Monarch butterfly and milkweed models the western U.S.

Today Matt Forister and I presented the results of our habitat modeling work for monarch butterflies and milkweeds for the western U.S. I've been involved in this research now since January 2017 and am very excited about having some preliminary maps and models ready. The work is an extension of the 2015 habitat modeling effort by Madeline Steele while she was at the U.S. Fish & Wildlife Service. Our work is a collaborative effort by University of Nevada Reno, the Xerces Society, the U.S. Fish & Wildlife Service, and others. I'm excited to get to work with such an iconic species, and I hope that through conservation and societal awareness of pesticides we can halt the decline of one of the most beautiful insect species in the world.

New version of Maxent software and paper out

It looks like Phillips and company has a new open-source version of Maxent out as well as an accompanying paper - http://biodiversityinformatics.amnh.org/open_source/maxent/

Sarah Barga wins 2nd place in GSA poster contest

It is always great news when students win awards!  Recently Sarah Barga, an EECB Ph.D. student advised by Beth Leger, won 2nd place in UNR's Graduate Student Association student poster contest for her poster entitled "Creating maps of potential habitat for Great Basin forbs using herbarium data". I'm proud to say that I helped her get started with the Maxent modeling, but really, Sarah has done a tremendous amount of interesting work and is currently working on finalizing two manuscripts at the same time.  The maps on the right are straight off of her poster. The different colors correspond with the amount of overlap for different top models. You can view her poster by clicking HERE and her research by visiting her website HERE .

Comments on Prepare Rasters for Maxent versus Find ArcGIS Rasters and Project to Template in MGET

On my page for the Prepare Rasters for Maxent Tool for ArcGIS I made a comment about its application versus another tool called Find ArcGIS Rasters and Project to Template in the Marine Geospatial Ecology Toolbox.  To access that page click HERE .  Here is the text of that comment:

The Marine Geospatial Ecology Tools for ArcGIS (http://mgel.env.duke.edu/mget) is an excellent toolbox that includes much of the functionality of the Prepare Rasters for Maxent toolbox. I highly recommend checking it out. In particular, check out the Find ArcGIS Rasters and Project to Template tool which will batch project and clip all rasters to a template. One primary difference between that tool and Prepare Rasters for Maxent is that Prepare Rasters for Maxent has built-in steps to resample and fill in missing data. If you know that you have gaps in your data then Prepare Rasters for Maxent may be your best bet. If you don't think that you have gaps or they are minor then consider Find ArcGIS Rasters and Project to Template as a speedier alternative. You can create a template raster by multiplying all of your rasters together in the Raster Calculator. By default this will take the intersection of all of the rasters resulting in a template that is smaller than any one of the inputs.

An under-appreciated method for partitioning data into training and test

Perhaps one of the most under-appreciated tools in ArcGIS is subset features. While there are numerous examples of how to generate random points, which can then be sorted and used to generate fields for subsetting your data why not just do it directly? The subset features tool in the Geostatistical Analyst toolbox does this. Another alternative to check out is the Marine Geospatial Ecology Tools. MGET has numerous tools for doing modeling and is a great resource for quickly running models and bring the results back into ArcMap for display.

Map algebra formula for calculating proportion of cover at different scales

Calculating the proportion of cover of different categorical classes (lets use a classified vegetation map as an example) isn't rocket science, however there are more steps involved that one might expect. Converting a categorical map to cover classes has a couple of advantages. One advantage is that it converts a categorical variable into a continuous variable making some models easier to develop and easier to interpret. Another advantage is that it allows the variable to be calculating at multiple spatial scales. When it comes to species distributions we usually don't expect the value of a single cell to be the primary driver of habitat selection. Rather it is the amount of different habitat types in some (usually unknown) neighborhood size that we expect to control habitat selection.

I've already reported on a tool for scaling up from categorical classes to continuous proportions of cover HERE.

In this post I provide a simple map algebra statement that can used in the raster calculator in ArcGIS to convert a single category in a categorical map to a percent cover map at some spatial scale.  This formula will do the trick:

(Float((FocalStatistics((Con("phase.tif" == 1, 1, 0)),NbrRectangle(33,33),"SUM")))) / (Float((FocalStatistics((Con("phase.tif" == 1, 1, 1)),NbrRectangle(33,33),"SUM"))))

All you need to do is swap out "phase.tif" with your raster, change the value after the == sign to represent the class that you are interested in, and change NbrRectangle(33,33) to a neighborhood size and shape of your choice.

If you have no gaps or holes in your raster dataset you are done.  However in many cases we are likely to have areas of NoData beyond which our raster will also be NoData.  To rectify this situation we can apply a second map algebra statement:

Con((IsNull("prpn.tif") == 1), 0, "prpn.tif")

Just swap out "prpn.tif" with the name of the raster resulting from the above step.  The IsNull tool ensures that NoData areas are identified and the conditional statement sets those areas to 0.

