UV Meta-analysis

Supplementary Information

GitHub Repository
The Rmarkdown file can be downloaded from the Code drop down menu (top right).

This supplementary file contains the R workflow for processing and analysing the raw data, and creating figures for the manuscript ID: GCB-23-0861 titled “Sub-lethal consequences of ultraviolet radiation exposure on vertebrates: synthesis through meta-analysis”.

---

Supplementary Information

PRISMA

Fig. S1. PRISMA flow diagram for the systematic data-collection process. n = number of papers remaining after each stage of selection. k = number of effect size after processing individual data, and n is number of observations/records after processing data. Searches were grouped by taxonomic groups.

---

Dataset

Calculate effect size, \(lnRR\), and sampling variance, \(v_i\)

The raw data for analysis is available on GitHub. This section is the workflow to import and clean the raw data for analysis. Effect size as the natural log-transformed response ratio, lnRR (1), and the sampling variance, \(v_i\) (2), was calculated following Hedges and Olkin (2014):

\[ \begin{equation} lnRR = \ln \left( \frac{\bar{X}_{1}}{\bar{X}_{2}}\right), \tag{1} \end{equation} \] \[ \begin{equation} v_i = \frac{SD_{1}^2} {\bar{X}^2_{1}n_{1}} + \frac{SD_{2}^2} {\bar{X}^2_{2}n_{2}}, \tag{2} \end{equation} \]

where \(\bar{X}_1\), \(SD_1\), and \(n_1\) are the mean, standard deviation, and sample size of the exposed group, respectively, while \(\bar{X}_2\), \(SD_2\), and \(n_2\) are the mean, standard deviation, and sample size of the control group, respectively. The standard error, \(se_i\) (3), was calculated as:

\[ \begin{equation} se_{i} =\sqrt{v_i}. \tag{3} \end{equation} \]

The inverse of \(se_i\) or precision (4) was calculated as:

\[ \begin{equation} p_{i} = \frac{1}{se_i}. \tag{4} \end{equation} \]

To account for unbalanced sampling, we used the ‘effective sample size’ instead of the sample size (\(n_1\) + \(n_2\)) for the meta-analysis of lnRR (Nakagawa et al., 2022). The effective sample size (5) is formulated as:

\[ \begin{equation} 4\tilde{n_{i}} = \frac{4n_{1i} n_{2i}}{n_{1i} + n_{2i}}. \tag{5} \end{equation} \]

When \(n\) = \(n_1\) = \(n_2\), the formula reduces to 2\(n\). Indeed, the inverse of \(\tilde{n_{i}}\) is a part of sampling variance in lnRR (6):

\[ \begin{equation} \frac{1}{\tilde{n_{i}}} = \frac{n_{1i} + n_{2i}}{n_{1i} n_{2i}} = \frac{1}{n_{1i}} + \frac{1}{n_{2i}}. \tag{6} \end{equation} \]

All conversions for the standard error (\(se_i\)), sampling variance, precision (the inverse of \(se_i\)), effective sample size (\(4\tilde{n_{i}}\)) and the inverse of \(\tilde{n_{i}}\) (\(\frac{1}{\tilde{n_{i}}}\)) followed Nakagawa et al. (2022). Data were also prepared to test for publication bias, and sources of non-independance such as shared controls, study effect, species effect, and phylogeny.

# Load and clean raw data
raw_dat <- read.csv("/Users/nicholaswu/Library/CloudStorage/OneDrive-WesternSydneyUniversity/UV meta-analysis/Vertebrate_UV_MetaAnalysis_FINAL.csv") %>%
  tibble::rowid_to_column("es_ID") %>% # add effect size ID
  dplyr::mutate(vi            = t_SD ^ 2 / (t_n * t_mean ^ 2) + c_SD ^ 2 / (c_n * c_mean ^ 2), # Sampling variance (v)
                sei           = sqrt(vi), # Standard error (SE)
                sei_inv       = 1 / sei, # Precision (inverse of SE)
                eff_n         = (4 * c_n * t_n) / (c_n + t_n), # Effective sample size
                inv_eff_n     = (c_n + t_n) / (c_n * t_n), # Inverse of eff_n
                sqrt_inv_eff  = sqrt(inv_eff_n), # Square root of the inverse of eff_n
                year_centre   = pub_year - mean(pub_year),  # Mean-centring year of publication)
                ln_delta_dose = log(delta_dose_J_day),
                ln_dur_day    = log(t_duration_day),
                species_OTL   = as.factor(species_OTL),
                life_stage    = factor(life_stage, levels = c("Embryo", "Larva", "Juvenile", "Adult")),
                region        = factor(region, levels = c("Tropical", "Subtropical", "Temperate")),
                classification = as.factor(classification),
                source_cat    = as.factor(source_cat),
                medium        = as.factor(medium),
                trait = as.factor(trait),
                metric        = as.factor(metric),
                uv_type       = as.factor(uv_type)) %>% 
  dplyr::select(-c(reference, notes))

# Account for shared controls
M <- metaAidR::make_VCV_matrix(data = raw_dat, V = "vi", cluster = "shared_control2", m = "c_mean",
                     sd = "c_SD", n = "c_n", type = "vcv", vcal = "ROM", obs = "es_ID")

Data summary

The raw_dat used for the meta-analysis comprises of 895 individual effect sizes from 73 studies across 47 species (detailed in Supplementary Table S1).

Table S1 - Summary

Table S1 - Responses. Trait categories and the specific metrics extracted for the study. Number of effect sizes, studies and species are shown. Responses with asterisks were corrected for direction. CAT = catalase, GST = glutathione S-transferase, LDH = lactate dehydrogenase, LPO = lactoperoxidase, ROS = reactive oxygen species, SOD = superoxide dismutase, U_crit = critical swimming speed.

