Create aggregrated antecedent weather available for use in wildfire-feedback cheatgrass loop

[kpalmqui]

In addition to fire probability based on cheatgrass biomass (see #324), we would like antecedent weather in the previous three years to be utilized for additional adjustment of the fire probability that is calculated based on cheatgrass biomass.

@menglandjohns and I are still working on the logic and code implementation for this (more details to follow), but we would like @chaukap to begin working on the code that creates the aggregated antecedent weather variables of interest. These variables were identified in Pilliod et al. 2017:
Pilliod et al. 2017.pdf

For each year, we want:

1. Mean spring (April-June) preciptitation in the three previous years.
2. Winter (October-March) precipitation in the previous year.
3. Spring (April-June) precipitation in the current year.

Using these variables, we will need to essentially check to see if the following are true:

1. Is mean spring precipitation in the previous three years higher than average?
2. Is winter precipitation in the previous year higher than average?
3. Is spring precipitation in the current year lower than average?

Because of these comparison to "average" conditions, we will also need running means for the above variables.

For year 1, we have no previous weather data to go on, so this implementation should be skipped. For years 2-3, I suggest we utilize the spring precipitation in the previous year and mean spring precipitation in the previous two years, respectively.

Much more detail to come here as we finalize:

1. How to adjust wildfire probabilities based on weather (after adjustment based on cheatgrass biomass)
2. The rules for determining whether previous weather or this year's weather is higher or lower than average.

@chaukap let me know if you have initial questions and feel free to break this down into smaller issues if you would like.

[chaukap]

I've been thinking though this issue and I believe the cleanest solution would be to create a new struct for the cheatgrass information:

struct cheatgrassPrecipitation {
  double[3] previousSpringPrecipitation;
  double winterPrecipitation;

  double meanSpringPrecipitation;
  double meanWinterPrecipitation;
}

That way all of the information is together.

While we are at it, I think it would also be a good idea to create a header file for Mortality called ST_mortality.h. We would put all of the globally accessible functions and variables there, and no longer need to redeclare or extern anything from mortality.

@kpalmqui how does that sound?

[kpalmqui]

@chaukap I think making a new struct for cheatgrass makes sense.

However, while the header file for Mortality seems like a good idea, shouldn't this be implemented in a different milestone/development branch?

[chaukap]

@kpalmqui It would be a little messy to create the header file so I agree. I'll get to work on this issue right away.

[chaukap]

@kpalmqui When you say "last winter" do you mean the previous year's October-December precipitation plus the current year's January-March precipitation, or do you mean the October-December precipitation from 2 years ago and the January-March precipitation from 1 year ago?

I'm assuming the first but I want to be sure.

[kpalmqui]

@chaukap thanks for following up about this. The Pilliod et al. 2017 study characterized a year as: October-September (see attached Figure).

Thus, with regard to your question it is actually the second: October-December precipitation from 2 years ago and the January-March precipitation from 1 year ago

[chaukap]

@kpalmqui Following commit 4f1a375 I believe all of the information you requested is available. Here's a summary of what I've done:

• Created a new struct, CheatgrassPrecip, to hold the information
• Instantiated the struct in ST_mortality.c as cheatgrassPrecip.
• Created a function for updating all fields of cheatgrassPrecip called _updateCheatgrassPrecip(int year)
• Incorporated this feature into gridded mode.

_updateCheatgrassPrecip(int year) is called at the top of mort_EndOfYear(). This means that all of the information is up to date before any decisions are made about disturbances.

To access the information:

• cheatgrassPrecip->prevSprings[i] accesses the Spring precipitation from the previous 3 Springs. prevSprings[0] refers to last year and prevSprings[2] refers to 3 years ago.
• cheatgrassPrecip->lastWinter refers to last Winter's precipitation.
• cheatgrassPrecip->currentSpring refers to this Spring's precipitation.
• cheatgrassPrecip->springMean refers to the running average of Spring precipitation. The average begins to a accumulate in year 2.
• cheatgrassPrecip->winterMean refers to the running average of Winter precipitation. The average begins to accumulate in year 3.

Testing

You can test my implementation using the print statement I left below the call to _updateCheatgrassPrecip(int year). It will give you a year-by-year print out of the information like this:

If there's anything else you need let me know!

[chaukap]

The antecedent weather is available to the wildfire code and therefore this issue is complete. If any modifications to the weather data need to be made I will open a new branch.