names and coordinates
•DEM from viewshed analysis
•Rotation of the camera in degrees
•Location of the camera (x, y, z)
•Focal length of camera in cm
•Resolution of camera in dpcm (dots per centimeter)
•Film height and width
•Offset X and Y coordinates
•Central pixel of the photographic image
Output:
Ageoreferenced Geotiff Image
2. ArcGIS 9.3.1 Spatial Analyst and ArcScene:
Extensions to perform the viewshed analysis on the DEM and to determine the camera point of
view (POV).
Input:
DEM model
Output:
DEM containing points visible from camera location
POV
3. IDL Virtual Machine
Platform required to run Corripio's georeferencing application. Freeware which comes with the
trial version of IDL 7.1. Downloaded from:
http://www.ittvis.com/Downloads.aspx
4. PCI Geomatica-- Focus
Required for image pre-processing, intpretation, line drawing and other editing functions.
APPLICATION: GOOGLE EARTH PRO
Google Earth Pro has many tools available that are not included in the base version of Google Earth. This
allows for the importation of spatial data in a variety of formats, including ESRI shapefiles and DEM images.
We found Google Earth Pro to be invaluable in determiningthe first approximation of POV, the azimuth of the
photograph from the POV, and even some ability to view the original and retaken oblique photographs overlaid
upon a DEM. We were also able to use Google Earth Pro to determine the field of view in each photograph.
This program is also good for showing the geographic location of the images using overlays (figure1).
Many of these findings are included in the results discussed below. Google Earth Pro, however, is not suitable
to any more than a rudimentary analysis of photos, as the DEM it contains is far too coarse to be of practical
value. While DEM files of higher resolution can be imported for viewing in Google Earth Pro, they remain as
“virtual” DEMs that can be seen and displayed. They are only images floating above the base DEM built into
the program.
Figure 1: Overlay of Jumping Pound Station Image 9-4 (Higgs, 2009) on Google Earth Pro imagery.
Test Landscape
The four criteria to be met for a test landscape to be chosen for this study:
1. Original photographs had to be clear, and allow fora consistent level of vegetation classification
2. Repeat images had to meet the same criteria as 1.
3. Multiple photo stations covering the same area of the landscape
4. Must be within areas for which Alberta SRD had already collected and processed LiDAR data at 1m
resolution.
The Mountain Legacy Project photo archives and data files were examined and yielded five areas, each
represented by a separate survey, meeting all of the above criteria. These surveys and locations were:
Bridgland 1918, Ya-Ha Tinda Survey
This survey covers the area surrounding the Ya Ha Tinda Ranch, and has 4 photo stations within areas of
available LiDAR data coverage. There are a further 3 stations outside the LiDAR coverage. Many of the
images are hazy in the midground and distance. There are significant blindspots on the landscape.
Bridgland 1927, North Saskatchewan Survey
Area surrounding Nordegg. This area was only repeated in 2009, and many of the images have not yet been
processed. There is broad coverage of the landscape, but numerous photo stations have not yet been repeated
as of yet, and this leaves several important blindspots on the landscape, in addition to there being only partial
LiDAR coverage.
Lambert 1927, Willmore Wilderness Area Survey
This survey focused on the south end of Willmore Wilderness in the area north of Twintree Lake, Jasper
NP). This survey is highly detailed, the landscape is well covered with 7 stations. For the purpose of testing
the hypothesis of whether or not MPB susceptibility has significantly changed over a century, this area is not
within an area of significant current MPB infestation, and therefore is less than ideal for this study.
Wheeler 1923, Willmore Wilderness Area
This survey covers the area where Jackpine and Meadowland Creek intersect. There are 7 photo stations,
good quality images, and excellent ground coverage. The area is within a currently heavily MPB infested
region. However, there is a lack of available AVI data within the Willmore to compare vegetation
classification taken from the oblique images, which could be vital if the next phase of this project is to go
ahead (classifying changes in MPB susceptibility on the landscape).
Wheeler Irrigation Survey, 1895-1897 (this area is the one chosen)
This survey covers area in Kananaskis Country, southeast of Lake Minnewanka. The Wheeler Irrigation
Survey was chosen as the final candidate as it had 14 photos stations, each with multiple photos all meeting
the aforementioned requirements. The area centered around Moose Mountain is heavily covered by no fewer
than 4 photo stations with overlapping photo coverage. The images are of excellent quality, LiDAR is
available, AVI is available to reference classification schemes, and it is an area of current management
concern to AB SRD.
The Wheeler Irrigation Survey 1895-1897 images centered on the western slopes of Moose Mountain were
chosen for this study. The repeat photographs for the area were taken in 2009. The Mountain Legacy
photographs in our dataset were collected from 3 camera stations located in the Rocky Mountains, roughly 56
km South-West of Calgary, Alberta. The three camera stations are: Jumping Pound Summit, Moose
Mountain Center and Moose Mountain West (see figure 2).
There were 5 photographs used from the Jumping Pound Summit Station, 3 from Moose Mountain West, and 1
from Moose Mountain Centre. Examples of the fields of view (FOV) for each of these photos are displayed in
figure 3a (Jumping Pound Summit images 9-4 and 10-4). The field of view can be deceptive, however, as many
features of the landscape are not visible due to being on the other side of hills/ridges. A viewshed analysis
(described below) shows what portion of the FOV is actually visible in the photograph. The viewsheds for
these two images are summarized in figure 3b. These images are displayed in Google Earth Pro.
Figure 2: Study Location and photo stations used for the test landscape.
Figure 3a and 3b: Field of view (FOV) for images 10-4 (red) and 9-4 (yellow) at Jumping Pound Summit Station. The corresponding viewshed
analysis is displayed in the right hand image with blue shaded polygons. The DEM did not cover all the mountain peaks, therefore the
viewshed does not include ground beyond the extent of the DEM.
CORRIPIO APPLICATION: METHODS AND EVALUATION
Data sets and tools used
Numerous data sets needed to be assembled to run the Corripio Application. Appendix B describes step-by-
step, button-click by button-click how to create the necessary data inputs, and how to run the Corripio
Application. The following is a more abbreviated version.
Data Processing and Preparation
The specific data inputs required for running the Corripio Application are described below:
1. DEM - in tiff format.
2. Photo - The photo to be georeferenced. 3 band in tiff format.
3. Viewshed - In tiff format.
4. Ground Control Points (GCPs)
5. Camera Coordinates - UTM and the elevation of the camera.
6. Target Coordinates - UTM and elevation of the centre pixel of the photo.
7. Focal length - From the Camera specifications
8. Image resolution –Scaled to account for cropping
9. Sensor width/ height - Scaled to account for cropping.
The procedure for creating these inputs is described below and illustrated in Figures 4 and 5:
1. Generate Shapefile Points of the Camera Stations from the Metadata.
2. Create DEM
3. Crop Photographs and Calculate Parameters.
A visual basic (VB) application was developed which calculates the scaled down values for the camera sensor/film
width(m) and height(m) and resolution(dpcm) based on the original size of the digital photographs (39mpx) and the
camera sensor dimensions.
4. Mark Centre and GCPs
5. Determine Z (elevation) value of Camera Point/Point of View (POV).
6. Identify the UTM Coordinates of Center and GCPs
To determine whether the original photo could be parameterized using the same values as the repeat image, the
original photo was overlaid with the repeat photo center and GCP vector layer for assessment. The center points were
located in the same location in all pairs.
7. Create Viewshed
8. Create Settings file:
The required inputs were entered into the application and saved in a settings file so they could be easily reloaded.
Figure 4: Data preparation workflow diagram (part 1).
Figure 5: Data preparation workflow diagram (part 2).
Running the Application and Managing Outputs
With the settings file as input, the application is ready to be run. The process is described in Figure 6 below,
and in full detail in Appendix B:
Figure 6: Process for running the Corripio Application
The “preview”function of the Corripio Application is used to view the perspective projection of the GCPs. In
some cases, matching GCPs is challenging and it is suspected that there are potential lens aberrations and
perspective distortions occurring. The application produces a georeferenced 3 band tiff image which has the
extents of the DEM.
It is important to note that due to a 10m DEM being used as input, the image is resampled to a 10m grid, and
therefore significant detail is lost in the photograph, which is readily apparent in the foreground of the
georeferenced images. Figures 7a-7h show screen captures from ArcScene with the input photograph (non
referenced) on top, and the Corripio output photographbelow:
Figure 7a: ArcScene screen capture showing non-referenced repeat photo (inset above) and the spatially referenced output from the Corripio
application (image Moose Mountain West, 2-4).
Figure 7b: ArcScene screen capture showing non-referenced repeat photo (inset above) and the spatially referenced output from the Corripio
application (image Moose Mountain West, 4-4).
Figure 7c: ArcScene screen capture showing non-referenced repeat photo (inset above) and the spatially referenced output from the Corripio
application (image Moose Mountain West, 3-4).
Figure 7d: ArcScene screen capture showing non-referenced repeat photo (inset above) and the spatially referenced output from the Corripio
application (image Jumping Pound Summit, 9-4).
Figure 7e: ArcScene screen capture showing non-referenced repeat photo (inset above) and the spatially referenced output from the Corripio
application (image Jumping Pound Summit, 10-4). Note: distant mountains visible in inset are beyond the extent of the DEM used in the
referenced photo, and therefore” do not exist”.
Figure 7f: ArcScene screen capture showing non-referenced repeat photo (inset above) and the spatially referenced output from the Corripio
application (image Jumping Pound Summit, 12-4).
Figure 7g: ArcScene screen capture showing non-referenced repeat photo (inset above) and the spatially referenced output from the Corripio
application (image Jumping Pound Summit, 14-4).
Figure 7h: ArcScene screen capture showing non-referenced repeat photo (inset above) and the spatially referenced output from the Corripio
application (image Jumping Pound Summit, 15-4).
The outputs can still be viewed without an ArcScene license by using Google Earth Pro.The output image can
be viewed draped across a DEM using either ArcScene (above in Figures 7a-h) or in Google Earth Pro (see
Figures 8a-b). The larger “shadows” with no data that are visible in Fig. 8a and 8b are due to the differences in
the DEM between the referenced image, and the base DEM of Google Earth Pro. Google Earth Pro uses a
variable DEM. In some locations in the world it uses a highly detailed 1m or less DEM, in other locations it is a
very coarse 50m DEM. This is one of its most significant drawbacks, it is impossible to know what the
resolution of the DEM is at a particular locationas this information is not available to the user.
