Statistical Methods, Standards and Guidelines
106
Point feature
A point feature represents a single location. It defines a map object too small to show as a line or
area feature. A special symbol of label usually depicts a point location.
Line feature
A line feature is a set of connected, and ordered coordinates representing the linear shape of a map
object that may be too narrow to display as an area such as a road or feature with no width such as a
contour line.
An area feature is a closed figure whose boundary encloses a homogeneous area, such as a state country soil type or lake.
1.2 Map Characteristics
In addition to feature locations and their attributes, the other technical characteristics that define maps and their use include: Map Scale Map Accuracy Map Extent and Data- Base Extent
1.3 Scale
To show a portion of the Earth's surface on a map, the scale must be sufficiently adjusted to cover
the objective. Map scale or the extent of reduction is expressed as a ratio. The unit on the left
indicates distance on the map and the number on the right indicates distance on the ground. Map
scale indicates how much the given area has been reduced. For the same size map, features on a
small-scale map (1:1, 000, 000) will be smaller than those on a large-scale map (1:1,200). A map
with less detail is said to be of a smaller scale than one with more detail. Cartographers often divide
scales into three different categories:
Small-scale maps have scales smaller than 1: 1,000,000 and are used for maps of wide areas where not much detail is required.
Medium-scale maps have scales between 1: 75,000 and 1: 1,000,000.
Statistical Methods, Standards and Guidelines
107
Large-scale maps have scales larger than 1: 75,000. They are used in applications where detailed map features are required.
1.3.1 Scale in Digital Maps
With digital maps, the traditional concept of scale in terms of distance does not apply because
digital maps do not remain fixed in size. They can be displayed or plotted at any possible
magnification.
In digital mapping, the term scale is used to indicate the scale of the materials from which the map
was made. For example, if a digital map is said to have a scale of 1:100,000, it was made from a
1:100,000-scale paper map.
1.4
Types of Information in a Digital Map
Three general types of information can be included in digital maps:
Geographic information, which provides the position and shapes of specific
geographic features.
Attribute information, which provides additional non-graphic information about
each feature.
Display information, which describes how the features will appear on the screen.
1.5
Map Resolution
Map resolution refers to how accurately the location and shape of map features can be depicted for
a given map scale. Scale affects resolution. In a larger-scale map, the resolutions of features more
closely match with real-world features because the extent of reduction from ground to map is less.
As much map scale decrease, the map resolution diminishes because features must be smoothed
and simplified, or not shown at all.
1.6
Map Accuracy
Many factors besides resolution, influence how accurately features can be depicted, including the
quality of source data, the map scale drafting skill and the width of lines drawn on the ground.
Data Automation Computer Aided Mapping has its limitations. Goal of GIS is not only to prepare a good map but also perform map analysis. Maps are the main source of data for GIS
Statistical Methods, Standards and Guidelines
108 2.1 Source Data Source maps should meet the minimum mapping requirements as s et forth by the Surveys and Mapping Division of Tanzania (SMD). If possible, only source maps which meet the SMD and those that are in good condition (preferably originals or archival copies) should be used. If a source map meeting these requirements cann ot be obtained, it is recommended that source data be drafted onto base maps which meet these standards, or onto a transparent overlay which has been punch registered to an appropriate base map
For regional land based analysis, the scales at or near 1:10 ,000 shall be the primary base scale for input into the GIS database. Agencies may use smaller scales if deemed appropriate for the project and its intended analysis.
For local or municipal land based analysis, scales of 1:2,500 or less shall be the prim ary base scale for input into the GIS database.
Large scale source maps should be in satisfactory condition and have a minimum of 4 tics coordinate in order to achieve an accurate registration (Georefencing).
2.2 Data Capture
2.2.1 Scanning Scanning of source data must be done at the high resolution of 300mp, and 24 bits; and should be saved in Tiff format
2.2.2 Registration (Georeferencing) A minimum of 4 coordinate tics at the extremes of the data (i.e. map corners) should be used to register a map. The registration Root Mean Square Error (RMSE) value should be less than or equal to 002. All RMSE values must be recorded in appropriate documentation forms and shapefiles.
Where new coordinate tics are to be established, tic locations should be entere d through keyboard entry of exact geographic coordinates (i.e. UTM or Longitude -Latitude coordinates). Addition of new tics should be evenly spaced. A tic match tolerance should be set in the tolerance file for coverage
Statistical Methods, Standards and Guidelines
109
2.2.3 Digitizing
A map manuscript should initially be digitized on screen
Features should be captured using a minimum number of coordinates needed to accurately
represent the cartographic feature within a scale of 1:10,000 Points should be entered by geographic
coordinates, if available. When manually digitizing points from a source map, one is required to
enter the exact center of the point or map symbol.
2.3 Cartographic Accuracy
Positional accuracy should be as follows:
90% of the planimetric features on the digital map must be within 0.010 inch of the
centerline of that feature on the source map when plotted at the original scale
100% of all features must be within 0.020 inch
The 0.010 inch interval is equivalent to a standard 0.010 plotter pen width. When a proof hardcopy plot of a digital map is overlaid on the original base map, discrepancies will be seen as an open space between the plotted feature and the original source map.
