Preliminary analysis

Tidy up data

Read in data from 2009 to 2020. Convert data to long format.

dat_phenophase <- tidy_phenophase()

setwd("phenologyb4warmed/")
usethis::use_data(dat_phenophase, overwrite = T)
setwd("..")

I have stored processed data in the R package. Use it directly.

Table: Table 1: Phenophase status.

site	canopy	heat	heat_name	water	water_name	block	plot	species	barcode	cohort	year	doy	phenophase	status	notes
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	82	budbreak	0	NA
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	82	oneleaf	0	NA
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	82	mostleaf	0	NA
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	82	senescence	0	NA
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	82	leafdrop	0	NA
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	84	budbreak	0	NA

Calculate time of first yes for each site, canopy, treatment, block, plot, species, and phenophase combination.

dat_phenophase_time <- calc_phenophase_time(dat_phenophase)

Table: Table 2: Starting time of phenophase.

site	canopy	heat	heat_name	water	water_name	block	plot	species	barcode	cohort	year	doy	phenophase	notes
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	137	budbreak	originally only barcode recorded
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	154	oneleaf	NA
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	168	mostleaf	originally only barcode recorded
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	263	senescence	NA
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2010	277	leafdrop	NA
cfc	closed	_	ambient	_	ambient	a	a3	abiba	3148	2008	2011	151	budbreak	this is not an aceru, but likely contains the data from the missing 3148 abiba. both reps present, but 3148 abiba missing

Change in phenophase time in ambient plots over years

p_phenophase_change <- plot_phenophase_change(dat_phenophase_time)

• Overall delay in time of budbreak. Check climate trend in the area.
• First five years seem to have more outliers. Check if I used the correctly cleaned data.

Difference in phenophase time induced by treatments

Using average time of first yes in ambient plots in each site, canopy, block, species, phenophase, and year as baseline, I calculate the difference between each individual’s time to the baseline. This difference in phenophase time represent the effect of treatments.

dat_phenophase_diff <- calc_phenophase_diff(dat_phenophase_time)
p_phenophase_diff <- plot_phenophase_diff(dat_phenophase_diff, x_label_short = T)

• Warming advanced budbreak in both ambient and reduced rainfall.
• Reduced rainfall dampened the effect of warming.

• For most deciduous species, warming delayed senescence in both ambient and reduced rainfall.
• Reduced rainfall may have dampened the effect of warming, but more tests need to be done.
• Evergreen trees’ senescence did not respond much to warming or drying.

Relationship between effect size and baseline

I try to reproduce Montgomery et al. (2020) in PNAS (paper). I first pool individuals in plots together to estimate the time of phenophase for each site, canopy, treatment, block, species, phenophase, and year. I compare the phenophase time to the baseline in ambient plots again to get a difference in phenophase time.

dat_phenophase_time_summ <- summ_phenophase_time(dat_phenophase_time)
dat_phenophase_diff_summ <- calc_phenophase_diff(dat_phenophase_time_summ)

I then look at if the effect of warming treatment differ between warm and early years and cold and late years. Instead of using temperature on the x-axis, I use the baseline phenophase time in ambient plots on the x-axis, because I don’t have environmental data yet.

In Montgomery et al. (2020), they used a community-level effect size by averaging 10 species’ effect size. I try to do the same, using data in 2009-2013.

p_diff_baseline_old <- plot_diff_background(dat_phenophase_diff_summ, year_range = "old", pool_sp = T)

• Effect of warming is stronger in early years compared to late years.
• Relationship only marginally significant, probably because phenology im ambient plot is not a perfect proxy for temperature.
• Effect of 3.4°C warming was roughly twice of that of 1.7°C.

I extend the same analysis to 2009-2020.

p_diff_baseline_full <- plot_diff_background(dat_phenophase_diff_summ, year_range = "full", pool_sp = T)

• Relationship still holds qualitatively.
• Less significant probably because in the first five years, there were some extremely early or late years, but later on we saw some intermediate years, with large variations in treatment effect.

I’m not too sure if pooling all species is fair, so I look at the relationship by species.

p_diff_baseline_nopool <- plot_diff_background(dat_phenophase_diff_summ, year_range = "old", pool_sp = F)

• The relationship between block-level effect and baseline may have been driven by some species such as red maple, white spruce, and red oak.
• Some species such as trembling aspen and white pine showed the opposite relationship.