Você está na página 1de 13

Georeferencers

Final Project Week 4


Deliverables -- Final Report, Maps and Metadata

Team Georeferencers
Week 4 Leader: Susan Williams
Members:
Dave Mattens
Brady Scribner
Kirstin Twardon
GEOG 484 GIS Geodatabase Development
Winter 2012
Pennsylvania State University

Description of the Database Design


Team Georeferencers is a group of Geographic Information System (GIS) professionals
specializing in GIS analysis, geodatabase development, map projections, georeferencing, and
digitizing. The team is competing for a contract from the Albemarle Charlottesville Historical
Society (ACHS) to create GIS data layers from the early 20th-century Sanborn maps of the City
of Charlottesville, Virginia. The ACHS has stated that the requested GIS will be a component of
a land use study which will analyze changes in land use over time. The GIS will also have
genealogy applications which will allow the ACHS to display the historical locations of people
and/or businesses. Lastly, the ACHS would like the GIS to provide a three-dimensional
visualization of the city of Charlottesville from the Sanborn map information.
The geodatabase consisted of a feature data set containing three feature classes: 1) a buildings
table (Buildings); 2) a blocks table (Blocks); and 3) a table for reference street data (StRefData).
Other data files used to produce the reference, thematic and bivariate maps were the
Charlottesville East, VA Topographic Map (o38078.tif) originally produced in 1973 and revised
in 1987, the Digital Orthophoto Quarter Quadrangle (DOQQ.jpg) from 1994, and four scanned
Sanborn maps.
All of the tables had the following common fields: OBJECTID, SHAPE, and SHAPE_Length,
which was either the polygon perimeter distance or the street polyline length. The Buildings and
Blocks tables also had the SHAPE_Area field.

The Buildings table provided the majority of information or the foundation for producing the
maps for analysis. The BldgID field was used as the primary key to provide uniqueness for each
record. Each GIS analyst was assigned a range of 1000 numbers (1000, 2000, etc.) to ensure
record uniqueness when maps were combined and edge-matched. Three fields, BldgUse,
BldgeStoryHt, and Descriptions were used to record the critical information needed to construct
the maps. Data entry for the BldgUse field was controlled by coded domain, which referenced
four categories of building use; Commercial, Public, Residential, and Unknown. The use of a
coded domain assured the accuracy of data entry.
The BldgStoryHt field records the number of stories for each building. Data entry was controlled
by setting a range domain from 1 to 4 with the ability to record story buildings.
The Descriptions field was used to record descriptive information about the building, such as
what type of commercial building it was (bakery, restaurant, drugstore, etc.) or how the building
was used for public use (church, school, post office, etc.).
Standard postal address fields were also used to record street address information, such as prefix
direction and type, street number and name, and suffix direction and type.
In addition to the common fields, the Blocks table contained one field which also served as a
primary key field, the BlockNum field. The BlockNum field contained the block numbers which
were gathered from the Sanborn maps.
Finally, the StRefData table contained the same standard postal address fields as listed above, but
also included fields used to provide a range and network of address, such as FromLeft, ToLeft,
FromRight and ToRight.
The georeferenced and digitized maps provide the types of analysis the ACHS has requested. The
best example of land use change over time can be demonstrated by the overlay of orthorectified
aerial photo with the two scanned Sanborn maps 1920_02A.jpg and 1920_02B.jpg. After a
period of 60 to 70 years, the eastern portions of both of these maps tend to line up remarkably
well, whereas the western portions bear little resemblance to each other. For example, an entire
neighborhood of homes can be viewed on the western area of the digitized 1920 Sanborn maps
but in 1994 the same area is now a major highway with warehouse-sized commercial buildings.
Conversely many portions of the Sanborn maps 1920_3A and 1920_3B appear unchanged. The
City Park and a number of churches can be seen on both the 1920 Sanborn maps and the more
current digital aerial photo.
The street reference data table provides enough information to facilitate the creation of an
address locator within the submitted GIS. This address locator will allow the ACHS to perform
geocoding for genealogical research for citizens interested in locating long lost relatives. Also,
the buildings table would provide the exact locations of historical addresses.

