ALE Data exploration
Kenzie Cooke
2024-09-30
1 Quality Control of carbonate chemistry
Alkalinity bottle sample quality. Each sample is run in replicate for quality control. If the replicates are within an acceptable threshold (usually < 5umol difference), then the mean of the replicates is used as the final TA value for that sample. The difference in umol between each replicate (deltas) for all collected samples are plotted below
1.1 Instrument precision (replicate deltas)
1.1.1 AllBottles_delta(umol)
Histogram depicts deltas between replicates of all 526 alkalinity samples collected for from Febuary 7th to March 19th, 2024
1.1.2 Deltas by treatment
1.1.3 deltas > 5 umol by treatment
1.2 Precipitation
1.2.1 PrecipitatedSamples_treatment
number of samples with confirmed visual precipitation in bottles post collection. Not confident that all samples with precipitatin were noted down.
1.2.2 Number of samples with suspicious pCO2 values
Carbchem parameters generated using all bottle samples (DIC and TA)
Datapoints where computed pCO2 (using DIC and TA in seacarb) produced suspicious values (likely due to visible precipitation in bottle samples post collection)
1.3 Stability of treatments over time
1.3.1 AmbientTreat_time
Ambient treatment over time using all bottle sample data
1.3.2 ElevatedTreat_time
1.3.3 HighTreat_time
1.3.4 XhiTreat_time
1.3.5 TA_Time (precipitation excluded, >5umol delta excluded)
Total samples collected over course of the experiment excluding those with precipitation and any with delta >5 umol> Left with 399 samples. Down from 550, eliminated 151 samples.
Break down of remaining samples:
ambient 128
elevated 114
hi 87
xhi 67
Updated plots below. Can see that in the hi and extra high treamtents, gets as low as four samples per time point towards the end of the experiment (down from 12 per tiempoint per treatment).
2 QC Growth rates
Took timepoint 0 on Feb 3rd. Corals were placed in vessels and ramped up to treatment levels for four days. Were at full treatment dosage on Feb 7. Noticed the lights were not set to the same parameters on Feb 15th. Corrected and took new datapoint on Feb 19th (T1). Took another point on March 8th (T2) and endpoint on March 23rd (T3).
2.1 Growth rates over time
2.1.1 Total calcification growth rates across timepoints
total calcificaiton growth rates across timepoints
Using initial timepoint zero to end (left) vs. using second timepoint zero to end (right)
2.1.2 Linear extension
2.1.2.1 Standards
Measured a standard in three photos from each timepoint to check precision/accuracy of imageJ method across timepoints
Across all timepoints, the ImageJ method has a standard deviation of ± 0.39 mm and a standard error of ± 0.11 mm.
Timepoint two’s standard is measuring nearly 0.75 mm shorter than timepoints zero and one. Can I apply an offset or something to correct for it?
2.1.2.2 Linear extension growth rates across timepoints
You’ll notice a a few negative rates from t2 to t3. The -0.4 mm/day rate comes from a coral that died three days after timepoint and it had a -1.47 mm difference in total length between tp2 and 3. I am going to maybe re-analyze the photos from timepoint 2.
linear extension growth rates across all timepoints. Going to investigate the super negative rates.
Using initial timepoint zero - end (left) vs. using second timeppoint zero to end (right)
2.1.3 Linear extension with offset correction
Growth results with linear extension offset applied to each timepoint. Took the mean standard length of timepoint 1 and (24.1 mm) and calculated offset for each timepoint that would bring the mean to 24.1 mm.
t0 = offset of +0.02 mm
t1 = no offset
t2 = offset of + 0.873 mm
t3 = offset of + 0.383 mm
Comparing growth rates withdout and with the offset applied (gr_t0t1 vs grt0t1e) “e” standing for edited.
3 Results
Data analysis using bottle data excluding deltas > or equal to 5 umol and any with precipitation.
3.1 Carb chem parameters
Carb chem parameters generated using seacarb with measured DIC and TA values. Includes samples from 2/3 - 3/23.
3.1.1 ANOVA_treatments
Was alkalinity statistically different between treatments?