Categorised trait	Specific metric	Effect size (k)	Studies (n)	Species (n)
Behaviour	Sheltering Behaviour	14	4	7
Behaviour	Time Active	7	2	2
Cardiorespiratory Function	Heart Rate	25	4	3
Cardiorespiratory Function	Hematocrit	20	7	5
DNA Damage	DNA Damage*	42	6	8
DNA Damage	Photolyase Gene Expression	15	1	1
DNA Damage	Pyrimidine Dimers*	4	1	1
Energy Metabolism	Aerobic Scope	3	2	1
Energy Metabolism	LDH Activity	21	1	1
Energy Metabolism	Maximum Metabolic Rate	11	3	3
Energy Metabolism	P:O Ratio	1	1	1
Energy Metabolism	Respiratory Control Ratio	1	1	1
Energy Metabolism	Resting Metabolic Rate	3	2	1
Energy Metabolism	Routine Metabolic Rate	24	11	7
Energy Metabolism	State 3 Respiration	1	1	1
Energy Metabolism	State 4 Respiration	1	1	1
Energy Metabolism	Tissue Level Respiration	1	1	1
Growth	Final Length	112	17	16
Growth	Final Mass	45	18	17
Growth	Growth Rate	2	1	1
Growth	Length (change)	4	1	1
Growth	Specific Growth Rate	9	2	2
Hormonal Stress Response	Corticosterone	7	3	2
Hormonal Stress Response	Cortisol	31	7	5
Locomotion	Burst Swimming	40	6	5
Locomotion	Routine Swimming Speed	24	1	1
Locomotion	Ucrit	5	4	2
Neuromusclular	Acytlcholinesterase	6	1	1
Oxidant	H2O2 Levels*	8	1	1
Oxidant	ROS Production*	27	3	2
Oxidative Damage	Alanine Aminotransferase*	8	1	1
Oxidative Damage	Asparate Aminotransferase*	8	1	1
Oxidative Damage	LPO Activity*	76	7	7
Oxidative Damage	Malondialdehyde*	2	2	2
Oxidative Damage	Protein Carbonyl*	5	4	2
Oxidative Damage	Protein Peroxide Levels*	1	1	1
Oxidative Damage	Thiobarbituric Acid Reactive Substances*	6	1	1
Oxidative Stress Response/Defense	CAT Activity	65	8	8
Oxidative Stress Response/Defense	CAT Gene Expression	23	2	2
Oxidative Stress Response/Defense	DT-diaphorase	1	1	1
Oxidative Stress Response/Defense	Glutathione	10	3	2
Oxidative Stress Response/Defense	Glutathione disulfide reductase	7	2	2
Oxidative Stress Response/Defense	Glutathione Peroxidase	7	2	2
Oxidative Stress Response/Defense	Glutathione Transferase	1	1	1
Oxidative Stress Response/Defense	GST Activity	42	2	3
Oxidative Stress Response/Defense	SOD Activity	65	9	9
Oxidative Stress Response/Defense	SOD Gene Expression	8	1	1
Oxidative Stress Response/Defense	SOD1 Gene Expression	15	1	1
Oxidative Stress Response/Defense	SOD3 Gene Expression	15	1	1
Oxidative Stress Response/Defense	Total Antioxidant Capacity	4	1	1
Oxidative Stress Response/Defense	Total Peroxidase	12	2	2

Life stage

Table S1 - Life stage. Continuation of data summary with additional information on the number of effect size, studies, and species in the database by life stage. Presentation in order by the number of effect size in the data.

Life stage	Effect size (k)	Studies (n)	Species (n)
Juvenile	285	22	16
Larva	262	32	22
Embryo	180	7	5
Adult	168	16	12

Species

Table S1 - Species. Continuation of data summary with additional information on the number of effect size, and studies in the database by species. Presentation in order by the number of effect size in the data.

Class	Species	Effect size (k)	Studies (n)
Fish	Danio rerio	97	4
Fish	Dicentrarchus labrax	95	2
Fish	Sparus aurata	75	2
Fish	Amphiprion clarkii	64	1
Fish	Coryphaena hippurus	61	2
Fish	Solea senegalensis	60	1
Fish	Clarias gariepinus	58	2
Amphibian	Litoria caerulea	57	1
Fish	Gambusia holbrooki	34	4
Fish	Coregonus lavaretus	29	3
Fish	Girella laevifrons	28	5
Amphibian	Limnodynastes peronii	23	4
Fish	Rutilus rutilus	19	2
Amphibian	Boana pulchella	15	2
Fish	Coregonus albula	15	2
Amphibian	Rana temporaria	14	6
Fish	Salmo salar	14	2
Fish	Pleuronectes platessa	12	1
Fish	Oncorhynchus mykiss	11	3
Amphibian	Bufo bufo	10	2
Fish	Amatitlania nigrofasciata	10	2
Fish	Cyprinus carpio	9	2
Amphibian	Ambystoma maculatum	8	1
Reptile	Psammodromus algirus	7	1
Amphibian	Ambystoma gracile	6	1
Bird	Gallus gallus	6	3
Fish	Engraulis mordax	6	1
Reptile	Sceloporus occidentalis	6	1
Amphibian	Rana arvalis	5	2
Fish	Scomber japonicus	5	1
Fish	Colossoma macropomum	4	1
Fish	Oryzias latipes	4	1
Reptile	Sauromalus ater	4	1
Amphibian	Limnodynastes ornatus	3	1
Amphibian	Rana cascadae	3	2
Amphibian	Ambystoma macrodactylum	2	1
Fish	Graus nigra	2	1
Fish	Pomacentrus amboinensis	2	1
Fish	Thalassoma lunare	2	1
Reptile	Eublepharis macularius	2	1
Reptile	Furcifer pardalis	2	1
Amphibian	Rana aurora	1	1
Amphibian	Rana blairi	1	1
Fish	Arapaima gigas	1	1
Fish	Esox lucius	1	1
Fish	Pimephales promelas	1	1
Fish	Seriola lalandi	1	1

Traits

Table S1 - Traits. Continuation of data summary with additional information on the number of effect size, studies, and species in the database by by traits. Presentation in order by the number of effect size in the data.

Trait	Effect size (k)	Studies (k)	Species (n)
Oxidative Stress Response/Defense	275	15	12
Growth	172	32	27
Oxidative Damage	106	11	8
Locomotion	69	11	8
Energy Metabolism	67	16	10
DNA Damage	61	8	10
Cardiorespiratory Function	45	11	8
Hormonal Stress Response	38	10	7
Oxidant	35	4	3
Behaviour	21	6	9
Neuromusclular	6	1	1

UV type

Table S1 - UV type. Continuation of data summary with additional information on the number of effect size, studies, and species in the database by UV radiation type and taxonomic class. Presentation in order by the number of effect size in the data.

UV type	Class	Effect size (k)	Studies (k)	Species (n)
UVA	Amphibian	25	7	4
UVA	Bird	5	2	1
UVA	Fish	245	12	9
UVA	Reptile	2	1	1
UVB	Amphibian	123	23	13
UVB	Bird	1	1	1
UVB	Fish	475	33	23
UVB	Reptile	19	5	5

Source

Table S1 - Source. Continuation of data summary with additional information on the number of effect size, studies, and species in the database by collection source. Presentation in order by the number of effect size in the data.

Source	Effect size (k)	Studies (k)	Species (n)
Captive	512	30	24
Wild	377	42	26
Not Specified	6	1	2

---

Which data are missing?

The following data were not included due of poor reporting by the studies extracted: age (hours post-fertilisation and days post-hatching), body length, body mass, latitude, and longitude. Threshold was set at Good = 0-5%, Okay = 5-40%, Poor = 40-80%, Scarce = 80-100%.

Fig. S2. Frequency of missing data for each variable extracted.