Figure 8a: Google Earth Pro screen capture showing the spatially referenced output from the Corripio application
(image Moose Mountain West, 2-4).
Figure 8b: Google Earth Pro screen capture showing the spatially referenced output from the Corripio application
(image Moose Mountain West, 4-4).
Assessment of Accuracy: Corripio Application
In order to evaluate the accuracy of the georeferencing process, the non-referenced retake photos and aerial
orthophotos covering the same area were used. Distinctive features that were clearly visible in both the
oblique and aerial orthophotos were located and the spatial location was marked on the photos. The distance
between the oblique photo point and corresponding aerial orthophoto points were measured. In total, 33
reference points were chosen in 6 photographs from the Jumping Pound Summit and Moose Mountain
Centre stations. The mean error in the distance between the Corripio Application referenced photo and its
real-world location (as determined from the aerial orthophotos) was 37.7m (standard error = 4.9m). Figures
9a-d show the error checking points from the Jumping Pound Station used for this process.
Figure 9a: Image Jumping Pound Station, 9-4 error checking points. The georeferenced image check points are in red, and the orthophoto
derived control points are in white. In this image they are nearly indistinguishable.
Figure 9b: Image Jumping Pound Station, 10-4, error checking points. The georeferenced image check points are in red, and the orthophoto
derived control points are in white.
Figure 9c: Image Jumping Pound Station, 14-4, error checking points. The georeferenced image check points are in red, and the orthophoto
derived control points are in white.
Figure 9d: Image Jumping Pound Station, 15-4, error checking points. The georeferenced image check points are in red, and the orthophoto
derived control points are in white.
In addition to this test, polygons were drawn on both sets of photographs to determine if the area measured
(and therefore relative distances between points) was accurate. 30 polygons were drawn in the same 5 photos
using cutblocks or meadows as distinct objects to measure. While sizes did not perfectly agree, they were
never more than 20% in disagreement, and there was no directional bias leading to the conclusion that there
was no significant difference in the sizes of cutblocks or meadows measured in the oblique versus aerial
photographs. An example of polygon size and locations derived from the two data sources is shown in figure
10.
Figure 10: Area testing of polygons in image Jumping Pound Summit 9-4. The purple polygon is the “real” location as per the orthophoto, and
the white polygon is that which was drawn from the referenced photo.
Another example of the accuracy and relative distance between objects was tested examining the change in
treeline in image Moose Mountain Centre, 12-3. The original and repeat photos are displayed below with
respective interpreted treelines in Figures 11a and 11b. Figure 11c shows the difference in location of the
treeline in the images between c.1895 and 2009 after referencing the images in the Corripio Application.
Figure 11a: Treeline vector in image Moose Mountain Centre 12-3 from Wheeler 1895-1897.
Figure 11b: Treeline vector in image Moose Mountain Centre 12-3 from Higgs 2009.
Figure 11c: Zoomed in image showing change in treeline vector location (within the green box drawn in Figs 9a and 9b) between c.1895 (black
line) and 2009 (orange line) after spatial referencing of images MMC 12-3 from Wheeler and Higgs. Vectors are displayed on 1m LIDAR DEM
with lower elevation at bottom of image and ridge top at the top of the image. There is a clear advancement of treeline.
PROJECT CONCLUSIONS AND RECOMMENDATIONS
The objectives for this phase of the project were to find a method to georeference the Mountain Legacy
Photos in such a way as to allow for quantitative analysis of the historical changes in landscape conditions
observable within the photos. We determined that the Corripio Application proved to be the most effective
tool for georeferencing the photographs. The process is still labour intensive, but considerably less so than
the approaches taken by Watt-Gremm (2004) and Rhemtulla (1999). The detailed methodology to
georeference MLP images is included as Appendix B. The automation tools (3 ArcToolboxTool models and
one visual basic application, submitted to Foothills Research Institute) referred to in Appendix B are
available from Foothills Research Institute, or by contacting the lead author of this study.
The Corripio Application itself runs quickly. On a powerful computer, the application will reference an image
in less than 5 minutes. However, this is dwarfed by orders of magnitude by the time required to select and
interpret images, assemble data, calculate input parameters, and analyze the output data. It would be misleading
to provide estimates for how long it takes to do these other steps, as it is wholly dependent upon the objective
for the study, and the landscape in the images themselves. The process becomes considerably more efficient
with larger numbers of images to georeference.
CONCLUSIONS
The Corripio Application offers a simple, accurate, and precise method to spatially reference oblique
angle photographs, provided key data inputs are known.
oSpatial accuracy of the method located points within 37.7m with a standard error of 4.9m of their
real-world locations in our test images.
oThe accuracy and precision of the method are directly affected by the resolution of the input
Digital Elevation Model, and the precision of the location of the Point of View (POV), and centre
point of the image being referenced.
oWe observed that the larger the field of view in each photograph, the greater any referencing errors
were. We did not analyze enough images to test this for statistical significance, but this has also
been noted by other users of the Corripio Application.
Interpretation of the photographs needs to occur in advance of the georeferencing process due to loss of
image resolution during referencing.
All images must be cropped from the top to remove most of the sky and distant background. This
effectively lowers the centre point of the image and brings it closer to the POV,improving accuracy.
oIf the centre point is on or above the horizonthe Corripio Application will crash.
GIS software is needed to create the inputs required for the Corripio Application and to analyze the
outputs.
oThe GIS software used must be able to view Digital Elevation Models in 3D, and to conduct
viewshed analyses.
Landscape complexity within images has a large impact upon the time required for referencing
oThe more complex the landscape in any particular image, the easier it is to identify the centre point
and POV as there are numerous features that aid the user in establishing locations, the converse is
also true.
oThe more complex the landscape, the longer it takes to interpret the image and analyze the output.
The source code of the Corripio Application can be altered with a paid IDL license (contact ITTVIS).
Google Earth Pro is useful for viewing inputs and outputs, but the base DEM is too coarse to be used for
accurate assessment of POV, centre point, or for analysis of the referenced images.
We have not described the steps required for image interpretation and data analysis: these will be unique
to the purpose the images are being analyzed for, and the hypotheses being tested.
RECOMMENDATIONS
To speed up data preparation and to improve accuracy of the referenced product, crop, scale, and resize
photographs:
oChoose an obvious feature to use as the image centre, and then crop the photo equally around the
nominal centre point.
oLimit the size of the landscape to be referenced in any single image.Divide images with large fields
of view into multiple smaller images to enable more accurate referencing.
oIf the image has complex terrain with slopes in the foreground and other slopes in the distance, with
large gaps (as revealed by the viewshed analysis), break the whole photograph into multiple smaller
extents and evaluate each smaller image.
Memory limitations of the Corripio Application can be overcome by limiting the area to be referenced as
indicated above, as this will allow a more detailed DEM tobe used.
ACKNOWLEDGEMENTS
The lead author would like to thank the Foothills Research Institute for funding this work via the Mountain
Pine Beetle Ecology Program. I would also like to thank the British Columbia Institute of Technology for
allowing me to sponsor Katelyn Loukes and Oliver Clovis for the GIS work done on this project. Their
innovation, dedication, and amazing attention to detail were truly impressive. So too was their patience during
countless DimDim and Skype conversations and meetings. I would also like to thank Debbie Mucha, the
Program Lead for the GIS Program at Foothills Research Institute for pointing me in the direction of the BCIT
Advanced Diploma in GIS program. I would like to thank Eric Higgs for his support through this project. And
last, but not least, I wish to thank Graham Watt for countless walks and talks about this project. He thought he
had moved beyond the MLP, sorry Graham for dragging you right back in.
REFERENCES
Alberta Sustainable Resource Development, 2009. Mountain Pine Beetles in Alberta. 2009. [online]. Government of
Alberta. Available from: http://www.mpb.alberta.ca [Accessed 25 November 2009].
Alberta Sustainable Resource Development, 2009. Mountain Pine Beetle Aerial Overview 2009. [online]. Government
of Alberta. Available from: http://www.mpb.alberta.ca/Files/MPB-AerialOverview-2009.pdf [24 August 2010].
Alberta Sustainable Resource Development, 2009. Mountain Pine Beetle Long Term Strategy. [online]. Government of
Alberta. Available from http://www.mpb.alberta.ca/AlbertasStrategy/Longtermstrategy.aspx [24 August 2010].
Anderson, S; Fonstad, MA; and B Vogt, 2009. Extending quantitative analysis of historical photographs. Available at:
http://communicate.aag.org/eseries/aag_org/program/AbstractDetail.cfm?AbstractID=31841. [Accessed Dec. 20
2009].
Arno, SF, 1980. Forest fire history in the Northern Rockies. Journal of Forestry 78 ( 8 ) : 460-465.
Aschenwald, J; Leichter, K; Tasser, E; and U Tappeiner, 2001. Spatio-Temporal Landscape Analysis in Mountainous
Terrain by Means of Small Format Photography: A methodological approach. IEEE Transactions on Geoscience &
Remote Sensing. 39, 880-885.
Barrett, SW, 1996. The historic role of fire in Waterton Lakes National Park, Alberta. Contract report Prepared for
Parks Canada. 27 pp.
Carroll, AL; Taylor, SW; Régnière, J; and L Safranyik, 2004. Effects of climate change on range expansion by the
mountain pine beetle in British Columbia. Pages 223–232 in Shore TL, Brooks JE, Stone JE, eds. Mountain Pine Beetle
Symposium: Challenges and Solutions. Victoria (Canada): Natural Resources Canada, Canadian Forest Service, Pacific
Forestry Centre. Information Report BC-X-399.
Corripio, J, 2004. Snow surface albedo estimation using terrestrial photography. 2004. International Journal of
Remote Sensing, 25 (24), 5705-5729.
De Beer, CM, and JW Pomeroy, 2009. Modelling snow melt and snowcover depletion in a small alpine cirque,
Canadian Rocky Mountains. Hydrological Processes 23, 2584–2599.
Feller, MC; Plett, M; White, C; and J Dordel, 2005. Forest fires, mountain pine beetle, and forest structure
relationships in Canada’s southern Rocky Mountains. 48ppg.
Geobase, 2008. [online]. Available from: http://www.geobase.ca [Accessed 28 Nov 2009].
