Emily Williams

For each city we find a set of temperatures that, when surpassed, create uncomfortable conditions for residents using the following methodology:

1. We observe Google search interest by census place/month for keywords that represent uncomfortable heat/cold conditions. We define separate keywords for heat/cold waves.
2. We record the temperatures of the coldest/hottest 5 sequential days by census place/month
3. For each census place - we pair standardized search interest from 1. for each heat/cold keyword with temperature recorded in step 2 above.
4. We then fit a piece-wise model consisting of two linear segments with one kink point of standardized search interest against temperature for each census place.
5. We define the kink point as the threshold temperature, at which comfortable conditions turn into uncomfortable conditions.
6. For each census place, we define the heat wave/cold wave threshold temperature as the maximum/minimum kink point temperature across all heat/cold keywords
7. For each county, we define the heat/cold wave threshold temperature as the average across all cities, and we document the most frequently appearing keyword