Finally, ArcScene was used to generate three-dimensional bivariate maps to show the 3-D
relative heights of the buildings along with the type of building and/or land use.
The collected data has made an assortment of different maps available. For example, reference
maps that show the exact locations of buildings with their accompanying address information,
two-dimensional thematic maps for both land use and building height, and 3D bivariate maps
showing both building heights in three dimensions and building use thematically were created
from the geodatabase.
By adding a year field to the buildings table, decades of time-change information in the
geodatabase can be provided. Because this information will be contained in a single table within
the geodatabase it will afford the means to perform any number of queries related to time,
building, and land use change. Although the year field would need to be repeated, its no
different than repeating zip codes for address within the same area.
Other than adding a year field and changing the default building height from one (1) to two (2),
we believe we have a sound geodatabase for conducting the types of analysis the Albemarle
Charlottesville Historical Society requires.

Time Estimate For The Project


As a team, the total number of hours required to georeference, digitize, and attribute spatial
features created from the four Sanborn maps was approximately 72 hours. Individually,
depending on the map being georeferenced and digitized, the average level of effort was
approximately 18 hours. Varying methods used to digitize map features resulted in differing
amounts of time logged per team member. For example, digitizing an entire building footprint
would be a greater time commitment than digitizing the same building without the building
attachments.
If we are awarded the bid, it would take an estimate of 468 hours to georeference and digitize the
remaining 26 maps for the 1920 period. More or less time may be needed if some of the Sanborn
maps are challenging to georeference and/or digitize. Altogether, georeferencing and digitizing
the 30 maps in one of the 10-year intervals would take an estimate of 549 hours, give or take a
few hours. To georeference and digitize the 30 maps in all 10 of the 10-year intervals, it would
take an average of 5,491 hours.

Averages/Estimates
Task

Average time
spent/task/map

Estimate for 30 maps


in a decade dataset

Estimate for 10
decade datasets

Geodatabase Design
and Maintenance
Georeferencing
Digitizing
Attributing

1 hours
9 hours
3 hours

Appending and
Edgematching

1 hours

40 hours

398 hours

Error Checking and


Quality Control

2 hours

64 hours

638 hours

38 hours
549 hours

375 hours
5,491 hours

Creating Shapefiles
and Metadata
Totals:

1 hours
18 hours/map

20
32
263
94

hours
hours
hours
hours

200
319
2,625
938

hours
hours
hours
hours

Analysis and Discussion of Inherent Level of Error


GIS databases will always contain a certain degree of error due to the human element. While
errors are impossible to eliminate entirely, they can be minimized to an acceptable level if the
GIS analyst takes certain precautions and follows methods meant to identify and minimize error.
Spatial Errors
One potential source of error is the very process of creating and/or scanning the Sanborn maps.
The original Sanborn maps were drawn by hand decades ago (nearly a century in the case of the
1920 maps we worked with) without the aid of GPS, computer, or other precision tools. As such,
the source maps will most likely contain inherent errors. This problem is compounded if the
source maps have been stretched, warped, carelessly handled or stored, or otherwise damaged.
Indeed, several of the maps that Team Georeferencers worked with seemed to have small
wrinkles or damaged corners and edges.
Other issues that may arise from the source maps are due to the resolution of the scanned image.
While these particular scans were sufficient for our work, increased resolution would certainly
have simplified and expedited the task. For example, when a team member would have to zoom