Higgs, ES; Bartley, G; and AC Fisher, 2009. The Mountain Legacy Project. Second ed. pp. 80
Hawkes, BC, 1979. Fire history and fuel appraisal of Kananaskis Provincial Park. Report prepared for: Resource
Assessment and Management Section, Planning and Design Branch, Parks Division, Alberta Recreation and Parks.
190ppg.
Jevons, SR, and SG Donelon, 2008. Historical fire frequency in subalpine forests of the Spray and Kananaskis
watersheds, Alberta. Alberta Ministry of Tourism, Parks and Recreation, 2008. 38 ppg.
Petrie, G, (1977), Orthophoto maps. Contemporary Cartography. 2(1), pg. 49-70.
Raffa, KF; Aukema, BH; Bentz, BJ; Carroll, AL; Hicke, JA; Turner, MG; and WH Romme, 2008.Cross-scale drivers of
natural disturbances prone to anthropogenic amplification: The dynamics of bark beetle eruptions. Bioscience 58:501–
517.
Rhemtulla, JM, 1999. Eighty Years of Change: The Montane Vegetation of Jasper National Park. Thesis (MSc).
University of Alberta.
Rogeau, M.-P, 1999. Fire history study of the central Rockies. Ecosystem InterAgency North Saskatchewan Unit. Banff
National Park 61 pp.
Rogeau, M.-P, 2004. Fire regime study of Kananaskis district, Alberta. Contract Report prepared for Spray Lake
Sawmills Ltd and Alberta Community Development, Parks and Protected Areas. 111 pp.
Rogeau, M.-P. 2005. Fire regime study of C5 FMU. Contract report prepared for Alberta Sustainable Resource
Development, Forest Protection Division.
Safranyik, L; Carroll, AL, 2006. The biology and epidemiology of the mountain pine beetle in lodgepole pine forests.
Pages 3–66 in Safranyik L, Wilson B, ed. The Mountain Pine Beetle: A Synthesis of Its Biology, Management and
Impacts on Lodgepole Pine. Victoria (Canada): Canadian Forest Service, Pacific Forestry Centre, Natural Resources
Canada.
Stockdale, C; and D Andison, 2009. Using oblique historical photos to determine past mountain pine beetle
susceptibility. Research Proposal. Foothills Research Institute.
Watt-Gremm, DG, 2004. Taking a good long look: disturbance, succession, Landscape change and repeat photography
in the upper Blakiston Valley, Waterton Lakes National Park. Thesis (MSc). University of Victoria.
APPENDIX A:
NOTE: this document has been edited from its original form to remove all financial statements that were in the
original proposal and to reflect changes in contact information of the lead author
Project Title: Using Oblique Historical Photos to Determine Past Mountain Pine
Beetle Susceptibility
Foothills Research Institute (FRI) Natural Disturbance (ND) Program
Principal Investigator (PI)
Chris Stockdale
Principal, Burning ecoLogic
(780)446-1015
Chris.Stockdale@burningecologic.com
Collaborators
Dr. David Andison, Bandaloop Consulting and Program Lead, FRI, ND Program
Dr. Eric Higgs, University of Victoria, principal investigator Mountain Legacy Project
Rick Arthur, ASRD, Wildfire Prevention Officer, member Mountain Legacy Project
Marie-Pierre Rogeau, Wildland Disturbance Consulting
Bill Riel, Canadian Forest Service, Victoria BC, MPB Spread Modeling expert
Potentially one or more of the following:
Students of BCIT Advanced Diploma in GIS Program
Students of University of Calgary Masters in GIS Program
Students of University of Victoria Geomatics Program
1. Background
Mountain Pine Beetle is present in Alberta today from Waterton Lakes National Park in the south, to Peace
River in the north, and as far east as Slave Lake. This is without known historical precedent. To the best of
our knowledge, it is unmatched in terms of duration, intensity and geographic range. While BC has had
continual endemic MPB presence, and has encountered repeated epidemic level MPB infestations in the past,
it has only known to have been present in Alberta on two occasions: in the Bow Valley in the 1940s, and
both there and in the Crowsnest Pass in the 1970-1980s. In both cases (1940s and 1970-1980s), these events
were “spillovers” from virulent infestations in British Columbia. Furthermore, as infestations waned in BC,
these populations in Alberta disappeared. It is suspected by leading scientists that these past Alberta MPB
infestations were only possible due to immigration from BC, and that the beetle was unable to breed and
maintain itself on the east side of the Rocky Mountains.
There are only two possible factors that can account for the virtual absence of MPB in Alberta until now,
either the climate was not favourable, some threshold combination of landscape conditions did not exist, or
both.
The climate barrier theory is supported by the lack of physical historical evidence north of Banff. The fact
that the MPB can survive today in these locations also supports the climate barrier theory given the warmer
winters across the west over the last decade or so.
However, there is also strong evidence supporting the landscape structure threshold theory as well. Evidence
from a variety of sources suggests that the landscape condition of the east slopes today is well beyond its
historical, natural range. For example, several historical landscape simulation studies on east slopes
landscapes suggest the amount of old forest is at the high end, or in some cases beyond the natural range.
Similarly, a comprehensive systematic repeat photography survey, the Mountain Legacy Project, shows
dramatic changes in the landscape between roughly 1910 and today. At the beginning of the 20th century,
forested areas were less prevalent, less dense, and less homogenous in structure and composition than they
are today at the start of the 21st century. Not surprisingly, coinciding, and quite possibly the result of this
past-century shift in landscape condition is the epidemic scale arrival of the MPB in Alberta.
The best way of resolving this question is to find physical evidence of either past MPB outbreaks, or
historical suitable landscape conditions. Definitive evidence of past MPB activity is difficult to obtain: snags
killed by mountain pine beetle do not tend to remain standing for more than a few decades, and the
distinctive signs of beetle activity (blue stain and galleries) do not last long on downed woody debris. As a
result, investigators have tried to search for signals in surviving trees that neighboured the trees that may
have been killed by MPB. These signals are sometimes detected in tree rings, but they are not conclusive
without known MPB epidemics to calibrate the method. There are numerous other factors that can cause
similar signals, including low intensity fires, sun damage, and other insects. A final issue with finding MPB
evidence is time.To search for evidence back into the time period of the early 1900s and 1800s, stands need
to be found today that would have been of susceptible age back at that time. They would have had to have
been a minimum 80 years old in 1900, therefore at least 200 years today. Furthermore, these sites need to be
located in likely locations of past MPB activity: south to southwest facing slopes, and in the low elevation
passes where MPB can overcome cold high elevation climates. There are very few such stands on these
aspects and in these valleys today as there have been significant landscape scale wildfire events through most
of these areas in the 1890s and 1930s.
A more promising angle of investigation is to look at landscape structure. For MPB to reach epidemic
proportions, numerous thresholds must be overcome at the tree, stand and landscape level. If the landscape
does not have large amounts of contiguous mature pine, an outbreak cannot be sustained, no matter how
favourable the climate. In other words, if we cannot find historical evidence of the necessary compositional
and structural thresholds, then even if climate was amenable, MPB could not have thrived in Alberta.
We propose using the historical oblique photos to allow us to quantitatively measure the historical MPB
susceptibility of the landscape in the early 1900’s to determine if it was even possible for there to have been
landscape scale MPB infestations in the past. No one has used these photographs yet to answer questions
regarding the past MPB susceptibility of the landscape, and this proposal is meant to pioneer these methods.
It may be revealed that historically there have been insufficient quantities and arrangements of mature pine
on the landscape for MPB to become established in the past regardless of climate. It may also be revealed in
which locations there may have been the possibility of previous epidemics, providing sampling guidance to
future dendro MPB studies in targeted locations.
Stepping back even further, despite a significant investment in these historical photos, their potential for
addressing stand, landscape, wetland, and aquatic studies in a quantitative manner remains largely untapped.
For example, this proposal is designed to test and identify the most promising method of using these photos
to allow for spatial analysis, and to provide information to forest managers to determine how unusual this
infestation is. If this Alberta infestation is indeed an unprecedented phenomenon, and if historical landscape
conditions can be quantified and proven to have limited the spread of MPB in the past, this will provide
management with quantitative measures and targets that they can use for managing this infestation. By fully
exploring the potential of these photos using innovative techniques, we may be opening up an entirely new
area of decision-support for land managers.
2. Qualifications
We propose running this project through the FRI Natural Disturbance Program. This allows us access to
significant data and information resources, access to a proven communications and extension network, low
overhead costs, and the expertise of, and input from, the current partnership base of the ND program. The
project is well within the mandate of the ND Program as per our long-term plan (Andison 2009).
As the principal investigator on this proposal, Mr. Stockdale’s expertise lies mainly in the areas of
disturbance ecology and interactions of MPB and fire. He has managed the Alberta prescribed fire program
and has 9 years experience in the field of mountain pine beetle and fire ecology. As a Forest Health Officer
with the Canadian Forest Service he conducted statistical analyses and examining stand dynamics of
lodgepole pine following MPB infestation, and the interrelationships between MPB and fire on the
landscape. He has a MSc in Forest Science from Oregon State University and a BSc Honours degree in
Biology from the University of Victoria.
Dr. Andison’s expertise lies mainly in the areas of natural disturbance patterns, biodiversity monitoring, and
disturbance simulation modelling throughout western Canada. He has published several relevant articles in
peer-reviewed journals and books, but is perhaps better recognized locally as the lead for the FRI Natural
Disturbance Program.
The ND program activity team to which Dr. Andison and Mr. Stockdale are responsible has successfully
provided consistent leadership over the last 14 years on a broad variety of natural pattern research, education,
and decision-support priorities. The ND activity team currently includes two members of forest industry, two
members of the Alberta government, one from national parks, and one from the FRI. All sitting members
represent agencies that work in the eastern slopes.
Eric Higgs will provide assistance in selecting appropriate imagery and providing advice regarding lessons
learned in his years of analyzing the Mountain Legacy photos. The Mountain Legacy Project facilities will
be at the full disposal of this project for no charge. This includes advanced imaging high resolution
workstations, a 3 TB data server hosting all images, and graduate students contributing feedback and advice.
Eric Higgs will also be fully engaged with implementing the solutions, and in the event that we use a Co-op
Student from the UVic Geomatics Program, will be the lead advisor to that student.
