Examples

• Metaweb's Freebase, Spatio-temporal semantic system enhancement (GIS Website LLP)
• City of Berkeley, California, Neighborhood Mapping (GIS Website LLP)
• UK Department for Environment, Food and Rural Affairs: Spatial Information Repository (IBM)
• US Department of Agriculture, Farm Service Agency: Business Process Reorganization (IBM)

• UK Rural Payments Agency: Rural Land Registry System Upgrade (GIS Website LLP)
• Oakland Unifed School District: School Finder and Map Center (Local Knowledge Consulting)
• City of Oakland: GIS Data Conversion (Local Knowledge Consulting)
• Boston, Massachusetts, Real Estate Mapping: Data Consolidation and Visualization (GIS Website LLP)

Metaweb Inc. was a startup company graced with 57 million dollars of venture capital to build a Creative Commons semantic dataset of "the world's knowledge." Metaweb was acquired by Google for an undisclosed amount in July, 2010. Metaweb calls their dataset "Freebase" and has opened it to public input since 2007. As Freebase's dataset grows, three types of information have emerged as pervasive "glue" types that unite otherwise discrete data elements: People, Business and Location.

Initially, Freebase supported only point geometries, representing a location as a latitude and longitude and enabling searches within simple user-defined regions (e.g. latitude < 38.5 and > 36.0, etc.). GIS Website assisted Metaweb with the extension of Freebase's spatial functionalty to store linear and polygonal geometries and to support spatial queries against those geometries. Metaweb encourages public participation, so endorsed a geospatial architecture that reused existing open-source software with an active user-base, specifically the Postgresql + Postgis spatial database. Freebase users can now query using standard geospatial comparison functions involving distance, contiguity, intersection and so on, both for geometries created on-the-fly or for comparisons between existing Freebase geometries.

The video at right provides an 8 minute summary of Freebase and its spatial capabilities, as delivered by GIS Website's Jonathan Lowe at the 2008 O'Reilly Where 2.0 conference.

Freebase users can now query using standard geospatial comparison functions involving distance, contiguity, intersection and so on, both for geometries created on-the-fly or for comparisons between existing Freebase geometries. Specific to a semantic datastore, users can also extend a geographic search to include what Metaweb calls "relevance". In other words, the search results can be filtered not only by geometry but by the other data elements' degree of similarity to a source topic.

The below two illustrations are OpenLayers and Google mashups for browsing Freebase geospatial data for an area of interest, such as Berkeley, California. The Freebase topics containing geometries appear as tag clouds below the maps; semantic types with the most topics appear with larger font sizes than types with just a few topics.

The above mash-up combines data from Freebase and Google and spatial javascript libraries from OpenLayers. The selected type is "/architecture/building" so the map displays the locations of Freebase's noteworthy buildings for that region of Berkeley, California. The tag cloud's variations on architecture types -- building, house, structure -- illustrates the flexibility of a semantic data model: any individual geometry may be linked to more than one type.

Following the integration of complex spatial types and functions into Metaweb's architecture, users can now store, retrieve and search within polygonal regions of any shape, such as the extent of the Oakland Hills firestorm shown above. The small sliver at the bottom of the burn area shows where airborne cinders leapt over the freeway to start a new burn area.

• At a technical level, preserving the existing (non-relational) data store's retrieval performance speeds was the ultimate challenge. Scott Meyer's strategy was to handle geometric data independently from the core semantic data store, as geometries could contain any number of vertices and hence unpredictable processing times.
• Establishing consensus about the best data model for location data was more difficult in the organizational chaos of a start-up than in a corporate culture's top-down direction. This was compounded by the need to agree the model not only internally, but with the user community.

• All of Freebase's data is required to fit the Creative Commons definition if it is loaded into the official database. Finding geographic data of this stripe is not easy after finishing with TIGER datFinding geographic data of this stripe is not easy after finishing with TIGER data. As Tim O'Reilly commented in Where 2.0, "It's a great idea, but it still needs more data to be really sueful."

For more information, please contact us.

Berkeley Real Estate map plotting listing and sales histories with (some) facade photos from 2004 to present.

Berkeley Hydrology (watersheds and creeks) map displaying which segments of each creek qualify as surface flows and thereby constrain the building permits for intersecting and nearby properties.

• Presenting multiple overlays to public users but keeping each thematic grouping cartographically readable.
• Aggregating data that grows over time (such as crimes) for visual clarity, but enabling users to investigate details (such as invidual incidents).

• Building a browser-based interactive spatio-temporal site without spending any money on software.
• Enabling a variety of navigation and query capabilities using a minimum of controls (seven in this case).
• Converting disparate data sources to common quality levels

The Berkeley Neighborhood Mapping application is for demonstration purposes only, so is not normally accessible to the public. If interested in test driving the application, please contact us.

