A Data Model for Representing Cities in Three Dimensions - Esri

A Data Model for Representing Cities 

in Three Dimensions 

Harvard University 

Graduate School of Design 

Presented by Paul Cote 

ESRI User Conference August 2008



Abstract 

This talk will describe an ArcGIS Geodatabase model 

and model-builder toolkit for organizing and 

visualizing three-dimensional city models. This 

simple relational database schema provides means of 

organizing 3d model assets and retrieving logically 

consistent visualizations of the city at different time 

periods, proposed development scenarios, and at 

appropriate levels of detail. The presentation will 

discuss the extension of this model with reference to 

the emerging Building Information Spatial Data Model 

BIM Standards and CityGML. 

Paul Cote. ESRI UC 2008 

2



Findings: 

To build a data model for buildings, campuses and 

cities, requires that we build a schema to represent 

many different types of features that are inter-related. 

We should: 

Avoid the usual home-grown, seat-of-the-pants 

problem solving strategy and participate in 

community-based data modeling work 

Anticipate the need to exchange data between 

different specialized software tools 

Encourage our vendors to support community 

based data modeling efforts and vendor neutral 

formats for data exchange! 


3



Acknowledgement 

This work was made possible by funding by the Town 

of Brookline, Massachusetts Department of 

Management Information Systems. Feng Yang, 

Director. 


4

Evolution and Sustainability 

We’ve seen plenty of demos, but how can we evolve 

our municipal GIS infrastructure to incorporate 3d 

building models that are a solid, sustainable part of our 

infrastructure? 

Sustainable evolution of municipal infrastructure will: 

Make use of existing GIS layers 

Build on the scalable relational (table-based) data 

structure of GIS 

Extend GIS data models to deal with true 3d data 

Require interoperability with 3d modeling tools and 

new Building Information Modeling Tools 

Reliable cross-vendor interoperability will depend on 

stable open interchange standards 




5

Today’s GIS Infrastructure 

Traditional layers in a municipal GIS infrastructure are 

two dimensional: 

Property Parcels 

Building Footprints 

Orthophotography 

Edge of Pavement 

Photogrammetric Contours 




6

2.5 Dimensional Models From 2d GIS 

Height Information can be added to 2D GIS data to 

develop two and a half dimensional models for 

visualization 

Buildings may be Extruded 

Contours and Edge of Pavement may be turned 

into ground surface models as Triangulated 

Irregular Networks (TIN) 

Orthophotography or thematic information may be 

draped on surfaces 




7

GIS as Basis for 3D Models 

GIS tools like ArcGIS are not suited to making true 3d 

Models, but GIS information infrastructure is very 

useful as a basis for detailed 3d modeling. 

Represents consistent, controlled geometry over 

broad area 

Provides rough 2.5-D massing context 




8

GIS -> 3D Modeling Interoperability 

Transferring data (interoperability) between GIS and 

3d Modeling tools has been a very difficult process 

until a few years ago. 

Common CAD formats (DWG/DXF) does not pass 

information for extrusion 

ESRI has no way of interchanging TIN information to 

DWG / DXF 




9

Sketchup Interoperability 

Sketchup is a 3d modeling tool that deals with 

georeferencing and interoperability with geospatial 

tools. 

Capable of getting terrain with draped aerial 

photos from Google Earth or ArcGIS 

Capable of getting extruded massing models from 

ArcGIS 

Very rich capabilities for making 3d models 

Models may be encapsulated as COLLADA files 

COLLADA files may be dealt with as singular data 

objects in a database or Google 3d Warehouse 

Vendors of 3d modeling tools are rushing to 

incorporate generic COLLADA import and export. 




10

Incorporating true 3d models with GIS not only 

provides a rich visualization tool, but also a means of 

managing large collections of models. 