---

Prepare phylogeny for analysis

This section provides the workflow to extract the phylogenetic tree from the Open Tree of Life (OTL; see phylogenetic reconstruction in main text), match the OTL names with the data set, and create a correlation matrix for subsequent analysis.

sp_all <- sort(unique(as.character(raw_dat$species_OTL)))  # generate list of species (as character format)
taxa <- rotl::tnrs_match_names(names = sp_all)  # match taxonomic names to the OTL

# check if species list match OT identifier taxa[taxa$approximate_match ==
# TRUE,] # none so far

# retrieving phylogenetic relationships among taxa in the form of a trimmed
# sub-tree
tree <- rotl::tol_induced_subtree(ott_ids = rotl::ott_id(taxa), label_format = "name")

# Change OTL species name to match raw_dat
tree$tip.label[tree$tip.label == "Oncorhynchus_mykiss_(species_in_domain_Eukaryota)"] <- "Oncorhynchus_mykiss"

# Compute branch lengths
tree <- ape::compute.brlen(tree, method = "Grafen", power = 1)
tree <- ape::multi2di(tree, random = TRUE)  # use a randomization approach to deal with polytomies

phylo_cor <- ape::vcv(tree, cor = T)

---

Meta-analysis

Model selection

Prior to the meta-analysis, we performed model selection on all moderators via the dredge function. Due to the long processing time, we used the saved RDS file when creating the html file.

eval(metafor:::.MuMIn)
options(na.action = "na.fail")  # required for dredge to run

mumin_dat <- raw_dat %>%
    dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_"))
mumin_dat <- mumin_dat[!apply(raw_dat[, c("life_stage", "source_cat", "region", "medium",
    "classification", "trait", "uv_type", "exposure_temp", "ln_dur_day", "ln_delta_dose")],
    1, anyNA), ]  # remove rows with any NA value on one of these columns

# Run multi-level model
mumin_model <- metafor::rma.mv(yi = lnRR, V = vi, mods = ~life_stage + source_cat +
    region + medium + classification + trait + uv_type + exposure_temp + ln_dur_day +
    ln_delta_dose, random = list(~1 | es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL),
    R = list(species_OTL = phylo_cor), method = "ML", test = "t", dfs = "contain",
    data = mumin_dat)

# candidate_models <- MuMIn::dredge(mumin_model) # generate all possible
# combinations of moderators saveRDS(candidate_models, file =
# 'candidate_models.rds')
candidate_models <- readRDS("/Users/nicholaswu/Library/CloudStorage/OneDrive-WesternSydneyUniversity/UV meta-analysis/candidate_models.rds")

options(na.action = "na.omit")  # set back to default
subset(candidate_models, delta <= 2)  # display all models within 2 values of AICc

## Global model call: metafor::rma.mv(yi = lnRR, V = vi, mods = ~life_stage + source_cat + 
##     region + medium + classification + trait + uv_type + exposure_temp + 
##     ln_dur_day + ln_delta_dose, random = list(~1 | es_ID, ~1 | 
##     study_ID, ~1 | species, ~1 | species_OTL), data = mumin_dat, 
##     method = "ML", test = "t", dfs = "contain", R = list(species_OTL = phylo_cor))
## ---
## Model selection table 
##     (Int) cls exp_tmp lif_stg ln_dlt_dos ln_dur_day trt uv_typ df logLik AICc
## 781     +                   +      -0.02              +      + 20   -440  922
## 773     +                   +                         +      + 19   -441  922
## 789     +                   +                 0.021   +      + 20   -441  922
## 797     +                   +      -0.02      0.020   +      + 21   -440  922
## 775     +      0.0072       +                         +      + 20   -441  923
## 783     +      0.0068       +      -0.02              +      + 21   -440  923
## 774     +   +               +                         +      + 20   -441  924
## 782     +   +               +      -0.02              +      + 21   -440  924
##     delta weight
## 781  0.00  0.197
## 773  0.03  0.193
## 789  0.74  0.136
## 797  0.82  0.131
## 775  1.39  0.098
## 783  1.43  0.096
## 774  1.94  0.074
## 782  1.96  0.074
## Models ranked by AICc(x)

MuMIn::sw(MuMIn::model.avg(candidate_models, subset = delta <= 2))  # display weight of each moderators

##                      life_stage trait uv_type ln_delta_dose ln_dur_day
## Sum of weights:      1.00       1.00  1.00    0.50          0.27      
## N containing models:    8          8     8       4             2      
##                      exposure_temp classification
## Sum of weights:      0.19          0.15          
## N containing models:    2             2

Meta-analysis

Conduct phylogenically controlled, multi-level meta-analysis. The global_model tested the overall effect of UV exposure on effect size, and tested for publication bias. The overall_model (detailed in the ‘Meta-regression’ section) tested the overall effect with all relevant moderators based on the model section process described above. All models contained the following random effects: Effect size ID, study ID, species ID, study ID, and phylogeny.

The global multi-level, meta-analytic model was fitted as follows:

\[ \begin{equation} Y_i = \beta_0 + \beta_1\sqrt{\frac{1}{\tilde{n_{i}}}} + \beta_2c(\text{year}_j) + \alpha_{k[i]} + sp_{k[i]} + s_{j[i]} + e_i + m_i, \\ \alpha_{k[i]} \sim \text{Normal}(0, \sigma^2_{\text{phylogeny}}\textbf{A}), \\ sp_{k[i]} \sim \text{Normal}(0, \sigma^2_{\text{species}}), \\ s_{k[i]} \sim \text{Normal}(0, \sigma^2_{\text{study}}), \\ e_i \sim \text{Normal}(0, \sigma^2_{\text{residual}}), \\ m_i \sim \text{Normal}(0, v_i), \tag{7} \end{equation} \]

where \(Y_i\) is the \(i\)th effect size, \(\beta_0\) is the weighted overall meta-analytic mean, \(\beta_1\sqrt{\frac{1}{\tilde{n_{i}}}}\) is the inverse of effective sample size, \(\beta_2c(\text{year}_j)\) is the mean centring of publication year, and \(\alpha_{k[i]}\) is the phylogenetic effect (a random effect) for species \(k\) applied to effect size \(i\) (Cinar et al., 2022). Phylogenetic effects are assumed to be normally distributed with a mean of zero and variance \(\sigma^2_{\text{phylogeny}}\) [i.e. \((0, \sigma^2_{\text{phylogeny}}\textbf{A})\)], where \(\textbf{A}\) is a phylogenetic correlation matrix derived from a phylogenetic tree. \(sp_{k[i]}\) is the species-specific effect for the \(k\)th species applied to effect size \(i\), \(s_{k[i]}\) is the study-specifc effect for the \(j\)th study applied to effect size \(i\), \(e_i\) is the effect size-spefic effect (within-study effect, or residuals) for the \(i\)th effect size, and \(m_i\) is the sampling effect for the \(i\)th effect size, resulting from varying precision for each effect size, \(v_i\) (which is known as the sampling variance for the effect).

