According to our USGS colleagues, the best way to implement the number of ignitions/day as a function of FWI is with a 'zero-inflated Poisson distribution'. This is basically a log-linked Poisson with two parameters, a and B, which takes the form:
Number of ignitions = e ^ (a + B*FWI)
When I fit this equation to Tahoe and run it with historic FWI for a single year, it produces
So we need to implement this for lightning and human accidental ignitions per day, allowing the user to define the two parameters. for each ignition type. I'm still working on the shape of prescribed burns, which is a little different.
According to our USGS colleagues, the best way to implement the number of ignitions/day as a function of FWI is with a 'zero-inflated Poisson distribution'. This is basically a log-linked Poisson with two parameters, a and B, which takes the form:
Number of ignitions = e ^ (a + B*FWI)
When I fit this equation to Tahoe and run it with historic FWI for a single year, it produces
So we need to implement this for lightning and human accidental ignitions per day, allowing the user to define the two parameters. for each ignition type. I'm still working on the shape of prescribed burns, which is a little different.