11

Workflow for Developing City Model 

Selection of Model 

Elements 

Terrain, Massing, 3d 

Buildings 

Based on Date or Scheme 

Export to CAD or 

Rendering System 

Rendering: 

Hi Quality (e.g. 3d Studio) 

Web-Based (e.g. Google 

Earth) 

Creating new model 

elements 

Upload of new model 

elements to repository 




12

Concern about Vendor-Specific Exchange 

Formats 

Nobody should build an infrastructure that depends 

on vendor specific arrangements for interchanging 

proprietary data formats! 

Despite the best intentions, interoperability may 

cease to work as software features change 

COLLADA is an exchange format maintained by a 

consortium (Khronos) that is regulated by a 

consortium of vendors, its specification is open 

and has stable well documented versions. 

At version 9.3 ArcGIS will be able to import a 

COLLADA file and convert it to its own 3d Data 

format (multipatch). YAY! 




13

Vendor Neutral Exchange: COLLADA 

COLLAborative Design Activity 

Standards Consortium: 

Khronos Group 

Communities: 

Visualization 

Gaming 

Capabilities: 

Very rich visual and physical simulation: photorealistic 

shaders, cameras, movement 

No semantic model 

No georeferencing 

Modeling / Encoding 

UML / XML 

Adoption: 

Many 3d authoring tools, 3d Studio, Sketchup … 

The vehicle for textured models in Google Earth 


Graduate School of Design Paul Cote. ESRI 14

GIS Management of 3D Building Models 

The availability of sketchup and the interoperability 

with Google Earth and ArcGIS is leading to the 

development of lots of georeferenced 3d building 

models. We need a systematic means of organizing 

these assets 

Visualizing large models requires different levels 

of detail for distant or up-close rendering: 

Cities change, we need to be able to render 

temporal scenarios buildings based on year built 

and year demolished 

Need fictitious design schemes with proposed 

buildings and demolitions 




15

Building Model Management Schema 

Abstract Buildings 

Building ID 

Attributes… 



1 

n 

n 

Schemes 

Scheme ID 

Name 

Skin Parts 


Part ID 

Attributes… 

Massing Parts 


Part ID 

Attributes… 

1 n 

n 

n 

1 

Scheme Parts 

Scheme_ID 

Part ID 

Render 


1 

16

New Table: Abstract Buildings 

The Parcels table can form a table of information about 

buildings. Unique Building IDs are created from Parcel IDs. 

This table has no geometry associated with it. Any building 

associated with a parcel are designated Building 0. This 

lumping is inaccurate, but sufficient for an initial buildings 

table. 

B0 

Cam_P364234 




Territory Cambridge 

Building Cam_P364234_B0 

Owner Harvard 

Built Date 1870 

Demo 

Date 

Null 

Address 49 Quincy St 

Stories 5 

17

New Layer: Building Massing Parts 

Where individual building parts vary in terms of their attributes, they may be 

distinguished with unique IDs and individual build and demo dates, etc. The 

Building Massing Part IDs are formed by appending a Massing Part ID to the 

Building ID. 

B0_M1 

Cam_P364234 

B0_M0 



Building Massing Parts 

Part_ID Cam_P364234_B0_M1 

Owner Null 

Built 

Date 

Demo 

Date 

1998 

Null 

Address Null 

Stories 2 

18

3d Building Massing Parts Model 

The Massing Parts Layer forms a complete model of every building in the 

metro area at a low level of detail (CityGML LOD2). Building parts may have 

more specific attributes from their parent Abstract Buildings. 

B0_M0 

B0_M1 




Part_ID Cam_P364234_B0_M1 

Owner Null 

Built 

Date 

Demo 

Date 

1998 

Null 

Address Null 

Built Yes 

Stories 2 

19

3d Building Skin Models 

Using a 3D authoring tool, models of building exterior skins can be 

encapsulated as sketchup or collada models and placed into relational tables 

as georeferenced objects. Like Massing Parts, Skin Parts have unique IDs 

linking them to abstract buildings and may have more specific attributes. 

