Strengths:
A well motivated investigation, carried out with good
Suggestions:
The outlier removal decisions seem reasonable. However, the county data reveal that some of these outliers are shared across counties, which makes it harder to explain them as data collection or processing errors.
The MIF2 diagnostics show that some of the searches essentially failed (green and cyan in the local search). It might be clearer to remove them from subsequent analysis.
It would be useful to compare to an ARMA or log-ARMA benchmark likelihood, to give some indication of whether substantial additional changes are required reach the point of having a mechanistic model with good statistical fit.