Cold Acclimation Green Crab Experiment Part 38
Water temperatures for WA and Julia’s experiments
I need to create figures depicting the water temperature conditions for each experiment! I previously analyzed temperature conditions for the MA experiment, so I proceeded to do the same thing for the WA experiment and Julia’s data.
WA experiment
I conducted all analyses in this R Markdown script. I previously started the analysis, but I stopped because there were issues with certain HOBO loggers dying and needing to be replaced!
The first thing I needed to do was calculate the average temperature conditions within tanks and treatments. I initially created a for loop to calculate conditions after the ramp down had occurred:
for (i in 1:(length(hobos)/2)) {
treatmentTempStats[i,2] <-
mean(rbind(hobos[[i]][,2][115:length(hobos[[i]][,2])],
hobos[[i+8]][,2][115:length(hobos[[i+8]][,2])]), na.rm = TRUE)
treatmentTempStats[i,3] <-
sd(rbind(hobos[[i]][,2][115:length(hobos[[i]][,2])],
hobos[[i+8]][,2][115:length(hobos[[i+8]][,2])]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStats
I made these calculations starting at the 115th row (after 7 a.m. on 2024-07-23). However, some of the HOBO loggers were put in after that point since the original logger batteries died! I individually calculated average conditions for the tanks where loggers were replaced:
treatmentTempStats$meanTemp[1] <- mean(rbind(hobos[[1]][,2][115:length(hobos[[1]][,2])],
hobos[[1+8]][,2][1:length(hobos[[1+8]][,2])]), na.rm = TRUE) #Tank 1 revised mean
treatmentTempStats$sdTemp[1] <- sd(rbind(hobos[[1]][,2][115:length(hobos[[1]][,2])],
hobos[[1+8]][,2][1:length(hobos[[1+8]][,2])]), na.rm = TRUE) #Tank 1 revised SD
treatmentTempStats$meanTemp[2] <- mean(rbind(hobos[[2]][,2][1:length(hobos[[2]][,2])],
hobos[[2+8]][,2][115:length(hobos[[2+8]][,2])]), na.rm = TRUE) #Tank 2 revised mean
treatmentTempStats$sdTemp[2] <- sd(rbind(hobos[[2]][,2][1:length(hobos[[2]][,2])],
hobos[[2+8]][,2][115:length(hobos[[2+8]][,2])]), na.rm = TRUE) #Tank 2 revised SD
treatmentTempStats$meanTemp[5] <- mean(rbind(hobos[[5]][,2][1:length(hobos[[5]][,2])],
hobos[[5+8]][,2][1:length(hobos[[5+8]][,2])]), na.rm = TRUE) #Tank 5 revised mean
treatmentTempStats$sdTemp[5] <- sd(rbind(hobos[[5]][,2][1:length(hobos[[5]][,2])],
hobos[[5+8]][,2][1:length(hobos[[5+8]][,2])]), na.rm = TRUE) #Tank 5 revised SD
treatmentTempStats$meanTemp[8] <- mean(rbind(hobos[[8]][,2][115:length(hobos[[8]][,2])],
hobos[[8+8]][,2][1:length(hobos[[8+8]][,2])]), na.rm = TRUE) #Tank 8 revised mean
treatmentTempStats$sdTemp[8] <- sd(rbind(hobos[[8]][,2][115:length(hobos[[8]][,2])],
hobos[[8+8]][,2][1:length(hobos[[8+8]][,2])]), na.rm = TRUE) #Tank 8 revised SD
The calculations were saved here.
I then calculated treatment-level conditions. The colder conditions were 2.137014 ± 0.3486562 and the cold conditions were 5.199698 ± 0.2157631. Due to having two extra tanks in the room, the conditions were not as cold as the MA experiment (~1.66ºC). However, the SE vaguely overlap, so there’s that. I’ll probably need to include some sort of section in the discussion about this, but I don’t think a ~0.5ºC difference would lead to failure to right in one population but not another.
