Site Modeling in Context

Broad-Scale 3d Site Models with AcrGlobe

Developing Rough 3D Site Models with GIS Data

What do we do when considering a design project on a site? We probably go and visit the site, and take some photographs, and get a feel for the place and its surroundings. Next we may go back to the office and gather together whatever information we can eaily find about the place, its context, its history, etcetera, and we try to organize and assimilate these pieces of information for our own reference, and as a means of discussing aspects of the site with others. Our compiled information will also serve as a base upon which we may sketch out scenarios.

Our representations of potential design scenarios in the context of what is and was there will help us to understand the potential and limitations of the site in terms of its scale, and its connections to its surroundings in terms of circulation and views. Not only will these initial sketches help us to understand these things, but they should help us present these possibilities effectively to others as well.

This tutorial takes a look at the initial compiled data for a project at Fields Point, in Providence Rhode Island. We have compiled layers of information from various sources, with careful attention to their spatial referencing systems, such that they all coincide in a single system. This system lets us look at the relation of land cover, buildings and terrain in three dimensions. Having arranged this information, and having explored the site on the ground, we should be able to prepare representations of the site that help us to point out the important aspects of the site -- including its scale and its capacity for various types of development and its limitations and areas of critical concern.

The Tools

The tools we will look at in this tutorial are ArcGIS and ArcGlobe. ArcMap was used for transforming the datasets, and we will use ArcMap to create and edit information. ArcGlobe is the tool that portrays data in three dimensions and permits us to move around the scene to capture images and animations. The following software manuals will be indispensible for understanding the procedures coverdd in this tutorial:

Compiling Information

The work of site representation can be done with scaling xerox machine, exacto-knives, rubber cement and chipboard and trace paper. There is also a wealth of information available in digital form and a few tools from the world of Geographic Information Systems that are useful for compiling spatial data, either in paper or digital form to prepare rough three dimensional models of sites, their context, the things that we find there, or propose to build. Though these GIS tools are rough compared to more sophisticated three-dimensional modeling packages, GIS is a logical place to begin in the process of bringing together and transforming spatial data from various sources. ANd we can use this GIS-compiled data as a starting place for the development of more sophisticated 3d site models.

Working with data sources in ArcGlobe is very similar to ArcGIS. There are a couple of important new things to learn, including the difference between Draped, Floating and Elevation Layers, and how to set the extrusion property for layers.


Data Sources

We will begin by taking a look at various resources that we can scrounge together. Right-Click and Open this Link to download the compiled sample dataset. Extract the contents of this zip archive to your C:\temp\your_username folder, . This data collection is organized according to the principles for organizing a place-based data collection for a collaborative project. It includes a lot of resources we have gathered to help understand the site.

You should look into the gis/pbcote/data/stb_sox folder and open the two documents in the docs folder to see the data in ArcMap and in ArcGlobe.

Scanned MapsWe begin with the basics -- paper maps, such as we may find in the library. These maps will help us to understand the contrext of our site, and something about its history. In this case, the site is Fields Point on the Narraganset Bay on the south side of Providense Massachusetts. We can see from some historic maps for 1894 and 1939 and the present, that Fields Point has not always been there, but has resulted from successive epochs of filling. The current USGS map was obtained from the Natural Resources Conservation Service. This site offers seamless coverage of the united states at a scale of 1:24,000 and the maps are georeferenced which makes them line up with opther properly referenced data sources. The historic maps were taken from the Dimond Library at the University of New Hampshire. To get these historic maps to line up with eachother, they had to be georefernced Many sources of geographic images can be found at the GSD's GIS manual page of Sources of Geographic Images.

Aerial Photographs: It is expensive to hire a helicopter to help you gather aerial photos. And photos taken from the air always have so much distorion that it is impossible to tile together a mosaic of photos of a site that has the sort of scale as we would expect from a plan. Luckily for us, the United States Geographical Survey collects fairly high-resolution aerial photograph which it transforms into planietrically correct maps and makes them available abailable at low cost, or for free from Micorsoft's Terraserver Note that there is a lot more imagery available from the USGS site, if you don't find what you need from terraserver.

