1

Model the Outbreaks

We model the number of infestations, commonly referred to as "spots," rather than modeling the beetles themselves. This approach provides more accurate predictions for forest management.

2

Statistical Analysis

We use zero-inflated regression models to account for the fact that most locations in most years do not experience outbreaks, providing robust predictions across diverse conditions.

3

Zero Inflation

Because most locations in most years do not experience an outbreak, a very large number of zeroes occurs in the data. Zero-inflation is designed for precisely this kind of data.

4

Test Input Variables

Predictions incorporate spring trapping data, previous year outbreak history, climate indices like PDSI, and winter severity to provide comprehensive risk assessments.