Meta-regression

Conduct phylogenically controlled, multi-level meta-regression. The overall_model follows the same formula (7) but applies multiple moderator variables (\(\beta\)):

\[ \begin{equation} Y_i = \beta_0 + \displaystyle\sum_{l = 1}^{p} \beta_l x_{l[i]} + \alpha_{k[i]} + sp_{k[i]} + s_{j[i]} + e_i + m_i, \tag{8} \end{equation} \]

where \(\displaystyle\sum_{l = 1}^{p} \beta_l x_{l[i]}\) is the sum of all effects for all moderators \(x_i\) (fixed effects; \(p\) = number of slopes). All other notations are the same as described above (7).

global_model <- metafor::rma.mv(yi = lnRR, V = M, mods = ~1 + sqrt_inv_eff + year_centre,
    random = list(~1 | es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL), R = list(species_OTL = phylo_cor),
    method = "REML", test = "t", dfs = "contain", data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

# Moderators based on best model
overall_model <- metafor::rma.mv(yi = lnRR, V = M, mods = ~1 + life_stage + trait +
    uv_type + ln_delta_dose, random = list(~1 | es_ID, ~1 | study_ID, ~1 | species,
    ~1 | species_OTL), R = list(species_OTL = phylo_cor), method = "REML", test = "t",
    dfs = "contain", data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

global_r2 <- orchaRd::r2_ml(global_model) * 100
overall_r2 <- orchaRd::r2_ml(overall_model) * 100

Heterogeneity analysis

Heterogeneity is a persistent problem in ecological and evolutionary meta-analysis (Borenstein, 2019; Noble et al., 2022; Senior et al., 2016). Here, heterogeneity is reported as \(I^2_\text{total}\) and is calculated as:

\[ \begin{equation} I^2_\text{total} = \frac{\sigma^2_\text{phylogeny} + \sigma^2_\text{species} + \sigma^2_\text{study} + \sigma^2_\text{residual}}{\sigma^2_\text{total}}, \tag{9} \end{equation} \]

where \(\sigma^2_\text{total} = \sigma^2_\text{phylogeny} + \sigma^2_\text{species} + \sigma^2_\text{study} + \sigma^2_\text{residual} + \sigma^2_\text{sampling}\) is the total effect size variance and \(\sigma^2_\text{sampling}\) is the sampling error variance calculated as:

\[ \begin{equation} \sigma_{\text{sampling}}^2 = \sum w_{i}\left( k-1\right) / \left[ \left( \sum w_{i}\right)^2 + \sum w_{i}^2\right] \ \tag{10} \end{equation} \]

where \(k\) is the number of studies and the weights, \(w_i = 1 / v_i\), can be calculated using the inverse of the sampling variance (\(v_i\)) for each effect size \(i\). Lastly, we also calculated how much variance is explained by the model as \(R^2\) which can be separated by \(R^2_{\text{marginal}}\) (fixed effects only) and \(R^2_{\text{conditional}}\) (fixed and random effects):

\[ \begin{equation} R^2_{\text{marginal}} = \frac{\sigma^2_\text{fixed}}{\sigma^2_\text{fixed} + \sigma^2_\text{phylogeny} + \sigma^2_\text{species} + \sigma^2_\text{study} + \sigma^2_\text{residual}}, \\ R^2_{\text{conditional}} = \frac{\sigma^2_\text{fixed} + \displaystyle\sum_{r = 1}^{u} \sigma^2_r}{\sigma^2_\text{fixed} + \sigma^2_\text{phylogeny} + \sigma^2_\text{species} + \sigma^2_\text{study} + \sigma^2_\text{residual}}, \tag{11} \end{equation} \]

where \(\sigma^2_\text{fixed}\) is the variance fo the fixed effects expressed as \(var(\displaystyle\sum_{l = 1}^{p} \beta_l x_{l[i]})\) (Nakagawa and Schielzeth, 2007). The difference between \(R^2_{\text{marginal}}\) and \(R^2_{\text{conditional}}\) is whether the random effect variance is included in the numerator. This is expressed as \(\displaystyle\sum_{r = 1}^{u} \sigma^2_r\), where \(u\) is the number of random factors, and \(\sigma^2_r\) is the variance component of the \(r\)th random factor.

Model output

Table S2 - Model average

Table S2. Summary output of the model averaging of seven candidate models with AIC delta <= 2.The model average was presented as either the full average, which assumes that a moderator is included in every model, or the conditional average, which averages over the models where the moderators appears.

## 
## Call:
## model.avg(object = candidate_models, subset = delta <= 2)
## 
## Component model call: 
## metafor::rma.mv(yi = lnRR, V = vi, mods = ~<8 unique rhs>, random = 
##      list(~1 | es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL), data = 
##      mumin_dat, method = ML, test = t, dfs = contain, R = list(species_OTL = 
##      phylo_cor))
## 
## Component models: 
##       df logLik AICc delta weight
## 3467  20   -440  922  0.00   0.20
## 367   19   -441  922  0.03   0.19
## 3567  20   -441  922  0.74   0.14
## 34567 21   -440  923  0.82   0.13
## 2367  20   -441  923  1.39   0.10
## 23467 21   -440  923  1.43   0.10
## 1367  20   -441  924  1.94   0.07
## 13467 21   -440  924  1.96   0.07
## 
## Term codes: 
## classification  exposure_temp     life_stage  ln_delta_dose     ln_dur_day 
##              1              2              3              4              5 
##          trait        uv_type 
##              6              7 
## 
## Model-averaged coefficients:  
## (full average) 
##                                        Estimate Std. Error z value Pr(>|z|)  
## intrcpt                                -0.02704    0.38522    0.07     0.94  
## life_stageLarva                        -0.27644    0.17975    1.54     0.12  
## life_stageJuvenile                     -0.00542    0.19022    0.03     0.98  
## life_stageAdult                        -0.30038    0.20757    1.45     0.15  
## ln_delta_dose                          -0.01005    0.01406    0.71     0.47  
## traitCardiorespiratory Function         0.07111    0.26142    0.27     0.79  
## traitDNA Damage                         0.02602    0.25684    0.10     0.92  
## traitEnergy Metabolism                  0.35124    0.25430    1.38     0.17  
## traitGrowth                             0.24152    0.19466    1.24     0.21  
## traitHormonal Stress Response           0.03221    0.26825    0.12     0.90  
## traitLocomotion                         0.21291    0.19571    1.09     0.28  
## traitNeuromusclular                     0.46248    0.28568    1.62     0.11  
## traitOxidant                           -0.26971    0.25435    1.06     0.29  
## traitOxidative Damage                   0.09531    0.24665    0.39     0.70  
## traitOxidative Stress Response/Defense  0.40133    0.24487    1.64     0.10 .
## uv_typeUVB                             -0.17372    0.08209    2.12     0.03 *
## ln_dur_day                              0.00559    0.01312    0.43     0.67  
## exposure_temp                           0.00137    0.00456    0.30     0.76  
## classificationTissue/Organ System      -0.01868    0.12452    0.15     0.88  
## classificationWhole Animal Performance -0.07202    0.18771    0.38     0.70  
##  
## (conditional average) 
##                                        Estimate Std. Error z value Pr(>|z|)   
## intrcpt                                -0.02704    0.38522    0.07     0.94   
## life_stageLarva                        -0.27644    0.17975    1.54     0.12   
## life_stageJuvenile                     -0.00542    0.19022    0.03     0.98   
## life_stageAdult                        -0.30038    0.20757    1.45     0.15   
## ln_delta_dose                          -0.02019    0.01388    1.45     0.15   
## traitCardiorespiratory Function         0.07111    0.26142    0.27     0.79   
## traitDNA Damage                         0.02602    0.25684    0.10     0.92   
## traitEnergy Metabolism                  0.35124    0.25430    1.38     0.17   
## traitGrowth                             0.24152    0.19466    1.24     0.21   
## traitHormonal Stress Response           0.03221    0.26825    0.12     0.90   
## traitLocomotion                         0.21291    0.19571    1.09     0.28   
## traitNeuromusclular                     0.46248    0.28568    1.62     0.11   
## traitOxidant                           -0.26971    0.25435    1.06     0.29   
## traitOxidative Damage                   0.09531    0.24665    0.39     0.70   
## traitOxidative Stress Response/Defense  0.47128    0.19351    2.44     0.01 **
## uv_typeUVB                             -0.17372    0.08209    2.12     0.03 * 
## ln_dur_day                              0.02093    0.01800    1.16     0.25   
## exposure_temp                           0.00703    0.00819    0.86     0.39   
## classificationTissue/Organ System      -0.12589    0.30160    0.42     0.68   
## classificationWhole Animal Performance -0.48520    0.19211    2.53     0.01 **
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Table S3 - Global model