We obtained several images form terraserver and mosaicked them together to provide a low-resoution (8mtr) image of the site context, and a higher-resolution image of the immediate environs of the site. These images are georeferenced, and if we understand how to identify their projections correctly, then these photos form a highly accurate and up-to-date reference for what is on and around our site. According to Terraserver, these images were taken on April 4, 2002

Terrain Information: so far, everything we have gathered is two-dimensional, though the lay of the land IS implied in the contours of our USGS topographic maps. These days, the USGS also makes available information about elevations in the form of Digital Elevation Models, which are rasters, the value each pixel representing its elevation. Elevation models like this are available for most of the world. To learn more, look at the GSD's On-Line manual page Obtaining and Transforming Digital Elevation Models. This elevation information can be used to add another dimension to the pixels of our 2d aerial photograph, by setting the Elevation Property of the image to Float above the globe surface on the elevation surface provided by our terrain model.

Creating a Reference Image for ArcGlobe

ArcGlobe can take vectore datasets georeferenced images and drape them onto a terrain surface defined as a raster elevation model. Doing this with images is much less taxiing on your computer's graphics card than it is with vector geometry. This is why we will take the basic reference image representing the ground condition and compose it in ArcMap and then export it as a high-resolution image that will cover our terrain model in ArcGlobe. This is a simple matter organizing the map the way you need it, then exporting it as a JPEG file of approximately 3000x3000 pixels with a world file. I recommend using an aerial photo that is tinted with a land and water mask as shown in the demo dataset. Putting contours on the map will make it much easier to find the critical spots in the terrain when you are flying around. This file will be georeferenced so it will line up with other layers in arcmap, but to make it line up with ArcGlobe, you need to define its coordinate system, which will be the same as that of your data frame.Its worth noting here that you can make copies of this imsge with arccatalog and you can modify the copies in photoshop to reflect changes you may be thinking about. These new images will remain georeferenced so long as you don't lose the .prj file. You also may need to rebuild the pyramid layers of the copy after you alter it in photoshop, otherwise your changes may not show up at different scales.


Vector Representations of Buildings and Trees: In most places, the local utility companies of the local government has done a recent photogrammetric survey to record the locations of buildings, and edge of pavement, and sometimes trees. In the case of Fields Point, we found some of this data which had been digitized in AutoCAD for an earlier project. This information had to be shifted about 49000 feet to the west, but once we figured out the right number, it lined up perfectly with our other information! We then called upon a nice topological function of GIS to turn the unclosed polylines from CAD into nice closed buildings. We did a similar thing with the trees, and for reasons that will become clear later, we used a generalize function on the tree polygons.

The elevation of our building footprints can be derived from the elevation model, just as we did with the airphoto. The buildings may then be extruded to a constant height. These representations of the houses to the west of Fields point have a problem of logical inconsistency. Even though we don't know the exact height and color of each of these buildings, we can be almost certain that these houses aren't all the same height and color. So we can use GIS to assign random values for the heights in the attribute table for the height, and color, and then these attributes can be used to vary the display of the buildings. A similar technique could be used to make a variety of different tree symbols from the tree canopy data. This is left as an exercise.

Adjust the Display properties of ArcGlobe

One of the neat things about ArcGlobe is that it can display very broad-scale 3d models with lots of detail. This is accomplished by managing the level of detail in the scene. Our model is not really very big, so we can adjust the options for ArcGlobe to display all of the detail in our digital terrain model and in our image to allow the maximum amout of detail to be shown.


  • Use Tools->Options->Cache to turn the level of detail for Images and Elevations all the way up.
  • In the Cache properties for your stb_sox_1 raster elevation model, uncheck the option to Compress Value Range to 16 Bits
  • IN the Cache properties fo your stbsox groundplan image, turn off the 16bit compression option and also the lossy compression option.