in to determine building height, the print became very grainy and pixelated; thereby leaving the
analyst to a somewhat subjective interpretation of the data.
Georeferencing is another potential source of error. While the Sanborn maps were drawn to
scale (one inch on the map equals fifty feet on the ground), they were not defined by a specific
projected coordinate system and therefore the RMS Error, which is essentially a measure of how
two layers fit together, was of no diagnostic value. Our team members instead relied on a careful
visual inspection to determine accuracy and alignment with the orthophoto.
The Sanborn maps were basically rubbersheeted, a process in which a map is stretched to align
with an orthorectified aerial photo for which the map projection is known. This process also
contains inherent errors as different team members may use different control points during the
georeferencing process. Our experience was complicated by some control points that were less
than optimal, including areas in shadows, as well as areas that were missing landmarks entirely
due to significant changes in landuse over time. Maps 2B and 2A were particularly difficult to
georeference as large sections of the map were missing known landmarks and street intersections
which caused the control points to be unevenly dispersed across the map image. The control
points could not therefore be ideally spaced across the image. The process can also be
influenced by the analyst if fatigue, bad vision, or poor manual dexterity results in poor
placement of the control point pairs. The amount of alcohol the analyst downs in an evening
while digitizing late at night could also affect accuracy, with sometimes remarkable results.
Georeferencing errors could be problematic as additional maps from earlier time periods may
have to be georeferenced with even fewer known landmarks and ideal control points.
Digitizing may contain some error as it is highly-dependent on user skill. Team members
minimized potential errors by setting the correct snapping environment, toggling selectable
layers and utilizing various Edit Tools (ie- auto-complete polygon, cut polygon), among other
techniques.
Digitizing also required a degree of interpretation on each team members part as some lines
could be rather ambiguous. Additionally, there was no definite agreement as to what the
Albemarle Charlottesville Historical Society (ACHS) considered a building. For example,
some team members digitized an entire building footprint while others opted to omit any
attached structures such as porches and decks. This could lead to quite a bit of inconsistency
over time for other subsets of maps. It is important that the team have standardized methods in
place to make certain that error is minimized and that the database will meet the needs of the
client.
Team members performed a visual inspection after digitizing in order to check for spatial errors.
Missing or incorrectly-shaped or placed polygons, street intersections not properly split or

joined, streets digitized in the wrong direction, and missing or duplicated streets or buildings
were some of the spatial elements analyzed for errors.
Edgematching was also performed to give the map a better look and consistency. However, the
process may introduce a degree of spatial error as edgematching decisions can be fairly
subjective. Nevertheless, this degree of error would be trivial when compared to the overall goal
of features appearing visually consistent.
Attribute Errors
A well-constructed database is the first line of defense against attribute errors, and the
Georeferencers geodatabase was designed to be as user-friendly and intuitive as possible while
still allowing room for future expansion. Data entry is assisted by multiple Coded Domains, a
range domain, default values and other design choices as described above. This prevents issues
such as multiple spellings (ie- Street, St, or S), misspellings, and provides a means for
standardized data entry.
In order to obtain attribute data, team members had to refer to several Sanborn map legends.
Despite the availability of several Sanborn map information sources, attributing features was
sometimes a subjective decision-making process. At times the Sanborn maps weren't always
clear and the data on the maps was not easily read. Other times, the Sanborn map legends were
confusing and open to interpretation.
Team members performed an inspection of one another's attribute tables after attributing spatial
features in order to check for inaccurate or inconsistent feature attributes. Utilizing a variety of
methods, missing or incorrect street address ranges, building addresses, building use codes,
building height codes, and unique feature IDs were analyzed for errors.
Conclusion
Although some level of error is to be expected due to the nature of the source (Sanborn) maps
and the differing skills among team members, the Georeferencers team was able to significantly
minimize the occurrence of spatial and data errors. By utilizing a user-friendly database,
implementing topology and other tools, and following methodologies to standardize work among
varying analysts, the onset of errors was diminished. Furthermore, team members avoided
working when fatigued or distracted to ensure spatial and attribute data integrity. Each member
of the Georeferencers team provides a range of talent and skills that allowed the Sanborn map
project to be completed successfully while minimizing the level of error inherent to the project
overall.
Despite the fact that the Georeferencers team based our design and study on only a subset of the
maps for a single time period, the database and resulting maps are to be used for visual

representation of the city concerning land use change studies, address location and genealogy,
rather than for surveying or engineering purposes. Therefore, it is our assessment that the level
of error is acceptable for the intended purpose and the geodatabase will satisfy the ACHS
requirements.