- Attribute Coding
3.1 Item Definitions The NBS GIS Database Design should be consulted for item definitions in a new thematic layer. If a new layer is not identified in the Database Design document, or if there are questions relating to item definitions, please contact Geo-Information staff.
Item definitions must be consistent from coverage to coverage within a layer. Item names and codes should be unique from Layer to Layer
3.2 Attribute Integrity Attribute values must be 100'% correct when verified against source map attribute values, or compared with a list of source attribute values.
Statistical Methods, Standards and Guidelines
110 3.3 Map Coordinates
3.3.1 Universal Transverse Mercator Universal Transverse Mercator (UTM), Zone 35 to 37 souths (depending on the location) is the projection used in Country‟s library The map coordinate system of the Census Library is Universal Transverse Mercator (UTM). Units are meters. The datum is Arc 1960.
3.4 Edge matching Features must be accurately edge matched to adjoining features.
-
- Spatial Topology Data submitted to the Tanzania country wide GIS shall be topologically clean and free of error.
A topologically clean coverage will contain:
No slivers
No overshoots
No undershoots
No open polygons
No unlabeled polygons
No more than one label per polygon
No unresolved line segment intersections
4.0. Testing for Accuracy Transform an empty coverage to test positional accuracy of tics before digitizing features. Positional accuracy of features must be tested with verification plots after automation. Attribute values should also be plotted with features at a legible scale and verified against the source map, or printed and checked against the digital database.
A final plot of the digital map, or a legible verification plot, as well as a printout of its attributes should be kept on file as reference.
The accuracy of attribute values must also be tested by comparing them to source data.
Statistical Methods, Standards and Guidelines
111 An accuracy rating for each layer must be provided in the layer's documentation, as well as a description of the rating method used. The type of accuracy test performed, a description of the method used, and the result of the test must be indicated in the Layer‟s documentation.
The specification of accuracy and the extent to which that accuracy is assessed is subjective. Several factors will determine how accurate the data must be and the rigor of the tests perfor med. How the information will be used, the consequences of inaccuracies and project constraints, such as costs and time limitations, can all play a role in evaluating the degree of testing.
The cost of accuracy assessment must be weighed against the benef its of the accuracy information. The assessment method used to test a layer's accuracy should be based on its intended use. Where consequences of error are less critical, verification plots and an assignment of an accuracy level (rating) may suffice withou t any further testing. For other layers, it may be important to report an accuracy measurement which indicates the likely error for an individual point and also includes a level of confidence.
4.1 Verification Plots Positional accuracy of features must be tested by overlaying a plot of the digital data in inches with the source map from which it was digitized.
A proof hardcopy of the digital map must be plotted in inches prior to transforming the coverage into its final projection. When a proof plot is o verlaid on the original base map, discrepancies will be seen as an open space between the plotted feature and the original source map. If more than 10 percent of the total features are found to be more than 0.010 inches from the centerline of the source map features, then they must be re -digitized. All features found to be more than 0.020 inch away from source map features must be corrected. A pen size of .25 mm or .010 inch must be used. A light table is recommended to facilitate verification.
- 2 Normal Distribution Where consequences of error are critical, a more rigorous test, such as a normal distribution model, should be performed to indicate probability of error.
The normal distribution mathematical model has been shown to be a good predictor of er ror distribution for digital spatial data. To model the normal distribution, the average positional errors
Statistical Methods, Standards and Guidelines
112 must be found for a set of test points. Test points should be easily verifiable (i.e. road intersections, corners of large buildings, etc.) and must be verified against an independent source of higher accuracy, such as field verifications. The variation of error values for those points must be determined using the standard deviation. The standard deviation is then incorporated into the normal distribution model.
To quote an accuracy measurement of a data set, a confidence level must be chosen. The higher the level of confidence is, the lower the accuracy will be. For example, at a confidence level of 80%, the predicted error of a data set might be 1.75 meters. But at 95%/0, the predicted error may be 2.15 meters. This last statement means that 95 percent of the time features will be within 2.15 meters of their true position and 5 percent of the time, features will be more than 2.15 meters away from their true position.
- 3 Tic Coordinate Registration If a transformation RMSE value is higher than .005, additional measures should be performed to lower the RMSE before proceeding.
Several steps can be performed to lower this value and/or increase positional accuracy:
i) Re-mount the source map and flatten it against the digitizing board. Secure it adequately
with tape at the corners and at several other areas around the edge. Overlay the map with a
piece of clear mylar or translucent paper. Re-register the map carefully.
ii) Additional tics, such as 2.5 minute interior tics, can be digitized but should be evenly
spaced. Using 7 or 8 tics will reveal more information about which tics(s) are bad. A "bad" tic
coordinate can be thrown out.
iii) If new coverage tics are needed, the coordinates should be entered by keyboard only.
iv) Tics should only be drafted onto source maps by those experienced with this technique.
5.0. Documentation 5.1 Lineage Report The accuracy, processing history, and lineage of a layer must be recorded at the time of data capture, and must be available to other users in the form of a Lineage Report for evaluation of the data. This report will be attached to the database.