A sense of the Place

So we begin to get a picture of this Fields Point by compiling together some data from various sources, and traansforming it in various ways using GIS. What we would like to do is use these data to get a sense of what it is like to be in and around our site. A sense of the scale and the capacity of the site. For example -- how large a visual presence is that big oil tank (or whatever type of tank that is) when viewed from various points around our site? How big a building could we put on this site? What would the visual impact be of a new minor-league baseball park if it was loacted on this site?

Site Photos: Of course, no amount of desktop information can substitute for a visit to the site! Because of the scholarly photographic activities of people like Alexey Sergeev we have found some very useful site photos: (1), (2) on the web by doing a google image search using the term "Squantum Point" which is accross the river from our site. Alexey says that these photos were taken March 7 2004 from the East Bay Bike Path.

Another search for providence river turned up this oblique image which seems to have been taken much earlier, but has a good view of the gantries (the booms used to load ships.)

Click on these links to open up the images at theior full size: Squantum1.jpg, squantum2.jpg, port_oblique.jpg.

So now we have an idea of how tall the tanks are. Obviously, we are going to have to add those tanks to our model, and if we can set up our camera right where Alexey stopped to take his picture, we can use Alexey's image to help us get the tanks the right size. At the same time, it will help us get a feel for the various tools for setting our camera where we want it. Our first approach to camera location will seem a little awkward, but it will get more fluid once we learn the basics.

Our object is to recreate the views as seen by Alexey. This will start by placing our point of view in ArcGlobe at the X,Y location where he parked his bike to take a picture. We also want to make sure that our height is approximately eye-height. For this purpose, we have set up a point featureclass called Observers and have added a couple of points to it, and have adjusted the symbology of the points to be little balls, floating 2 meters above the terrain surface.


TO begin setting up this view, we need to find the spot where Alexey was standing Ehen he took his photographs. We have placed a point there in our Observers shape file, and this point has been extruded to 2 meters, to serve as a landmark to the exact position of his camera. Using the Zoom to Target tool we can get close, and then orbit the scene to get a subterrainnean view, to see if we need to adjust the offset of the observer points in the layer's Elevation Properties. NExt, we will get back above grouns, use the zoom to target tool to zoom right to the point where Alexey meets the ground. Then use the Pan Tool to move the center of the scene down, in order to climb to the top of Alexey's head. Now that we have got the center of our scene at alexey's foot, and out camera positioned at his head, we can use the Navigate Tool to look around.

Now that we know how to place the camera and point it where we want, the next thing is to learn to control our frame -- using the Narrow and Expand Field of View buttons, to effectively change the focal length of our camera's lense. We might understand this process a little better if we pause here to consider that the buttons that ArcGlobe calls Zoom actually move the observer; while the Narrow and Expand field of View buttons actually have the effect of a Zoom Lense.

Positioning the ArcGlobe Camera

Positioning the camera in ArcGlobe is accomplished with the zoom and set target tools and with the pan tool.

  1. Move your cursor over the tools in the ArcGlobe and find the Navigate tool, and the Zoom to Target tool and the Pan Tool and the Zoom Tool and the Navigation Mode tool.
  2. Right Click on your Site_1m to zoom out to the extent of our hi-res orthophoto.
  3. now use your Navigate tool, which in other programs might be called the orbit around center tool, to pivot your scene into perspective.
  4. Now click the Navigation Mode Toggle to switch to globe mode. Notice how this affects the behavior of your Navigate Tool Now switch it back. Because ArcGlobe starts in a full globe view this is the default setting for ArcGlobe
  5. Now use the Zoom to Target Tool to zoom to Alexey's foot.
  6. Orbit the view back and adjust the Elevation Offset property of the observer's layer to fine tune his elevation if necessary.
  7. Now use the Pan Tool to move the top of Alexey to the center of your view.
  8. Now use the Navigate tool to rotate the scene around Alexy's head.
  9. If you find that your view is pivoting around an inconvenient point, You can use the Center Target tool to choose a new pivot point. You must click on the elevation model, and then move yourself up with the hand tool.
  10. Here is the place to point out that in a perspective scene, if we aren't looking right at the horizon, then things get wonky.
  11. So, after you have orbited the scene a little bit you may need to use the hand tool a little more to adjust the relative position of Sergey's head with respect to the actual horizon, which may actually be beneath the slightly elevated skyline.
  12. Finally, use the Field of View tools to zoom in and out.
  13. Create a Bookmark so that you can easily return to this viewpoint.

