Re-consider drought periods

[teixeirak]

I'm wondering if it would be worth it to change how we're calculating the drought year. Foster et al specifically did a October-September timeframe to calculate TMAX and other values, and it reinforces to me that while we identify 1966, 1977, and 1999 as drought years, the PDSI values don't concisely match those years (e.g. the 1966 drought went from 1965 - a bit in 1967). As in, taking a mean of PDSI values over all years reveals slight derivations in these three, but it's not the same as taking the values for the drought itself.

• I say this especially given 1977 I believe has positive PDSI values for the first part of the year and then drops.

Foster et al also bring in potential temperature variations in the future using RCPs 4.5 and 8.5 from the IPCC report. I think that could be interesting but I also wonder if that's going too far for this particular analysis.

Also of note, it seems many tree ring analyses use temperature as a fixed effect.

_Originally posted by @mcgregorian1 in https://github.com/SCBI-ForestGEO/McGregor_climate-sensitivity-variation/issues/9#issuecomment-479545029_

[teixeirak]

Since 1901, these are the 10 largest droughts, defined as May-June PET - PRE: 10 largest PET-PRE: 82.63, 81.9, 112.43, 84.57, 89.43, 83.87, 83.37, 86.97, 80, 82.13 Corresponding drought years: 1911, 1914, 1930, 1936, 1944, 1964, 1966, 1977, 1999, 2007

Here's a figure showing species ring width indexes in response to these droughts.

Our droughts (and droughts in general) clearly differ in terms of previous conditions, onset speed, severity, etc. Perhaps it makes sense to analyze them all separately, at least to get a sense of whether different factors contribute during different droughts?

Note that 1964-1966 is all drought. This might bias/ weaken the 1966 response (relative to previous 5 years).

[mcgregorian1]

Ok. This is where we start running into discrepancies, because from the tree cores I'm using, there were no pointer years (years where at least 50% of trees had at least 30% negative growth compared to prior years) for 1914, 1930, 1936, 1944, and 2007.

• as a reminder, I discounted any pointer years that had <5 cores (to maintain a majority)
• Changing the metrics above will obviously give us different results and entirely different models. I'm not sure if we want to do this -- Should we talk about this more? Changing metrics could reveal more significant effects than what we're seeing.

[teixeirak]

I'm not suggesting that we use all those years, just posting the full list FYI. I don't think it makes much sense to go back too far (particularly with canopy position); I'd exclude pre-1950 droughts. 2007 was dry by this particular metric, but didn't seem to put much stress on the trees.

[mcgregorian1]

In sum of running the models per year, the top model for each is below

year	position	fixed effects in model	R2	AIC diff. of next best model
1964	best	position, tlp, rp, elev, dbh	0.195	0.5
1964	second	position, tlp, rp, dbh	0.19	>3
1966	best	tlp, rp, dbh	0.26	>1.5
1977	best	position, tlp, rp, elev, dbh	0.183	0.9
1999	best	position, rp, elev, dbh	0.219	0.26
1999	second	position, tlp, rp, elev, dbh	0.227	0.17
1999	third	position, rp, dbh	0.214	0.5
all	best	position, tlp, rp, dbh	0.132	0.52

1964

r-squared for top 2 models is 0.195 and 0.19, respectively.

1966

r-squared for the top model is 0.26

1977

r-squared for top 2 models is 0.183 and 0.181, respectively

1999

r-squared for top 3 models is 0.219, 0.227, and 0.214, respectively

all years

r-squared for top model is 0.132

[teixeirak]

I like what you've done here-- breaking out by drought and also showing the full model. One modification, though: please be sure to include year in instances of the full model. The top two results shown above don't include year and have to be thrown out.

[mcgregorian1]

Ah you're right, that's my bad from that iteration of running the code