To create a suitability map, data that defines the criteria must first be prepared, transformed to a common scale and weighted and combined. From a suitability map, the best location to site or preserve can be identified. While this weighted additive process is well established, it is subject to inherent constraints. The Evaluate environment provides a way to explore these constraints, so you can then address them. This will give you more confidence in the results when you apply them to make decisions in your modeling.
The Evaluate environment is composed of a series of interacting panes, maps, and plots.
There are four tabs in the Evaluate Pane (the lower center in the image below) when the suitability map is selected on the Evaluate tab (the Suitability Modeler pane in the image below) and two panels when the locate map is selected. The Evaluate Pane is divided into two panels. The left panel, for the Overview tab, it is identified by Explore model input and output in the image below, controls the Evaluate suitability map and the right panel, for the Overview tab, it is identified by Explore influence of criteria in the image below, controls the Evaluate criteria map. Use the left panel to explore the composition of the weighted transformed criteria values that produce the suitability values. Use the right panel to explore how a specific criterion contributes to the model. You can explore the constraints that are inherent to the suitability modeling process through a series of statistics.
For information on the mechanics of using the Evaluate environment, see the Evaluate tab in Suitability Modeler and Evaluate Pane in Suitability Modeler topics.
Addressing the constraints
In the Evaluate environment in the Suitability Modeler, you use a series of tabs and panels to identify and address the constraints in suitability modeling. Each tab and panel are covered in the following sections. The purpose of the tab and panels, when to use them, what constraint they address, and the specific statistics to apply are covered, among other details.
It is important to understand how criteria interact, whether you are exploring individual cells, parcels, observations, or the resulting regions from Locate. To facilitate this, a series of statistics is included in each tab and panel.
Overview tab
Use the Overview tab to evaluate any model. This tab addresses the constraint that the inputs, transformations, and weights can be subjective. Explore the spatial distributions of the base and weighted transformed criteria values as well as the final suitability and locate maps to determine if the specified inputs capture the preferences of the subject.
Overview constraints
There are only statistics in the right or Explore influence of criteria panel on the Overview tab. The following list describes various actions that can be taken to explore the constraints addressed by the Overview tab:
- Examine the spatial distributions of each of the base and weighted transformed criteria values in the left or Explore model input and output panel.
- Determine if any of the transformations cancel each other out (locations with high suitability in one criterion are assigned low suitability in another) or if they all prefer certain locations over others.
- Temporarily remove criteria and examine the effect on the spatial distribution in the Explore influence of criteria panel. Examine the results from various perspectives with the Suitability values, Remaining percent, and Percent contribution statistics.
- Compare the results of removing criteria to the original suitability values.
Criteria tab
Use the Criteria tab to explore how the transformed criteria values interact to create the suitability values.
Standard statistics
The general premise that the higher the suitability the better may not always be true. The original base and transformed criterion values that were entered are lost once they are combined. Two locations can have the same suitability values, and therefore, in the weighted additive approach, they will be equal in preference. Either location can be selected resulting in the same satisfaction. However, knowing the composition of the weighted transformed criteria values that produce the suitability values can have a profound impact on the location you select.
Example
Consider an example of a suitability model for a solar power farm installation with three criteria. There are two viable potential locations that have overall suitability values of 22. These values could be arrived at from the weighted transformed values from three criteria in the following ways:
| Location | Distance from electric lines | Slope | Solar radiation gain |
|---|---|---|---|
One | 10 | 10 | 2 |
Two | 7 | 7 | 8 |
The two locations are both assigned the same overall suitability and may seem equal in preference. However, the composition of the individual weighted transformed criteria values that produced the suitability values are significantly different. If you select location one, the project might not produce as much power as it could.
However, consider the impact if the composition for location one was to be altered to the individual suitability values shown in the table below:
| Location | Distance from electric lines | Slope | Solar radiation gain |
|---|---|---|---|
One | 2 | 10 | 10 |
Two | 7 | 7 | 8 |
The weighted transformed value for the distance from electric lines criteria now contributes the 2 and the solar radiation gain contributes 10. As a result, location one maybe the more preferred location.
Determine the composition
To understand the composition of the weighted transformed criteria values creating the previous scenarios, the following questions need to be answered:
- What is the range of the input weighted transformed criteria values?
The range is large in location one and small in location two. However, a small range may not always be good. The low range maybe of low criteria values.
- What are the highest and lowest values?
Location one has a high of 10 but a low of 2.
- Which criterion contributes the highest and lowest values?
