Metadata Education Project

Metadata education suggestions and materials for:

Learning Material | Learning Outcomes | Preparatory topics | Complementary topics | Vocabulary

• What is metadata?
• Why is metadata important?
• Overview of the metadata content standard

• Data sources: determining fitness-for-use
• Good and bad examples of metadata

Vocabulary definitions

• Data Quality Section
• Source scale
• Process Steps
• Spatial Organization Information
• Spatial Reference Information

• Why document data as it is created/automated?
• What should you document as you create/automate data?
• Example exercises to demonstrate the importance of metadata

Not all existing data, however, is readily available nor will it always suit the requirements of the application (see the topic on Data sources: determining fitness-for-use). Many organizations, government offices, businesses, and students are still faced with the need to create their own data in order to accomplish the needed analyses/solutions with a GIS system.

Metadata is a key element in respect to using existing data sources. The quality of output from a GIS is only as good as the quality of the data input into the system. The more you know about your data, the more confidence you will have in your results.

But if you are creating the data yourself, instead of using existing data created by others, is metadata all that important?

Yes! Metadata is an investment in your data. If data creation and/or manipulation are approximately 70% of the cost of implementing GIS within an organization, why would you not document this investment with metadata? Having this documentation reduces time and confusion in the future, if there is staff turn-over, if changes are made to the data, if procedures need to be replicated, if codes are forgotten, if questions are raised about the data's quality... and so on.

What if the data is to be created for a one-time only GIS project? Is metadata still worth the investment? - Most likely. Questions can still be raised about the results of the project (remember: garbage in = garbage out). Other projects might benefit from the data created for your project, if only they had some documentation of the data's quality, or even knew of the data's existence through a metadata catalog.

Another argument for creating metadata as you create data is that the well-defined data quality elements that are a part of the metadata content standard can provide a guideline for creating data using quality assurance procedures, including checking the data for errors and providing an estimate of its accuracy.

Here are two examples of datasets that were created without metadata or quality control/quality assurance measures. When the data was put to use, embarrassing errors were discovered in it!

• Legal descriptions converted to coordinate locations
• GPS locations of facilities

At the very least, it is important to document the sources and procedures used to create or automate the data. After the data creation/automation process is completed, it is also helpful to document other aspects of the data's quality, such as its estimated positional and thematic (attribute) accuracy, its completeness, and its logical consistency (see the topic on Data quality for more details.)

This documentation will have two direct benefits:

• it will provide evidence against the potential question of "garbage in = garbage out" concerning the results of the GIS analysis or solution
• it will provide a record of sources and procedures used in case replication is ever required (to prevent the need to "reinvent the wheel")

Most data is automated from existing sources, such as aerial photographs, satellite images, hardcopy maps or existing databases with a spatial description (such as a legal description or a street address). Describing these sources is the first step in documentation. Where did they come from? When were they created? What was their source scale or resolution?

If you are creating data "from scratch", as if often the case when data is collected by GPS receivers, it is important to provide information about the instruments and methods used to collect the data. This information is documented in the form of process steps. For instance, what sampling method was used to select the features to be collected within the study area? What brand and grade of GPS receiver was used? How many satellites were used to fix positions? What was done about interference (buildings, trees)? How were the locations corrected, if differential correction was required?

Describing process steps is important in all types of data automation and data manipulation:

• Digitizing using a tablet
  What source map(s) were used (paper/mylar; condition; date; source projection)? What type of registration marks were used and how many? What RMS error was used and was it consistent? What type of digitizing method was used (normal or stream mode; spaghetti or discrete digitizing)? What tolerances were used (snapping)? Did one person do all the digitizing or more than one person - did they follow the same procedure? Were checks/corrections made to data and if so where?
• Digitizing on-screen
  What kind of imagery or photography was used and what was its associated accuracy? (see also factors for tablet digitizing above)
• Scanning
  Was any pre-processing required on hardcopy before scanning (color separation)? What type of scanner used? What resolution (dots per inch) used to scan? Any filtering applied to the scanned image? Any clean-up applied to the scanned image such a noise removal? What methods used to register the image, how many registration marks used and their RMSE values? What methods used to test for and remove distortion from image? Were vectors automatically traced or hand-digitized from image?
• Conversion from legal descriptions
  Were legal descriptions of all features recorded to the same level or different levels (e.g. to quarter sections, or quarter/quarters, or lot descriptions?) What type of algorithm was used to generate latitude/longitude of centroid position - does it take individual section shapes/sizes and distortions into consideration for its calculations?
• Conversion from street addresses (geocoding)
  What kind of algorithm and parameters were used for geocoding? How many unmatched addresses remained after geocoding and what methods were used to match them? What source data was used to provide locations (scale and quality of roads and other features); were any random or systematic checks performed to verify correct geocoding of addresses?
• Conversion from satellite imagery
  What kind of sensor was used? What are its characteristics? What date(s) were the imagery collected and conditions present? How were images registered, what registration marks were used? How were distortions removed? Was the imagery resampled or filtered? What type of classifications (manual, supervised or non-supervised) were used? Was training data used in classification process? Were final categories checked for thematic accuracy?
• Conversion of existing spatial data
  If data was converted from vector to raster, what type of algorithm was used (e.g. majority, minority, weighted) and what cell size? If data was converted from raster to vector, what algorithm was used and what tolerances were set to reduce stair-step effects or filters to reduce noise? Were checks performed to remove incorrect intersections and other inconsistencies?
• Interpolation
  What kind of algorithms were used to convert sample points into surfaces/contours, or contour lines into continuous surfaces (e.g. global/local; exact/approximate)? Were barriers used in the interpolation process? What parameters were used (sill, range, nugget) in kriging and how were these values arrived at? What output resolution was selected? Were the original values described (raw data, averages, other classifications)?
• Appending or merging datasets
  What method of edgematching was used? What tolerances were used or decision rules?
• Other manipulation of datasets
  What was the precision of coordinates before performing transformations such as projections or overlays? What other types of manipulations were performed such as dissolves or elimination of slivers?

For specific information about entering or reading metadata on Sources and Process Steps, see the Data Quality section of the metadata content standard.

This is an excerpt from a class exercise created for the University of Northern Alabama's Advanced Digital Techniques in Geography course (fall 1998) by Lisa Keys-Mathews (web-site).

Your team is asked to complete the following tasks:

• Design the proper database including a detailed data dictionary (They want to see how well you know your utilities database requirements)
• Divide the dataset into four sections, assigning one section to each team member (They want to see how well your team works together)
• Heads-up digitize the data layers (They want to see your interpretation skills)
• Design and create metadata for all new files created (They want to see how well you understand metadata and its necessity)
• Edgematch the four sections into one large file (They want to see your company's edgematching capabilities)
• Combine the metadata files into one file for each layer in the database (Again a teamwork example)
• Document the process and give it to the Chairman of the Board (They want to see you demonstrate and document the process)
• Explain the process and demonstrate the data to the Chairman of the Board (They want to see your team-presentation abilities)

Advanced material

Abstract:
Nowadays one of the most successful applications of GIS is the management of a land-use cadastre. A lot of corporate GIS databases are in development, they support the legal management and distribution of cadastral maps. However, the propagation of geographical updates toward cadastral databases is still a methodological and technical problem to address in the context of large applications with many different users. This paper proposes a model based on lineage metadata that supports the management of geographical changes in the context of a corporate cadastre application. Geographical and cadastral changes are identified from an analysis of the French cadastre which acts as a case study for the development of our model. The lineage metadata model is based on the application of a direct acyclic graph that permits the management of the evolution of geographical objects and the generation of historical queries. The proposed model is specified and validated with the O_{2} object-oriented database management system.

Related Material

NCGIA 1990 Core Curriculum: Unit 7: Data Input

Back to Course Topics