## Df Sum Sq Mean Sq F value Pr(>F)
## treatment 3 21.650 7.217 174.9 <2e-16 ***
## Residuals 36 1.485 0.041
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Tukey multiple comparisons of means
## 95% family-wise confidence level
##
## Fit: aov(formula = avg_ta_mmol ~ treatment, data = grdata)
##
## $treatment
## diff lwr upr p adj
## ELEV-AMB 1.1974367 0.9527735 1.4420999 0.0000000
## HI-AMB 1.7490378 1.5043746 1.9937010 0.0000000
## XHI-AMB 1.8508255 1.6061623 2.0954887 0.0000000
## HI-ELEV 0.5516011 0.3069379 0.7962643 0.0000032
## XHI-ELEV 0.6533888 0.4087256 0.8980520 0.0000001
## XHI-HI 0.1017877 -0.1428755 0.3464509 0.6794773All treatments statistically different from eachother (p < 0.000) except for the extra high and high treatments (p = 0.68)
3.1.2 Stability of treatment over time
3.1.2.1 TA by treatment barplot and boxplot
## # A tibble: 4 × 3
## Treatment mean_ta sem_ta
## <fct> <dbl> <dbl>
## 1 AMB 2469. 12.8
## 2 ELEV 3653. 26.4
## 3 HI 4204. 74.5
## 4 XHI 4368. 70.2
#### Treatment stability over time
3.1.2.2 TreatStability_Time_AllDataPoints
3.1.3 Blanks
Total alkalinity in blanks by treatment
3.2 TotalCalcification_Alkalinity
3.2.1 Linear mixed effect models
3.2.1.1 T0 -> T3 (before lights fixed)
to check for random effects (tank, genotype) BW from t0 to t3, growth ~ ta, 1|geno, 1|tank
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: gr_t0t3 ~ avg_ta_mmol + (1 | geno) + (1 | tank)
## Data: grdata
##
## REML criterion at convergence: -28.6
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.49722 -0.86908 -0.05244 0.72798 1.65417
##
## Random effects:
## Groups Name Variance Std.Dev.
## tank (Intercept) 0.006953 0.08339
## geno (Intercept) 0.001861 0.04314
## Residual 0.019610 0.14003
## Number of obs: 40, groups: tank, 4; geno, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 0.003842 0.120247 33.871501 0.032 0.975
## avg_ta_mmol 0.178375 0.029403 33.153219 6.067 7.79e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## avg_ta_mmol -0.895## Type III Analysis of Variance Table with Satterthwaite's method
## Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
## avg_ta_mmol 0.72169 0.72169 1 33.153 36.803 7.791e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## ANOVA-like table for random-effects: Single term deletions
##
## Model:
## gr_t0t3 ~ avg_ta_mmol + (1 | geno) + (1 | tank)
## npar logLik AIC LRT Df Pr(>Chisq)
## <none> 5 14.319 -18.638
## (1 | geno) 4 14.052 -20.104 0.5343 1 0.46480
## (1 | tank) 4 11.978 -15.956 4.6827 1 0.03047 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Backward reduced random-effect table:
##
## Eliminated npar logLik AIC LRT Df Pr(>Chisq)
## <none> 5 14.319 -18.638
## (1 | geno) 1 4 14.052 -20.104 0.5343 1 0.46480
## (1 | tank) 0 3 11.871 -17.742 4.3620 1 0.03675 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Backward reduced fixed-effect table:
## Degrees of freedom method: Satterthwaite
##
## Eliminated Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
## avg_ta_mmol 0 0.71011 0.71011 1 35.332 34.111 1.212e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Model found:
## gr_t0t3 ~ avg_ta_mmol + (1 | tank)
Appears there is a tank effect on bw growth when using t0 to t3 growth points… Makes sense because of my lights!
3.2.1.2 T1 -> T3 (after lights fixed)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: gr_t1t3 ~ avg_ta_mmol + (1 | geno) + (1 | tank)
## Data: grdata
##
## REML criterion at convergence: -15.8
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.4794 -0.7872 -0.1146 0.6177 2.0609