Table S3. Summary output of the global_model. intrcpt = overall effect, sqrt_inv_eff = square root of the inverse of effective sample size (small sample effect), year_centre = mean centring of the publication year (decline effect). Heterogeneity is also presented, where I2_Total = total I², I2_es_ID, within study effect I², I2_study_ID = between study effects I², I2_species = species I², and 2_species_OTL = phylogeny I². Publication bias as fixed effects (\(R^2_{\text{marginal}}\)) explained 0.61% of the variation, while the random effects (\(R^2_{\text{conditional}}\)) explained 100% of the variation.

## 
## Multivariate Meta-Analysis Model (k = 895; method: REML)
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed       factor    R 
## sigma^2.1  0.1407  0.3751    895     no        es_ID   no 
## sigma^2.2  0.1586  0.3983     73     no     study_ID   no 
## sigma^2.3  0.0000  0.0000     48     no      species   no 
## sigma^2.4  0.0000  0.0000     47     no  species_OTL  yes 
## 
## Test of Moderators (coefficients 2:3):
## F(df1 = 2, df2 = 70) = 0.7325, p-val = 0.4844
## 
## Model Results:
## 
##               estimate      se     tval   df    pval    ci.lb    ci.ub     
## intrcpt        -0.3051  0.1058  -2.8823   44  0.0061  -0.5184  -0.0918  ** 
## sqrt_inv_eff    0.1695  0.1529   1.1084  892  0.2680  -0.1306   0.4696     
## year_centre    -0.0030  0.0061  -0.4986   70  0.6196  -0.0151   0.0091     
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

##       I2_Total       I2_es_ID    I2_study_ID     I2_species I2_species_OTL 
##            100             47             53              0              0

Table S4 - Overall model

Table S4. Summary output of the overall_model. intrcpt = overall effect. Moderator variables include life stage (embryo, larva, juvenile, adult), trai (see Table S1), UV type (UVA, UVB), exposure temperature, and the natural logarithm of daily dosage (J day). Heterogeneity is also presented, where I2_Total = total I², I2_es_ID, within study effect I², I2_study_ID = between study effects I², I2_species = species I², and 2_species_OTL = phylogeny I². Fixed effects (\(R^2_{\text{marginal}}\)) explained 15.84% of the variation, while the random effects (\(R^2_{\text{conditional}}\)) explained 100% of the variation.

## 
## Multivariate Meta-Analysis Model (k = 808; method: REML)
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed       factor    R 
## sigma^2.1  0.1156  0.3400    808     no        es_ID   no 
## sigma^2.2  0.1719  0.4147     62     no     study_ID   no 
## sigma^2.3  0.0000  0.0000     42     no      species   no 
## sigma^2.4  0.0000  0.0000     41     no  species_OTL  yes 
## 
## Test of Moderators (coefficients 2:16):
## F(df1 = 15, df2 = 46) = 6.8930, p-val < .0001
## 
## Model Results:
## 
##                                         estimate      se     tval   df    pval 
## intrcpt                                   0.1316  0.2860   0.4602   25  0.6494 
## life_stageLarva                          -0.2954  0.1603  -1.8424  792  0.0658 
## life_stageJuvenile                       -0.0243  0.1760  -0.1382  792  0.8901 
## life_stageAdult                          -0.3231  0.1975  -1.6359   46  0.1087 
## traitCardiorespiratory Function           0.0646  0.1683   0.3837  792  0.7013 
## traitDNA Damage                           0.1108  0.1676   0.6614  792  0.5085 
## traitEnergy Metabolism                    0.3478  0.1672   2.0800  792  0.0378 
## traitGrowth                               0.1973  0.1409   1.4009  792  0.1616 
## traitHormonal Stress Response             0.0542  0.1804   0.3005  792  0.7639 
## traitLocomotion                           0.1624  0.1497   1.0848  792  0.2783 
## traitNeuromusclular                       0.4813  0.2050   2.3479  792  0.0191 
## traitOxidant                             -0.2454  0.1642  -1.4951  792  0.1353 
## traitOxidative Damage                     0.1112  0.1524   0.7298  792  0.4657 
## traitOxidative Stress Response/Defense    0.4075  0.1492   2.7307  792  0.0065 
## uv_typeUVB                               -0.2033  0.0796  -2.5556  792  0.0108 
## ln_delta_dose                            -0.0204  0.0137  -1.4926  792  0.1359 
##                                           ci.lb    ci.ub     
## intrcpt                                 -0.4575   0.7207     
## life_stageLarva                         -0.6102   0.0193   . 
## life_stageJuvenile                      -0.3698   0.3211     
## life_stageAdult                         -0.7206   0.0745     
## traitCardiorespiratory Function         -0.2657   0.3948     
## traitDNA Damage                         -0.2181   0.4398     
## traitEnergy Metabolism                   0.0196   0.6759   * 
## traitGrowth                             -0.0792   0.4739     
## traitHormonal Stress Response           -0.2999   0.4083     
## traitLocomotion                         -0.1314   0.4562     
## traitNeuromusclular                      0.0789   0.8836   * 
## traitOxidant                            -0.5677   0.0768     
## traitOxidative Damage                   -0.1880   0.4105     
## traitOxidative Stress Response/Defense   0.1146   0.7004  ** 
## uv_typeUVB                              -0.3595  -0.0471   * 
## ln_delta_dose                           -0.0472   0.0064     
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

##       I2_Total       I2_es_ID    I2_study_ID     I2_species I2_species_OTL 
##            100             40             60              0              0

Fig. S3 - Funnel plot

metafor::funnel(raw_dat$lnRR, raw_dat$vi, yaxis = "seinv", ylab = "Precision (1/SE)",
    xlab = "Effect size (lnRR)")

Fig. S3. Funnel plot with the inverse of the standard error (i.e. precision) as the measure of uncertainty for the effect size (lnRR).

