###########################################################################################################
#
#      A landscape-scale assessment of tropical mammals reveals the effects of habitat and anthropogenic disturbance on community occupancy
#     
#      N. Cavada, R. Havmøller, N. Sharff, F. Rovero      
#
#      Appendix 1: R and Jags model code
#
###########################################################################################################

#set working directory
setwd("/your/path")

# load data 
 
load("multiregion_occupancy.RData")

#number of regions (i.e. forest blocks)
R<-5
#number of augmented species
M<-100
#number of camera trap sites in each region R
nsite<-c(51,34,26,25,28)
#number of camera trap days
nrep<-nrepmatrix
##total number of detected species in forest block
nspec<-c(29,33,24,25,24)
#total number oof augmented individuals for each forest block
nz<-c(71,67,76,75,76)

#Jags model
                     
cat("
model{ 
# Hyperpriors
  mu.omega ~ dnorm(0,0.01) #occupancy parameter
  mean.omega <- 1/(1+exp(-mu.omega))
  
  
  mu.lpsi~ dnorm(0,0.01) #community mean of occupancy (logit) - intercept
  mu.sd.lpsi ~ dunif(0,10) # variability among species in logit(psi)
  
  beta1 ~ dunif(-10,10) #slope of logit(psi) on logmass 
  
  mu.beta2.global ~ dunif(-10,10) #slope of logit(psi) on distance to nearest river (dRiv)
  tau.beta2.global <- pow(sd.beta2.global,-2)  
  sd.beta2.global ~ dunif(0,4)
  
  mu.beta3.global ~ dunif(-10,10) #slope of logit(psi) on distance to reserve border (dRes)
  tau.beta3.global <- pow(sd.beta3.global,-2)  
  sd.beta3.global ~ dunif(0,4)
  
  mu.beta4.global ~ dunif(-10,10) #slope of logit(psi) on moist forest cover (forCov)
  tau.beta4.global <- pow(sd.beta4.global,-2)  
  sd.beta4.global ~ dunif(0,4)

  mu.lp ~ dnorm(0,0.01)# community mean of detection
  mu.sd.lp ~ dunif(0,10) # variability among species in logit(p)
 
  alpha1 ~ dunif(-5,5) #slope of logit(p) on logmass
  mu.alpha2.global ~ dunif(-10,10) #slope of logit(p) on distance to reserve border (dRes)
  tau.alpha2.global <- pow(sd.alpha2.global,-2)
  sd.alpha2.global ~ dunif(0,4)


 
#priors for species trait - logmass
 
 for(r in 1:R){
         for(k in 1:M){
                 logmass[k,r] ~ dnorm(mu.logmass[r],tau.logmass[r])
         }
 }
 
 for(r in 1:R){
         mu.logmass[r] ~ dnorm(0,0.01)
         tau.logmass[r] <- pow(sigma.logmass[r],-2)
         sigma.logmass[r] ~ dunif(0,10)
 }
 
 

 
 for(r in 1:R){
        lpsi[r]<-mu.lpsi #region lpsi means constrained to overall mean
        sd.lpsi[r] <- mu.sd.lpsi
        tau.lpsi[r] <- pow(sd.lpsi[r],-2)
        
        lp[r] <- mu.lp #region lp means constrained to overall mean
        sd.lp[r] <- mu.sd.lp 
        tau.lp[r] <- pow(sd.lp[r],-2)
        
        
        
 
  for(k in 1:M){
        alpha2[k,r] ~ dnorm(mu.alpha2[r],tau.alpha2[r])
        beta2[k,r]~ dnorm(mu.beta2[r],tau.beta2[r])
        beta3[k,r]~ dnorm(mu.beta3[r],tau.beta3[r])
        beta4[k,r]~ dnorm(mu.beta4[r],tau.beta4[r])
  }