##
## Random effects:
## Groups Name Variance Std.Dev.
## tank (Intercept) 0.001767 0.04203
## geno (Intercept) 0.004355 0.06599
## Residual 0.029426 0.17154
## Number of obs: 40, groups: tank, 4; geno, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -0.11670 0.14118 34.77430 -0.827 0.414
## avg_ta_mmol 0.21147 0.03587 33.77813 5.895 1.21e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## avg_ta_mmol -0.931## Type III Analysis of Variance Table with Satterthwaite's method
## Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
## avg_ta_mmol 1.0225 1.0225 1 33.778 34.747 1.208e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## ANOVA-like table for random-effects: Single term deletions
##
## Model:
## gr_t1t3 ~ avg_ta_mmol + (1 | geno) + (1 | tank)
## npar logLik AIC LRT Df Pr(>Chisq)
## <none> 5 7.9005 -5.8010
## (1 | geno) 4 7.2820 -6.5639 1.23711 1 0.2660
## (1 | tank) 4 7.7342 -7.4685 0.33255 1 0.5642## Backward reduced random-effect table:
##
## Eliminated npar logLik AIC LRT Df Pr(>Chisq)
## <none> 5 7.9005 -5.8010
## (1 | tank) 1 4 7.7342 -7.4685 0.33255 1 0.5642
## (1 | geno) 2 3 7.1643 -8.3286 1.13988 1 0.2857
##
## Backward reduced fixed-effect table:
## Eliminated Df Sum of Sq RSS AIC F value Pr(>F)
## avg_ta_mmol 0 1 0.99823 2.2732 -112.71 29.753 3.186e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Model found:
## gr_t1t3 ~ avg_ta_mmol
Using the timepoint after the lights were fixed eliminates the tank effect.
Simplest linear model:
##
## Call:
## lm(formula = gr_t1t3 ~ avg_ta_mmol, data = grdata)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.29030 -0.15385 -0.01992 0.14237 0.43907
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.09713 0.14210 -0.683 0.498
## avg_ta_mmol 0.20772 0.03808 5.455 3.19e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.1832 on 38 degrees of freedom
## Multiple R-squared: 0.4391, Adjusted R-squared: 0.4244
## F-statistic: 29.75 on 1 and 38 DF, p-value: 3.186e-06## Analysis of Variance Table
##
## Response: gr_t1t3
## Df Sum Sq Mean Sq F value Pr(>F)
## avg_ta_mmol 1 0.99823 0.99823 29.753 3.186e-06 ***
## Residuals 38 1.27494 0.03355
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Total alkalinity does have an effect on total calcification (p <
0.001)
R-squared: 0.439
slope: 0.2077
3.2.2 Plots
3.2.2.1 TA_total calcification (Linear regression, T0-T3)
Effect of total alkalinity on calcification (standardized to initial surface area) from Feb 3rd to Mar23rd. (p < 0.000
total calcification growth rates by treatment from t0 to t3
3.2.2.2 TA_total calcification (Linear regression, T1-T3)
Effect of total alkalinity on calcification (standardized to initial surface area) from Feb 19th to Mar 23rd. p < 0.000
total calcification growth rates by treatment from t1 to t3
3.3 LinearExtension_Alkalinity
3.3.1 Linear mixed effect models
3.3.1.1 T0 -> T3
Mixed effect model to check for genotype and tank effects on linear extension (used linear extension growth rates with the offsets)
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: le_t0t3e ~ avg_ta_mmol + (1 | geno) + (1 | tank)
## Data: grdata
##
## REML criterion at convergence: -141.9
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.24567 -0.45090 0.09097 0.48883 2.23927
##
## Random effects:
## Groups Name Variance Std.Dev.
## tank (Intercept) 0.000e+00 0.00000
## geno (Intercept) 3.684e-05 0.00607
## Residual 1.147e-03 0.03387
## Number of obs: 40, groups: tank, 4; geno, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 0.046475 0.026571 37.799022 1.749 0.0884 .
## avg_ta_mmol 0.012092 0.007048 36.639295 1.716 0.0947 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## avg_ta_mmol -0.970
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')## Type III Analysis of Variance Table with Satterthwaite's method
## Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
## avg_ta_mmol 0.0033773 0.0033773 1 36.639 2.9435 0.09467 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## ANOVA-like table for random-effects: Single term deletions
##
## Model:
## le_t0t3e ~ avg_ta_mmol + (1 | geno) + (1 | tank)
## npar logLik AIC LRT Df Pr(>Chisq)
## <none> 5 70.956 -131.91
## (1 | geno) 4 70.871 -133.74 0.16931 1 0.6807
## (1 | tank) 4 70.956 -133.91 0.00000 1 1.0000## Backward reduced random-effect table:
##
## Eliminated npar logLik AIC LRT Df Pr(>Chisq)
## <none> 5 70.956 -131.91
## (1 | tank) 1 4 70.956 -133.91 0.00000 1 1.0000
## (1 | geno) 2 3 70.871 -135.74 0.16931 1 0.6807
##
## Backward reduced fixed-effect table:
## Eliminated Df Sum of Sq RSS AIC F value Pr(>F)
## avg_ta_mmol 1 1 0.0032942 0.047891 -267.11 2.8069 0.1021
##
## Model found:
## le_t0t3e ~ 1
no tank of genotype effect, but also no significant effect of TA on linear extension.