Figures for main text

Figure 1 - Geographical bias

Figure was generated with the coordinates of the wild sourced studies and the mean ultraviolet-B irradiation (UVB) of the highest month was obtained from Beckmann et al. (2014).

## OGR data source with driver: ESRI Shapefile 
## Source: "/Users/nicholaswu/Library/CloudStorage/OneDrive-WesternSydneyUniversity/UV meta-analysis/ne_50m_land/ne_50m_land.shp", layer: "ne_50m_land"
## with 1420 features
## It has 3 fields
## Integer64 fields read as strings:  scalerank

---

Figure 2 - UV type and region

UV type

uv_model <- metafor::rma.mv(yi = lnRR, V = M, mods = ~uv_type - 1, random = list(~1 |
    es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL), R = list(species_OTL = phylo_cor),
    method = "REML", test = "t", dfs = "contain", data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

Table S5. Summary output of the uv_model.

## 
## Multivariate Meta-Analysis Model (k = 895; method: REML)
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed       factor    R 
## sigma^2.1  0.1400  0.3742    895     no        es_ID   no 
## sigma^2.2  0.1574  0.3967     73     no     study_ID   no 
## sigma^2.3  0.0000  0.0000     48     no      species   no 
## sigma^2.4  0.0000  0.0000     47     no  species_OTL  yes 
## 
## Test of Moderators (coefficients 1:2):
## F(df1 = 2, df2 = 893) = 8.6509, p-val = 0.0002
## 
## Model Results:
## 
##             estimate      se     tval   df    pval    ci.lb    ci.ub      
## uv_typeUVA   -0.1033  0.0709  -1.4575  893  0.1453  -0.2425   0.0358      
## uv_typeUVB   -0.2239  0.0552  -4.0559  893  <.0001  -0.3323  -0.1156  *** 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Region

region_model <- metafor::rma.mv(yi = lnRR, V = M, 
                                mods = ~ region-1,
                                random = list(~1 | es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL),
                                R = list(species_OTL = phylo_cor),
                                method ="REML",
                                test = "t", 
                                dfs = "contain",
                                control = list(rel.tol = 1e-8), # model converge 
                                data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

Table S6. Summary output of the region_model.

## 
## Multivariate Meta-Analysis Model (k = 895; method: REML)
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed       factor    R 
## sigma^2.1  0.1404  0.3748    895     no        es_ID   no 
## sigma^2.2  0.1644  0.4054     73     no     study_ID   no 
## sigma^2.3  0.0000  0.0002     48     no      species   no 
## sigma^2.4  0.0000  0.0000     47     no  species_OTL  yes 
## 
## Test of Moderators (coefficients 1:3):
## F(df1 = 3, df2 = 44) = 4.4606, p-val = 0.0081
## 
## Model Results:
## 
##                    estimate      se     tval   df    pval    ci.lb    ci.ub    
## regionTropical      -0.2286  0.0997  -2.2922   44  0.0267  -0.4297  -0.0276  * 
## regionSubtropical   -0.1578  0.0839  -1.8808  892  0.0603  -0.3224   0.0069  . 
## regionTemperate     -0.2062  0.0840  -2.4539  892  0.0143  -0.3710  -0.0413  * 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Figure production

---

Figure 3 - Taxa and life stage

Taxa

taxa_model <- metafor::rma.mv(yi = lnRR, V = M, mods = ~class - 1, random = list(~1 |
    es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL), R = list(species_OTL = phylo_cor),
    method = "REML", test = "t", dfs = "contain", data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

print(taxa_model)

Table S7. Summary output of the taxa_model.

## 
## Multivariate Meta-Analysis Model (k = 895; method: REML)
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed       factor    R 
## sigma^2.1  0.1405  0.3748    895     no        es_ID   no 
## sigma^2.2  0.1620  0.4025     73     no     study_ID   no 
## sigma^2.3  0.0000  0.0000     48     no      species   no 
## sigma^2.4  0.0000  0.0000     47     no  species_OTL  yes 
## 
## Test of Moderators (coefficients 1:4):
## F(df1 = 4, df2 = 43) = 3.8077, p-val = 0.0098
## 
## Model Results:
## 
##                 estimate      se     tval  df    pval    ci.lb    ci.ub    
## classAmphibian   -0.2608  0.0984  -2.6508  43  0.0112  -0.4593  -0.0624  * 
## classBird        -0.4754  0.2959  -1.6064  43  0.1155  -1.0722   0.1214    
## classFish        -0.1646  0.0697  -2.3604  43  0.0229  -0.3051  -0.0240  * 
## classReptile     -0.0465  0.2038  -0.2282  43  0.8206  -0.4576   0.3645    
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Life stage

stage_model <- metafor::rma.mv(yi = lnRR, V = M, mods = ~life_stage - 1, random = list(~1 |
    es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL), R = list(species_OTL = phylo_cor),
    method = "REML", test = "t", dfs = "contain", data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

print(stage_model)

Table S8. Summary output of the stage_model.

## 
## Multivariate Meta-Analysis Model (k = 895; method: REML)
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed       factor    R 
## sigma^2.1  0.1397  0.3737    895     no        es_ID   no 
## sigma^2.2  0.1621  0.4026     73     no     study_ID   no 
## sigma^2.3  0.0000  0.0000     48     no      species   no 
## sigma^2.4  0.0000  0.0000     47     no  species_OTL  yes 
## 
## Test of Moderators (coefficients 1:4):
## F(df1 = 4, df2 = 69) = 4.8917, p-val = 0.0016
## 
## Model Results:
## 
##                     estimate      se     tval   df    pval    ci.lb    ci.ub 
## life_stageEmbryo     -0.0332  0.1226  -0.2708  891  0.7866  -0.2737   0.2073 
## life_stageLarva      -0.2657  0.0712  -3.7306  891  0.0002  -0.4055  -0.1259 
## life_stageJuvenile   -0.1098  0.0825  -1.3303  891  0.1838  -0.2718   0.0522 
## life_stageAdult      -0.2517  0.1159  -2.1723   69  0.0333  -0.4829  -0.0205 
##                         
## life_stageEmbryo        
## life_stageLarva     *** 
## life_stageJuvenile      
## life_stageAdult       * 
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Figure production

---

Figure 4 - Trait

Trait

trait_model <- metafor::rma.mv(yi = lnRR, V = M, mods = ~trait - 1, random = list(~1 |
    es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL), R = list(species_OTL = phylo_cor),
    method = "REML", test = "t", dfs = "contain", data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

Table S9. Summary output of the trait_model.

