Lab: Cost Surface Analysis
A cost surface facilitates the analysis of propagation, spreading and movement processes in 2D, unconstrained by (transportation) networks.
Cost distance analysis is a generic toolset somewhat similar to surface runoff, but not primarily driven by gravity forces. Rather, any kind of spreading or travel across a surface characterized by multiple 'friction' factors (land cover, slope, etc) is modeled with an algorithm based on local optimization.
Study area
Before we get started, you are invited to swipe through the study area overview below:
Cost surface
As a first step, explore the > Cost Distance Analysis basics and have a look at the algorithm discussion in this > How it all works document. You will want to get back to this after having completed your first analysis to fully understand how local minimization of cost leads to an overall optimization of service areas, paths and connectivity.
Before applying cost surface analytics to any multi-thematic problem, the relevant cost factor layers (like land cover, slope gradient and others) need to be aggregated into one cost surface reflecting the overall 'friction', i.e. the 'cost' of moving across the landscape. This typically is done through Weighted Overlay of these factor layers - if you have not yet completed the > Weighted Overlay Lab , this might be a good time to do so!
After gaining a general overview by scrolling through this Lab you can decide whether you want to establish your own case study with suitable data, or using the layers from this lab for your exercise and assignment - these can be accessed and/or downloaded as a > layer package .
If you really need a quick start skipping the creation of a cost surface through weighted overlay, this layer package also contains a ready-to-go cost surface generated through the steps outlined below:
Creating the Cost Surface reflecting the ease (of difficulty) of moving across the landscape, a total of 4 criteria were considered. To model the resistance or friction slowing down movement, all categories per criterion (=layer) are rated on a scale 1 to 10, with 1 representing 'easy going' and 10 'very hard to get across'.
Land Cover is considered the most important criterion with 45% of the overall weight, and categories are rated according to the screenshot at the right, with water defined as an absolute obstacle - NODATA.
Slope is considered the second most important factor with a 25% overall weight. Very steep slopes are considered a no-go and gentler slopes of course easier to navigate. In this exercise no distinction is made between uphill vs. downhill travel.
Curvature in this example is assigned a weight of 20%, with classes rated to reflect that movement is easier along convex landforms like ridges, and harder along likely wet and overgrown concave areas like creeks and ravines.
Elevation finally is given a low weight of just 10%, expressing that it might be harder and more strenuous to move at higher elevations. Which are not really that high in our current study area, though :-)
Keep in mind that all ratings ('points' assigned to categories) always should use the selected scale range of 1 to 10, in particular use 10 for the 'worst' category, to avoid implicit biasing of the Weighted Overlay analysis.
Explore the above created cost / friction / resistance across the study area resulting from the Weighted Overlay Analysis
Cost distance
Now we get to the core of our cost distance analysis, in the context of a case study involving cross-country movement by rescue teams. Very obviously, they want to follow the best route, incurring the least cost (most likely measured in time) on the way to their target location. Remember, this is a throwback scenario to the days without any roads ...
In this area, there are three Search-and-Rescue (SAR) stations, located in the towns of Seewalchen, Zell am Moos and Thalgau.
Based on the previously created cost surface, as a next step we need to calculate the > Cost Distance from these centers.
Due to the friction effects aggregated on the cost surface, the cost distance differs from euclidean distance and does not show precise concentric circles. On this cost distance surface, each SAR station is allocated their 'service area' (> Cost Allocation ), areas which can be reached faster than from alternative stations:
As a next step we want to identify the best path to incidents attended to from SAR locations. This > Cost Path analysis tracks a least cost route from SAR stations to incident locations - of course never crossing service area boundaries:
Each path is attributed with length and cost (try an 'identify click' on these paths), which e.g. could be units of time if properly chosen during Cost Distance analysis.
After this quick tour and showcasing of anticipated results you now can focus on working through this exercise or your own data set following the steps outlined above. Make sure you at least identify you own locations for whatever scenario you establish. You will notice that > Cost Allocation is the most powerful tool, as it includes the Cost Distance calculation as well as the Cost Backlink layer required for Cost Path all in one step.
Discussion
To fully appreciate the power of the cost distance method and tool set, we need to keep in mind that a cross-country walk is not the only use case. Cost distance is a widely usable general instrument e.g. covering as well the spreading of a contaminant from a source, as well as anisotropic propagation or movement through the > Path Distance tools where directional effects (wind, up-/downhill etc) can be included. A fuller exploration of algorithms, the definition of metrics and units for 'cost', as well as the many optional parametrizations not discussed here is a worthwhile effort for everyone 'diving deeper'!
In some application scenarios, e.g. in ecological habitat or migration analysis there is no clear origin or destination, but we intend to establish the connectivity between core regions, and look at broader areas for potential movement instead of strictly optimized paths. In this case the least cost > Corridor function adds up multiple cost distance surfaces.
Assignment
From quick and easy to more involved:
- Use the prepared cost surface, define a minimum of 2 centers, create a cost distance surface and follow the general logic outlined above. Make sure you understand units, how the 'spreading' algorithm creating cost distance works, and how paths to one or several destinations are created.
- Create your own cost surface through a weighted overlay based on a use case and related assumptions, then continue with (1).
- Independently set up your own application scenario including cost factors and work out the required information products from there.
Of course the value of your results are considered 3 > 2 > 1 ;-)
Not sure if that's the solution when measuring cost distance in minutes and hours ...