Rick Arthur has been involved with the Mountain Legacy Project for several years and has studied the
images for evidence of historical fire patterns. Erica Lee will provide advice and guidance to ensure the
questions being asked meet the business needs of the ASRD Forest Health Program. M.P. Rogeau will
provide advice and assistance regarding image interpretation. Bill Riel is the leading expert in Canada in
Mountain Pine Beetle spread dynamics and will be invaluable in providing advice regarding data structure
and questions that can be asked. Mr. Riel and Erica Lee will provide the link between stand conditions and
MPB threat. The entire team will participate in the interpretation of the results, and Stockdale and Andison
will work together on disseminating the results.
3. Goals and Objectives
One of the core goals of the FRI ND Program is to understand and describe historic disturbance regimes in
Alberta and beyond. This information is needed to develop decision support tools to help land management
partners choose and defend more sustainable disturbance options. The goal of this project is to take the lead
in implementation of spatial analysis of the Mountain Legacy Project photographs and to quantify historical
MPB susceptibility of the landscape.
The objectives of this project are to:
3. Identify the most pragmatic method of rendering the oblique Mountain Legacy Photos into a spatial
format
4. Select a test landscape for this study.
5. Interpret vegetation into classes of MPB Susceptibility
6. Describe past MPB susceptibility of the landscape, and to compare it to the situation today.
7. Discover whether landscape structure was one of the primary limiting factors preventing previous
MPB outbreaks.
8. The FRI Natural Disturbance Program
It is important to understand that this is by no means a stand-alone project. This proposal fits within a
deliberate package of leading-edge research, experimentation and decision-support tool development. As
described above, the FRI Natural Disturbance Program has been functioning for 14 years. As described in
the FRI ND long-term plan (Andison 2009), this project links directly with a number of existing research,
communications, and integration projects, as well as the FRI Communications and Extension Program. More
specifically, this project fits under the existing “Disturbance Regimes of East Slopes of Alberta” project
approved by the FRI ND activity team this year (see attached Appendix A).
4. Methods
4.1. Rendering Mountain Legacy Photos into Spatial Products.
There are numerous potential methods by which the oblique photos can be made spatial. We intend to use
students from the Advance Diploma in GIS Program at the British Columbia Institute of Technology, or
students in the Masters in GIS Program at the University of Calgary, or students in the University of Victoria
Geomatics Program to conduct a complete literature, technology and software review to identify the most
pragmatic method by which to do this.
Contacts have been made with all departments, and there are promising leads developing. The most suitable
program will be engaged once this project is started.
All of these programs provide students with opportunities to learn new skills and to test emerging cutting
edge technological applications to real world problems. All students coming into these programs at BCIT or
University of Calgary have either undergraduate or graduate degrees, and are highly skilled to get into these
competitive programs. At the UofC and BCIT students devote one semester to studying the problem
(literature review, discussion, and meeting with professionals) and writing up their findings. They then
choose the best course of action to meet the study objectives (described in 4.3-4.5 below), and then spend a
10-12 week session working full time in executing the chosen solution, fully documenting their
methodology, and delivering the data to the project coordinator (Chris Stockdale).
The University of Victoria students are 3rd or 4th year undergraduates, and have training in both computer
science and geography. A co-op term lasts 16 weeks, and the student is employed full time.
The chosen provider is hereafter to referred to as the Geomatics provider.
4.2. Select a test landscape.
Previous known historic MPB outbreaks have occurred in the low elevation passes between BC and Alberta:
the Crowsnest Pass, and the Bow Corridor. In consultation with the Foothills Research Institute Mountain
Pine Beetle Ecology Program, and with Eric Higgs, one of these locations will be chosen to study. The
landscape in question must have historical and current (retaken) photographs spatially located with the point-
of-view (POV) for each photograph known. Further, a series of high quality historical photos must exist to
allow for accurate interpretation. Many high quality images exist, but there are some historical photographs
that were obscured by poor lighting or visibility. The potential size of the study area will depend completely
on 4.1 above, the method chosen for rendering these photos into spatial images.
4.3. Image Interpretation.
Images will likely be subsampled, as there are known issues with “stretch” when rendering oblique photos to
a planimetric (flat overhead) view. Once the images have been rendered into a format for interpretation the
chosen landscape will be interpreted and classified into categories of MPB risk consistent with the methods
used by Alberta SRD using the Shore Safranyik Risk Rating System using the Climate Factor developed by
Allan Carroll. Allan Carroll will be contacted to determine whether any modifications are needed to the
climate factor to account for the early 1900 AD time period involved.
While the ASRD method has been developed using AVI as its primary input, the historical images do not
have the resolution required to create accurate measures of stand density, age, or percent pine. Estimates of
these can be made using the repeat oblique image, and comparing the known current MPB risk of each
spatial unit of analysis. Rather than a linear function such as the SSI, we will break the rating system into
categories. From each image, the most detailed areas will be chosen for subsampling. Marie-Pierre Rogeau
will be consulted for advice to assist with intpretation.
4.4. Compare Past and Current Levels of Mountain Pine Beetle Susceptibility
The repeat images will be classified the same way as the historical photos to ensure consistency of
classification methods. The two time periods will then be compared to see what the overall changes in MPB
susceptibility over the past century have been.
4.5. Determine Whether Landscape Structure Has Been the Limiting Factor in MPB
Spread in the Past in Southern Alberta
It is suspected that past Alberta infestations were only possible due to immigration from BC. However, we
do not know if climate alone, or host availability also prevented MPB from becoming firmly established in
Alberta following these historical outbreaks. With historical and current landscapes classified to levels of
MPB susceptibility, we will call upon Bill Riel’s expertise to run landscape spread simulations of MPB for
the two time periods in question to evaluate the ease of epidemic spread relative to the modern day.
5.1. Schedule, Contingencies, and Project Deliverables
Project Initiation
Fall 2009
This project is connected to the East Slopes Disturbance Regime Project, FRI ND Program is willing to
engage the potential collaborators to confirm their interest and availability to help with this project. The PI
will discuss this project with BCIT and the University of Calgary to determine whether they are able to
participate in this study. In the event that neither of these programs has the resources to assist in project
delivery, the PI will discuss this project with Alberta Tourism, Parks, and Recreation, Alberta Sustainable
Resource Development and other agencies to determine if there is internal capacity within any of these
organizations to do this work. If the work cannot be done with in-kind contributions, costs will be
determined.
Deliverable:
The PI will report back to the MPB EP Activity Team with the findings of this portion of
the study and identify who (if any) will be providing Geomatics support. If no suitable
Geomatics provider can be engaged, this project will not proceed any further.
Engagement of GIS Support and Selection of Imagery
Fall 2009 –Spring 2010
Potentially as early as October 2009, contingent upon when the Geomatics provider needs to be engaged.
Deliverable:
Upon completion of items 3.1 and 3.2 an interim report will be prepared and submitted to
the Activity Team to show progress to date. A meeting will be held where the findings
and data from these sections will be described and demonstrated.
Image Interpretation and Analysis
Summer 2010 –Spring 2011
With successful completion of items 3.1 and 3.2 the interpretive and analytical work can begin.
Deliverables:
1) The findings of the project will be compiled in a final report and presented to the
MPB EP Activity Team.
2) 4 FRI Quicknotes will be released upon completion of the analysis.
3) Journal article(s) will be explored to document the findings of 4.1, 4.3, 4.4 and 4.5.
4) A presentation will be made to the FRI MPB EP Activity Team and to any other
interested parties.
6. Benefits and Desired Outcomes
This research program will allow managers to evaluate landscape management scenarios to identify tradeoffs
between trying to reduce current MPB susceptibility, and maintain the landscape within its demonstrated
historical equivalent.
Furthermore, the outputs created from step 6.1 will also be directly transferable for examining, among other
things: habitat issues, fire management issues, timber assessments, hydrological implications, and other land
management matters.
Anecdotally we know that forested areas were less prevalent, less dense, and less homogenous in structure
and composition. This translates into lower fire and MPB risk, greater biodiversity, greater ecosystem
resilience, and a broader range of aesthetic and recreational experiences.
Modern day management objectives to return portions of the landscape to their historical condition, and thus
lower these risks face significant challenges not limited to: 1) public perception that the landscape of today is
“natural” and has always been as it is, 2) not having quantitative targets to shift the landscape towards. This
pilot study will open up new methods of looking at past landscape condition and be of great benefit in
management, research and public communication for the forest and land management communities, and
geomatics profession.
7. Timelines
Fall 2009 –Identify test landscape, secure involvement of BCIT Geomatics Provider
Winter 2009/2010 –Students will conduct literature, software and technology review
Spring, 2010 –Students will begin image analysis and spatial rendering
Summer/Fall, 2010 –Image interpretation
Winter 2010/2011 –Modeling and compilation of results.
Spring 2011 –Project wrap-up, communication of results, completion of papers and articles.
Literature Cited
Andison, D.W. 2009. Foothills Research Institute Disturbance Dynamics Long-Term Research plan,.
Version 10.0. FRI Technical Report. Hinton, Alberta.
BRITISH COLUMBIA INSTITUTE OF TECHNOLOGY
APPENDIX B
INSTRUCTION MANUAL
For Georeferencing Oblique Terrestrial Photographs from the Mountain Legacy Project
Oliver Clovis and Katelyn Loukes
This document details the process of georeferencing matched historical/recent day terrestrial oblique photographs from the
Mountain Legacy Project www.mountainlegacy.ca. Completed as part of a research project sponsored by the Foothills Research
Institute and supervised by Chris Stockdale from Burning Ecologic Consulting.
2
Contents
Introduction
............................... ................................ ................................ ................................ ................................ .............
4
Data: ............................... ................................ ................................ ................................ ................................ .....................
4
Software: ............................... ................................ ................................ ................................ ................................ ..............
4
Directory Structure:
............................... ................................ ................................ ................................ .............................
4
Naming
Conventions: ............................... ................................ ................................ ................................ ...........................
5
Part A. Crop Photographs and Calculate
Parameters ............................... ................................ ................................ ..............
6
1. Cropping the repeat
photograph ............................... ................................ ................................ .................................
6
2. Calculate the center pixel location
............................... ................................ ................................ ..............................
6
3. Calculate the Virtual Camera
Parameters............................... ................................ ................................ ....................
6
4. Mark the Center point and Ground Control Points
............................... ................................ ................................ .....
7
5. Convert 1band image to 3
bands ............................... ................................ ................................ .................................
8
6. Crop the Original Photo
............................... ................................ ................................ ................................ ...............
8
7. Display Cropped Original with Repeat Photo Center and
GCP’s............................... ................................ ..................