## 
## Multivariate Meta-Analysis Model (k = 895; method: REML)
## 
## Variance Components:
## 
##             estim    sqrt  nlvls  fixed       factor    R 
## sigma^2.1  0.1250  0.3535    895     no        es_ID   no 
## sigma^2.2  0.1241  0.3523     73     no     study_ID   no 
## sigma^2.3  0.0000  0.0000     48     no      species   no 
## sigma^2.4  0.0000  0.0000     47     no  species_OTL  yes 
## 
## Test of Moderators (coefficients 1:11):
## F(df1 = 11, df2 = 884) = 12.9307, p-val < .0001
## 
## Model Results:
## 
##                                         estimate      se     tval   df    pval 
## traitBehaviour                           -0.5313  0.1208  -4.3974  884  <.0001 
## traitCardiorespiratory Function          -0.2537  0.0884  -2.8718  884  0.0042 
## traitDNA Damage                          -0.5110  0.0822  -6.2196  884  <.0001 
## traitEnergy Metabolism                    0.0615  0.0806   0.7624  884  0.4460 
## traitGrowth                              -0.1741  0.0617  -2.8205  884  0.0049 
## traitHormonal Stress Response            -0.3452  0.1076  -3.2081  884  0.0014 
## traitLocomotion                          -0.2706  0.0770  -3.5153  884  0.0005 
## traitNeuromusclular                       0.0131  0.1639   0.0801  884  0.9362 
## traitOxidant                             -0.5796  0.0958  -6.0528  884  <.0001 
## traitOxidative Damage                    -0.2440  0.0705  -3.4616  884  0.0006 
## traitOxidative Stress Response/Defense   -0.0041  0.0654  -0.0634  884  0.9495 
##                                           ci.lb    ci.ub      
## traitBehaviour                          -0.7684  -0.2942  *** 
## traitCardiorespiratory Function         -0.4272  -0.0803   ** 
## traitDNA Damage                         -0.6723  -0.3498  *** 
## traitEnergy Metabolism                  -0.0968   0.2197      
## traitGrowth                             -0.2952  -0.0530   ** 
## traitHormonal Stress Response           -0.5564  -0.1340   ** 
## traitLocomotion                         -0.4216  -0.1195  *** 
## traitNeuromusclular                     -0.3086   0.3349      
## traitOxidant                            -0.7676  -0.3917  *** 
## traitOxidative Damage                   -0.3824  -0.1057  *** 
## traitOxidative Stress Response/Defense  -0.1325   0.1242      
## 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Figure production

---

Supplementary figure

# Taxonomic group and UV interaction
taxa_uv_model <- metafor::rma.mv(yi = lnRR, V = M, 
                                mods = ~ class + uv_type-1,
                                random = list(~1 | es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL),
                                R = list(species_OTL = phylo_cor),
                                method ="REML",
                                test = "t", 
                                dfs = "contain",
                                data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

taxa_uv_model2 <- orchaRd::mod_results(taxa_uv_model, group = "study_ID", mod = "class", by = "uv_type", weights = "prop", data = raw_dat)

taxa_uv_plot <- orchaRd::orchard_plot(taxa_uv_model2, xlab = "", g = FALSE,
                                      branch.size = 1.5, trunk.size = 4, condition.lab = "UV type") + 
  geom_point() + # mean estimate
  scale_shape_manual(values = c(21, 24)) +
  scale_colour_manual(values = c("#e6e58e", "#EDA200", "#D24E71", "#a733a4")) + 
  scale_fill_manual(values = c("#e6e58e", "#EDA200", "#D24E71", "#a733a4")) + 
  ylab("Effect size (lnRR)") +
  ggtitle("Taxonomic group") +
  mytheme() + theme(legend.position = "bottom")

# Life stage and UV interaction
stage_uv_model <- metafor::rma.mv(yi = lnRR, V = M, 
                                mods = ~ life_stage + uv_type-1,
                                random = list(~1 | es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL),
                                R = list(species_OTL = phylo_cor),
                                method ="REML",
                                test = "t", 
                                dfs = "contain",
                                data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

stage_uv_model2 <- orchaRd::mod_results(stage_uv_model, group = "study_ID", mod = "life_stage", by = "uv_type", weights = "prop", data = raw_dat)

stage_uv_plot <- orchaRd::orchard_plot(stage_uv_model2, xlab = "lnRR", angle = 45, g = FALSE, 
                                       branch.size = 1.5, trunk.size = 4, condition.lab = "UV type") + 
  geom_point() + # mean estimate
  scale_shape_manual(values = c(21, 24)) +
  scale_colour_manual(values = c("#EDEF5C", "#82CC6C", "#3CB08F", "#007E7D")) + 
  scale_fill_manual(values = c("#EDEF5C", "#82CC6C", "#3CB08F", "#007E7D")) + 
  ylab("Effect size (lnRR)") +
  ggtitle("Life stage") +
  mytheme() + theme(legend.position = "bottom")

prow <- cowplot::plot_grid(taxa_uv_plot + theme(legend.position = "none"), 
                   stage_uv_plot + theme(legend.position = "none"), 
                   labels = c("a", "b"), align = "hv")

legend <- cowplot::get_legend(taxa_uv_plot + theme(legend.position = "bottom"))

cowplot::plot_grid(prow, legend, ncol = 1, rel_heights = c(1, .1))

Fig. S4. Difference in UVA and UVB radiation effects by (a) taxonomic groups and (b) life stage.

trait_uv_model <- metafor::rma.mv(yi = lnRR, V = M, 
                                mods = ~ trait + uv_type-1,
                                random = list(~1 | es_ID, ~1 | study_ID, ~1 | species, ~1 | species_OTL),
                                R = list(species_OTL = phylo_cor),
                                method ="REML",
                                test = "t", 
                                dfs = "contain",
                                data = raw_dat %>%
        dplyr::mutate(species_OTL = stringr::str_replace_all(species_OTL, " ", "_")))

trait_uv_model2 <- orchaRd::mod_results(trait_uv_model, group = "study_ID", mod = "trait", by = "uv_type", weights = "prop", data = raw_dat)

orchaRd::orchard_plot(trait_uv_model2, xlab = "lnRR", angle = 45, g = FALSE, 
                      branch.size = 1.5, trunk.size = 4, condition.lab = "UV type") + 
  geom_point() + # mean estimate
  scale_shape_manual(values = c(21, 24)) +
  scale_colour_manual(values = c("#8f6798", "#7f669f", "#7a7cab", "#7290af", "#6aa1b0", "#64b1af", "#4eb99d", "#56c57f", "#87d662", "#BBDF27", "#FDE725")) + 
  scale_fill_manual(values = c("#8f6798", "#7f669f", "#7a7cab", "#7290af", "#6aa1b0", "#64b1af", "#4eb99d", "#56c57f", "#87d662", "#BBDF27", "#FDE725")) + 
  ylab("Effect size (lnRR)") +
  ggtitle("Traits") +
  mytheme() + theme(legend.position = "bottom")

Fig. S5. Effect of UVA and UVB radiation on 11 biological traits.

ggplot(raw_dat) + geom_hline(yintercept = 0, linetype = "dashed", alpha = 0.5) +
    geom_point(aes(x = metric, y = lnRR, colour = trait), show.legend = FALSE) +
    viridis::scale_colour_viridis(discrete = TRUE) + xlab(NULL) + ylab("Effect size (lnRR") +
    coord_flip() + facet_wrap(~trait, ncol = 3, scales = "free_y") + mytheme() +
    theme(axis.text = element_text(size = 5), strip.text.x = element_text(size = 6),
        legend.position = "bottom")

Fig. S6. Effect of ultraviolet radiation on specific responses across different biological and functional traits. Data presented as individual effect sizes (lnRR).