These questions are answered through a standard set of statistics in the Evaluate environment.
Define the standard statistics
There are five statistics that are used to tease apart the combination of the weighted transformed criteria values that produce the final suitability values to address the situations described above. These standard statistics are used on the Criteria tab to explore the combination within cells. They are used on the Summarize within tab to determine the composition of the weighted transformed criteria values within polygons (such as parcels), in the observations on the Validate tab, and the resulting regions in the Locate panels. These five statistics are the standard statistics in each tab or set of panels.
Use the Range statistic to know the range of the values for each location. Look out for situations where a criterion with high suitability is compensating for a criterion contributing low suitability. A high range indicates the suitability values vary greatly. This may not be desired.
The Highest value and Lowest value statistics show what the highest and lowest values in the range are. Low values in areas with a high range may be problematic. It typically means you have at least one criterion that has low suitability.
Identify which criteria are producing the highest and lowest values with the Highest criterion and Lowest criterion statistics. Examine the spatial distributions of each. You may be most concerned with areas where the criterion that is most critical to the model is contributing the lowest values. These two statistics indicate which criteria are driving the analysis and where.
Specific statistics
There are some specific statistics that are distinct to the Criteria tab. Those statistics are listed in the right or Explore criteria values panel.
The Percent contribution statistic shows how much a criterion contributes to the final suitability values. The From mean statistic shows how much a criterion is above or below the mean value. If a location has a high percent contribution, the criterion is above the mean, and it is highly suitable, then this criterion is significant to the location and to the subject.
For a specified criterion, the Lowest in highest suitability statistic shows where, for a specific criterion, the low weighted transformed values are in the higher suitability locations. It may be problematic if the criterion that is the most important to the analysis is not contributing significantly to the highly suitable areas. The Highest in highest suitability statistic shows where, for a specific criterion, the high weighted transformed values are in high suitability locations. These locations maybe more preferred.
The Highest in lowest suitability statistic shows where, for a specific criterion, the highest weighted transformed values are in the low suitability locations. If the specific criterion is important to the subject, unfortunately, these locations, will probably not be selected. The Lowest in lowest suitability statistic shows where, for a specific criterion, the lowest weighted transformed values are in the lowest suitability locations. These locations will not be considered, and because of the composition of the criteria, they should not be.
Summarize within tab
Use the Summarize within tab when the decision unit is something other than raster cells. For example, if the decision units are based on parcels, you would want to identify which parcel is the best value for the suitability gained.
The following sections describe actions that can be taken with this tab to address constraints.
Standard statistics
As described in the Criteria tab section above, there are several standard statistics for exploring the composition of the weighted transformed criteria values that produce the suitability values. Instead of exploring the composition of the weighted transformed criteria values for cell locations as on the Criteria tab, on the Summarize within tab, the standard statistics are used to tease apart the composition of the weighted transformed values within each decision unit such as within each parcel.
The standard statistics are available in the left or Explore suitability within zones panel for analyzing the parcels. They are Range, Standard deviation, Highest value, Lowest value, Most influential criterion, and Least influential criterion.
Use standard statistics
Consider a solar farm model, where solar radiation gain might be the criterion of interest. To determine if all areas in a parcel will receive good solar gain, use the Range or Standard deviation statistics. If the range is low in high suitable areas, most of the parcel should receive good solar radiation gain.
Specific statistics
There are some specific statistics that are distinct to the Summarize within tab.
A primary constraint may be based on area. For each parcel, you may want to see how much suitability is gained. To do this you can calculate the Sum statistic in the left or Explore suitability within zones panel. However, larger parcels will typically contribute more suitability. Use the Zone area option to view the area of each parcel.
To compare the parcels relative to one another, you may want the sum of the suitability to be normalized by the area. Use the Sum normalized by area (mean) statistic to calculate the mean suitability value for each parcel. Parcels with higher means suggests, on average, which parcels contain the higher suitability values.
In addition to area, value may be a significant constraint. For example, if you have a budget and want to determine which parcels cost the most, use the Zone value option.
To adjust the parcel value and total suitability with area use Value per suitabiilty unit statistic. When the constraints for identifying the best parcel include several factors such as minimum area needed, maximum budget, and of a certain quality, use the Area, value, mean thresholds statistic in the Explore suitability within zones panel. This statistic will identify candidate parcels that meet the three specified thresholds.
