Error, Accuracy

1
Error, Accuracy Precision
2
Importance

• Unchecked error can make results of GIS analysis
  almost worthless
• Error comes from strength of GIS ability to
  combine many types of data based on location
  integrate many datasets within one system
• But every time a new dataset is used, its error
  is inherited
• This can mix with error already in the dataset in
  unpredictable ways

3
Avoiding error

• Planning
• Estimating its effects
• Understanding its sources
• All will make you more aware of potential
  limitations of GIS for reaching impossibly
  accurate and precise solutions

4
Definitions

• Accuracy
• Degree to which information matches true or
  accepted values
• Pertains to quality of data, number of errors in
  a dataset or map
• Refers to
• horizontal and vertical accuracy
• Attribute, conceptual and logical accuracy

5
Definitions

• Precision
• Level of measurement and exactness of description
  in database
• May measure position to fraction of a unit
• Remember precise data may be inaccurate.
  Surveyors make mistakes, data is entered
  incorrectly

6
Precision

• Level of precision required varies greatly
  (engineering vs marketing)
• Highly precise data can be costly and difficult
  to collect (and time-consuming)
• High precision does not indicate high accuracy
  high accuracy does not indicate high precision.
  Both are expensive
• Additional terms
• Data quality refers to relative accuracy and
  precision of particular GIS database
• Error covers both imprecision of data and its
  accuracies

7
Types of Error

• Positional accuracy and precision
• Applies to both horizontal and vertical positions
• Function of the scale at which a map (paper or
  digital) was created
• USGS specifications requirements for meeting
  horizontal accuracy as 90 percent of all
  measureable points must be within 1/30th of an
  inch for maps at a scale of 120L or larges, and
  1/50th of an inch for maps at scales smaller than
  120K.

8
Accuracy standards

• 11200 /- 3.33 feet
• 12400 /- 6.67 feet
• 14800 /- 13.33 feet
• 110,000 /- 27.78 feet
• 112,000 /- 33.33 feet
• 124,000 /- 40.00 feet
• 163,360 /- 105.60 feet
• 1100,000 /- 166.67 feet
• Means a point, line, etc. is in a probable
  location

9
Types of error

• Attribute accuracy precision
• Non-spatial data can also vary greatly in
  precision
• Mistakes come from many directions
• Conceptual accuracy precision
• GIS depend upon abstraction and classification of
  real-world phenomenon
• User determines how much info is used, and how
  its categorized

10
Types of error

• Logical accuracy precision
• Info can be used illogically
• Example not comparing zoning maps with
  floodplain maps
• Data in a GIS db must be used and compared
  carefully to yield useful results

11
Sources of error

• It is the users responsibility to identify and
  prevent sources of error. Some of the obvious
• Age of data. Past collection methods may be
  unknown, nonexistent, unacceptable
• Areal cover. Eg, soil maps incomplete at borders
  and transition zones, portray reality
  inaccurately. Lack of remote sensing data, lack
  of vector data in certain parts of the world.

12
Sources of error

• Map scale. Restricts type, quantity, and quality
  of data
• Density of observations. (eg, 40ft contours)
• Relevance. Using surrogate data instead (eg,
  aerial photographs for habitats, forest types)

13
Sources of error

• Format.
• Transmission, storage, processing can all
  introduce error
• Conversion of scale, projection, raster-gtvector,
  pixel size can all affect format error
• Multiple conversions can exaggerate error
• Accessibility. Whats open and available here may
  be restricted in other countries
• Military, inter-agency rivalry, privacy laws,
  economic factors
• Cost. Good data is expensive.

14
Sources of error

• In addition to obvious, there are errors due to
  natural variation or original measurements
• Positional accuracy. Dependent on the types of
  data being used/observed. Easier to place
  discrete (buildings, roads) than more nebulous
  (soil types, relief, drainage)

15
Sources of error

• Accuracy of content. Correct labeling and
  presence of specific features (can be missed by
  oversight, design, faulty instrumentation)
• Natural variation. (salinity)

16
Sources of error

• Processing. Hard to find, must be looked for,
  require knowledge of info and systems used to
  process data.
• Numerical. Different capability to perform
  mathematical operations. Processing error.
• Topological analysis. Can result in slivers,
  overshoots, dangles.
• Classification/generalization. Bad definition of
  classes, poor/inappropriate generalization.
• Digitizing/geocoding, raster-gtvector.
• Operator error.

17
Propagation Cascading

• Propagation one error leads to another. Eg, bad
  point in one map, used to register a second
  coverage
• Cascading erroneous, imprecise and inaccurate
  information can skew analysis when info is
  combined into new layers
• Difficult to predict and manage
• Both of these can affect horizontal, vertical,
  attribute, conceptual and logical accuracy and
  precision

18
Documentation

• Without it, you have no idea of precision and
  accuracy of a dataset
• Ask when youre buying, borrowing, downloading
  data
• Prepare your own data quality reports for data
  you create

19
Documentation

• Without metadata, ask
• What is the age of the data?
• Where did it come from?
• In what medium was it originally produced?
• What is the areal coverage of the data?
• To what map scale was the data digitized?
• What projection, coordinate system, and datum
  were used in maps?
• What was the density of observations used for its
  compilation?

20
Documentation

• Ask
• How accurate are positional and attribute
  features?
• Does the data seem logical and consistent?
• Do cartographic representations look "clean?"
• Is the data relevant to the project at hand?
• In what format is the data kept?
• How was the data checked?
• Why was the data compiled?
• What is the reliability of the provider?

21
Dealing with error

• Tag questionable items
• Add notations (this area not field checked),
  cautionary statements, legal disclaimers
• Document your sources and their documentation