---

References

Beckmann, M., Václavík, T., Manceur, A. M., Šprtová, L., Von Wehrden, H., Welk, E. and F, C. A. (2014). glUV: A global UV-B radiation data set for macroecological studies. Methods in Ecology and Evolution 5, 372–383.

Borenstein, M. (2019). Heterogeneity in meta-analysis. In The handbook of research synthesis and meta-analysis (ed. Borenstein, M.), M, C. H.), Hedges, L. V.), and Valentine, J. C.), New York, US: Russell Sage Foundation.

Cinar, O., Nakagawa, Shinichi and Viechtbauer, W. (2022). Phylogenetic multilevel meta‐analysis: A simulation study on the importance of modelling the phylogeny. Methods in Ecology and Evolution 13, 383–395.

Hedges, L. V. and Olkin, I. (2014). Statistical methods for meta-analysis. London, UK: Academic Press Inc.

Nakagawa, S. and Schielzeth, H. (2007). A general and simple method for obtaining R\(^{2}\) from generalized linear mixed-effects models. Methods in Ecology and Evolution 4, 133–142.

Nakagawa, S., Lagisz, M., Jennions, M., Koricheva, J., Noble, D. W., Parker, T. H., Sánchez-Tójar, A., Yang, Y. and O’dea, R. E. (2022). Methods for testing publication bias in ecological and evolutionary meta-analyses. Methods in Ecology and Evolution 13, 4–21.

Noble, D. W., Pottier, P., Lagisz, M., Burke, S., Drobniak, S. M., O’dea, R. E. and Nakagawa, S. (2022). Meta-analytic approaches and effect sizes to account for “nuisance heterogeneity” in comparative physiology. Journal of Experimental Biology 225, jeb243225.

Senior, A. M., Grueber, C. E., Kamiya, T., Lagisz, O., Malgorzata, A, S. E. S. and Nakagawa, S. (2016). Heterogeneity in ecological and evolutionary meta-analyses: Its magnitude and implications. Ecology 97, 3293–3299.

---

Session Information

R version 4.2.2 (2022-10-31)

Platform: aarch64-apple-darwin20 (64-bit)

attached base packages: stats, graphics, grDevices, utils, datasets, methods and base

other attached packages: ape(v.5.7-1), rotl(v.3.0.14), rgdal(v.1.6-7), raster(v.3.6-20), sp(v.1.6-1), MuMIn(v.1.47.5), data.table(v.1.14.8), DataExplorer(v.0.8.2), cowplot(v.1.1.1), pander(v.0.6.5), orchaRd(v.2.0), lubridate(v.1.9.2), forcats(v.1.0.0), stringr(v.1.5.0), dplyr(v.1.1.2), purrr(v.1.0.1), readr(v.2.1.4), tidyr(v.1.3.0), tibble(v.3.2.1), ggplot2(v.3.4.2), tidyverse(v.2.0.0), flextable(v.0.9.1), metafor(v.4.2-0), numDeriv(v.2016.8-1.1), metadat(v.1.2-0), Matrix(v.1.5-4.1), metaAidR(v.0.0.0.9000) and knitr(v.1.43)

loaded via a namespace (and not attached): ggbeeswarm(v.0.7.2), colorspace(v.2.1-0), ellipsis(v.0.3.2), estimability(v.1.4.1), httpcode(v.0.3.0), rstudioapi(v.0.14), farver(v.2.1.1), mvtnorm(v.1.2-2), fansi(v.1.0.4), mathjaxr(v.1.6-0), xml2(v.1.3.4), codetools(v.0.2-19), cachem(v.1.0.8), jsonlite(v.1.8.5), latex2exp(v.0.9.6), shiny(v.1.7.4), rentrez(v.1.2.3), compiler(v.4.2.2), httr(v.1.4.6), emmeans(v.1.8.6), fastmap(v.1.1.1), cli(v.3.6.1), formatR(v.1.14), later(v.1.3.1), prettyunits(v.1.1.1), htmltools(v.0.5.5), tools(v.4.2.2), igraph(v.1.4.3), coda(v.0.19-4), gtable(v.0.3.3), glue(v.1.6.2), Rcpp(v.1.0.10), jquerylib(v.0.1.4), fontquiver(v.0.2.1), vctrs(v.0.6.2), crul(v.1.4.0), nlme(v.3.1-162), xfun(v.0.39), networkD3(v.0.4), timechange(v.0.2.0), mime(v.0.12), lifecycle(v.1.0.3), pacman(v.0.5.1), XML(v.3.99-0.14), terra(v.1.7-29), scales(v.1.2.1), ragg(v.1.2.5), hms(v.1.1.3), promises(v.1.2.0.1), parallel(v.4.2.2), fontLiberation(v.0.1.0), yaml(v.2.3.7), curl(v.5.0.1), gridExtra(v.2.3), gdtools(v.0.3.3), sass(v.0.4.6), stringi(v.1.7.12), fontBitstreamVera(v.0.1.1), highr(v.0.10), zip(v.2.3.0), rlang(v.1.1.1), pkgconfig(v.2.0.3), systemfonts(v.1.0.4), rncl(v.0.8.7), evaluate(v.0.21), lattice(v.0.21-8), labeling(v.0.4.2), htmlwidgets(v.1.6.2), tidyselect(v.1.2.0), magrittr(v.2.0.3), bookdown(v.0.34), R6(v.2.5.1), generics(v.0.1.3), pillar(v.1.9.0), withr(v.2.5.0), crayon(v.1.5.2), gfonts(v.0.2.0), uuid(v.1.1-0), utf8(v.1.2.3), tzdb(v.0.4.0), rmarkdown(v.2.22), officer(v.0.6.2), viridis(v.0.6.3), progress(v.1.2.2), grid(v.4.2.2), digest(v.0.6.31), xtable(v.1.8-4), httpuv(v.1.6.11), textshaping(v.0.3.6), openssl(v.2.0.6), stats4(v.4.2.2), munsell(v.0.5.0), viridisLite(v.0.4.2), beeswarm(v.0.4.0), vipor(v.0.4.5), bslib(v.0.4.2) and askpass(v.1.1)

---

1. Hawkesbury Institute for the Environment, Western Sydney University, NSW 2753, Australia, nicholas.wu.nz@gmail.com↩︎