Raster GIS Tutorial (Earth Shelter)
This tutorial serves several purposes: First, it demonstrates a methodology for turning questions and intentions regarding the real world into conceptual models; how the concepts in a model may be represented with data and functions in a data model; and how a data model may serve as a laboratory for experimentation. Most importantly, this process will end with a mentality and a method for understanding and discussing the level of confidence we have in the model, and for evaluating the utility of a model, either as a means of generating useful infomration about a place and alternative futures, or at least as a means of understanding something about the challenge and potential for making useful models of the data-world that help us to know useful things about the real world.
This slideshow is currently under construction!
On a secondary level, this tutorial takes us through several layers of technology: An understanding of raster data structures and the functional components and grammar of map algebra, that transform and derive associations from raster layers, cells, zones and neighborhoods -- to represent spatial relationships among concepts. Finally, we will look at some ways theat map algebra functions can be linked together to make elaborate chains of logical operations that may be tweaked and re-run to perform experiments for investigating alternative strategies for underswtanding or changing the landscape.
- Right-Click Here to Download the Concord Sample Dataset Unzip the contents to c:\projects, then open the map document earthshelter\docs\compilation.mxd in ArcMap 10.0 or later.
- Place Based Models in Scholarship and Decision Support
- Understanding and Critique of Place-Based Data
Purpose and Question
In beginnning any exploration of models, it is important to have a clear statement of purpose. Without a clearly stated purpose it is pointless to try to evaluate anything.
Power in the Landscape: The Pilgrim Pueblos Project
The rising cost of fuel and the problems caused by burning fossil fuels and nuclear power have led us to an increased appreciation for renewable energy, and in particular passive solar design. After some research, we have determined that some sites are better for passive solar homes than others. We want to buy up select sites around the country to build housing developments that will have the greatest passive solar potential. We plan to create a system that will be national in scope. Ew will employ a Mashup similar to HousingMaps.com, which will automatically process real-estate parcels that come on the market. We will take the location of the parcel and use data of national extent to flag filter and flag parcels that have high earthshelter potential. Flagged parcels will be examined more closely by a trained map interpreter befor being recommended by a site-visit and potential offer.
We begin with a few simple criteria for establishing the potential of a site:
We seek sites having the following properties
- Building Sites must Have Sufficient Slopes to take advantage of eath-sheltered design and passive cooling in the summer.
- Slopes should Be More or Less South-Facing for best solar heat gain in the winter.
- Should Have Forested Areas Upwind in order that trees can provide shade and slow down the winter winds
- Building sites that Are Directly Up-Slope from Drainage Features will not be rated so highly.
- But building sites Having Potential Water Views will recieve higher marks.
- Building sites that have High Accessibility to Commercial Areas will be flagged for special scrutiny (for our more urbane customers)
Note that each of the terms highlighted in Bold in the conceptual model involves a term of fact, e.g. Commercial Areas and a term of relationship, e.g. High Acessibility. Part of our task will be to find data to represent the static facts, and procedures to represent the relationships in theis model. This arrangement of data and procedures will be our Data Model Our data model becomes a laboratory for experimentation when we alter aspects of the facts and the relationship and think logically about the impacts of various decisions.
Understanding of Models, and Choice of Errors
Part of our goal is to understand how well we can build a model that will work on a national scale, using the concept of a mashup (see HousingMaps.com. This mashup would automatically find listings of real-estate for sale, and then would evaluate each of them for potential for one of our developments, and would flag potential properties for further on-site evaluation. This will requuire somewhat consistent data that is national in scope. But before implementing this national program, we need to develop a pilot test case using the best quality local data that we can find. This calibration dataset will allow us to evaluate the sorts of error we may be facing when we try to buid a model with coarser-grained national data.
Closely examining both models in the same pilot area, will help us develop a level of confidence in the national model. It is important to tink about two important types of error. First is Errors of Omission, which would be a failure to identify a site in our data model as having potential, when on the ground, it actually does. We also can expect Errors of Comission, in which we identify sites in our data model as having potential, but investigation on the ground reveals that they don't in reality, meet our criteria. As it happens, in the implementation of the model, the decisons we make will lead to more or less of each of these types of errors. Choosing which type of error you would rather have is a key element in the design of data models!
Because our model is the first, automated, stage in a site evaluation process, we would rather have a model that is biased toward errors of comission. The second stage of evaluating sites is to use a trained human being to evaluate sites that are recommended by our model. If our model fails to flag sites that possibly have potential, then our analyst will not have as much work, but we may lose an opportunity to examine a site which may, in reality, have potential. The feedback from the analyst about what sites are flagged, either rightly or wrongly, will help us to fine-tune the model over time.
At each step in this procedure we will create functions that transform input data into a scheme of logical values that are driven by our purpose. Some of our criteria will be derived by looking at associations among our transformed layers. At eac step is is crucial to examine the outputs that are generated to make sure that they make logical sense based on an examination of the topographic map and aerial photograph. It is very easy to make a mistake that will generate a ridiculous result. THe absurdity of the result may be obvious when we look at it directly, however it will be very difficult to figure out that we have made a mistake once we have combined the erroneous layer with other steps of our analysis. This dtep-by-step evalustion of our work wil lalso be a good time to think about which factors of data quality or decisions that we make will be critical factors leading to errors of omission and comission.
- An Introduction to Cells, Attributes, Layers and Zones in Discrete and Continuous rasters.
- A simple extraction of information dataset using a boolean operation.
- A simple ad-hoc aspect analysis of aspect with a geoprocessing tool
- Creating a re-useable model for evaluating earth-sheltering capability
- Setting up the modelbuilder environment
- Concepts of Fact: The pattern of Data Selection Reclassification of Values and Weighted Sum.
- Concepts of Relationships: direct association in space with Local Functions
- Re-Using and Sharing Models
- Raster Sampling a vector layer
- Association of Proximity and Accessibility: Incremental Functions
- Discussion of NoData in Incremental Functions
- Fancy Incremental Functions
- Cost Distance (considering surfaces of uneven resistance)
- Path Distance (considering slope)
- Discussion of NoData in the Merge operation. (Sometimes known as Cover)
- Functions of Visibility
- Focal Functions
- Counting Forested Cells Upwind of each location
- Zonal Functions
- Sorting Sites by potential size using RegionGroup (aka: Clump)
- Evaluating a property parcel with a zonal function
- Logical evaluation of the model
- Fine Tuning of the Model
- Experimentation -- sensitivity to varied data quality; evaluation of alternative futures.
Examine a Discrete Value Raster
Cartographic Modeling procedures make a lot of use of Raster Layers, that represent evenly-sapced, congruently-defined locations on the ground as Cells. In a single layer, each cell is tagged by a Value, which may be used to disctiminate various different discrete types of locations known as Zones; or surfaces that vary from cell to cell, in Continuous fashion. The regular relationships among cells allow for many powerful ways to create and use logical relationships among locations and their properties, as we shall see.
The New ENgland Gap Vegetation (Gap_Veg) layer is a discrete value raster. There are several ways to evaluate the raster dataset. For one thing, it has metadata that can be found in its folder (or click here to see the metadata) But even without any metadata at all, we can learn a lot about this layer by examining its properties and its logical consistency with other layers.