How can GIS evolve to interact more with process models to provide better insight into change over time and space?

It is imperative in this time of global challenges that GIS evolve to analyze phenomena over time and space. This week’s question is paraphrased from a paper written by Carl Steinitz of Harvard University’s Graduate School of Design. The paper asks a number of compelling questions and provides a framework for GIS modeling. The fact that it was written in 1993 makes it no less relevant today as this problem of linkages still exists.

No other tool can play as critical a role as GIS in understanding the complex interactions between Earth systems and human impacts. It’s becoming ever more clear that man has placed a great burden on the planet and atmosphere, and that these manmade pressures are only increasing. If we’re going to find expedient solutions to these issues, GIS will need to evolve to accept larger datasets, incorporate multiple and in-depth Earth system process models and address dynamic processes across large space and long time.

A Matter of Scale

The issue of scale is a very important part of this discussion as the tools and approaches should vary widely based on the scale of the problem that we’re trying to address.

Global challenges represent issues of massive scale that are very difficult to comprehend and rather hard to model. When we’re talking about problems such as coastline response to global warming, pollution in a watershed and atmospheric affects of carbon emissions, a large-scale model is needed. Also in problems of this size, it becomes necessary to involve a much broader data set from multiple disciplines, and to incorporate multiple analysis frameworks that trade on discipline expertise.

When the issue is to model smaller geographies, such as analyzing the effect of different planning outcomes on a neighborhood or a small city, the modeling burden isn’t so great and the problem isn’t quite as complex. Still there’s an issue of drawing together multiple datasets and involving a multidisciplinary approach.

When getting down to smaller geographies, such as buildings or intersections, a smaller model and individual inputs can be adequate. Here the use of a desktop tool can be quite adequate, and the need for collaboration and communication might not be so great.

Tracking Movement on Machines

The idea of modeling and analyzing dynamic processes over space and time has been around since the early days of GIS. There have been incremental steps along this route to address these issues, but limitations on computing power have typically been the choke point when dealing with issues on a large scale.

There simply isn’t a workstation around, and very few server installations, that can handle the enormous data handling loads of problems that involve both time and space over large geographies. Adding three-dimensional geographic space further increases the computing load, but is increasingly necessary as models improve.

Interestingly, gaming platforms have evolved nicely to handle 3D visualization and have driven down costs while providing increasingly powerful graphics processors. The move toward 64-bit computing and clustered computing has begun to make supercomputing capabilities affordable. There’s also the advent of grid computing with internetworked machines harnessed together to spread the processing loads.

By combining these advances, there’s great promise for taking the choke point of processing out of the equation, and making the processing problem a less-expensive obstacle.

Offline Analysis

Dynamic process modeling on a medium scale requires in-depth and complex geoprocessing and analysis. The deep thinking and complex programming involved in solving these issues makes it a problem that is outside of the realm of most users, yet the outputs of this analysis is needed for policy changes to improve our balance with nature.

The break between those that can build dynamic processing engines for analysis and those that need these outputs requires a new approach. We’ve seen interesting business models emerge for offline analysis in the Building Information Modeling space, where architects submit a model online to a Web-based software as a service (SaaS) outfit and in return receive detailed reports.

Green Building Studio, which was just purchased by Autodesk, is one such outfit that takes an architectural plan and returned in-depth energy analysis after running the model through a complex analysis engine, aligning the design with the LEED standard for energy efficiency. This SaaS approach takes the burden of the computing load and the expertise needed for the model away from the end user. This approach also ensures that the expertise of the analysis engine will constantly improve as more models are passed through the system and a better grasp of parameters emerges.

This approach is a compelling and reasonable alternative for geospatial modeling of dynamic processes at a neighborhood or small city scale. It concentrates the expertise in one organization for consistent improvement of the model, while taking away the unrealistic expectation that non-experts will be able to use a desktop tool for trusted analysis in areas that are outside of their professional expertise.

This approach can work well for large-scale geospatial problem solving in areas of land-use planning, hydrological modeling, transportation planning, environmental impact analysis and the like. It’s at this large scale where problems of process power and data analysis become very complex.

Better Simulation

What about tools to offer insight to dynamic issues on a local scale? When we’re talking about a very local issue of the interaction of one valve on a water network, one intersection in a transportation network or one transformer on a grid, these problems needs to be modeled in time and space, but don’t present an overwhelming processing burden. The issue of these local problems is more a need for simulation than it is of in-depth analysis.

Here’s where a better simulation model that incorporates 3D reality above and below ground could come into play. I viewed such a prototype simulation window at this week’s Autodesk event. The simulation was used to model the water load of a network both before and after a building went online. When the valve was shut off, the simulation displayed the affect on the network – showing how water loads along that node scaled back and eventually dissipated.

This local simulation across both time and space provided an incredibly compelling visual that imparted knowledge of how water reacts in this network and provided clear visual confirmation that the planned action would return the desired result.

Simulation at the local scale is a means to quickly put actions into context. Advances along these lines will give GIS practitioners a much better understanding of the time and space realities of their networks and assets. Simulation will also make apparent any deficiencies in the model, and will help push the need for better data quality right down to the individual user who can see the effect of poor data when looking at a model or simulation view.

The issues of time and space and process models is certainly not a trivial problem on any of these scales. The local simulation issue seems to be the closest to resolution for a software tool perspective as it’s the easiest one to coordinate and implement from at technical sense. Issues of medium scale are predominantly a matter of combining disparate domain expertise and modeling that I believe could be most easily addressed from a service perspective. When you get to the large scale of climate and watersheds, it becomes a matter of great scientific import that will be driven by policy decisions and investments from government. There’s certainly the expertise to solve these big-picture problems, what’s lacking is a mandate and funding to push us toward this goal.

Read what Jeff Thurston has to say on this subject here.