Before I could make the figure I had to merge my datasets together by approximate time. The roll argument in the data.table package was meant for this! The package website wasn’t very helpful, so I based most of my code on this Stack Overflow question and this blog. I also confirmed that I could roll join multiple tables based on this Stack Overflow question. After A LOT of trial and error with the code, I figured out that tables should be merged from smallest to largest. I went about the merges in three steps:
#Format all temperature data files as data.tables
tank1a <- as.data.table(tank1a)
tank1b <- as.data.table(tank1b)
tank2a <- as.data.table(tank2a)
tank2b <- as.data.table(tank2b)
tank3a <- as.data.table(tank3a)
tank3b <- as.data.table(tank3b)
tank4a <- as.data.table(tank4a)
tank4b <- as.data.table(tank4b)
tank5a <- as.data.table(tank5a)
tank5b <- as.data.table(tank5b)
tank6a <- as.data.table(tank6a)
tank6b <- as.data.table(tank6b)
tank7a <- as.data.table(tank7a)
tank7b <- as.data.table(tank7b)
tank8a <- as.data.table(tank8a)
tank8b <- as.data.table(tank8b)
temp <- tank3a[tank2b, on = "dateTime", roll = T, rollends = F][tank3b, on = "dateTime", roll = T, rollends = F][tank4a, on = "dateTime", roll = T, rollends = F][tank4b, on = "dateTime", roll = T, rollends = F][tank6a, on = "dateTime", roll = T, rollends = F][tank6b, on = "dateTime", roll = T, rollends = F][tank7a, on = "dateTime", roll = T, rollends = F][tank7b, on = "dateTime", roll = T, rollends = F][tank8a, on = "dateTime", roll = T, rollends = F][tank1a, on = "dateTime", roll = T, rollends = F] #Use data.table rolling join to match HOBO logger data together by nearest dateTime. Use only the loggers wiht at least 411 observations
temp2 <- tank5b[tank1b, on = "dateTime", roll = T, rollends = F][tank2a, on = "dateTime", roll = T, rollends = F][tank5a, on = "dateTime", roll = T, rollends = F][tank8b, on = "dateTime", roll = T, rollends = F][temp, on = "dateTime", roll = T, rollends = F] #Roll HOBO loggers with limited data. First file named is the smallest one, final file rolled is temp (largest file with the most data)
Each join step, I confirmed that the data for each logger started and stopped at the correct time. I could then select the relevant columns needed to make the figure!
temp2 %>%
dplyr::select(dateTime,
temp1a, temp1b,
temp2a, temp2b,
temp3a, temp3b,
temp4a, temp4b,
temp5a, temp5b,
temp6a, temp6b,
temp7a, temp7b,
temp8a, temp8b) %>%
rowwise() %>%
mutate(., colderTemp = mean(c(temp1a, temp1b,
temp2a, temp2b,
temp3a, temp3b,
temp4a, temp4b), na.rm = TRUE),
coldTemp = mean(c(temp5a, temp5b,
temp6a, temp6b,
temp7a, temp7b,
temp8a, temp8b), na.rm = TRUE)) %>%
ggplot(., aes(x = dateTime, y = colderTemp)) +
geom_line(color = plotColors[1]) +
geom_ribbon(aes(x = dateTime, y = coldMean, ymin = coldMean - coldSD, ymax = coldMean + coldSD), fill = plotColors[2], alpha = 0.15) +
geom_hline(yintercept = coldMean, colour = plotColors[2], linetype = 3) +
geom_line(aes(x = dateTime, y = coldTemp), color = plotColors[2]) +
geom_ribbon(aes(x = dateTime, y = colderMean, ymin = colderMean - colderSD, ymax = colderMean + colderSD), fill = plotColors[1], alpha = 0.15) +
geom_hline(yintercept = colderMean, colour = plotColors[1], linetype = 3) +
scale_x_datetime(name = "") +
scale_y_continuous(name = "Temperature (ºC)",
breaks = c(seq(0,5,1), seq(5,15,5))) +
theme_classic(base_size = 15)
#Combine data from all dataframes and select columns of interest. Use rowwise operations to calculate average temperature for each treatment at each timepoint. Plot colder temepratures for each timepoint. Add a ribbon with the overall average colder temperature and SD. Add a line for the overall average colder temperature. Repeat with the cold temperature. Modify x-axis label. Modify y-axis.
Figure 1. WA water temperature conditions
Re-analysis of WA data
I revised my WA genotypes, so I need to update the data analysis in this R Markdown script. I saved the revised average TTR and threshold analysis stats afterwards.
