Climate / extreme weather event: mechanism and hypothesis development

[sigmafelix]

graph LR;
  A[Climate change]-->B[Extreme weather];
  B-->C[Significant impact];
  C-->D[Health outcomes];
  A-->D;

• The link between climate change and health outcomes would be direct, but there would be steps where climate change gradually affects health outcomes. A problem is to operationalize/define a proxy of variables that are described in high-level (or vague) language, for example, "significant impact." This factor can either be a moderator or simply dichotomized meteorological variable(s).
• What is the existing mechanism for mediators between climate change to health outcomes?
• How is actual "hazard" of extreme weather measured?
  • FEMA hazard declaration and individual assistance
  • Emergency fund spent at jurisdiction level (e.g., county)
  • Small Business Association disaster loans

[sigmafelix]

• FEMA data: Ripple or spatial spillover effect of a local impact of extreme weather events
• NOAA HMS smoke data: actual spatial extent and its duration of wildfire events
• Combined: three-level climate effect -- frequency and duration of observable events, actual damage in immediate neighborhood (circular buffer or census tract), and in distant neighborhood (e.g., large circular buffer, nth order neighboring county)

TODO

• FEMA data cleaning: point-census tract-county level summaries by hazard types
• NOAA HMS data conversion: vector to 500m or 1km raster (keeping spatial and temporal information intact)

[sigmafelix]

• Smoke exposure has low variability in North Carolina, where most of the participants reside
• FEMA data include details of damage, housing type, insurance coverage status, etc.; however, records are meaningfully available in hurricane and severe storm (esp. 2010-2015) categories
• NOAA storm damage dataset includes start-end location and time, type of storm events, and spatial range of storm events (for example, a figure below on all storm events with >0 mi. radius in 2019).
• First Street Foundation data reflects potential risk; will be included in the main model but will not serve as a primary variable in hypothesis tests
• Literature often selected remarkable extreme weather events then operationalized exposures as frequency (i.e., event-based; once, twice, ...). It is viable to use spatiotemporal change in frequency and severity of storms.

TODO

• Classifying preexisting negative mental health status from PEGS cohort
• Storms are the most common and pronounced extreme weather events in NC; will focus on these
  • Hurricane as a separate binary variable or inclusive of all storm exposures?
• Thresholding storm events

[sigmafelix]

Also refer to: USFS Research Data Archive -- wildfire occurrence database