######## Ren et al. 2024 
######## Effects of manipulated rainfall and intraspecific variation within dominant species on community assembly: insights from a long-term grassland restoration experiment

# Loading necessary packages
library(car)
library(cati)
library(dplyr)
library(emmeans)
library(ggbiplot)
library(ggplot2)
library(ggpubr)
library(hypervolume)
library(lattice)
library(lme4)
library(lmerTest)
library(lsmeans)
library(MASS)
library(multcomp)
library(multcompView)
library(paran)
library(rasterVis)
library(vegan)

######## Data imput
eco1=read.csv ("TVD_env.csv", header=T,row.names=1)
toMatch <- c('DRY','MESIC','WET') 
eco1 <- filter(eco1, grepl(paste(toMatch, collapse="|"), eco1$ecotype))
toMatch <- c('2012','2014','2018','2019') # year which have rainout treatment
eco <- filter(eco1, grepl(paste(toMatch, collapse="|"), eco1$stage))
eco[,c("ecotype","block","stage","treatment","plot")] <- 
  lapply(eco[,c("ecotype","block","stage","treatment","plot")], factor)
ecotype <- eco$ecotype
block <- eco$block
stage <- eco$stage 
treatment <- eco$treatment
plot <- eco$plot
Richness <- eco$Richness
PDsesmpd <- eco$PDsesmpd
FDsesmpd <- eco$FDsesmpd

######## Hypothesis 1: GLMM & LMM modeling
# taxonomic diversity
richs<-glmer(Richness ~ ecotype*stage*treatment + (1|plot:block),  
             data=eco,family = poisson(link = "log"),
             control=glmerControl(optimizer = "bobyqa",
                                  optCtrl = list(maxfun = 2e5)))
Anova(richs)
marginal = lsmeans(richs,  ~ ecotype:stage) # multiple comparison
CLD.TD = cld(marginal,
          alpha=0.05,
          reversed = T, 
          type = "response", 
          Letters=letters,
          adjust="tukey")
CLD.TD

# phylogenetic diversity
pds<-lmer(PDsesmpd~  ecotype*stage*treatment  + (1|plot:block), 
          data=eco, REML=T, control=lmerControl(optimizer = "bobyqa", 
                                                optCtrl = list(maxfun = 1e5)))
Anova(pds,test.statistic = 'F')
marginal = lsmeans(pds,  ~ ecotype:stage)
CLD.PD = cld(marginal,
          alpha=0.05,
          reversed = T,
          type = "response", 
          Letters=letters,
          adjust="tukey")
CLD.PD

#FD
fds<-lmer(FDsesmpd~ ecotype*stage*treatment  + (1|plot:block), 
          data=eco, REML=T, control=lmerControl(optimizer = "bobyqa", 
                                                optCtrl = list(maxfun = 1e5)))
Anova(fds,test.statistic = 'F')
marginal = lsmeans(fds,  ~ ecotype:stage)
CLD.FD = cld(marginal,
          alpha=0.05,
          reversed = T,
          type = "response", 
          Letters=letters,
          adjust="tukey")
CLD.FD