Julia’s experiment
Time to tackle Julia’s data! I first exported and cleaned the raw data from the HOBO loggers, which can be found here. This was slightly complicated due to a temperature logger malfunction in tank 11, but I could cross-reference the anomalies with conditions in tank 13.
Then I created this R Markdown script to analyze the data. This data analysis was going to be a bit different. I needed to calculate average conditions by tank and treatment for the different parts of the experiment: pre-exposure, pulse 1, rest, and pulse 2. Before analysis, I rolled the data together:
#Format all temperature data files as data.tables
tank10a <- as.data.table(tank10a)
tank10b <- as.data.table(tank10b)
tank11a <- as.data.table(tank11a)
tank11b <- as.data.table(tank11b)
tank12a <- as.data.table(tank12a)
tank12b <- as.data.table(tank12b)
tank13a <- as.data.table(tank13a)
tank13b <- as.data.table(tank13b)
temp <- tank13a[tank10b, on = "dateTime", roll = T, rollends = F][tank10a, on = "dateTime", roll = T, rollends = F][tank13b, on = "dateTime", roll = T, rollends = F][tank11b, on = "dateTime", roll = T, rollends = F][tank12a, on = "dateTime", roll = T, rollends = F][tank12b, on = "dateTime", roll = T, rollends = F][tank11a, on = "dateTime", roll = T, rollends = F] #Roll data by closest dateTime. List begins with the shortest file and ends with the longest file
tempClean <- temp %>%
dplyr::select(dateTime,
temp10a, temp10b,
temp11a, temp11b,
temp12a, temp12b,
temp13a, temp13b) #Select necessary columns from rolled data
I then used a series of for loops to calculate conditions for each portion of the experiment:
treatmentTempStats <- data.frame("tank" = c(10:13),
"meanTempPre" = rep(0, times = 4),
"sdTempPre" = rep(0, times = 4),
"meanTempPulse1" = rep(0, times = 4),
"sdTempPulse1" = rep(0, times = 4),
"meanTempRest" = rep(0, times = 4),
"sdTempRest" = rep(0, times = 4),
"meanTempPulse2" = rep(0, times = 4),
"sdTempPulse2" = rep(0, times = 4)) #Create an empty dataframe
#Pre-exposure: 1:22
for (i in seq(2, 8, 2)) {
treatmentTempStats[(i/2),2] <-
mean(rbind(tempClean[[i]][1:22],
tempClean[[i+1]][1:22]), na.rm = TRUE)
treatmentTempStats[(i/2),3] <-
sd(rbind(tempClean[[i]][1:22],
tempClean[[i+1]][1:22]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStats
#Pulse 1: 31:108
for (i in seq(2, 8, 2)) {
treatmentTempStats[(i/2),4] <-
mean(rbind(tempClean[[i]][31:108],
tempClean[[i+1]][31:108]), na.rm = TRUE)
treatmentTempStats[(i/2),5] <-
sd(rbind(tempClean[[i]][31:108],
tempClean[[i+1]][31:108]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStats
#Rest: 112:291
for (i in seq(2, 8, 2)) {
treatmentTempStats[(i/2),6] <-
mean(rbind(tempClean[[i]][112:291],
tempClean[[i+1]][112:291]), na.rm = TRUE)
treatmentTempStats[(i/2),7] <-
sd(rbind(tempClean[[i]][112:291],
tempClean[[i+1]][112:291]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStats
#Pulse 2: 301:406
for (i in seq(2, 8, 2)) {
treatmentTempStats[(i/2),8] <-
mean(rbind(tempClean[[i]][301:406],
