new to maxent - please help

[tiggerkate]

Hi all,

I haven't used Maxent before (and GIS skill set is not too advanced)
so another first-timer. Having some difficulties so a little help
would be great, questions as follows...

1) I tried running a csv. file with climatic layers from World Clim
(all at 1km) but get a meddage about out of geographic bounds. I
understand this means the areas for my sample points do not match
those of the layers - not sure how to resolve this.

2) Does the samples file have to be in decimal degrees, to ensure the
same dimensions as the ASCII layers. As above, I understand they have
to cover the same region?

3) I was hoping to use NDVI (vegetation cover) and build a layer from
scratch - I have the values for each point - but not too sure how to
go about this, or even if vegetation cover qualifies as an
environmental layer for use in MaxEnt? I have read different accounts
of this being used as a variable (e.g. not temporal or climatic but
vegetation cover is still used by some).

Was also hoping to use distance measures as variables, e.g. distance
to waterbodies etc but again not sure if or how these can be used. Is
it just bioclimatic variables sourced from online resources such as
WorldClim that can be used? Seems that ecological \ interactive
variables can't be used for this method - perhaps suited for other
ENFA analyses?

4) As the layers are large data sets in ASCII formats, how do you run
correlations on these? I've assumed a standard stats package can't
process these and I've never run correlations outside of SPSS or
Minitab..

5) Assuming if you have to resample the layers in GIS to ensure the
same extent, the cell size would have to be specified the same for
all, e.g. 1000 for 1km resolution?

6) Can SWD and environmental (grid) layers be used at the same time?
Also a little confused when a background file should be used..

Thanks loads for any thoughts, much appreciated.

[Heather Peacock]

Hi,

I will try to give you some answers and suggestions based on my experience with Maxent.  Please let us know if you disagree or have different suggestions.

1) You need to make sure your point localities are in the same units as the WorldClim data, i.e. decimal degrees, and that no points in the species file lie outside of the extent of the WC data you are using.  You can use arc to test for this.

2)  I suppose this is answered in question 1.

3) You can use NDVI data for forest cover, I used a forest cover raster in my analyses.  I made a binary raster of forest/non-forest and resampled it to match the resolution of the WorldClim data layers.  You can do this for any feature though, water/not water, etc.  I am not sure how you would incorporate distances, I think you will just have to include ascii files of water features (binary) and the program will determine distances etc. and its relative importance (select the jackknife option in the setting window).

4)  As far as Maxent is concerned it can calculate relative importance/influence of each environmental variable used.  To me it appears robust enough to deal with some correlation between variables, however, if you want to run correlation analyses prior to running maxent you can probably use any stats software.

5)  Yes, you will have to resample your other environmental layers to match that of WC data.

Hope this helps!  Good luck!

Heather

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To post to this group, send email to max...@googlegroups.com.
To unsubscribe from this group, send email to maxent+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/maxent?hl=en.

[JC Riveros]

Hi,
It was a very detailed and comprehensive answer Heather. A couple of
comments to question 3.
In the case of the forest/no forest layer you must be sure to identify
this variable as categorical in Maxent. For the distance to water
features it would be necessary to create a distances buffer using the
Spatial Analyst extension in ArcMap. The resulting layer must be
resampled using worldclim data set as a template and cropped to your
study ares extension. You must be careful to replace the water
features in the final layer with a zero value.
Finally a word of warning about using NDVI to differentiate forest
from non forest areas. There are several vegetation types besides
forests that have a strong photosynthetic signal and could be
erroneously classified as forest. Even worst, some crops and
plantations are difficult to distinguish without a more detailed
analysis.
Hope it helps.
JC

--
Sent from my mobile device

JC Riveros

[tiggerkate]

Hi both,

Thanks for your suggestions and shared experience - a big help.

Re the distance variable: I am familiar with the point > distance tool
using Hawths Tools in ArcGIS - a simple way of calculating distance
between points / features between or within layers. Only I'm not so
sure how to then transform this extracted information (which of course
is by cooordinates for each species location point) to raster in order
to get it in the ASCII format. Is this where interpolation comes in?

The same issue still goes for NDVI values (currently a % for each
species point). Don't quite follow the distances buffer tool >
resampling with worldclim data template suggestion.. Assuming this
would involve rasterisation and resampling in GIS?

To test there are no outliers outside the WC extent, does this just
involve checking the layer CS details? Or do you mean there is a tool
available to do this?

Thanks again,
Kate

On Feb 14, 12:36 pm, JC Riveros <jcri...@gmail.com> wrote:
> Hi,
> It was a very detailed and comprehensive answer Heather. A couple of
> comments to question 3.
> In the case of the forest/no forest layer you must be sure to identify
> this variable as categorical in Maxent. For the distance to water
> features it would be necessary to create a distances buffer using the
> Spatial Analyst extension in ArcMap. The resulting layer must be
> resampled using worldclim data set as a template and cropped to your
> study ares extension. You must be careful to replace the water
> features in the final layer with a zero value.
> Finally a word of warning about using NDVI to differentiate forest
> from non forest areas. There are several vegetation types besides
> forests that have a strong photosynthetic signal and could be
> erroneously classified as forest. Even worst, some crops and
> plantations are difficult to distinguish without a more detailed
> analysis.
> Hope it helps.
> JC
>

> > <kate_allbe...@hotmail.com>wrote:

> JC Riveros- Hide quoted text -
>
> - Show quoted text -

[Heather Peacock]

You can do a simple select by location query to see if all points lie within the WC data boundary.

[John B]

Hi,

To produce a raster of distance to nearest feature, you can use the
Euclidean Distance tool (Spatial Analyst). Be sure to set your desired
processing extent, cell size, etc. in the Spatial Analyst options
(ArcGIS 9.x) or the Geoprocessing>Environments dialog (ArcGIS 10).
This is straight line distance ignoring slope etc. You can also look
at the Path Distance tool, which allows you to include cost rasters
and which identifies the distance from each cell to the nearest
feature along least cost paths. I haven't used this but am sure you
could use it with a DEM to get true distance taking slopes into
consideration.

More info: http://webhelp.esri.com/arcgiSDEsktop/9.3/index.cfm?TopicName=euclidean_distance

John

On Feb 15, 6:43 am, Heather Peacock <heather.peac...@gmail.com> wrote:
> You can do a simple select by location query to see if all points lie within
> the WC data boundary.
>