# Graphs on ecotype X year interaction
erico <- eco %>% # richness summary
  group_by(stage, ecotype) %>%
  summarise(
    richness = mean(Richness),
    se1 = sd(Richness) / sqrt(n())
  )
erico2 <- eco %>% # PD summary
  group_by(stage, ecotype) %>%
  summarise(
    pdsesmpd = mean(PDsesmpd),
    se2 = sd(PDsesmpd) / sqrt(n())
  )
erico3 <- eco %>% # FD summary
  group_by(stage, ecotype) %>%
  summarise(
    fdsesmpd = mean(FDsesmpd),
    se3 = sd(FDsesmpd) / sqrt(n())
  )
# richness graph
p <- ggplot(erico, aes(stage, richness, color = ecotype, shape = ecotype)) +
  geom_errorbar(
    aes(ymin = richness-se1, ymax = richness+se1),
    position = position_dodge(0.3), width = 0.3)+ labs(y="Richness") +
  geom_point(aes(size=I(3)), position = position_dodge(0.3)) + theme_classic() +
  geom_line(aes(x = stage, y = richness, color = ecotype,group = ecotype),
            position = position_dodge(0.3),linetype = "dashed", alpha = 0.6) +
  scale_colour_manual(values = c("red4","darkgreen","navy"), name = "ecotype", 
                        labels = c("DRY","MESIC", "WET")) +
  theme(legend.position = "right",axis.title.x=element_blank(),
        axis.text.x=element_blank(),axis.ticks.x=element_blank(),
        text = element_text(size = 13)) + 
  xlab("Year") + ylab("Richness")
# PD graph
q <- ggplot(erico2, aes(stage, pdsesmpd, color = ecotype, shape = ecotype)) +
  geom_errorbar(
    aes(ymin = pdsesmpd-se2, ymax = pdsesmpd+se2),
    position = position_dodge(0.3), width = 0.3) +
  geom_point(aes(size=I(3)), position = position_dodge(0.3)) + theme_classic() +
  geom_line(aes(x = stage, y = pdsesmpd, color = ecotype,group = ecotype),
            position = position_dodge(0.3),linetype = "dashed", alpha = 0.6) +
  geom_hline(yintercept=0, linetype="dotted", color = "grey",size=0.8, alpha = 0.5) +
  scale_colour_manual(values = c("red4","darkgreen","navy"), name = "ecotype", 
                        labels = c("DRY","MESIC", "WET")) +
  theme(legend.position = "none",axis.title.x=element_blank(),
        axis.text.x=element_blank(),axis.ticks.x=element_blank(),
        text = element_text(size = 13)) + xlab("Year") + ylab(bquote(PD[sesmpd]))
# FD graph
r <- ggplot(erico3, aes(stage, fdsesmpd,color = ecotype, shape = ecotype)) +
  geom_errorbar(
    aes(ymin = fdsesmpd-se3, ymax = fdsesmpd+se3),
    position = position_dodge(0.3), width = 0.3)+ 
  geom_point(aes(size=I(3)), position = position_dodge(0.3)) + theme_classic() +
  geom_line(aes(x = stage, y = fdsesmpd, color = ecotype,group = ecotype),
            position = position_dodge(0.3),linetype = "dashed", alpha = 0.6) +
  geom_hline(yintercept=0, linetype="dotted", color = "grey",size=0.8, alpha = 0.5) +
  scale_colour_manual(values = c("red4","darkgreen","navy"), name = "ecotype", 
                        labels = c("DRY","MESIC", "WET")) +
  theme(text = element_text(size = 13), legend.text = element_text(size=13),
        legend.direction = "horizontal") +
  xlab("Year") + ylab(bquote(FD[sesmpd])) +
  guides(color = guide_legend(nrow = 1)) 
# graph assembly
ggarrange(p,q,r,
          labels = c("A", "B", "C"), label.x = 0.02,
          ncol = 1, nrow = 3, font.label = list(size = 13),
          common.legend = T, legend = "bottom", align = "v",
          legend.grob = get_legend(r))

######## Hypothesis 2: the ITV's contribution to community trait variance

# loading dataset & trait test figure
cati=read.csv ("cati.csv", header=T,row.names=1)
cati[,1:5] <- lapply(cati[,1:5], log)
cati[,c("ecotype","spp")] <- 
  lapply(cati[,c("ecotype","spp")], factor)
tr.cati <- cati[,1:5]
ind.cati <- cati[,6]
spp.cati <- cati[,7]
res <- Tstats(tr.cati, ind.cati, spp.cati, nperm = 999, independantTraits = T)
plot(res,cex=0.001)

# PCA
# parallel analysis to retain no. of axis
library(paran)
paran(tr.cati,cfa=F)
sppcode <- cati[,8]
domsppcode <- cati[,10]
ecot.pca <- prcomp(tr.cati, scale. = TRUE)
ggbiplot(ecot.pca, obs.scale = 1, var.scale = 1,
         groups = sppcode, ellipse = TRUE, circle = T, labels = domsppcode, labels.size=4,
         varname.size = 0.001 ) +
  scale_color_discrete(name = '') +
  theme(legend.direction = 'horizontal', legend.position = 'none') +
  theme_classic() +
  theme(legend.position = "none") 
# species symbols should be added mannually in the PCA diagram