Cam_P364234 

B0_S0 



Building Skin Parts 

Shape Multipatch 

Model Sketchup or Collada 

Blob 

Part_ID Cam_P364234_B0_S0 

Owner Null 

Build 

Date 

Demo 

Date 

1870 

Null 

Built Yes 

20

3d Building Skin Parts 

Skin parts and massing parts for a given building may have independent dates. 

The Model field holds a Binary Large Object that can be downloaded to a 3d 

authoring tool for editing. 

Cam_P364234 

B0_S1 






Blob 


Owner Null 

Buid 

Date 

Demo 

Date 

2001 

Null 

Built Yes 

21

Built and Unbuilt Scenarios 

The Built attribute is set to Yes for buildings that were actually built. Unbuilt 

buildings can be represented in the schema to allow experimentation with 

proposed scenarios. 

Cam_P364223_B0_S0 






Blob 


Owner Null 

Built 

Date 

Demo 

Date 

2001 

Null 

Built Yes 

22

Built and Unbuilt Scenarios 

The Built attribute is set to Yes for buildings that were actually built. Unbuilt 

buildings can be represented in the schema to allow experimentation with 

proposed scenarios. 

Cam_P364223_B0_S1 






Blob 


Owner Null 

Built 

Date 

Demo 

Date 

2001 

Null 

Built No 

23

Relational Queries Create Views Based on Dates 

or Scenarios 


Year 

Demo 

Shape Multipatch Address 


Blob 

Stories 




Territory 

Building 

Owner 

Year Built 


Owner Null 

Built 2001 

Demo Null 

Built Yes 

Cambridge 

Cam_P364234_B0 

Harvard 

1870 

Null 

49 Quincy Part_ID St Cam_P364234_B0_M1 

5 


Owner Null 

Built 1998 

Demo Null 

Address Null 

Stories 2 

24

The Scheme View Tool 


Graduate School of Design Paul Cote, 2008 

A tool has been created 

with all of the SQL to 

create the three views 

necessary to portray a 

Date and Scheme 

Specific Scenario 

based on the user’s 

privileges. One needs 

only to double-click on 

the tool and fill in the 

blanks. 

25



Current Built View 

26



1998 Historic View 

27

Unbuilt Scheme: Turn Gund Hall Around 



28

View From Proposed Design Scheme 


Graduate School of Design

1 




Attributes… 

1 

Schemes 

Scheme ID 

Name 



Part ID 

Attributes… 



Part ID 

Attributes… 


Graduate School of Design Paul Cote. ESRI 

n 

n 

1 n 

n 

n 

1 

30 

Scheme Parts 

Scheme_ID 

Part ID 

Render 

1

The Scheme View Tool 



A tool has been created 

with all of the SQL to 

create the three views 

necessary to portray a 

Date and Scheme 

Specific Scenario 

based on the user’s 

privileges. One needs 

only to double-click on 

the tool and fill in the 

blanks. 

31

Extending to Interior Spaces 

This 

This 

3d 

3d 

Model 

Model 

Schema 

Schema 

is 

is 

compatible 

compatible 

with 

with 

the 

the 

Penobscot 

Penobscot 

Bay 

Bay 

/ 

/ 

ESRI 

ESRI 

data 

data 

model 

model 

for 

for 

interior 

interior 

spaces. 

spaces. 




32

1 




Attributes… 



1 

n 

n 

Interior Spaces 

Interior Spaces 



Part ID 

Part ID 

Attributes… 

Attributes… 



Part ID 

Attributes… 



Part ID 

Attributes… 

Paul Cote. ESRI 

n 

n 

33

Expect Problems with Interrelated Models 

Expect problems, and you will rarely be disappointed! 

What are some problems that we can expect with 

merging interior spaces with building exterior models 

Opportunity for audience participation! 




34

Building Information Models 

Building Information Modeling is essentially the 

application of GIS data modeling concepts to 

buildings. 