}

 for(r in 1:R){
         mu.alpha2[r] <- mu.alpha2.global
         tau.alpha2[r] <- tau.alpha2.global
         mu.beta2[r] <- mu.beta2.global
         tau.beta2[r] <- tau.beta2.global
         mu.beta3[r] <- mu.beta3.global
         tau.beta3[r] <- tau.beta3.global
         mu.beta4[r] <- mu.beta4.global
         tau.beta4[r] <- tau.beta4.global

}


# state process

 for(r in 1:R){
         for(k in 1:M){
                 w[k,r] ~ dbern(omega[r]) #Metacommunity membership indicator (data augmentation variable)

                
                 beta0[k,r] ~ dnorm(lpsi[r], tau.lpsi[r]) 
                 
                 alpha0[k,r] ~ dnorm(lp[r], tau.lp[r])
                 

#occupancy model
      
         for(i in 1:nsite[r]){
                 logit(psi[k,i,r]) <- beta0[k,r] + beta1 * logmass[k,r] + beta2[k,r] * dRiv[i,r] + beta3[k,r] * dSettl[i,r] + beta4[k,r] * forCov[i,r]
                 mu.psi[k,i,r] <- psi[k,i,r] * w[k,r]
                 z[k,i,r] ~ dbern(mu.psi[k,i,r])

#observation model

                 logit(p[k,i,r]) <- alpha0[k,r] + alpha1 * logmass[k,r] + alpha2[k,r] * dSettl[i,r]
                 mu.p[k,i,r] <- p[k,i,r]*z[k,i,r]
                 yaug[k,i,r] ~ dbin(mu.p[k,i,r], nrep[i,r])
          }
          }
 }
 
#derived quantities

 for(r in 1:R){
  logit(omega[r]) <- mu.omega
  Ntotal[r] <- sum(w[,r])    
  n0[r] <- sum(w[(nspec[r]+1):(nspec[r]+nz[r]),r]) 
   for(k in 1:M){
    Nocc.fs[k,r] <- sum(z[k,1:nsite[r],r])    
    } 
     for (i in 1:nsite[r]) {
      Nsite[i,r] <- sum(z[1:nspec[r],i,r])  #1:nspec[r]     # Number of occurring species at each site
     }
 }}", fill = TRUE, file="multiregion_occupancy.txt") 
 

#data 
jags.data <- list(M=M, nsite=nsite, R=R, nrep=nrepmatrix, nspec=nspec, nz=nz, yaug=yaug, 
                logmass=logmass, dRiv=dRiv, dSettl=dSettl, forCov=forCov)
                   

##initial values


 inits <- function() {list(z=zst, w=wst, mu.omega=runif(1,-0.8,-0.6), lpsi=rnorm(n=nspec+nz), 
                      lp=rnorm(n=nspec+nz), beta1=runif(1,0.1,0.4), alpha1=runif(1,-0.18,0.15), mu.alpha2.global=runif(1,-0.5,0.5), sd.alpha2.global=runif(1,0.5,1.5),
                      mu.beta2.global=runif(1,-0.5,0.5), sd.beta2.global=runif(1,0.5,1.5),
                      mu.beta3.global=runif(1,-0.5,0.5), sd.beta3.global=runif(1,0.5,1.5),
                      mu.beta4.global=runif(1,-0.5,0.5), sd.beta4.global=runif(1,0.5,1.5))  
   
     }                    
 
params <- c("mu.lpsi","mu.sd.lpsi" ,"mu.omega","mu.beta2.global","sd.beta2.global","beta1",
            "mu.beta3.global","sd.beta3.global","mu.beta4.global","sd.beta4.global","mu.lp",
            "mu.sd.lp","mu.alpha2.global","sd.alpha2.global","alpha1",
            "mu.logmass","sigma.logmass","n0","Ntotal","Nsite")

n.burnin <- 5000
n.iter <- 50000
n.thin <- 10
n.chains <- 3
  
library(R2jags)

out <- jags(data=jags.data,
            inits=inits,
            parameters.to.save=params,
            model.file="multiregion_occupancy.txt",
            n.chains=n.chains,
            n.iter=n.iter,
            n.burnin=n.burnin,
            n.thin=n.thin)