#************************************************************************************************
#Create capture histories from Manomet data
#The code contained in this script produces capture histories for analysis with both Cormack-Jolly-Seber and Multi-state mark-recapture models. The first section takes input data  and formats it so that it can be easily converted into matrices. The input data contains a unique record for each bird ,with columns for band number, date, time of capture, and weight. The second portion takes this formatted data and produces CJS capture histories.

#Code written by N.N.Dorian
#Last revised: January 22nd, 2019
#************************************************************************************************

##################################
#####Format data from Manomet#####
##################################

rm(list = ls()) #clear environment

library(lubridate) #load this package to make working with dates easier

setwd() #change this directory to reflect the location of your dataset

#pick which species you want to create matrix for
#you will likely have to change the code in this section to match your column names
df <-read.csv("VEER.csv")

#format csv to analyze
df$Date<-as.Date(df$Date, format = "%m/%d/%Y") #format dates
df$Time<-hm(df$Time) #extract time of captures
df$Year<-year(df$Date) #extract Year
df$DOY<-yday(df$Date)#create new column for day of year
df$Hour<-hour(df$Time)#create new column for hour of day
df$Weight<-as.numeric(as.character(df$Weight))#make weight continuous 

#check to make sure none of the bands are weird
#if they are, use the commented-out line
zeros<-which(df$Band==0)
#df<-df[-c(zeros),]

#create unique identifiers that account for individuals caught across multiple years
df$yearband<- paste0(df$Band, df$Year)#new column band-yearID
df$dyband <-paste0(df$Band,df$Year, df$DOY)#new column with band-day-yearID

#subset data for recaptures
capturefreq<-as.data.frame((table(df$yearband)))#freq of band captures
recap<-as.vector(capturefreq$Var1[which(capturefreq$Freq >1)])#recap band nums
dfrecap<- subset(df, yearband %in% recap)#data for recaptures from t0-t1
#dfrecap<-dfrecap[-c(which(table(dfrecap$yearband) ==1)),]

#subset data for birds captured only once
singlecap<-as.vector(capturefreq$Var1[which(capturefreq$Freq == 1)])
dfsinglecap<-subset(df, yearband %in% singlecap)

#dataframe of birds captured more than once in a day
multday<-as.data.frame(table(dfrecap$dyband))
multdaybands<-as.vector(multday$Var1[which(multday$Freq >1)])
dfmultday<-subset(dfrecap, dyband %in% multdaybands)

#subset by banding season
data.spring <-df[df$DOY >= 105 & df$DOY <= 166,] #pull out spring banded birds April 15 to Jun 15
data.fall <-df[df$DOY >= 227 & df$DOY <= 319,] #pull out fall banded birds, Aug 15 to Nov 15

####DO UP TO HERE EACH TIME REGARDLESS OF METHOD######

########################################################
#####Cormack-Jolly-Seber analysis capture histories#####
########################################################

#This function *makech_cjs* takes Manomet banding data and converts it into capture histories that can be analyzed using CJS models in a Bayesian context. It produces matrices for each time period and season (as specified) where each row corresponds to the sequence of captures (captured/non-captured = 1/0) over the course of each banding season. These capture histories need to be modified further in order to be analyzed using ProgramMARK.

makech_cjs<-function(data, startdoy, enddoy, startyear, endyear){
  doy<-seq(startdoy, enddoy, by =1)#sequence of dates when banding happened
  doysub<-data[data$DOY >=startdoy & data$DOY <= enddoy,] #subset data by day of year range

  years<-seq(startyear, endyear, by = 1) #sequence of years you are interested in 
  yearsub<-doysub[doysub$Year >=startyear & doysub$Year <= endyear,] #subset range by years too

  yearsub$bandyearID <-paste0(yearsub$Band,yearsub$Year)#combine band and year into a single ID 
  bands<-unique(yearsub$bandyearID)#pull out unique band nums
  bandyears<-substr(bands, nchar(bands)-3, nchar(bands)) #lets us use year as a covariate

  recaptures<-array(0, dim = c(length(bands), length(doy)+2)) #create blank matrix to populate with capture histories
  recaptures[,1]<-bands #fill first column with band numbers
  recaptures[,2]<-bandyears #fill second column with year

  colnames(recaptures)<-c("band", "year", doy) #name the columns

  #loop over all band numbers
  for(j in 1:length(bands)){
  
  subset<-yearsub[yearsub$bandyearID == bands[j],]#subset data by band number
  position<-as.vector(match(subset$DOY,colnames(recaptures)[3:ncol(recaptures)])) #get the day of year when the bird was captured/recaptured
  recaptures[j,position] <- 1 #assign "1" to the matrix on those days
}
  assign("matrix.surv",recaptures, envir = .GlobalEnv)
}

#Produce and export CJS capture histories for each season and period combination
#SPRING
mawach.spring.7015<-makech_cjs(df, 105, 166, 1970, 2015)
write.csv(mawach.spring.7015, "veerch.spring.7015.csv")

#FALL 
mawach.fall.7015<-makech_cjs(df, 227, 319, 1970, 2015)
write.csv(mawach.fall.7015, "veerch.fall.7015.csv")