In the image below on the left the vegetation class in black is converted into a proportion cover with darker blues indicating higher values. The light blue indicates that either 1) a different vegetation class was dominant or 2) none of the vegetation classes were present in the spatial neighborhood.

Paper accepted in Ecological Applications

My paper entitled "Multi-scale connectivity and graph theory highlight critical areas for conservation under climate change" has been accepted in Ecological Applications pending a few minor revisions!  The general idea of this paper is to combine measures of habitat connectivity at the rangewide-scale, metapopulation scale, and local scale to inform conservation decisions. We use a combination of graph theory, Circuitscape, and least-cost paths to analyze the effect of different renewable energy development scenarios and climate change scenarios on habitat connectivity of the Mohave ground squirrel.  This paper follows from our earlier effort to map habitat suitability for Mohave ground squirrel:

Inman, R. D., Esque, T. C., Nussear, K. E., Leitner, P., Matocq, M. D., Weisberg, P. J., ... & Vandergast, A. G. (2013). Is there room for all of us? Renewable energy and Xerospermophilus mohavensis. Endangered Species Research, 20(1), 1-18.

as well as our earlier report to the California Energy Commission that included an analysis of climate change effects on habitat availability, landscape genetics, and a more limited analysis of renewable energy development scenarios titled "Habitat Modeling, Landscape Genetics, and Habitat Connectivity for the Mohave Ground Squirrel to Guide Renewable Energy Development" downloadable HERE .

In addition to demonstrating a multi-scale approach for assessing habitat connectivity and informing conservation decisions, the paper also presents a novel methodological contribution to make graph theory operational for species with continuously-distributed habitat.

Illustration by M. A. Walden

 

New paper - Functionally relevant climate variables for arid lands: using climatic water deficit modeling to predict distribution of desert shrub vegetation

I'm pleased to announce that my paper "Functionally relevant climate variables for arid lands: using climatic water deficit modeling to predict distribution of desert shrub vegetation" got accepted in the Journal of Biogeography. The paper compares two different suites of climatic variables used for modeling the distributions (climatic envelopes, niches, habitat suitability) of eighteen shrub species in the Great Basin. The first approach uses the so-called bioclimatic variables commonly used today in species distribution modeling. The latter approach uses the climatic water deficit which uses a Thornthwaite water balance model to estimate potential and actual evapotranspiration and climatic water deficit (PET minus AET). We expanded upon the work of others by deriving new variables that more fully describe the  shape of the water balance climograph. Our study found that for these shrub species there was similar model performance using the two suites of variables. However, we believe that the climatic water deficit approach may prove to be more effective in studies that predict range shifts under climate change.

The paper compliments the Climatic Water Deficit Toolbox for ArcGIS which can be found HERE. Thank you co-authors Peter Weisberg, Jeanne Chambers, and Camie Dencker as well as those who provided critical reviews.

The paper can be downloaded HERE.

Updates to the Prepare Rasters for Maxent Tool - upscaling rasters

Maxent software (http://www.cs.princeton.edu/~schapire/maxent) is frequently used for presence-only species distribution modeling. Maxent requires, however, that input ASCII raster files be aligned with one another and have the same spatial extent. This tool pre-processes raster data in preparation for Maxent modeling to ensure that all rasters have the same extent, same cell size, and aren't missing data. You can download the tool HERE.

There is a new tool that can be used to upscale data. Upscaling is useful when simple resampling is insufficient. For example, if you have a 30 meter classified vegetation map simply using the majority resample to resample to 1000 meter cell size would be insufficient because it would just tell you about the majority vegetation type present. Better would be to count the number of pixels of the vegetation type of interest within the 1000 meter cells and calculate the proportion of that vegetation type. This also solves the problem of having to treat the data as categorical in Maxent. Vegetation type would have to be treated as a categorical variable whereas proportion of X vegetation type is a continuous variable.

This example illustrates the idea of upscaling.

In the image above we have a forest/non-forest GIS layer overlaid on a coarse-resolution grid. The need for the coarse resolution grid is due to uncertainty in the location of the occurrence points and the resolution of other datasets that will be used in Maxent (e.g. the MODIS satellite products have a cell size of 250 - 1000 meters depending upon the product). The image below shows tree cover at the 1000 meter scale derived from running the Upscale Rasters for Maxent Tool. An important pre-processing step is deciding which vegetation types should be included in the model and which should be omitted. Included vegetation types should be reclassified as 1 prior to running the tool and omitted vegetation types should be reclassified as 0.