Grid Cell Modeling: The Other GIS

June 1, 2008  - By
Most real-world datasets are continuous, and therefore more accurately displayed in a grid cell-based GIS than as points, lines, or polygons.

By Art Kalinski, GISP

Back in the mid-1980s, when I established the U.S. Navy’s first GIS, we used mapping software from a company called National Planning Data Corporation (NPDC). In the process, I had several interesting GIS-related discussions with NPDC’s founder, Peter Francese. His observation was that as we’ve grown in knowledge and sophistication, we’re actually substituting information for resources. He used the telephone as an example. If you’ve ever handled a ’20s-era telephone, you may remember that it weighed a very heavy 10 to 12 pounds because it was made from copper, brass, steel, and lots of Bakelite (one of the first synthetic thermosetting resins).

There were only three things you could do with that phone: dial, talk, and listen. By comparison, in the ’80s phones had evolved into one-pound devices made of lightweight copolymers and integrated circuit chips that featured memory, autodial, and speakerphone. What Peter observed is that we substituted our growing knowledge of plastics and integrated circuits for traditional materials. Today’s four-ounce cell phones continue that evolutionary model.

GIS has evolved in a similar way. With GIS, we are substituting spatial knowledge and analysis to use resources more efficiently, whether it is military effectiveness, forest management, mining, oil exploration, or transportation. Despite the growth of GIS and spatially enabled applications, surprisingly few people have augmented their traditional point, line, and polygon GIS with more sophisticated spatial tools and applications, such as grid cell modeling or raster-based GIS.

Most are familiar with raster image processing software such as IDRISI or ERDAS but few realize that they also contain strong modeling and analysis tools.  The majority of GIS users operate in an ESRI environment but only a few take advantage of grid cell modeling found in ArcInfo GRID or Spatial Analyst.

Polygon GIS vs. Grid Cell GIS.

Polygon GIS vs. Grid Cell GIS.

I agree that the original ESRI software GRID was not easy to use. I continue to be thankful to Chris Cappelli of ESRI who helped me learn ArcInfo 6.0 GRID when I was working on my master’s degree at UNC Charlotte back in ‘92.  Likewise, if you ever had to read Dana Tomlin’s book Cartographic Modeling, which was a key publication developing the rules of grid cell modeling and Map Algebra you may remember how deceptively simple it seemed and how the learning curve shot into the stratosphere half way through the book.

Why bother? you say.  The big reason is that most information you work with doesn’t have discrete borders. We constantly display demographic data, noise footprints, trade areas, soils, elevation, medical, environmental, biological and atmospheric and data sets as Points, Lines or Polygons. Yet in the real world the only certainty is death, taxes and the political boundary that defines the taxable footprint. Most datasets are continuous and don’t have clear discrete boundaries.  I can show you the edge of my property but I can’t show you a clear boundary of moisture content in my lawn.

Want to see the value of displaying continuous data as continuous data rather than a generalized polygon? Look at these polygons, now roll over the polygons to see the data as a continuous dataset. You can see how limited your understanding of the data is with simple polygons. Continuous data fills in the gray areas between and paints a more understandable picture.

Roll over the blue/green polygon to reveal the continuous gray tone eye.

Roll over the blue/green polygon to reveal the continuous gray tone eye.

Roll over the blue/green polygon to reveal the continuous gray tone eye.

Roll over the blue/green polygon to reveal the continuous gray tone eye.

Why can’t I use Points, Lines and Polygons to do my analysis? You can, and using tools such as joins, unions and intersects will do simple spatial data analysis. If you need to work with an area of continuous data the best you’ll be able to do is a series of buffer polygons that approximate the data.  But even more important, if the interaction of the datasets is a complex mathematical model, then a traditional GIS will reach its limit quickly.

Remember that in a traditional GIS not only is a polygon defined as a series of vertices and arcs but the software also has to keep track of the topological relationship of the features.  That’s a lot of overhead to maintain. By comparison, a grid cell based GIS is made up of a large matrix of cells that are consistent in size and location. Just like the computer screen you are viewing the only thing that changes are assigned values of each cell. This makes processing extremely fast, especially on large datasets

This is the critical difference between a polygon based GIS and grid cell based GIS.  Several years ago I remember seeing a community planning software called Index that appeared to use grid cells. The hope was that it could be used for MPO regional transportation planning. The problem was that it was a traditional polygon GIS that only looked like a grid based GIS because it used square polygons.  Since each cell had to carry all the topological baggage of a polygon, it was extremely slow and crashed on all except the smallest size city.

A true grid cell GIS is very fast and capable of digesting some very large datasets. I’ve seen some very effective site selection programs that take multiple layers of grid data to determine the optimal characteristics of successful sites and search a new region for locations that meet the same criteria. John Calkins, ESRI’s expert in GRID and Spatial Analyst cited numerous examples ranging from site suitability work for oil and gas exploration to an ingenious effort to combat terrorism using “Human Terrain” modeling that identifies locations of populations by religious, political and ethnic background. A similar effort was very successful in identifying drug traffic sites in US cities almost as soon as established.

Drilling through multiple layers - ESRI             2D or 3D “surface” from a mathematical function.

Drilling through multiple layers: ESRI.

2D or 3D “surface” from a mathematical function.

2D or 3D “surface” from a mathematical function.

But where grid cell modeling really shines is the ability to get the cells to react to adjoining or nearby cells based on simple or very complex mathematical functions. The bottom line is that if you can describe what you want to happen as a mathematical formula, grid cell modeling can do it. Simple formulas like gravity models used in location analysis or very complex relationships such as the behavior of forest fires are examples of grid cell modeling work currently being done.

So don’t be stuck in the Point, Line and Polygon GIS.  Dust off your old GIS text books and I’m sure you will find a chapter on grid cell or raster based GIS. The good news is that with programs like ESRI’s Model Builder the process is now much easier. As GIS users become more numerous and sophisticated we need to stay ahead of the curve.  Grid cell based GIS may be one way to do that add new visibility to your GIS operation.


This article is tagged with , and posted in GeoIntelligence Insider, Mapping, Opinions