tempClean[[i+1]][301:406]), na.rm = TRUE)
treatmentTempStats[(i/2),9] <-
sd(rbind(tempClean[[i]][301:406],
tempClean[[i+1]][301:406]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStats
I saved the conditions here. It’s weird to see that the second pulse for tank 13 was lower than every other tank. I am not entirely sure why that happened. I then used a similar approach to calculate overall treatment conditions:
treatmentTempStatsOverall <- data.frame("treatment" = c("control", "treatment"),
"meanTempPre" = rep(0, times = 2),
"sdTempPre" = rep(0, times = 2),
"meanTempPulse1" = rep(0, times = 2),
"sdTempPulse1" = rep(0, times = 2),
"meanTempRest" = rep(0, times = 2),
"sdTempRest" = rep(0, times = 2),
"meanTempPulse2" = rep(0, times = 2),
"sdTempPulse2" = rep(0, times = 2)) #Create an empty dataframe
#Pre-exposure: 1:22
for (i in seq(2, 4, 2)) {
treatmentTempStatsOverall[(i/2),2] <-
mean(rbind(tempClean[[i]][1:22],
tempClean[[i+1]][1:22],
tempClean[[i+4]][1:22],
tempClean[[i+5]][1:22]), na.rm = TRUE)
treatmentTempStatsOverall[(i/2),3] <-
sd(rbind(tempClean[[i]][1:22],
tempClean[[i+1]][1:22],
tempClean[[i+4]][1:22],
tempClean[[i+5]][1:22]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStatsOverall
#Pulse 1: 31:108
for (i in seq(2, 4, 2)) {
treatmentTempStatsOverall[(i/2),4] <-
mean(rbind(tempClean[[i]][31:108],
tempClean[[i+1]][31:108],
tempClean[[i+4]][31:108],
tempClean[[i+5]][31:108]), na.rm = TRUE)
treatmentTempStatsOverall[(i/2),5] <-
sd(rbind(tempClean[[i]][31:108],
tempClean[[i+1]][31:108],
tempClean[[i+4]][31:108],
tempClean[[i+5]][31:108]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStatsOverall
#Rest: 112:291
for (i in seq(2, 4, 2)) {
treatmentTempStatsOverall[(i/2),6] <-
mean(rbind(tempClean[[i]][112:291],
tempClean[[i+1]][112:291],
tempClean[[i+4]][112:291],
tempClean[[i+5]][112:291]), na.rm = TRUE)
treatmentTempStatsOverall[(i/2),7] <-
sd(rbind(tempClean[[i]][112:291],
tempClean[[i+1]][112:291],
tempClean[[i+4]][112:291],
tempClean[[i+5]][112:291]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStatsOverall
#Pulse 2: 301:406
for (i in seq(2, 4, 2)) {
treatmentTempStatsOverall[(i/2),8] <-
mean(rbind(tempClean[[i]][301:406],
tempClean[[i+1]][301:406],
tempClean[[i+4]][301:406],
tempClean[[i+5]][301:406]), na.rm = TRUE)
treatmentTempStatsOverall[(i/2),9] <-
sd(rbind(tempClean[[i]][301:406],
tempClean[[i+1]][301:406],
tempClean[[i+4]][301:406],
tempClean[[i+5]][301:406]), na.rm = TRUE)
} #For each tank, combine the treatment temperature data from both HOBO loggers, either find the average or SD, and save in treatmentTempStatsOverall
I saved the output here. Tank 13 brought down the average pulse 2 conditions for the treatment tanks. Again, the ranges overlap so maybe it’s ok? I started to calculate differences using an ANOVA for each portion of the experiment, but I’m not going to include those results since I think I need to use an approach that does better with uneven sample sizes.