Thematic Maps
The Georeferencers Team has created the following thematic maps to display precisely how the
functionality of our geodatabase fulfills the requirements set forth by the Albemarle
Charlottesville Historical Society: a plan view (2D) building height map [see Figure 1 and
Figure 2], a plan view (2D) building use map [see Figure 3], and a 3D bivariate map
representing both building height and land use [see Figure 4 and Figure 5].

Figure 1: A 2-Dimensional map of Charlottesville, Virginia created with the help of the 1920 Sanborn Fire Insurance Map. This map is a combination
of four individual Sanborn maps that were georeferenced to a 1994 orthorectified digital aerial photo and then digitized. This 2-D map shows the height,
or number of stories, of each building. The number of stories ranged from 1 to 4 with half stories of 1.5 and 2.5. However, this map is not an accurate
representation of the height of the buildings. Some buildings (not including the attached structures) had two different heights, so the taller height was
taken when adding the attribute data.

Figure 2: Thematic map showing building height in a portion of Charlottesville, Virginia in the 1920s. Four Sanborn maps were georeferenced to a
1994 rectified aerial photo (DOQQ.jpg), digitized from scanned Sanborn maps, and combined into a single map. The 2-dimensional map shows the
height of buildings in the number of stories ranging from one story to four stories by half stories.

Figure 3: Thematic map showing building use in a portion of Charlottesville, Virginia in 1920. The buildings were categorized as Commerical, Public,
Residential, or Unknown. The location of this specific area within the state of Virginia can be seen on the inset shown at right. This map is a
combination of four individual Sanborn Fire Insurance maps from 1920 that were scanned and georeferenced to a 1994 orthorectified digital aerial photo
and then digitized.

10

Figure 4

11

Figure 5: A 3-Dimensional, Bivariate Map of Downtown Charlottesville, VA produced from Sanborn Insurance Maps from the 1920s. Four Sanborn maps were
georeferenced to a 1994 aerial photo (DOQQ.jpg), digitized from scanned Sanborn maps, and combined into a single map. The 3-dimensional map shows the
height of buildings in the number of stories ranging from one story to four stories and the buildings are color-coded according to building use (Commercial,
Public, Residential, and Unknown). The map was produced using ArcScene, exported and saved as a .jpg, brought into ArcMap where the title, legend, picture,
and north arrow were added in Layout View. The map was exported again as a .jpg and placed in a MS Word document where street names were added using text
boxes. The map was oriented in ArcScene to a true north direction.

12

References
Report header [ACHS logo] courtesy of the Albemarle Charlottesville Historical Society. Used here for
educational purposes only. Retrieved March 19, 2012 from http://www.albemarlehistory.org/
Sanborn Map logo courtesy of the University of Virginia Library, Geostat Center. Used here for
educational purposes only. Retrieved March 18, 2012 from
http://fisher.lib.virginia.edu/collections/maps/sanborn/.
Sloan, Jim (2012). GIS Database Development. The Pennsylvania State University World Campus
Certificate/MGIS Programs in GIS. Retrieved March 2012 from https://www.eeducation.psu.edu/geog484/

ESRI Data & Maps. USA Counties ArcGIS Layer Package. Retrieved March 18, 2012 from
http://www.arcgis.com/home/item.html?id=a00d6b6149b34ed3b833e10fb72ef47b
All maps produced with ArcMap v10.0 by Environmental Systems Research Institute (ESRI). (2012)
ArcGIS 10 Help. Esri, Redlands, California.

This document is published in fulfillment of an assignment by students enrolled in an


educational offering of the Pennsylvania State University. The students, named above,
retain all rights to the document and responsibility for its accuracy and originality.

13

Você também pode gostar