Narrow region/times of interest

[mhoeger]

Factors for identification feasibility

• For peak identification, look at biowindows
• Depends on host + attack species
• Open question: Do we want to look into earlier stages? not peak identification stage?

Factors for impact

• What are the biggest threats?
• Watershed impact? Where?

[mhoeger]

Relevant Papers

do all attacks look the same? what can we tell from aerial survey data?

• Accuracy of USFS polygons: Accuracy of aerial detection surveys for mapping insect and disease disturbances in the United States
  • Damage type observations for mortality and defoliation had high-levels of accuracy
  • Accuracy declined as the observations went from less to more specific calls.
  • Aerial surveyors had high agreement when mapping the location of polygons.
  • Tree injury and mortality are mapped primarily during the summer months (June, July, August?)
  • 2004 aerial detection surveys costs were estimated at approximately $0.025 ha−1
  • Try to limit duplication of "last year's" damage
  • lodgepole pine had low commission error (maybe a good host species to focus on?)
• Supplementing the Forest Health National Aerial Survey Program with Remote Sensing during the COVID-19 Pandemic: Lessons Learned from a Collaborative Approach
  • ^How remote sensing/satellite imagery can help the USFS
• Remote Sensing in Forest Health Protection
  • Most forest damage first appears as a change in color of forest canopy
  • defoliation = red/brown or grey hue
  • coniferous bark beetle - color of fading trees and mortality distinct enough to most of the major
  • bark beetle species to permit reasonably accurate identification of the causal insect by aerial observation!!
  • pattern and color of southern pine beetle does not vary by host species attacked
  • ^"active heads"
  • pg 24 has a really helpful table on attack characteristics!!
  • darker can be from previous season's attack!!
  • pg 76 has biowindows!!!
  • mountain pine beetle, (on pine) western pine beetle (ponderosa pine)
• Individual tree map: US species maps (with characteristics, like density of forest (probably can use as a msk (maybe an input feature too?
  • https://www.fs.usda.gov/foresthealth/applied-sciences/mapping-reporting/indiv-tree-parameter-maps.shtml
  • https://www.arcgis.com/apps/webappviewer/index.html?id=4ebf103ddeeb4766a72e58cb786d3ee2
• Evaluating hydrologic impact from concurrent insect and fire disturbances
  • Insects damage a forest and can change the hydrology of the area (how much water is used, transpiration, etc.)
  • Care less about early stages (and maybe not even about individual species identification?)

[Respawnen1]

Random comments on Accuracy of USFS polygons paper:

1. Interesting that they mention southern-pine beetles have a "very distinct" tree mortality pattern
2. Bark beetles are the most common tree mortality agent
3. They mention aerial surveys are the current best method of detecting mortality - I'd wager this will change to satellite based monitoring when resolution and coverage eventually reach appropriate levels (this is basically the project motivation - see if its viable enough as of 2024..)
4. I am unsure how aerial surveyors are accurately judging mortality type without ground confirmation - there's likely some domain knowledge at play here
5. Paper mentioned incorporating geo-referenced base layers such as roadmaps (bodies of water and burn maps could be useful too) to avoid using irrelevant pixels. This might be a good idea for our work, though alignment may prove to be an issue.
6. In terms of not duplicating previous year's damages, I'm assuming we will only compare adjacent years - perhaps also getting average image statistics over all recorded years could be beneficial (EX: Compare average # of tree pixels from all years to 2023 or something)
7. From what I'm understanding the authors just went in person to areas included in USFS's datasets and compared, sounds simple enough
8. Why is commission error important compared to overall error? Additionally, other trees are reported to have 0 commission error
9. I'm assuming there's gonna be a tradeoff between picking a common mortality cause and the amount of available data
10. "Observations for tree genera or forest type had a high level of accuracy" - This is good to hear
11. Unless we have ground observations we'll likely have to assume the aerial data is correct

[Respawnen1]

Random comments on Supplementing the Forest Health National Aerial
Survey Program with Remote Sensing during
the COVID-19 Pandemic: Lessons Learned from
a Collaborative Approach paper:

In short, maybe look into a tool called "DeltaViewer", and maybe consider contacting an email address supplied below for access to meeting recordings.

1. Covid disrrupted aerial surveying of forests
2. Covid allowed a bunch of federal and private organizations to collaborate by combining data to better determine forest health
3. remote sensing historically used but not high enough resolution (yet)
4. in response to covid, a bunch of remote sensing workflows were developed
5. First webinar - Forest Health Survey Options and Opportunities
   During the COVID-19 Pandemic - dead link
6. Second webinar - dead link
7. DeltaViewer change detection method - rapidly respond to temporally discrete events in near-real time, such as storms
   or insect defoliation and mortality in near-real time. Used in google earth engine
8. The recordings from these meetings were distributed across the
   forest health community and are stored for future reference in a location where both state and federal partners can access recordings and all training material
   generated from the weekly topics of interest. If you
   would like access to these materials, please contact
   Ryan Hanavan, FHM coordinator (ryan.hanavan@
   usda.gov)