Summary of habitat working group April 11 2010

Attending - Andy, Nate, James, Karen

Questions

Habitat: What affects the likelihood of monarchs using a site, and abundance in site? 

Divide by 1) spring northward movement, 2) stationary time in July and early August, 3) fall movement south

Multi-scale selection: Combination of landscape/regional level (weather, land cover) and local scale (plant, patch [size and characteristics])

Life cycle phases:

Breeding: James, Karen, Nate, Leslie

What factors affect the probability that monarchs use a particular site?  Are the data good enough to do habitat selection?  We can do this on a plant by plant scale?

What factors affect their success (survival, density)?  Is this site specific?  Are specific sites good year after year? 

Where are MLMP sites vs. random sites?  How do they differ from the landscape as a whole?

NLCD and CEC vs. volunteer-collected data (problems with different scales - True multi-scale approach does more than just use one set of local features, then broader set of features - two separate models that answer questions at different scales.  Real multi-scale approach - models that include different variables, final composite model might have some landscape features and some fine-scale features

Milkweed: condition, species, dispersion
Nectar
Proximity to water? Moisture indices (does this change over the course of the season)
Size of patch
Proximity to other patches
Characterize area around site – matrix

Natural enemies?

Landscape and milkweed species characteristics change with region

NABA and BMN data show absence - if accurately located, could be used on landscape level.

What factors affect when monarchs start using a particular habitat in the spring?
What factors affect their use of habitat throughout season?

Problems - can we define habitat?  Makes large scale landscape study harder.  22 classes of landcover, can classify as appropriate or not - not forested, not pavement - can assign values to different habitat types.  Maybe use INVest software, assign value to edges specific to kind of edge.  How to deal with observer bias?  Chip's work in KS on mw presence in different habitats will feed into this - .  Does human-based classification equate with monarch habitat?  Monarch may see a few different kinds of habitat - but could still pick up something from remote data that help us define habitat.  Need to carefully select important classifications.

Fine scale vs. large scale location data

Expert elicitation - formalized way of assigning values without data (Erica will send references)

Response variables: Look at occurrence or density?  Selection alone, egg density (look at this as a function of the size of the site, what is the size of the site – this will be determined by the matrix?), survival in site (measure of habitat quality, need to clean data).

Merge data sets
Clean mlmp data - remove FL?  But a problem with continually ignoring FL data - could run a separate habitat use model.  Are they using the same kinds of habitat in FL?
Improve mlmp data – locations, site characteristics

Try Ohio data?  Will they add landscape level characteristics?  Are there different relationships that fall out? 

Over-representation of urban landscapes and yards - modeling approach to test effects of this, and also of lack of absence data.  There are approaches to dealing with this - maximum entropy: detect signals from noise - can generate importance of relationships between variables (threshold or continuous) and contribution of each variable in supporting the model.  Are there variables that we can measure at landscape scale that would be potentially important to monarchs but relatively insensitive to urban-rural gradient?

Combined Spring migration/breeding habitat selection model paper -  maybe a long range goal

south to north are rows, columns are time early to late

Need an a priori model – generate hypotheses based on IBM

• Define fronts based on JN data (could we include what determines the movement of the front? This is what determines the biggest habitat selection decision they make.) Define polygons of first JN observations, select a kernel within these polygons (95%) - will get rid of outliers).  Use MLMP data within that kernel.  Does performance of monarchs vary in each time-frame/region, and with habitat characteristics (abundance and survival [eggs to fifths])?
• Could look for multiple waves, and MLMP eggs outside (especially to the south of) JN polygon
• First milkweed data from JN and MLMP – correlate with JN movement.  Does this determine where monarchs are?
• diffusion model problem
  

Fall migration: Nate, Andy, Elizabeth
Selection of roost sites: correlation with latitudinal movement
Roost locations – zip code
Large scale geographic features, also look at notes associated with each sighting. Agricultural areas where islands of trees in large barren areas.
Possibly contact observers to give more exact details on locations – exact location, flowers, other things in area
How long does roost last?

How many monarchs?

What kind of trees?
Can look at random sites vs. roost sites.

Monarch as a generalist

Resources

JN data from spring migration

JN and MLMP first milkweed occurrence data

MLMP site characteristic data and monarch density

OH site characteristics and monarch sighting data

IL site characteristics and monarch sighting data

Results highlights

Relevance to other migratory species - multiple broods are different, selection among multiple sites for breeding (could contrast with birds, waterbirds make similar decisions among patches, but similar for other insects)

Planned Manuscripts

Breeding paper

Fall migration habitat papers:

1. Examine large-scale (landscape) features of roosts

2. Examine small-scale characteristics of roosts based on sighting notes

Tasks

• Access and clean databases (JN, MLMP, Ohio?, IL) - put JN and MLMP into GIS database, Karen remind all MLMP folks to update site characteristics and locations (need 1 km resolution). Karen will work with Leslie to get detail on habitat characteristics and GIS locations of transects in IL and OH.
  
• Karen sends James and Nate site characteristics, milkweed species, management
  
• Summarize biology and questions, interpretation - Karen
  
• Nate and James - modeling - data and fiddling before next meeting
  
• regular conference calls - monthly
  
• habitat classes - overlay with monarch locations
  
• Prioritize land cover data, not focus on weather for now (moisture gradient may be necessary)

• Categorize habitat characteristics of roost sighting data from sighting notes in JN - Andy
• Overlay roost sightings onto national land cover spatial data set, output landscape characteristics of roosts for analyses - Nate and Andy
• Think about expert elicitation (Bayesian belief networks) model for next time (conceptual framework for a quantitative approach): habitat quality nodes, enemy nodes, human effect nodes, management nodes - transition probabilities that combine expert opinion with known relationships.  Nate has done two workshops like this.  Problem that these aren't spatially explicit.