If you are concerned about the cost to build the solar farm, you may want to know if there are any steep slopes in the parcels you are considering which are being hidden by the transformations and weights. You can do this using the right or Explore criteria within zones panel. Select the base criterion of slope, and apply the Highest value statistic. Compare the parcels with high slopes to your candidate parcels from the Area, value, mean thresholds statistic described above.
You may want to explore a specific criterion as possibly being the most important to the subject. For example, what are the range of the weighted transformed values within each parcel? What are highest and lowest values? What is the sum of weighted transformed values the criterion contributes to the parcel? By exploring a criterion using the right panel, you can identify where that criterion will be most significant to the subject.
Validate tab
Use the Validate tab when you have field observation data of the subject that you can use to determine how good the model is.
The following sections describe actions that can be taken with this tab to address the constraints.
Standard statistics
As described in the Criteria tab section above, there are several standard statistics for exploring the composition of the weighted transformed criteria values that produce the suitability values. Instead of exploring the composition of the weighted transformed criteria values for cell locations as on the Criteria tab, on the Validate tab, the standard statistics are used to tease apart the composition of the weighted transformed values within each observation. You can learn which criteria contribute the most suitability to each point or polygon observation.
The standard statistics are available in the left or Explore suitability within observations panel for analyzing the observations. They are Range, Standard deviation, Highest value, Lowest value, Most influential criterion, and Least influential criterion.
Specific statistics
There are some specific statistics that are distinct to the Validate tab.
For point observations, such as in a bear suitability model, you would like to see if the known locations of bears are in areas of highest suitability. In the Explore suitability within observations panel, the Suitability values statistic can be used to determine this. The color of the observation points in the map corresponds to the suitability value at the location. Green colored points indicate the observation correspond to locations with high predicted suitability. This indicates your model is predicting well. Red colored points indicate the observations are in low predicted suitability. This indicates where the model might be wrong. However, there may be reasons the bears need to travel through these areas. Perhaps a bear was displaced and is moving to new habitat.
For point observations, if a set of random points is taken, you might like to know how the means between the observed and random points compare. In the Random points statistic, the mean of the observations should be higher than the mean of the random points. If it not, then the model may not be predicting well.
If you have bear observations for a bear model, and the observations are from three different bears, you may want to see if there is a difference in the mean suitability values between the bears. In the Explore suitability within observations panel, set the Field defining observations to the bear categories. With Category comparison statistic, you will be able to determine if certain bears, the ones with the higher mean, make better choices relative to your model.
You can see which criterion is the most important to each observation with Most influential criterion. This may provide insight into what the bears are reacting to.
Use Range or Standard deviation in the Explore suitability within observations panel to see if the bears choose locations with a variety of different weighted transformed criteria values. This may indicate your model predictions might be off. It is expected the bears should stay in the more suitable areas. However, there maybe be other circumstances influencing them. For example, they may need to go into lower suitable areas to reach higher suitable areas.
You can analyze the observations relative the specific criterion in the right or Explore criteria within observations panel. You may select the most influential criterion and analyze how it effects the bears as they move through the landscape.
Locate panels
Use the Locate tab to determine how good the final regions from the suitability model are.
The following sections describe actions that can be taken with this tab to address constraints.
Standard statistics
As described in the Criteria tab section above, there are several standard statistics for exploring the composition of the weighted transformed criteria values that produce the suitability values. Instead of exploring the composition of the weighted transformed criteria values for cell locations as on the Criteria tab, in the Locate panels, the standard statistics are used to tease apart the composition of the weighted transformed values within each identified region.
The standard statistics are available on the left or Explore suitability within regions panel for analyzing the regions. They are Range, Standard deviation, Highest value, Lowest value, Most influential criterion, and Least influential criterion.
Specific statistics
There are some specific statistics that are distinct to the Locate panels.
You can examine the distribution of the suitability values in each region with the plot on the Evaluate tab. The narrower the spread is for each line, the less variability there is in that region. This indicates that the region is homogeneous in preference and contains high values. This is generally the most preferred scenario.
For each region, you can see the number of cells in the regions through the Count statistic identified in the Statistics by region table on the Evaluate tab. You can also see statistics of the suitability values, those being the Total, Average, Median, Highest, and Lowest within each region.
In addition, for the resulting regions, you can see statistics for the CoreArea, CoreSum, and Edge. These values are also recorded in the Statistics by region table on the Evaluate tab. Depending on the needs of the subject, you can analyze which regions will be most preferred. For example, for a wildlife subject such as a bear, for security they may prefer regions with the greatest CoreArea or CoreSum.