9
8. Determine the UTM coordinates of the Camera
Station............................... ................................ .............................
9
Part B. Determine the UTM Coordinates of Center point and Ground Control Points
............................... .........................
10
1. Generate an
ArcScene............................... ................................ ................................ ................................ ................
10
2. Collect ground control points
............................... ................................ ................................ ................................ ....
10
3. Generate the GCP file
............................... ................................ ................................ ................................ ................
11
Part C. Create Settings
File ............................... ................................ ................................ ................................ ....................
12
1. Load input
parameters ............................... ................................ ................................ ................................ ...............
12
2. Load file
paths ............................... ................................ ................................ ................................ ........................... .
12
3. Save settings
............................... ................................ ................................ ................................ ..............................
12
Part D. Run the Application Preview the Ground Control Points
............................... ................................ ..........................
13
1. Preview the Ground Control
Points ............................... ................................ ................................ ...........................
13
2. Adjust
parameters............................... ................................ ................................ ................................ ......................
13
3. Run
............................... ................................ ................................ ................................ ................................ .............
14
4. Examine result in ArcMap and ArcScene
............................... ................................ ................................ ...................
14
5. Move the successful result
............................... ................................ ................................ ................................ ........
14
6. Clip result for display and convert to
kml ............................... ................................ ................................ ..................
14
Part E. Error
Analysis ............................... ................................ ................................ ..............................................................
14
1. Create 2 Shapefile Point Feature
Classes............................... ................................ ................................ ...................
16
2. Create shapefile points at control
locations ............................... ................................ ................................ ..............
16
3
3. Compare the distances between the photo points and the corresponding satellite
points ............................... .....
16
Appendix A. Preparing the
Viewsheds ............................... ................................ ................................ ................................ ...
17
1. Create a Viewshed
............................... ................................ ................................ ................................ .....................
17
2. Analyze the
Viewshed ............................... ................................ ................................ ................................ ................
17
Appendix B. Making a Shapefile from your Camera
Point............................... ................................ ................................ .....
18
Appendix C. Selecting a DEM and Satellite/Aerial photograph
............................... ................................ ............................
19
Generating a DEM Mosaic
............................... ................................ ................................ ................................ .................
19
Clipping a DEM in ArcMap/ArcCatalog
............................... ................................ ................................ ..............................
19
Appendix D. Filtering Historical
photo ............................... ................................ ................................ ................................ ..
20
Appendix E. Locating Ground Control and Center Points in Google Earth
............................... ................................ ............
21
Appendix F. ArcGIS Model Diagrams
............................... ................................ ................................ ................................ ....
22
4
Introduction
Data:
10 Meter DEM of study area obtained by resampling a 1m lidar DEM (DEM10m.tif)
Original Photographs
Repeat Photographs
10m resolution Spot Satellite Photograph of study area obtained from geobase.ca (SattelitePhoto.tif)
Software:
ArcGIS Desktop 9.3.1
ArcInfo License (ArcScene, ArcMap and ArcCatalog) with Spatial Analyst and 3D Analyst Extensions.
Note: Some of the methodology which utilizes ArcScene and the ArcInfo license may be duplicated using GoogleEarth pro.
PCI Geomatica Focus
Version 10.2, 2009
Cost = $2,950.00.
Available at: http://store.pcigeomatics.com/product_p/geo-dt.htm)
Possibly substitute with Adobe Photoshop or ENVI/ERDAS products
Notepad++
Cost = Free
Available at http://notepad-plus.sourceforge.net/uk/download.php
Javier Corrpio’s Georeferencing Application
Cost = Free
Included in Deliverables as Photogeoref.SAV.
IDL virtual machine
Cost = Free
Comes with the IDL trial version download http://www.ittvis.com/ProductServices/IDL/VirtualMachine.aspx)
Directory Structure:
Under the Deliverables Folder we have five main folders (Camera Stations,
Parameter Calculator, KML to SHP converter, DEM and Viewsheds) one
toolbox (GeoRefProject.tbx), the Corripio georeferencing application
(photogeoref.sav) and 1 tiff file (SatellitePhoto.tif).
SatellitePhoto.tif Is a 10m resolution orthophoto of the study area. We
used this to check the accuracy of our results and assist in recreating the
virtual camera view.
GeoRefProject.tbx Contains 3 ArcGIS models (Viewshed, ViewFromPhoto
and Error). Viewshed was used to generate viewsheds using the DEM, a
Camera Location point shapefile and a user defined height above the DEM.
ViewFromPhoto creates a polygon outline of the viewshed for use in
clipping the black background out of the final georeferenced photo. Error
was used to compare points on the georeferenced photo to points on the
satellite photo. To run these models, double click on them in ArcCatalog.
Parameter Calculator: Contains an application for calculating inputs used in Corripio’s Georeferencing Application. We have included
both a compiled version as well as the source code.
Error Analysis: Contains 2 directories (Jumping Pound and Moose Mountain Center). In each directory, we have saved 2 sets of
shapefiles per photo. One set contains points from the Satellite photo and the other contains corresponding points taken from the
georeferenced photo. For each photo, there are 2 .dbf files, dist and point (can be opened in excel) which contain comparisons
between corresponding distances and locations on the satellite photo and the georeferenced photo. In Jumping Pound there is an
5
.xlsx file containing a summary of the jumping pound error measures. As well, there is a .dbf file for each photo called
Sat_point_DistanceFromCameraloc which contains the distance of the satellite error points from the camera location. In Jumping
Pound there is an .xlsx file containing a summary of the jumping pound error measures.
KML to SHP converter Contains an application which can be used to convert kml files to shapefiles. This is part of an alternate
methodology which substitutes Google Earth for ArcScene (see Appendix E).
Viewsheds contains 1m, 10m and 20m viewsheds from each of the camera stations. For some photos, in order to produce
viewsheds which closely resemble the landscape seen in the image, the height of the camera above the ground was modified.
Naming conventions for viewshed are explained below.
Camera Stations contains 4 subdirectories, one for each camera station (Jumping Pound 1, Jumping pound 2, Moose Mountain
Center and Moose Mountain West). Each of these 4 directories contains a photo directory for each photograph taken at that
camera station.
Photo directories (eg. Jp10-4) contain 3 sub directories (Input, Output and Vector Viewsheds)
Input contains the photo specific files and information needed for input into the georeferencing application. This
includes a .set settings file can be uploaded to the georeferencing application to automatically enter in the saved
settings/paths, a .dat file containing the ground control point coordinates, two cropped photographs (recent and
original) in tiff format ending in H08 (retake) and W96 (original) and one text file ending in “_info” which contains
saved metadata and calculations pertaining to the photograph.
Output contains the georeferenced retake and original photographs (.tif, ending in H08, W96), clipped versions of
them which have been modified to exclude the black background (.tif, ending in _clip) and kml versions of the
clipped photographs.
Vector Viewsheds Contain the polygon shapefile version of the viewshed which was used to clip the black
background out of the georeferenced output.
Naming Conventions:
Images
Example: JP9_4H08.tif
The first letters indicate the photo site
JP...........Jumping Pound
MMC.........Moose Mountain Center
MMW........Moose Mountain West
The next two numeric values (separated by a dash) represent the photo identification eg (9-4)
The last characters indicate what the file contains:
H08 ..........Cropped retake photograph
W96..........Cropped original photograph
GCP..........DAT file containing the ground control points
Settings.... Saved parameters for the georeferencing application
Info...........Text file containing all of the parameters, calculations and additional information required to
georeference the photo (metadata).
Viewsheds
Example: JP1_10m_2
The first 3 values are the station site (as opposed to the photo site)
The next 3 digits indicate the DEM used (1m, 10m or 20m)
The last number indicates the height of the camera above the DEM the default we used was 2m
6
Part A. Crop Photographs and Calculate Parameters
In the following sections we will be using the 15-4 photograph from jumping pound station 1 as an example
1. Cropping the repeat photograph
a) Open up PCI Focus
b) On the top menu bar navigate to Tools > Clipping/Subsetting
c) On the popup window select the browse button and navigate
to the folder containing the repeat photographs from jumping
pound (not included in our data pack) C:\--Jumping
Pound\Repeat Masters\HIGG2008_W96-15-4.tif. You may
want to first create backup copies of your repeat and original
photos before you start working with them.
d) Click the checkbox next to Rasters
e) Drag the top of the red border/frame surrounding the image
down to the level at which you wish to crop the photo (see
Figure 1.0).
f) Open Notepad and save the blank file as C:\--
\Deliverables\Camera Stations\Jumping Pound1 \Jp15-4_Info.txt
Figure 1.0 Clipping the Photograph
g) In the text file record the line value (L) in the right text box under Upper Left (see right – 1549L)
h) In the same text file record both numeric values directly above the photo next to Output Raster Size (see below). These are
the number of pixel columns (P) and number of lines (L) in the cropped photograph (5612P 2339L). Save the text file.
i) Now in the bottom left output section, click the Browse button and in the pop up window navigate to C:\--
\Deliverables\Camera Stations\JumpingPound1>Jp15-4>Input and type in Jp15-4H08_Crop as the file name. In the Files of
Type drop down list under the file name select TIFF 6.0(.tif) and click the Save button.
j) Click the Clip button
2. Calculate the center pixel location
a) Open up Jp15-4_Info.txt in Notepad if it is not already open
b) Divide the number of pixel columns from 1-h by 2: 5612P/2 = 2806P
c) Divide the number of pixel lines from 1-h by 2: 2339L/2 = 1169.5L
3. Calculate the Virtual Camera Parameters
Since the images have been cropped the factory specifications for the camera parameters cannot be used. Instead the cameras
sensor width, height and resolution must be scaled before any calculations are done. The original pixel columns and lines in the
photographs taken with the Hasselblad H3D Camera before being down sampled and cropped are 7212 pixels, 5412 lines. The
original sensor/film width and height is: w= 0.049m, height =0.0367m. You will use these values and the pixel and line values from
step 1-h to perform the following calculations.