3.3.1.2 T1 -> T3 ’
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: le_t1t3e ~ avg_ta_mmol + (1 | geno) + (1 | tank)
## Data: grdata
##
## REML criterion at convergence: -140.4
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.64787 -0.49941 -0.00502 0.60640 1.97451
##
## Random effects:
## Groups Name Variance Std.Dev.
## tank (Intercept) 0.000e+00 0.000000
## geno (Intercept) 4.666e-05 0.006831
## Residual 1.190e-03 0.034499
## Number of obs: 40, groups: tank, 4; geno, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) -0.000715 0.027123 37.711310 -0.026 0.9791
## avg_ta_mmol 0.015738 0.007179 36.543885 2.192 0.0348 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr)
## avg_ta_mmol -0.968
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')## Type III Analysis of Variance Table with Satterthwaite's method
## Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
## avg_ta_mmol 0.00572 0.00572 1 36.544 4.8059 0.03481 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## ANOVA-like table for random-effects: Single term deletions
##
## Model:
## le_t1t3e ~ avg_ta_mmol + (1 | geno) + (1 | tank)
## npar logLik AIC LRT Df Pr(>Chisq)
## <none> 5 70.197 -130.40
## (1 | geno) 4 70.085 -132.17 0.2257 1 0.6347
## (1 | tank) 4 70.197 -132.40 0.0000 1 1.0000## Backward reduced random-effect table:
##
## Eliminated npar logLik AIC LRT Df Pr(>Chisq)
## <none> 5 70.197 -130.40
## (1 | tank) 1 4 70.197 -132.40 0.0000 1 1.0000
## (1 | geno) 2 3 70.085 -134.17 0.2257 1 0.6347
##
## Backward reduced fixed-effect table:
## Eliminated Df Sum of Sq RSS AIC F value Pr(>F)
## avg_ta_mmol 0 1 0.0056594 0.052141 -263.71 4.6267 0.0379 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Model found:
## le_t1t3e ~ avg_ta_mmol
Hooray! Signficiant effect of TA on linear extension with no tank or genotype effects !!!!
##
## Call:
## lm(formula = le_t1t3e ~ avg_ta_mmol, data = grdata)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.088059 -0.020747 -0.002731 0.024178 0.069129
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.0002348 0.0271335 -0.009 0.9931
## avg_ta_mmol 0.0156402 0.0072712 2.151 0.0379 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.03497 on 38 degrees of freedom
## Multiple R-squared: 0.1085, Adjusted R-squared: 0.08508
## F-statistic: 4.627 on 1 and 38 DF, p-value: 0.0379## Analysis of Variance Table
##
## Response: le_t1t3e
## Df Sum Sq Mean Sq F value Pr(>F)
## avg_ta_mmol 1 0.005659 0.0056594 4.6267 0.0379 *
## Residuals 38 0.046482 0.0012232
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## [1] 0.1085393## avg_ta_mmol
## 0.01564021Simplest linear model produces p = 0.038
r-squared: 0.11
slope: 0.016
3.3.2 Plots
3.3.2.1 TA_LinearExtension (T0-T3)
Effect of TA on linear extension from Feb 3rd to Mar 23rd. No statistically significant effect (p = 0.095)
linear extension growth rates by treatment from t0 to t3
3.3.2.2 TA_LinearExtension (T1-T3)
Effect of TA on linear extension from Feb 19th to Mar 23rd. Significant effect (P = 0.038)
3.3.2.3 Total_LE_Growth(mm)_TA
linear extension growth rates by treatment from t1 to t3
linear extension– total growth vs TA.
3.4 Final Figures
3.4.1 Buoyant weight
3.4.2 Linear Extension
