The spatial data model is the heart of a GIS and CAD system. It is more than data alone. More than format. The data model governs how well your software will perform useful functions because it represents your depth of understanding of the process(s) that you are attempting to apply spatial data and associated geoprocessing functions toward. It is not a random happening that modeling and simulation lie so close to GIS – they are a manifestation of this realisation. This also explains why data translation is an art – not a simple database operation.
There are raster and vector data models, but a spatial data model includes not only the model for storing data, but also a consideration of the structure, organisation and rules relating to the information for a particularly application. A connection exists between this organisation and the associated rules etc. That is, specific types of processes are associated with different events. A spatial data model that allows water to flow up hill is likely not a good model. However, one that allows water to flow downhill and takes hydrological processes into consideration, probably is. A while ago I had the opportunity to review the Arc Marine model which models the processes within marine environments. But not all data models are restricted to marine and hydrology.
CAD software also have data models. One interesting example of this can be found in the GenerativeComponents product from Bentley. The product is a modeller and is used for advanced building design. In concept this product allows different ‘what-if’ design scenario’s, but follows the logic and rules that building construction geometry generally follows (ie. a roof must be supported, certain designs are impossible to construct). The MicroStation product enables this software to run, serving as the platform.
There are spatial data models for many spatial processes. For example, transportation spatial data models may include the movement of physical assets (trains, cars, boats), but also couple population movement, weather and cost, into the overall model.
It takes time, resources, testing, understanding and financial investment to create a high quality model. This is why we see so few of them. As you might imagine, there is more to a spatial data model than the data alone – there is also the process associated with the data, which also needs to be considered.
However, in considering the process, we also need to ask, “is this data going to help us construct the model (thus reality) we are trying to create to represent reality?” Therefore, it is important to consider the type of information that is needed when building the spatial data model. For example, it is hard to model tree growth for forestry purposes without any information based on tree growth – also keeping in mind all trees do not growth the same way in all places. Or, to develop a spatial data model for oil transport through pipelines without an understanding of oil production for a region. It would be difficult to develop a windpower model without topography and data relating to wind.
Different spatial data models can perform better than others. Performance matters. And, this is one reason why we see different formats. If you were designing a spatial data model for agriculture, wouldn’t you want to differentiate your product from others? Wouldn’t you want to build the highest quality agricultural model possible? Probably, right?
One way to do that is by understanding the agricultural process well. Then setting out to develop the spatial data model that best represents all the processes relating to the farm operation involved. This might include the type of data, type of sensors to collect it, resolution of data, biological factors related to plants and even the weather. It could include tillage, seed sources and types, soil properties and other factors. With all this information a spatial data model called ’spring wheat’ could be built. But as you construct the model, it becomes apparent that certain kinds of data organisation, rules and relationships must be integrated into the spatial data model for it to operate efficiently. This leads you to pursue these enhancements more and more, until, you have developed a unique approach for storing and using the information.
The balance of performance against standardisation is where data translation and transformation comes in. It is not difficult to translate the ’spring wheat’ model between different formats. But, unless that transformation includes the soil properties, seed sources, data characteristics, weather and other factors – which it was constructed with – then the spatial data model loses fidelity, its truth.
This is also the reason why we don’t use GIS to design complicated building structures. The design modeling for structures follows specific rules, understanding and organisation, as well. Very, very few software can perform both GIS and CAD operations well, also for this reason. Generally, each imports data from the other and can work with it – and that might be all that is needed in some cases.
This integration of spatial data models is a hot topic these days. But it is not easily solved. At least we have moved beyond the format arguments alone, which is important, because it allows us for the first time to understand underlying processes arising from GIS and CAD environments – thereby connecting structural / infrastructure design to spatial geo-processing.
It is important to understand spatial data models in principle and what they mean, particularly when moving data and processes. Software are optimised for performance according to specific spatial data models.
What if a global modeling environment were to offer compatibility with multiple file formats?
Interview: CityGML – Modeling the City for the Future
Book Review: Arc Marine – GIS for a Blue Planet
1Spatial Conference – Stansted, UK: Data as Foundation
Interview: The Relationships of Spatial Processes to Data Quality
A well thought out spatial data model is critical to get the most out of geographic information systems (GISs), because it dictates how spatial data are stored and represented within the database, and the rules for how the data can be analyzed and manipulated. In addition to different data models to represent vector or raster data, the data model is also the means to create a common set of attributes, rules and workflows for specific application areas.
Determining your spatial data model is a foundational step, and increasingly this is accomplished collaboratively with a set of like-minded peers in the industry at large. It helps to look outside of your organization for assistance here, because standardized spatial data models will greatly enhance interoperability among many users and will ensure that you get the most from your system.
Models within the Model
The term model can get quite confusing, given its use in a number of different contexts. The word model crops up in the spatial domain when referring to how data is described and stored in the database or how a drawing represents reality in 3D. There are also many different models within a spatial data model that parse different elements of the data and how it can be used within a system.
A conceptual or semantic model describes the elements of significance for a specific purpose, including attribute characteristics and relationships between attributes. The logical model represents business requirements with definitions and examples that prioritize importance and how elements relate to each other. The physical model describes how the logical model is represented in a database, with tables, columns, rules, storage procedures, etc.
Key to Integration
Spatial data models can apply to industries or to specific problems or purposes. There are standard data models for such communities as agriculture, forestry, geology, local government, etc. There are data models for such data types as addresses, land parcels, base maps and hydrolology. And there are data models for specific analysis functions such as carbon footprint calculations and biodiversity.
The data model facilitates integration of data to and from other systems and collaboration with other users. Coming to consensus on a data model greatly enhances the utility of the model because it then meshes with other representations in other organizations. The commonality of the model assists advanced application, because analysis tools and workflows can also be standardized across domains leading to easier and more reliable analysis.
A good foundational spatial data model is important, but that doesn’t mean that the model needs to be static. The spatial data model can be extended for additional purposes down the road.
Coincident to our tackling this week’s question, ESRI announced the Building Interior Space Data Model. This announcement is illustrative of the importance of a data model for the extension of GIS functionality to new application areas. Specifically, this data model creates a means to ingest information about the interior of buildings within a GIS, and opens up a whole new area of application for facility information management. The data model was painstakingly crafted to ingest CAD and BIM data, to work at multiple scales, and to integrate with other systems.
How spatial data is modeled in the database has huge implications for the sustainability of your system, the effectiveness of the system, and your return on investment.
The GIS Spatial Data Model, GIS@UW
Download ESRI Data Models
ESRI Building Interior Space Data Model (BISDM)