***We have written an Application which automatically completes these calculations using the number of pixel rows and columns
supplied by the user (see C:/-/deliverables/ParameterCalculator).***
a) Open up Jp15-4_Info.txt in Notepad if it is not already open.
b) Calculate a ratio of reduction for the photo width and height
i. Width Ratio= 5312/7212 = 0.7366
ii. Height Ratio= 2339/5591 = 0.4322
c) Multiply the height and width ratios or reduction by the actual sensor height and width to obtain the height and width of
the virtual sensor/film:
7
Virtual sensor width
=0.7366(0.049)
=0.0361m
Virtual sensor height
=0.4183509211(0.0367)
=0.0159m
d) Record the virtual sensor width and height values in the Jp15-4_Info.txt file
e) Calculate the x and y resolution. X: divide the virtual sensor width from 2- by the number of pixel columns from 1-g Y: divide
the virtual sensor height from 2- by the number of pixel lines from 1-g. The smallest result will be the resolution used in the
application.
X resolution
2339/0.0159 /100= 1475 dpcm
Y resolution
5312/0.0361/100= 1471 dpcm
f) Record the smallest resolution in Jp15-4_Info.txt (Y resolution in this example)
4. Mark the Center point and Ground Control Points
You may also be able to use Photoshop; the main idea is to have a visible marker on the center pixel(s) of the photograph so that you
can detect that location on your 3D model.
a) Open PCI focus and click on the Launch the Add Layer Wizard icon on the top menu
b) Select RGB on the first pop up screen and click Next
c) Select Browse and navigate to the repeat photograph directory C:\--\Deliverables\Camera
Stations\JumpingPound1\Jp15-4\Input\Jp15-4H08Crop.tif and click Open
d) Now in the channels available box click on the first channel (1 [8u] B 1556) , the channel number should appear in the red
box
e) Now click on the second channel (2 [8u] B 1556) and then the third channel (3 [8u] B 1556), filling up the green and blue
boxes
f) Click Finish
g) Right click on NewArea in the table of contents and select New Vector Layer > Point
h) Now select the Cursor Control icon on the top menu and input the center point coordinates which you calculated in
the A2(b) and A2(c).
i) Once the cursor has moved to the center point, zoom in until you can see the individual pixels. Now Click on New Shapes >
Point and click on the center point. Make the point is a color which is easy to see (such as red) and save the vector layer as
a shapefile (Right click on vector layer>save) in Jumping Pound>Site1>Jp15-4 >CenterGCP Reference>CenterGCP and then
save the whole display as a .grp project file (file>save)C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-
4\CenterGCP Reference\CenterGCP.grp
j) Now insert points (preferably in a different color) at easily recognizable locations (such as mountain peaks and river bends).
You will use these as Ground Control Points
Also: When georeferencing photo pairs, it is helpful to find the center points of both photos first to assess how closely the photos
match. If the center points match, then identical input parameters may be used in the software.
8
5. Convert 1band image to 3 bands
The Corripio Application requires a 3 band image as input. This is not a problem with the color images as they have 3 bands (Red,
Green and Blue). However in order to generate a 3 band black and white image, we must copy our black and white band into an
existing color image, overwriting the 3 bands.
a) Open Windows Explorer and copy the uncropped color repeat photo that
matches the 1 band greyscale original which you are converting and
paste it into a new folder with a recognizable name (eg. Template.tif).
You will be replacing the information in this file with the information from
the original photograph (using it as a template).
b) Open PCI focus and click on the Launch the Add Layer Wizard icon
on the top menu
c) Select Grayscale on the first pop up screen and click Next
d) Select Browse and navigate to the directory containing your one band
original photographs C:\--\ Jumping Pound Summit\Original
Masters\W96-15-4.tif and click Open
e) Select the one channel in the Channels Available box (1 [8u] contents not specified) and click Finish
f) Click on the Files tab (right tab at the top of the table of contents next to Figure 5.0
the map tab) and select the plus sign next to Rasters beneath your W96-15-4.tif file path.
g) Right click on the image layer and select Export (save as)>to Existing File (see Figure 5.0).
h) Click on Browse next to Source File and navigate to your one band
original image C:\--\ Jumping Pound Summit\Original Masters\W96-
15-4.tif (see Figure 5.1).
i) Click Browse next to Destination file and navigate to your 3 band
color template file (Template.tif)
j) In the Source Layers box select the one layer (1 [8u] contents not
specified)
k) Now in the Destination Layers box click on the first layer (1 [8u] B
1556). Now select Overwrite and then click on Transfer layer.
l) Repeat this by again selecting the layer in the Source Layers box (1
[8u] contents not specified) and then clicking on the second layer in
the Destination Layers box (2 [8u] B 1556) and then selecting
Overwrite and Transfer layer.
m) Repeat with the third layer in the Source Layers box (3 [8u] B 1556).
n) Open windows explorer, right click on template.tiff and select rename
o) Rename the file JP15-4_W96_Uncropped.tif
Figure 5.1
p) Cut the file (Right click>Cut) and paste it into the C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input
directory
6. Crop the Original Photo
In order to use the same in put parameters for both the original and repeat photo the two photos must be cropped to exactly the
same dimensions.
a) Open up PCI Focus
b) On the top menu bar navigate to Tools > Clipping/Subsetting
c) On the popup window select the browse button and navigate to the folder containing the original repeat photo C:\--
\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\ JP15-4_W96_Uncropped.tif
d) Click the checkbox next to Rasters
e) In the Upper left text box type in the Line value from step 1-g (1549L)
f) Now in the bottom left output section, click the Browse button and in the pop up window navigate to Jumping C:\--
\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input and type in Jp15-4_W96 as the file name. In the Files of
Type drop down list under the file name select TIFF 6.0(.tif) and click the Save button.
9
g) Click the Clip button
h) Delete JP15-4_W96_Uncropped.tif
7. Display Cropped Original with Repeat Photo Center and GCP’s
This step is to determine whether the original and repeat photo is similar enough to use the same input parameters.
a) Open Notepad and load C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\ Jp15-4_Info.txt
b) Open up PCI Focus and Select File>Open and navigate to C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\
CenterGCP Reference\ CenterGCP.grp
c) Click on the Launch the Add Layer Wizard icon on the top menu
d) Select RGB on the first pop up screen and click Next
e) Select Browse and navigate to C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\ Jp15-4_W96.tif and click
Open
f) Now in the channels available box click on the first channel (1 [8u] contents not specified), the channel number should
appear in the red text box.
g) Now click on the second channel (2 [[8u] contents not specified) and then the third channel (3 [8u] contents not specified),
filling up the green and blue text boxes.
h) Click Finish
i) Assess whether the center point is located in the same location on the original as it is in the repeat
j) If it is drastically different, then repeat steps 2-4 with the original cropped photograph, creating a separate text file for the
original photograph parameters C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\ JP15-4_W96_info.txt
k) If the center point is in similar locations on both photographs, then use the parameters from the repeat photo in the
georeferencing application.
8. Determine the UTM coordinates of the Camera Station
The UTM coordinates of the camera positions from the field notes can be used however it is best to use a the value
collected from the DEM for the elevation.
a) Open up ArcMap
b) Open up C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\ JP15-4 _info.txt in Notepad
c) In ArcMap Click on the Add Data icon on the top menu and browse to the 10m DEM (C:--
\Deliverables\DEMs\10mDEM.tif) and click Add
d) In ArcMap Click on the Add Data icon and browse to the camera station shapefile (C:\--\Deliverables\ Jumping Pound
1\JPstation1) and click Add. See Appendices A and E for information about generating shapefiles using either ArcCatalog or
Google Earth)
e) In ArcMap Click on the Identify Tool on the Tools toolbar (if not already on display View>Toolbars>Tools)
f) In ArcMap on the Identify tool pop up window, select the 10MDEM layer on the Identify From drop down list
g) In ArcMap Click on the JpStation1 shapefile point on the map display
h) In JP15-4 _info.txt, record the X, Y, Z coordinates that show up in the Identify popup window (this is the camera station
location).
Part B. Determine the UTM Coordinates of Center point and Ground Control Points
***note: The Geographic Coordinates of the Center Points and the Ground Control Points could be located in GoogleEarth by
navigating to the camera view and picking and recording the point locations using the cursor (or putting in place marks). See
Appendix E for a suggested methodology. ***
1. Generate an ArcScene
In order to properly assess the accuracy of many of the steps in this process it is useful to have an ArcScene which closely
approximates the view in each image.
a) Open ArcScene, add your 10 m DEM to the scene (add data , browse, navigate to C:\--
\Deliverables\DEMs\10MDEM.tif)
b) Now add SatellitePhoto.tif to the scene (add data, browse C:\--\Deliverables\SatellitePhoto.tif)
c) Add a shapefile point of the camera location to the scene (add data,
browse, C:\--\Deliverables\ Jumping Pound 1\JPstation1)
d) For each layer, right click on it in the table of contents, select
Properties and in the properties menu click on the Base Heights tab.
e) Click on Obtain heights for layer from surface: and select
10mDEM.tif from the drop down menu.
f) Click on Raster Resolution and set the x and y to the cell size of your
DEM (x=10, y=10) (see Figure 1.0).
g) To navigate to your camera location, Right click on the JPstation1
shapefile layer in the table of contents and select zoom to layer.
h) Change your viewing direction using the Fly tool (bird icon , top
menu). Click once and then scroll around and then press escape when
you are facing the direction you want – don't click a second time.
Figure 1.0
i) To adjust your elevation (most photos are not taken at ground level), change your view until you are facing the ground and
then zoom out.
j) Continually scroll/adjust your view until it matches up with the oblique photo to be georeferenced. Once you are happy
with the view, go to bookmarks>save bookmark and name your location Jp15-4_cameraview. Name the scene C:\--
\Deliverables\Camera Stations\JumpingPound1\ CameraView.sxd
This can be extremely frustrating and time consuming at first, but it gets easier as you become accustomed to ArcScene's navigation
tools.
As well, sometimes ArcScene will crash if you set more than two layers to base heights. If this is the case, then add the satellite photo
and station shapefile first (not the DEM) and then set their base heights to the 10M DEM and zoom to the station shapefile layer.
