Simulations for frost risk, climate variability and cue use

[dbuona]

We discussed some ideas for simulations.

One:

One was to perhaps simulate climate data with 1) high inter-annual variability and relatively low intra-annual, and 2) with high intra-annual variation (also among years).

Then simulate phenology cues with a) low forcing (f*) b) high forcing cue c) chilling or photoperiod requirement + forcing cue d) other scenarios. Run simulations a lot and calculate how often a frost occurs after leaf out under each scenario as a metric of frost risk.

Two:

Use actual extracted climate parameters

@lizzieinvancouver is going to make an initial stab and some sims and @MoralesCastilla will jump in too.

[lizzieinvancouver]

Okay, 4+ months later and I got started ... see commit daab052e9deb3e194478bbab9e05b97ff4ab70c2 ... it's not ready, but something I can hopefully pick this Wednesday.

[lizzieinvancouver]

Fix the code: ranges_simclimphen.R

1. Set up a normal for mean climate
2. Create annual data (instead of sigmainter maybe think of it as mean annual, and creates MEAN spring and winter temperatures) and then use those values are the mean... draw winter and make spring one degree warmer
3. Add GDD
4. Record frost events

Eventual data frame, BIG change -- make sure dataframe grabs ANNUAL data

1. mean winter temperature and mean spring (2 columns)
2. sigma inter
3. sigma intra
4. Fstar
5. site (or rep)
6. year
7. number of frost events

[lizzieinvancouver]

A few more notes:

• when we say intra-annual we mean daily variation
• when we say inter-annaul we mean yearly variation (at one site)

High and low forcing... we're not sure what that means OSPREE cue-wise so we will just do a variety of Fstar values and leave it to @dbuona to work out.

[lizzieinvancouver]

Getting reasonable values for our simulations:

@AileneKane Two bits of code from decsens repo may help -- this one selects where we should get climate data for depending on where we had PEP data (and other constraints perhaps) and this one plots data that @cchambe12 had pulled and formatted a little ... so this later file and related data may be a good place to start? Not sure, see what you think.

[lizzieinvancouver]

@lizzieinvancouver Needs to work on getting rid of chill and, with @cchambe12 help, then adding in frost events, fixing DF etc.

[lizzieinvancouver]

@AileneKane @cchambe12 I updated the sim code changing to just GDD and trying to add frost events, it definitely needs another set of eyes @cchambe12 ....

I tried to flag where to edit for inter and intra annual variation, so that edits to that and the gdd/frost stuff can happen at the same time.

[cchambe12]

@lizzieinvancouver the frost event code looks great! Very elegant. I went through a few examples to double check the output and it all looks good to me.

[AileneKane]

@lizzieinvancouver @cchambe12 I've updated the simulation code to include inter and intra-annual variation. Please take a look and let me know if you have any questions or if I should change anything (or of course feel free to update yourself as needed!)

[AileneKane]

@lizzieinvancouver @cchambe12 I looked at the January temps for some betula pendula sites that had been pulled for decsens work, and quantified intra (daily) and inter-annual variation in january temperature, for some sense of what might be reasonable to use in our simulations here.

[AileneKane]

Intra-annual variation is within sites (17 different lats/longs) and years, Inter-annual is across years at each site

[lizzieinvancouver]

@AileneKane I thought I should add some plots of daily temp to the sim code, so that people are forced to see what's happening with that (since it's weird how I pasted it together). When I run it though ... I see (early on) flashes of the plot that look too perfect (I think). Could you run the code and see if they make sense to you? Maybe it's fine and as expected and I just don't understand, or maybe my plotting is off ... but otherwise it seems odd to me that we seem to get these estimates with zero noise (and they are not all in a row, they flash in every so often early in the sims).

[lizzieinvancouver]

Other than my one concern above ... related to original aim:

Then simulate phenology cues with a) low forcing (f*) b) high forcing cue c) chilling or photoperiod requirement + forcing cue d) other scenarios. Run simulations a lot and calculate how often a frost occurs after leaf out under each scenario as a metric of frost risk.

The sims can be adjusted for forcing ... and then we need to decide if we want to add chilling or photoperiod.