Skip to content

Latest commit

 

History

History
68 lines (56 loc) · 3.2 KB

tasks.md

File metadata and controls

68 lines (56 loc) · 3.2 KB
Raw

To do

Track Data

  • Clean 2017
    • Separate out multi-track files
      • Edits done in ArcMap, see log
    • Clean scat data
      • Check to see that scat sites aren't mislabeled
  • Clean 2016
    • Clean transect data
      • Separate out multi-export tracks
    • Clean scat data
      • Some scats found to be mislabeled. Fix.
  • Format for analysis
    • Get grid overlay over tracks, count up visits & times of visit if possible.
    • Get interval between visits as data. IMPORTANT, as we're adjusting theta to be site-specific daily deposition rate
    • Get index of transects as data, for transect-level effect. Not sure the utility for prediction, though, since other parts of ADK are unreferenced to transect

Scat Data

  • Make scats inherit location from nearest point in corresponding track.
  • Reference scats to grid cells, indicate whether first, second, third, etc. replicate.

Covariate Data

  • Format spatial habitat covariate.
    • Reduce resolution to 50m, pull out relevant habitat types. Do this during runtime for null model, and model with transect effect
  • Format distance to road covariate
    • Options include:
      • Distance to highway and distance to minor road as separate covariates
        • Extra parameters to estimate, potentially covarying
      • Distance to any road
        • May weaken the signal if response is different between road classes
      • Distance to any road, weighted by highway class
        • How to do this?
  • Format distance to water body covariate
    • Should definitely weight by water body type or separate. Lakes are different from rivers are different from ephemeral pools are different from wetlands.
      • Start with lakes, wetlands, due to cooling + forage. Moose & rivers not known to be associated.
  • Format distance to human settlement
    • Weight by population size? Potentially -
      • $(max(\text{Dist}, .01) * \text{Population})$ = Weighted population distance
      • Closer distance only matters if population is high.
      • One method is to calculate for all towns, and take the maximum value.
  • Format temperature maximum prediction
  • Format elevation (easy)
  • Potential covariate: mobile data availability, from the NYS Office of Information Technology Services- GIS Program Office. Latest data is 2014, so fairly applicable.
    • Rationale is that 4G will be available in all areas where people demand it (i.e. tend to spend more time), 2G/3G where people spend less time, and no signal where people spend no time. It is thus a proxy of time spenty as well as population size.

Grid Overlay

  • For spatial prediction, will need a 50m x 50m grid over all prediction areas. This means masking out the towns and water bodies. Do this in ArcMap.

Analysis

  • Make model work according to commentary in model.txt file
  • Format data. Need:
    • Index of transect ID as data
    • Interval between visits as data
  • Summarize model output

Writing

  • Revise intro based on Angela's comments
  • Methods
  • Results
  • Discussion
    • Don't forget commentary about model extensions