Bookmark your camera view and then remove the shapefile layer and add the DEM (setting its base heights)
2. Collect ground control points
Using the marked photograph from A-4 and the ArcScene view,
select and record the coordinates of points on the scene which
correspond to the marked center and ground control points on the
image.
a) Open up the PCI Focus project containing the photo and
center point vector. Open PCI Focus, select file>open and
then navigate to C:\--\Deliverables\Camera
Stations\JumpingPound1\Jp15-4\CenterGCP
10 Figure 2.0
11
Reference\CenterGCP.grp
b) Now open up ArcScene to your saved camera view scene C:\--\Deliverables\Camera
Stations\JumpingPound1\CameraView.sxd and select bookmarks>Jp15-4_cameraview.
c) In Notepad, open up C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\JP15-4_W96_info.txt to record your
coordinates. Try to position both of the windows so that you can view both at the same time (see Figure 2.0).
d) In ArcScene make sure your 10MDEM layer is on the top of the table of contents list (you can click and drag it to the top
with your mouse).
e) In ArcScene right click on the 10MDEM layer and choose Properties. In the properties menu, click on the Display tab and
input 100 into the transparency text box. This will allow you to select points with elevation values from the DEM, while still
being able to view the draped satellite photo.
f) Now, using your PCI Focus image as a reference, try to identify the point in your ArcScene camera view that corresponds
with your center point mark.
g) Click on the identify tool in ArcScene to select the point that matches the center point mark in PCI Focus. Record the X, Y, Z
coordinates in JP15-4_W96_info.txt (already open in Notepad).
Also: you may want to make a list of potential center point coordinates while you have both the files open. It often takes a lot of trial
and error before the best point is found. The more uncertain you are about which point in your 3D model is the center, the more
coordinates you should record.
3. Generate the GCP file
Corripio’s Georeferencing Application requires a .dat file in csv format containing at least one ground control point. Ground control
points are not used to georeference the photo, but as an accuracy check. The software has a preview option which maps only the
ground control points, rather than the whole DEM onto the photo. This lets you assess whether you need to change your parameters
before proceeding. The Corripio Georeferencing Application was written on a MAC which uses a different symbol to indicate the end
of a line. If you are running windows then you will need to use the “convert to MAC format” option of notepad++.
a) While you still have PCI Focus and your ArcScene camera view open, open Notepad ++ and select Format>Convert to MAC
format.
b) Now look for points on the PCI Focus display photo which you can identify in your 3D scene.
c) Using the identify tool in ArcScene, record the UTM coordinates of these points (X, Y, Z) and a corresponding name in your
notepad document. You can also make a new vector layer in Focus, to record where you have collected your ground control
points.
d) The format of your text file should be:
xcoord,ycoord,zcoord,labelA
xcoord2,ycoord2,zcoord2,labelB
e) Now save your file in C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\Jp15-4_GCP.DAT
Troubleshooting: if you receive an error when running the application (i.e. This file is not a valid ASCII file) add a blank line at the
end of the file.
12
Part C. Create Settings File
1. Load input parameters
a) In windows explorer Open up Deliverables folder and right click on Photogeoref.sav, select open with>Idl.
b) If you already have a .SET file then click on Load Settings and navigate to the .set file
c) If you do not already have a .set file then input the data as follows from the information text file C:\--\Deliverables\Camera
Stations\JumpingPound1\Jp15-4\Input\JP15-4_W96_info.txt :
i. Input in the camera station coordinate position in UTM X,Y,Z (whole numbers only, no decimals) See step A-8-h
ii. Input the target (center point) coordinate position in UTM in X,Y,Z ( no decimals) See step B-2-g
iii. Input the focal length in meters (0.03582)
iv. Input the resolution in dpcm See step A-3-f
v. Input the virtual camera film/sensor height and width in meters See step A-3(d)
2. Load file paths
Instead of manually typing in the paths, you can leave them blank, click on load data and navigate to each file using the
pop up windows (look at the top of each pop up to see which file it wants you to select).
OR, you can manually type in the file paths:
a) DEM (C:\--\Deliverables\DEMs\10MDEM.tif),
b) Image (C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\JP15-4_H08.tif)
c) visibility (C:\--\Deliverables\Viewsheds\jp1_10m_2.tif)
d) GCP (C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\Jp15-4_GCP.dat) files
3. Save settings
If you have just loaded your data
manually and/or made changes,
you should click on Save Settings
to save your own .SET file.
Sometimes this file does not save
properly; in which case, you can
modify an existing .set file by right
clicking on it and opening it in
Notepad++ (see Figure 3.0).
Figure 3.0 – This is the properly aligned .set file
If the settings file is not loading properly,
then open it and take a look. Sometimes
the paths get saved as 4 lines, instead of 2
and will not work (see Figure 3.1). To fix
this, place the DEM and image paths on
the same line separated by a comma and
ending with “, nan” the viewshed and GCP
paths on the same line separated by a
comma and ending with “, nan”. You
should end up with two lines of comma
separated paths, each ending with “, nan”.
Figure 3.1 – This .set file is improperly aligned
13
Part D. Run the Application Preview the Ground Control Points
1. Preview the Ground Control Points
After your data is loaded, rather than running the whole program and finding out that your parameters are incorrect, you can click on
“Preview GCP's”. This shows where the GCP points you picked from the ArcScene (or collected in the field) are expected to map onto
the photograph. The GCP points are displayed as white crosses and the exact center of the photograph is displayed as an orange
cross.
a) Click preview GCPs, an image of your photo should show up in the
view screen (see Figure 1.0)
b) Click Zoom
c) Scroll around and see whether the white crosses are in the correct
locations
2. Adjust parameters
During the preview the GCPs and target coordinates are displayed in their
absolute positions (in UTM coordinates) however only the target point is linked to
the photo as you know from Section B-2.
Figure 1.0
If the GCPs do not appear in the correct positions on the photo, altering the coordinates of the target will change the perspective of
the scene slightly, giving the effect of moving the GCPs on the photo. Since the GCPs are in places you’ve picked to be readily
identifiable whereas the target point is predetermined by merit of being in the centre of the photo, the target point likely has the
most error associated with it and so moving it will not affect the
accuracy of the methodology. You can view the changed point
in ArcMap or ArcScene to ensure it is still in a valid area.
Figure 2.0 – Shows an overhead look of the Corripio Preview layout.
While the Camera, target and GCPs can be visualized as being in
their 3D UTM coordinates the photo is still a 2D plane linked to the
target.
Figure 2.1 – Shows the effect of moving the target coordinates.
Because the perspective has changed the GCPs appear in a
different position on the yellow image than on the blue image
14
3. Run
a) When satisfied with the position of the GCPs, click on Run. This part typically takes a really long time, after this part of the
program has run you will see an image of the DEM points in red, mapped onto the image. You can zoom in and examine how
the DEM has been rotated and scaled to fit.
b) If the fit looks reasonable, click on Accept to finish the georeferencing process. If not, click on Start Again. You will now see
a display of the final georeferenced image (already saved) overlaid onto a black background with contour line).
The intermediate image with the red DEM points mapped on is automatically saved into the same directory as the image file and
named as <image name>coimg.tiff. The final georeferenced image is also in this directory but named as <image name>ref.tiff.
4. Examine result in ArcMap and ArcScene
a) Open up ArcScene and load C:\--\Deliverables\Camera Stations\JumpingPound1\ CameraView.sxd.
b) Click the Add Data icon and navigate to C:\--\Deliverables\Camera
Stations\JumpingPound1\Jp15-4\Input\JP15-4_H08ref.tif.
c) Right click on the JP15-4_H08ref.tif layer and choose properties.
Select the Base Heights tab and set the base heights to the
10MDEM.
d) Click on the display tab and choose Cubic Convolution as the
resampling method (default is set to nearest neighbour).
e) Compare the draped georeferenced image layer to the cropped
photo (see Figure 4.0).
f) Open up ArcMap and load C:\--\Deliverables\Camera
Stations\JumpingPound1\ CameraView.sxd and the Satellite
photo. Compare landmarks such as rivers borders and mountain
ridges to assess the fit.
5. Move the successful result
Figure 4.0
It is important to move or rename the photo; otherwise it is possible the application will save over it on a subsequent trial
a) In windows explorer navigate to C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Input\JP15-4_H08ref.tif
b) Cut JP15-4_H08ref.tif and paste it into to C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Output
6. Clip result for display and convert to kml
Because the application returns an image that is the full extent of the DEM it is desirable to clip out a large area of No Data values
which are displayed in black. This will also make the kml more presentable in Google Earth.
a) Open up ArcCatalog or ArcMap and click on the Toolbox icon to load the ArcToolbox menu.
b) In ArcToolbox, Navigate to Data Management Tools>Raster>Raster Processing>Clip
c) In the popup dialogue (see Figure 6.0) click on the folder icon next to Input Raster and navigate to your georeferenced
result C:\--\Deliverables\Camera Stations\JumpingPound1\Jp15-4\Output\Jp15-4_H08ref.tif
d) Click the folder icon next to Output Extent and navigate to your 10m viewshed polygon C:\--\Deliverables\Camera
Stations\Jumping Pound1\JP15_4\JP15_4\Vector Viewsheds\jp15_4_10m.shp
e) Check the Use Input Features for Clipping Geometry box
f) Click the folder icon next to Output Raster Dataset and navigate to C:\--\Deliverables\Camera
Stations\JumpingPound1\Jp15-4\Output and name the file Jp15-4H08.tif.
The clip tool will save the output as .img by default, so you must delete the .img extension in the Output Raster Dataset text
box and replace it with .tif.
15
g) Now open up ArcMap and add Jp15-4H08.tif to the Display.
h) Open up the Convert Layer to KML tool located in ArcToolbox Tools >Conversion Tools>To KML>Layer to KML.
Now select your Jp15-4 layer in the layer drop down menu and input 1 into the Layer Output Scale (you use the default output
name).Click ok
16
Part E. Error Analysis
1. Create 2 Shapefile Point Feature Classes
a) Create two new Shapefile point feature classes (see Appendices A and E). Name one Jp15-4SatPoint and the other Jp15-
4PhotoPoint.
2. Create shapefile points at control locations
a) Add your georeferenced photograph and the satellite photograph to an ArcMap display
b) Start an Edit Session in ArcMap. View>Toolbars>Editor from Editor Toolbar: Editor>Start Edit Session
c) On the editor toolbar click the pencil icon (Sketch Tool), make sure “Create New Feature” is listed in the Task drop down
menu and “Jp15-4PhotoPoint” is listed in the Target drop down menu.
d) Looking at the georeferenced photograph, Pick out unique landmarks (such as cut block corners and river bends) which can
also be identified on the satellite image.
e) Add points to these locations by clicking on them with your mouse, noting the order in which points are created
f) When complete Switch the Task drop down menu from “Jp15-4PhotoPoint” to “Jp15-4SatPoint”
g) Repeat steps d and e using the satellite photo instead of the georeference image. Make sure that the points are created in
the same order as the points in Jp15-4PhotoPoint.