In this illustration, the IBM Websphere Portal calls an enhanced Web Mapping Service API to draw both the backdrop image, the layer of interest (Countryside Stewardship Agreements) and the user's selection. Another API for transforming data between database and flat-file formats provides an estimate of the output shapefile size and download time.

In this illustration, the SPIRE portal displays a portlet interface to an underlying location based services API, in this case a gazetteer enabling input of a placename and output of any matching candidates. Selecting a listed candidate causes the map portlet to zoom and recenter on that place's geographic location.

• Integrating a wide variety of software within constraints of standards frameworks (e.g. JSR 168 portal standard, OGC WMS specification, etc.)
• Designing web services APIs that withstand changes in requirements with minimal need for redesign.
• Designing and implementing a temporal model for spatial features, spatial metadata and system metrics.
• Wiring IBM Websphere portlets bi-directionally.

• Establishing user agreement on spatial data quality levels and validation methods for assessing quality.
• Coordinating a team of a dozen IBM IT specialists and as many third-party contractors and partners over a multi-year, multi-million dollar (pound) project.
• Building a security model relating departments and individuals to each dataset in the SPIRE collection and to capabilities in the SPIRE portal.

Due to data sensitivity and licensing restrictions, the UK DEFRA SPIRE portal is currently an intranet-based application not available for public viewing. A subset of publicly available data from the SPIRE collection is available at the MAGIC website, an ESRI ArcIMS-based site.

Because the FSA has multiple programs, each of which may apply subsidy calculations to the same overall farm property, a significant piece of their data modeling work involves clarification of "the rules" across space, time and seasonal policy changes. This illustration shows our approach for confirming rules, in this case, on valuing crop rotation.

The FSA hoped to simplify their existing data collection processes (mainly paper forms and field office representatives) by putting farm data online and encouraging farmers to maintain their own spatial data directly. This illustration shows a sample of the design proposals for such an assisted ditization interface.

• Although constrained to the same land, different FSA policymakers calculate subsidy payments based on different models, defying consolidation in a single model.
• Improvements in spatial technology are not always welcomed by users; farmers who had been subsidized for 10 acres of corn were not interested in adopting a more accurate system if the more accurate acreage of 9.5 acres reduced the FSA payout accordingly. A rules-based system tied to the data model was our approach to managing such parallel views of the world.
• Collecting full business requirements before building any working applications takes dedication on the part of both the consultants and the clients.

• Policy lifecycles do not start and end evenly; there are overlaps even within departments. The temporal model had to handle comparisons of events that did not perfectly match.
• The FSA's constituents are farmers first and technology users second; any public-facing interfaces had to be highly intuitive and/or contain embedded storyboard-style documentation.
• No single person at the FSA can unilaterally approve (and later vouch for) a design proposal, so design decisions must be clearly recorded and circulated among a changing group of stakeholders.

The resulting deliverable to USDA FSA was a 72 page report: USDA FSA Modern Farm Records Requirements and Architecture Target.

A sample page from the RPA's Conceptual Data Model document, authored and illustrated by GIS Website LLP. This page explains the assigment of County-Parish-Holding identification numbers to land parcels based upon the 5- or 10-mile rule.

A sample page from the RPA's Logical Data Model initially established by GIS Website LLP. Upgrading the GIS happened in parallel with an upgrade and integration of the RPA's workflow system -- the gray areas distinguish GIS, workflow and associative entities from each other.

A sample page from the RPA's Target Business Process Model, authored by GIS Website LLP, shows the series of actions and decisions that the RPA's Triage Team must take when receiving a request to modify a GIS land parcel polygon to reflect some change in the real world. To improve readability of the flow, the "happy path" or ideal flow is highlighted in gray.

Conceptual illustrations, entity-relationship diagrams and process flow models are all useful and necessary to explaining a design, but are not always accessible to those unfamiliar with their notation, particularly at an executive level. Storyboards, however, have universal appeal because, if done properly, they translate long lists of dry, written requirements into a vision that the project's sponsors, users, developers and managers can all directly understand. This storyboard excerpt, authored by GIS Website LLP, illustrates the RPA's new process model and GIS Toolkit.

• The biggest challenge by far at the RPA was overcoming beaurocratic inertia to turn written design documents into live working applications.

• The RPA's spatial, temporal and associative data was rife with errors. Identifying them was relatively straightforward, but correcting them was much more difficult due to the network of legacy applications and partner agency processes depending upon RPA data on a regular basis. Data migration and cleansing activities consequently took on greater significance than usual.