A new Scenario

So now we have a reasonable representation of our site, lets see if we can use it to learn anything new. What would a minor league baseball park look like on our site? We will consider The PawSox Park in Pautucket This Map shows us the basic shape of the park, and using the fact that the distance between Home Plate and Second base is 127 feet, we can judge that PawSox park rill roughly fit in a rectangle that is 5 x 127 by 4x 127, or approximately 200 meters by 175 meters.


So, the reason we have put all of this together is not only so we can understand the existing condition on the site, though this might be a sufficient reason for building the model, but we also would like to use the model to investigate alternative scenarios and how they might relate to the site and its context.

So here is a rough representation of the PawSox minor league baseball stadium as it would fit on our site. Here is how that stadium might look from across the river at Squantum Point.

Click hee to see an amimation

Click here to see how to embed an avi animation into your pdf document!

I have created a pawsox_3d geodatabase in the stb_sox folder (which you now Within the Scene feature dataset, you will find 4 feature classes that have been made to store buildings, object_points, camera_tracks and observers. YOu may edit these to add your own buildings, with their heights, the points, added to the object_points layer, with their 'name' attribute used to assign a symbol, such as any of the 3d symbols available under More Symbols option in the Symbology Properties of the arcGlobe layer. Take a look at how your object_points layer defines the 3d symols for the house and tea house.

The new Pawsox hill scenario includes its own raster elevation model, named stbsox_1, and a new buildings layer, named my_buildings, and a new groundplan image, named stb_sox1_groundplan.jpg. You can look at these with your Table of Contents set at Type mode, to see how they are defined as Elevation layers and Draped layers, apropriately. Back in Display Mode, you can see how the Group Layer, Pawsox Hill allows you to turn on and off this scenario and that the elevation layer for the new scenario supercedes the larger terrain model of the existing condition.

YOu can edit the my_buildings feature class in arcMap, but remember, Once you edit the features and their attributes, remember to stop editing, and reload the layers in ArcGlobe by right-clicking on the layer and choosing Reload.

Creating Animations

See Notes on Making Animations and Troubleshooting Problems with ArcGlobe Your challenge is to make an animation that moves the camera from one place to another and swaps the pawsox grop layer on and off to show a perspective with and without baseball park and the new hill.

We want to help our clients understand how the new proposal mediates views into the site, across the site, and creates prospects within the site. To do this, we will create bookmarks for the specific viewpoints we are concerned with. Normally, these would be places where we would make a perspective drawing. One problem with perspective drawings is that sometimes it is difficult for the client to understand where the perspective viewpoint has been taken. So we will use an animation to assist with this task.

First we will plan our shot. The idea will to be to start with a plan view that we will capture as a still image by exporting the ArcGlobe scene as a jpg. This can be marked up with diagramatic details that explain the concept of where the viewpoint is, and where the hill is, and how the hill blocks the view of the tank. Our animation will start from this exact same viewpoint. Then we will define five Camera Track Keyframes that swoop gracefully to our viewpoint, then pan from left to right and back again, then to fly back to the plan viewpoint. Ending the animation where you started is a nice toucn that keeps your audience oriented and lets you begin your next animation always where the last one leaves off. We will also create three Layer Track keyframes that turn on and off the Paw Sox group layer. We want the pawsox scenario to begin Off, then turn on, then turn off again. The next thing we will do is use the Time Track to choreograph the layer keyframes and the Camera keyframes so that the pawsox layer turns on and off at the apropriate times.