3. Compare the distances between the photo points and the corresponding satellite points
a) Open up the ErrorAnalysis Tool using ArcCatalog. C:\-\Deliverables\GeoRefProject\ErrorAnalysis
b) Now click browse next to the Photo Collected Points and navigate to Jp15-4PhotoPoint
c) Click browse next to Control Collected Points and navigate to Jp15-4SatPoint.
d) Chose locations and names for the point to point and distance to distance comparison tables. The point-to-point
Comparison Table will generate a table which shows the distance between the Satellite and georeferenced photo control
points in the Distance column. The distance to distance Comparison table shows the distance between every possible pair of
points within each of the shapefiles. The Dist-Dif column shows the distance difference between corresponding point pairs
in the Satellite and Georeferenced photo shapefiles (eg. compares distance A-B in Satellite photo to corresponding distance
A-B in Georeferenced Photo .
e) Click on Show Help for details about the inputs and outputs
f) Click OK to run the model
Now repeat Part D using the Original photograph (Jp15-4_W96.tif) as the image input if you are using the same parameters. If
you are using different parameters for the original photo, go back to step A-3 and modify the W96 info file and precede to steps
B, C and D.
17
Appendix A. Preparing the Viewsheds
In order to account for the height of the camera above the ground the viewshed tool in the ArcToolbox requires you to add a field to
the camera point called OFFSETA with a value representing the height to add to the height value from the DEM at that point. For
example if the tripod was set at 2m the value in the OFFSETA column would be 2. Since creating and editing this column is tedious we
created a model which adds the column and populates it.
Note: You may want to consult the field notes for the station to see if any note about the tripod elevation has been made.
1. Create a Viewshed
a) Open ArcCatalog
b) In the Table of contents navigate to the GeoRefProject toolbox. C:\--\Deliverables\GeoRefProject.tbx
c) Double click the Viewshed model.
d) In the pop up dialog click on the folder icon next to Observer Point and navigate to your camera point shapefile C:\--
\Deliverables\Camera Stations\JumpingPound1\JPStation1.shp
e) Next enter the height of the camera above the DEM in the Height Above DEM textbox. The textbox will default to 2.
f) Now click on the folder icon next to DEM and navigate to the DEM you want to use C:\--\Deliverables\DEM\10mDEM.tif.
g) Lastly, specify an output location for the viewshed. Click on the folder icon next to Viewshed(output) and navigate to the
folder where you will store the viewshed C:\--\Deliverables\Viewsheds name the viewshed JP1_10m_x where x is the
height you entered in step e).
2. Analyze the Viewshed
a) Open ArcScene
b) Load the viewshed, the satellite photo and the camera station and set their Base Heights as in Part B 1.
c) Orient the view to look in the same direction as the photo as in Part B 1.
The viewshed should appear as a solid colour. Using the pan tool grab near
the top of the image and pull down. You should see gaps open up between pieces
of the viewshed, these gaps represent non-viewable areas and their size is relative
to the distance between the top of the ridge and the lowest visible part of the next
ridge. View the photo in any program (PCI Focus or Windows Picture viewer). The
outline of the pieces of the viewshed which are viewable in the ArcScene should
correspond in relative size and shape to features such as ridges and mountain
peaks that are visible in the photograph.
Some feature may not appear if they are smaller that the cell size of the DEM used.
If more is visible in the viewshed than in the photo then Observer Point may have
been set too high. Alternatively, if the viewshed shows less than is seen in the photo the Observer Point may have been set too
low. If one side of the viewshed appears to show too much and the other not enough you will need to move the location of the
camera point parallel to the to the orientation of the photo (left or right as you look at the photo).
18
Appendix B. Making a Shapefile from your Camera Point
When deciding what your study area extent is going to be, you need to be able to see where the photo was taken from.
a) Open up ArcCatalog; right click the folder where you want to keep your shapefiles. Select New>Shapefile to make a new
shapefile. Name it and click on Edit underneath the Description box
b) Now you can either import a coordinate system (click on import and select a file, such as the DEM which already has the
coordinate system defined)
c) To do it manually, click Select > Projected Coordinate Systems > UTM > NAD1983 >NAD 1983 UTM Zone 11N.prj
d) Click OK twice, you should now see your new shapefile
e) Open up ArcMap and add your shapefile point (add layer button)
f) Start an edit session View>Toolbars>Editor>Start Editing
g) Make a new point by selecting the sketch tool (pencil) and then right clicking on the blank map display and selecting
absolute x, y from the context menu. Enter in the UTM coordinates of your camera location, click to set down a point and
then save your edits.
You can also enter your XY UTM coordinates into an excel file and convert it to a shapefile
a) Open ArcCatalog, click on the excel file and select a worksheet
b) Right click on the worksheet and select Create Feature Class >From XY Table
19
Appendix C. Selecting a DEM and Satellite/Aerial photograph
a) Use the Camera station UTM coordinates (found in the field notes) and Google Earth to get a general idea of where your
study site is
b) Decide on what area you will need to cover
c) We recommend first generating a 3D camera view model in ArcScene, selecting the peripheral point locations using the
investigate tool and then creating a shapefile polygon from these points (see part A steps 1 and 2).
d) Clip and/or mosaic DEM's until you arrive at a suitable site extent
If you do not already possess a DEM, they are available from www.geobase.org. However, more detailed DEMs will yield better
results and the center point selection process will be quicker and easier. If a very large area is to be covered then very fine DEM’s 1-
5m may crash the georeferencing application. You will probably need to mosaic more DEM's together to increase coverage and then
clip excess portions out. If you have recorded the orientation (direction) in which the photo was taken, you can immediately narrow
down your coverage by excluding the portions of the DEM which lie behind the photographer.
Generating a DEM Mosaic
DEMs are often very large which makes them difficult to use. Often they are stored in small sections so that you need only load the
portions needed. In our case we need a single DEM which covers a large area so we need to mosaic these sections together.
a) Open ArcCatalog
b) Navigate to ArcToolbox Tools > Data Management Tools>Raster>Raster Dataset >Mosaic to New Raster
c) Input the DEM's (or air photos) you wish to attach together and select an output destination.
This may take several hours with high resolution DEM's. If you are working with 1m DEM's be prepared to leave it overnight.
Clipping a DEM in ArcMap/ArcCatalog
a) First you must create a shapefile polygon which represents the extent of your study area. There are at least 3 different ways
to do this.
1. Create a new Shapefile in ArcCatalog and then open an edit session in ArcMap and create a new feature (you can input
points from ArcScene, or else digitize over the air photo or DEM).
2. Draw the polygon in Google Earth pro (handy as you can view in 3D as you digitize)
3. Using ArcToolbox tools with the Spatial Analyst Extension, reclassify a raster which covers only the area you want to
clip to a single value and then convert it to a polygon.
b) Navigate to ArcToolbox Tools > Data Management Tools > Raster Processing > Clip
Input your DEM as the input raster, input your shapefile as the output extent and then check the box that says use input
features for clipping geometry.
20
Appendix D. Filtering Historical photo
Some of the historical photographs had a lot of speckle on them which needed to be smoothed out. This is easily done using PCI Focus
(Though other raster processing software such as ENVI could be used). The output grayscale image may then be converted to a 3-
band image and cropped for input into the application.
a) Open up PCI Focus, Click on the Launch the Add Layer Wizard, select grayscale
b) Now browse to the historical photograph, select it, click ok and then click on the file path in the top Files Available Box and
then in the bottom Channels Available box, select the channel (should be 1 [8u] contents not specified).
c) Now click Finish the photo should now be displayed.
d) In the table of contents, right click on the layer and select Filter. Make sure you are on the low pass tab. You can use the
default 3X3 filter size and average first. Now click apply to view.
e) Examine the photograph and continue adjusting the filter size and alternate between the average, median and mode
options until you have smoothed out all the speckle (remember to use the apply to view option, followed by the remove
from view option to reset the view before trying another filter).
f) Once you choose a filter size and filter that gives a good result then click the "apply to file" option. Your file path should
already be in the top box. Now in the select layer drop down menu, make sure the selection is not new layer, but instead
the image channel with data in it (there should only be one).
g) Now click ok.
21
Appendix E. Locating Ground Control and Center Points in Google Earth
This is an alternative method if you do not have ArcScene.
a) Navigate to your camera location in GoogleEarth
b) Use your marked photograph to locate the Center point and Ground Control Points
c) Create a Placemark at Each Point (see Figure 1).
d) Click on each Placemark and Save Place As >KML (change from KMZ in drop
down menu)
e) Download DNR Garmin 5.4.1 from
http://www.dnr.state.mn.us/mis/gis/tools/arcview/extensions/DNRGarmin/
DNRGarmin.html (also provided under Deliverables).
f) Install DG Garmin (click on dnrgarminsetup54.exe and run through the
installation wizard)
g) Open DG Garmin (Start>AllPrograms> DNGarmin>DN Garmin) and choose
file >load From> File >change “files of type” drop down selection from .txt to
Google Earth Format(.kml) and navigate to your saved
.kml point file and click open.
h) In the pop up window, select point.
i) To set the Projection, navigate to File>Set Projection and
choose NAD 83 datum and UTM Zone 11 then select ok.
j) Go to Edit>Project Coordinates
k) Choose File>SaveTo>File and in the “files of type” drop
down menu select “ArcView shapefile(Projected)”. Name
the file and click Save.
l) Now Open up ArcMap and add the newly created
shapefile from step 10 and your 10mDEM.
m) Making sure that your shapefile point is at the top of the table of contents, click on the identify tool, and select the
10MDEM layer on the Identify From drop down list.
n) Click on your camera station shapefile point on the map display
o) In your info text file, (eg JP15-4 _info.txt), record the X, Y,Z coordinates that show up in the Identify popup window (this is
the camera target location). Use the same methodology for the Ground Control points, except save the points in a .dat file.
22
Appendix F. ArcGIS Model Diagrams
Figure 1: Viewshed This model creates a viewshed from a user defined height above the DEM.
Figure 2: ViewFromPhoto This model creates a polygon outline of the viewshed for use in clipping the black background out of the
final georeferenced photo. The visibility outline is a triangular polygon which represents the camera field of view. We used a
different visibility outline for each photo.
Figure 3: Error This model generates two .dbf tables which 1) Display distances between a set of georeferenced photo points and a
set of control points taken from a satellite photo and 2). Display the differences between corresponding distance segments
generated from the georeferenced photo points and those generated from the control points.