For more information, please contact us.

OUSD's staff were swamped with calls from the public asking within which school boundary a particular street address fell. This School Finder application dramatically reduced call volumes.

In addition to detailing OUSD student-to-school assignment, the Map Center made data from the City of Oakland's municipal GIS available to the public, as shown here with landuse data.

Innovation in display to solve the obscured choropleth problem and to teach children about the differences in scale between imagery and abstract maps. This map stacks population over an aerial photo.

• Developing a browser-based mapping application with bleeding edge (at the time) technology.
• Joining out-of-the-box spatial components (ESRI's MapObjects) with custom-coded high-performing C modules for image modification not initially supported by ESRI.

• Designing a user interface simple enough for school children to understand intuitively.
• Presenting opaque choropleth maps (such as population density) without obscuring the underlying reference features (such as roads and annotations).

To explore the live application, please visit the Oakland Unified School District's Map Center, live since 1999.

School Sites

Sewer Pipes

Land Parcels

Topographic Contours

Fire Hydrants

Census Blocks

1. Aerial locations
2. Aerial transmission pathways
3. Aspect of land surface in relation to a compass bearing
4. Assessor's Parcels--City of Oakland
5. BART station architecture
6. BART station locations
7. BART surface rail
8. Breaklines for DTM generation
9. Calls for Service in 1995
10. Census blocks and block groups (US, 1990)
11. Census Tracts (US, 1990)
12. Child Care (Head Start) Centers
13. City Council Districts
14. City limits of Oakland and Piedmont
15. Community development districts
16. Container Routes permitted in harbor areas
17. Contour lines--index: 25' intervals
18. Contour lines--intermediate: 2' interval, or 5' in hills
19. Digital orthophoto tile reference grid
20. Fire Assessment Districts
21. Fire Battalions
22. Fire Hydrants

23. Fire Stations
24. Fire Still Districts (alias: Fire Response Radii)
25. Flood plains
26. Freeways with on and off ramps
27. Harbors
28. Hazardous Materials Facilities
29. High-volume transit corridors
30. Hospital and Medical Care Sites
31. Land and Water Bodies for Alameda County
32. Liquifaction Zones
33. Major Container Freight Stations
34. Manhole locations
35. Meteorological data collection stations in Bay Area
36. Parks
37. Patrol beats of Oakland Police Department
38. Patrol districts of Oakland Police Department
39. Patrol sectors of Oakland Police Department
40. Plan Areas
41. Port Gantry Cranes
42. Rail lines

43. School districts--elementary schools
44. School districts--high schools
45. School districts--junior-high schools
46. School locations, K - 12th grades and day-schools
47. Select Part One Crimes in 1995
48. Sewer and Storm Drainage Basins and Sub-basins
49. Sewer grid from NAD27--historic reference
50. Sewer pipes (underground network)
51. Shelter Sites
52. Slope or "angle of repose" of land in relation to horizontal plane
53. Soil composition by series and type
54. Streams
55. Streets (center-of-pavement lines) including transit capacity data
56. Streets resurfaced within the past 5 years
57. Trails & Paths
58. Underground storage tank locations
59. Vegetation class in wild land areas
60. Zoning

To explore a live application based on Oakland's GIS data, please visit the City of Oakland GIS data in a Flash interface designed and deployed by The Urban Explorer. Much of the OUSD Map Center's data comes from the City of Oakland's collection.

This view illustrates the use of a Google Maps backdrop mashed up with real estate listings point locations color-coded by the length of time they have been listed, circles marking a 15-minute walk from grocery stores (purple) and desirable neighborhoods (orange). Selecting a listing opens a speech bubble that allows individual users to request a tour of the property or to reject it. All users see each other's choices.

Demonstrating use of a different backdrop (Google aerial photography) with a superimposed Web-Map-Services (WMS) layer to reveal the assessor's parcel boundaries (with faint purple lines). While some property listings provide site maps, they rarely show the lot lines in the context of the surrounding neighborhood as this application does. The red and green circles around the yellow diamond in the lower left show a listing that both the realtor and one of his clients think would be worth seeing.

• Although very useful once mastered, OpenLayers has limited documentation, so has a steep learning curve. Developers are expected to learn from the Javascript libraries themselves or to study a collection of examples.

• One of OpenLayers' strengths is its reuse of tiled commercial reference maps such as Google, Yahoo, Microsoft and OSM. However, these backdrops' "Spherical Mercator" projections require special treatment if the data being superimposed on them is a different custom projection, as is the case the MassGIS data in this application.

The Boston Real Estate Mapping application was built for a private client, so is not on public display. However, if interested in test driving the application, please contact us.