[Marxan] Liberia Conservation Prioritization

Trevor Wiens tsw.web at gmail.com
Wed Dec 11 15:05:23 EST 2013


I'm sure some on the list can provide more in-depth answers to some of
your questions but I'll try to answer some of them at least generally to
get you going.

I've inserted my answers below. Feel free to ask follow up questions as


On Tue, Dec 10, 2013 at 8:27 AM, Jessica Junker
<jessica_junker at eva.mpg.de>wrote:

>   Dear all Marxan users
>  I am new to the group and have a few conceptual questions regarding
> Marxan.
>  I would like to prioritize areas for future conservation across Liberia.
> My planning units are 1.8*1.8 square km pixels, of which Liberia includes
> approximately 29000. Rather than using individual species-distributions, I
> have for each pixel modelled nationwide data on chimpanzee abundance, large
> mammal species richness and tree species richness (all based on transect
> data from a nationwide survey), which I would like to include as
> conservation feature types. For each of these I was going to set minimum
> targets of 30% of the total of all pixels in Liberia. If the algorithm is
> able to meet these targets under the conditions I specify, including some
> amount of spatial clumping and management costs, I will try and maximize
> these targets within the predefined spatial limits ( i.e. the local
> authority in Liberia is willing to protect 30% of Liberia's remaining
> forest cover).

A while ago (within the last 2 months) on the list there was a very good
discussion about using diversity measure which I encourage you to find the
archive and read through it. In brief, diversity is not always a good
measure for numerous reasons, one of which being that diversity does not
necessarily equate with areas in need of conservation.

>  Additionally, I would like to include positive costs into my conservation
> features, namely altutude and slope. Here, I argue that high altitude
> areas, which have more slope, are more difficult to logg and thus cheaper
> to purchase for conservation purposes. Here my target would be all areas
> above 500 meters have low cost and those below 500m have high cost. So my
> conservation target would be >500. How can I specify this in my file?

You could approach this in a number of ways. Each PU has cost associated
with it so areas so you need to first intersect your PU layer with your
elevation layer and then assign costs accordingly.

>  As costs, if have three different categories, which, if I understand the
> program correctly, I need to combine as one somehow. These are my costs:
>  1) snare abundance - a proxy for hunting (unit=number of snares per
> pixel/ planning unit)
>  2) empty gunshell abundance - a proxy for hunting (unit=number of empty
> gunshells per pixel/ planning unit)
>  The above two costs are easy to combine as they are measured in the same
> units
>  3) Locations of mineral deposits ranked 0-3 (0=no mineral deposit, 1-3=
> occurrence of least to most precious mineral deposit). I include this as a
> measure of potential future threat, as precious minerals attract illegal
> miners.

There are a lot of ways to combine costs, but lets assume for a moment that
you think all these costs should have the same importance. In this case I
would scale them all to match the scale of the ordinal variable and then
add them. In your case however I would start with elevation as the base
case of 1 or 2 and add to that.
Please note that having PUs with no cost is not good so if you are coming
up with a complex costs layer, make sure you are not ending up with areas
of no cost; this will provide Marxan no reason to not remove them and
introduce noise into the solution.

>  I also include all observations of threatened mammal species (GPS
> locations) buffered by their mean home range size into the conservation
> area network (i.e. status=2).
>  I include all existing protected areas into the conservation area
> network (i.e. status=2).
>  I exclude all main roads buffered by 2 km on each side from the
> conservation area network (i.e. status=3).
>  My questions are:
>  1) Seeing that I don't have the usual conservation feature types, such as
> individual species distributions or habitat types, do you think Marxan is
> the appropriate program to use?
Marxan can be used with some caution, you just need to be very conscious
that you are dealing mostly with coarse filters not fine filters and make
sure you understand that they are valid proxy measures for what you are
trying to achieve.

>  2) Can I try and maximize my conservation feature targets by
> experimentation as described in my first paragraph? Is this legitimate?
Marxan is not a target maximization tool, but a cost minimization tool. You
can try a series of target ranges, say 30, 40 and 50% and see what the
differences are in cost. There are other theoretical things that can be
done as well such as turning costs into features and features into costs,
but that is not standard use; I've only done that in one case as a
theoretical experiment and would not advise that path as other tools exist
for the Maximum Coverage problem.

 3) I would like to make my map pixels for which I have values for chimp
> abundance, species richness, etc. my planning units. These exist in a
> raster format only and I do not know how to convert these into a shapefile
> format so that each pixel constitutes one polygon (this doesn't seem to be
> possible in ArcMap). And if I am not mistaken, for the planning unit file,
> I need my units in the shapefile format?!

In Arc you can use Zonal Statistics to collate the raster files much faster
than converting them to shape files first. This functionality also exists
in QGIS 2.0 now, if you don't have the extension needed in Arc.

>  4) How can I combine my cost layers and what values should I give them?

See suggestions above.

>  5) What will the program do if I have areas that have both status 2 and 3
> (e.g. areas within the home range of threatenend species of which we have
> observation records, which are at the same time within the 2km buffer of a
> road)? Would it be better then to include roads and surrounds in the cost
> layer as well? If yes, how would I be able to combine this information with
> the other cost layers?

Each PU can only be one status and this needs to be determined up front. I
would come up with a logical set of rules on how to decide if it was 2 or
3. Perhaps you could look at the amount of area in each PU and select the
status based on greatest intersected area.

>  6) In addition to the above mentioned rules, I would also like to
> prioritize pixels that are close to protected areas in neighboring
> countries. Do I need to pre-select potential pixels close to these areas
> and include them in the boundary length file somehow? The conservation
> areas in the neighboring countries are actually outside my target area
> (i.e. they are not part of my planning units!). Is there any way I can do
> this?

You have a number of options:
1. Expand your area to include those PUs and then lock them in.
2. Lock in those PUs on the border area that are adjacent to PAs to force
Marxan (with a calibrated BLM value) to build on those areas
3. Adjust the cost of those PUs so that they are inexpensive to select

>  Sorry for all these questions - I would be grateful for your input and
> ideas on this!
>  Thank you so much for the advice.
>  Regards,
>  Jessi
>  Jessica Junker
> jessica_junker at eva.mpg.de
> PhD candidate
> Max-Planck Institute for Evolutionary Anthropology
> Department Primatology
> Deutscher Platz 6
> 04103 Leipzig, Germany
> Phone: +49 (0)341 3550 805
> Cell: +49 (0)160 905 94996
> http://www.eva.mpg.de/primat/people/jessica-junker/home.html
> ****************************************************
> _______________________________________________
> Marxan mailing list
> Marxan at lists.science.uq.edu.au
> http://lists.science.uq.edu.au/mailman/listinfo/marxan
> The Marxan Archives http://lists.science.uq.edu.au/pipermail/marxan/

Trevor Wiens
Apropos Information Systems
Calgary, Alberta
Ph. 403-973-5901
Fax 780-666-4580
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.science.uq.edu.au/pipermail/marxan/attachments/20131210/3229c86c/attachment.html 

More information about the Marxan mailing list