Finally I made the experimental condition figure! I wanted to include averages and SD ribbons for the two pulses. To do this, I employed geom_segment for the line (based on this Stak Overflow post), then defined the particular data window for geom_ribbon (based on this blog post) to ensure that they were concentrated at the pulses and not across the whole figure:
tempClean[4:406,] %>%
dplyr::select(dateTime,
temp10a, temp10b,
temp11a, temp11b,
temp12a, temp12b,
temp13a, temp13b) %>%
rowwise() %>%
mutate(., controlTemp = mean(c(temp10a, temp10b,
temp12a, temp12b), na.rm = TRUE),
treatmentTemp = mean(c(temp11a, temp11b,
temp13a, temp13b), na.rm = TRUE)) %>%
ggplot(., aes(x = dateTime, y = treatmentTemp)) +
geom_line(color = plotColors[1]) + #Average treatment temperature
geom_segment(aes(x = tempClean$dateTime[31], xend = tempClean$dateTime[108],
y = treatmentTempStatsOverall$meanTempPulse1[2], yend = treatmentTempStatsOverall$meanTempPulse1[2]),
color = plotColors[1], linetype = 3) + #Pulse 1 treatment segment
geom_ribbon(data = tempClean[31:108],
aes(x = dateTime,
y = treatmentTempStatsOverall$meanTempPulse1[2],
ymin = treatmentTempStatsOverall$meanTempPulse1[2] - treatmentTempStatsOverall$sdTempPulse1[2],
ymax = treatmentTempStatsOverall$meanTempPulse1[2] + treatmentTempStatsOverall$sdTempPulse1[2]),
fill = plotColors[1], alpha = 0.15) + #Pulse 1 treatment ribbon
geom_segment(aes(x = tempClean$dateTime[301], xend = tempClean$dateTime[406],
y = treatmentTempStatsOverall$meanTempPulse2[2], yend = treatmentTempStatsOverall$meanTempPulse2[2]),
color = plotColors[1], linetype = 3) + #Pulse 2 treatment segment
geom_ribbon(data = tempClean[301:406],
aes(x = dateTime,
y = treatmentTempStatsOverall$meanTempPulse2[2],
ymin = treatmentTempStatsOverall$meanTempPulse2[2] - treatmentTempStatsOverall$sdTempPulse2[2],
ymax = treatmentTempStatsOverall$meanTempPulse2[2] + treatmentTempStatsOverall$sdTempPulse2[2]),
fill = plotColors[1], alpha = 0.15) + #Pulse 2 treatment ribbon
geom_line(aes(x = dateTime, y = controlTemp), color = plotColors[2]) + #Control treatment temperature
geom_segment(aes(x = tempClean$dateTime[31], xend = tempClean$dateTime[108],
y = treatmentTempStatsOverall$meanTempPulse1[1], yend = treatmentTempStatsOverall$meanTempPulse1[1]),
color = plotColors[2], linetype = 3) + #Pulse 1 control segment
geom_ribbon(data = tempClean[31:108],
aes(x = dateTime,
y = treatmentTempStatsOverall$meanTempPulse1[1],
ymin = treatmentTempStatsOverall$meanTempPulse1[1] - treatmentTempStatsOverall$sdTempPulse1[1],
ymax = treatmentTempStatsOverall$meanTempPulse1[1] + treatmentTempStatsOverall$sdTempPulse1[1]),
fill = plotColors[2], alpha = 0.15) + #Pulse 1 control ribbon
geom_segment(aes(x = tempClean$dateTime[301], xend = tempClean$dateTime[406],
y = treatmentTempStatsOverall$meanTempPulse2[1], yend = treatmentTempStatsOverall$meanTempPulse2[1]),
color = plotColors[2], linetype = 3) + #Pulse 2 control segment
geom_ribbon(data = tempClean[301:406],
aes(x = dateTime,
y = treatmentTempStatsOverall$meanTempPulse2[1],
ymin = treatmentTempStatsOverall$meanTempPulse2[1] - treatmentTempStatsOverall$sdTempPulse2[1],
ymax = treatmentTempStatsOverall$meanTempPulse2[1] + treatmentTempStatsOverall$sdTempPulse2[1]),
fill = plotColors[2], alpha = 0.15) + #Pulse 2 treatment ribbon
scale_x_datetime(name = "") +
scale_y_continuous(name = "Temperature (ºC)",
limits = c(14.9,33),
breaks = c(seq(15,30,5), seq(31,33,1))) +
theme_classic(base_size = 15)
#Combine data from all dataframes and select columns of interest. Use rowwise operations to calculate average temperature for each treatment at each timepoint. Plot colder temepratures for each timepoint. Add a ribbon with the overall average colder temperature and SD. Add a line for the overall average colder temperature. Repeat with the cold temperature. Modify x-axis label. Modify y-axis.
Figure 2. Temperature conditions for Julia’s experiment
Going forward
- Create an experimental setup figure
- Outline introduction
- Outline discussion
- Genotype Catlin’s samples
- Re-analyze Catlin’s TTR data
- Analyze Catlin’s HOBO data
- Add Catlin’s methods and results to manuscript
- Determine methods for comparing population responses
- Troubleshoot lipid assay protocol
- Conduct lipid assay for crabs of interest