Building information developed and used in 

Design, Construction, Facilities Operation, 

Demolition is integrated into a single model 

Many tools are capable of creating and analyzing 

BIM 

The ability to exchange information among these 

tools allows toolmakers to specialize and users to 

have a choice of tools 




35



IFC 

In order to have interoperability between the various 

specialized tools used in the Architecture Engineering, 

Construction and Facilities Management requires 

exchange of structured information. 

The emerging exchange standard for 

interoperability in the BIM world is IFC: Industry 

Foundation Class 

IFC is a stable, open specification of a consortium 

of vendors and large users of building data. 

Even though IFC is very thorough and 

complicated, simple profiles for interior spaces 

have been developed 


36

Vendor Neutral Exchange: IFC 

Industry Foundation Classes 


International Alliance for Interoperability (IAI 

BuildingSmart) 

Communities: 

Architecture Engineering and Construction 

Facilities Management 

Capabilities: 

Very rich semantic model for building systems and 

relationships 


EXPRESS / STEP, XML 

Adoption: 

Very broad adoption in authoring tools for Building 

Information Modeling (BIM) 

Requirement for many big building customers e.g. U.S. 

General Services Administration 




37

IFC Model Views may be Very Simple 

Slide taken from NBIMS V1 



38



Solution: 

The makers of BIM authoring tools like Autodesk, 

Graphisoft and Bentley and others have implemented 

import and export of IFC for exchange of interior 

space data. 

ESRI should make an IFC importer to populate the 

interior spaces model. 

BIM authoring tools should export COLLADA 

models of building skins at different levels of 

detail with attributes, ready for import into GIS 

data models. 


39



More Problems 

What are some more problems that we can expect in 

integrating building models with other elements of city 

models? 

Building models and terrain models need to meet 

each other at the right place witout awkward 

intersections 

We need to model terrain at different levels of 

detail 

Need to model vegetation at different levels of 

detail. 

Buildings and roads and other outdoor features 

have important relationships. 

Many different communities and specialized tools 

will be necessary to build these models. 


40

Vendor Neutral Exchange: CityGML 

City Geographic Markup Language 


Open Geospatial Consortium (Best Practice Specification) 

Communities: 

Photogrammetry 

Municipal Geospatial Infrastructure (Mainly in Europe) 

Capabilities: 

Very rich semantic model for city objects and relationships 

5 predefined levels of detail 

Image textures supported 


UML / XML / GML3 Profile 

Adoption: 

Berlin, Bonn and many other German cities 

Compatible with OGC Web Feature Services 




41

CityGML is Adapted for Modeling Semantics of 

Observable Objects 

Robot: 

Mobilerobots.com. 

Photo: 

DirectionsMagazine 

www.aeromapss.com/lidar.htm 



Paul Cote. ESRI UC 2008 42

CityGML: Engineered objects relate to various 

terrain surfaces 

Image from OGC CityGML Discussion paper, Kolbe, Groeger, Czerwinski 




43

CityGML: Objects May Honor Specific Levels of 

Detail 





44

CityGML can provide a Spatially Exhaustive 

Semantically Rich model of the city 





45

CityGML: Transportation Objects 





46



Conclusion: 

To build data models for buildings and cities, we will 

build a container for many different types of features 

that are inter-related. 

We should: 

Avoid the usual home-grown, seat-of-the-pants 

problem solving strategy and participate in 

community-based data modeling work 

Anticipate the need to exchange data between 

different specialized software tools 

Encourage our vendors to support community 

based data modeling efforts and vendor neutral 

formats for data exchange! 


47

References and Resources 

Demo Data and Full Documentation: 

http://www.gsd.harvard.edu/pbcote/research/brookline 

Building Information Spatial Data Model: 

http://www.bisdm.org 

Industry Foundation Classes (IFC): 

http://www.iai-international.org 

http://www.facilityinformationcouncil.org/bim/ 

City Geography Markup Language: 

http://www.citygml.org/ 

Contact: pbcote@gsd.harvard.edu 




48

A Data Model for Representing Cities in Three Dimensions - Esri

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?