In a former posting I discussed how wilderness is not only about politics, religion, philosophy and legal instruments. Unless we force it into a practical context, the term “wilderness” remains an intangible size. Geographers have a long history for making representations of the intangible – be it disease (John Snow), social justice and injustice, demography and more. To my knowledge one of the first impressions of wilderness or “the wild” is what we can find in some older maps. “Hic sunt dracones” (here be dragons) is an expression which can be found on the “Hunt-Lenox Globe” (c. 1503–07). Other maps bear similar indications of uncharted or remote areas. We, the geographers, have moved on. Today we paint our dragons in more sophisticated ways.
The use of technology to “find” or delimit wilderness has a long history in Norway and other countries. I will continue this tradition and do an analysis which in many ways is similar to those done in Norway. The encroachment types will be slightly different. So will the technology used to do the analysis.
In this posting I will look at how FME can be used to establish a wilderness areas data set. The results will be presented in a separate posting.
How it has been done in Norway
The understanding of wilderness in Norway has been areas without encroachments (INON). The zones were split into three categories; areas more than 5 kilometers away from encroachments (zone W), areas 5 to 3 kilometers away from wilderness (zone 1) and areas 3 to 1 kilometers (zone 2). The area from 0 to 1 kilometers from an encroachment is referred to as wilderness near areas. In addition the Norwegian model was based on historical data since around 1988 providing measurements and change statistics every 5 year. The map below is based on WMS-data from NEA (NEA 2014).
The last analysis of this kind was done using a combination of Python scripts and Arcpy. It was based on analyzing raster data, not vector data like I do in this project.
The Norwegian wilderness analysis is but one way of illustrating the human land use footprint. Many other exist. From a geographers perspective measuring distance only is a straightforward way of doing such an analysis. One could of course consider using information about slope, line of sight, species vulnerability to specific encroachment types, species ranges and more. This would however complicate the analysis – as well as the communication of the analysis/indicator with the general public considerably.
The possibilities for distance categories, encroachment categories and more are endless. Discussing them based on philosophical, political, biological issues is beyond the scope of this article – although admittedly that part also has its sides. Giving room for philosophical considerations was duly done in a former posting.
Wilderness analysis using FME
Doing a wilderness analysis first hand is technically interesting. Doing it using normal desktop tools takes time and patience. I have done this many times and will hopefully not have to do it manually again later. Using FME for the same job is a breeze.
- FME is a tool for converting (primarily spatial) data between standards. Currently – FME supports over 400 different data transformers. The user interface is built around what I would call a mature and open ended version of the ArcGIS model builder. It is flexible and has a relatively good user interface. It replaces the need for mundane python programming supporting loops, array/list handling, tests and more. Should one still need more sophisticated python-coding it allows embedding of python calls.
As a habitual user of FME I would really like to see the system support auto-layout of the different transformers and sources/destinations. yEd graph editor from yWorks has managed to do this – and I am sure FME in due time will be able to as well. Should you not have access to FME an alternative could be the Graphical modeler in QGIS or GeoKettle.
Even though working with FME is less demanding than doing the job with ArcGIS it took some time. Quite a bit actually. So I think it is well worth sharing it 🙂
The model requires two categories of input vector files:
- encroachment features
- polygon(s) delimiting the analysis areas(s)
The output of the model is:
- a layer with the zones w, 1, 2 and n is created
The resulting codes for the areas generated are as follows:
|Code||Name||Distance from encroachment|
|1||wilderness, (zone 1)||3-5 kms|
|2||wilderness (zone 2)||1-3 kms|
|n||near encroachment||0-1 kms|
Admittedly this analysis will probably be faster and in many ways easier if done based on raster imagery. But for smaller areas (like the Bulgaria analysis) with not too many encroachment objects, the consumption of time is tolerable.
The FME-projects area available on github under the project SAWE – System for Analysis of Wilderness Encroachment. A reader versed with FME should be able to download the FME project files and use do an analysis based on own own input data.
The files are being developed as this series continues. Contributions and comments are as always more than welcome.
The following is a list of presentations of wilderness analysis using the presented code:
Currently I do not have much time to use on a spare time project like this. But here are a couple of ideas which I would love to work more on:
- Country level analysis for more countries
- Regular calculation of the wilderness status for countries with a good OSM data basis using batch procedures
- Statistics generation – also batch based
- Regular change analysis (see forthcoming posting) giving regular degradation statistics
The flipside of the coin…
Using FME or other tools doing an analysis like this is not really a big thing. It’s pretty straightforward. The challenge is to know when not to use a tool. Not only will the tool require a relevant scientific (ecological/biological) or political context, it will also require access to good data.
Decision makers could use a data set like this as part of an environmental impact assessment process. Starting with wilderness will give an indication of which areas are less disturbed by humans than others. Still, this is a national indicator, and in actual projects one should look at the data used as basis for the analysis – just to make sure.
The better the data is the more this is becomes wilderness map. The less good the data are the more this becomes a map showing where data is not available. Again – this analysis requires complete vector data for an area to produce a credible wilderness map.
In the next part of this series I will look at how using this tool with OpenStreetMap data from Bulgaria, Guinea and possibly also other countries. I will hopefully also be able to look at a smaller area with presumably good data.
NEA. 2014. Map catalog. Website accessed 2014.12.25.