################################################################
############ CODE DISEASE SPREAD vs VACCINATION ################
################################################################

# Author: Sebastian Napp (CReSA-IRTA)
# date: 2018

# Install different packages that we will need
install.packages("maps")   
install.packages("mapdata")  
install.packages("maptools")  
install.packages("sp")    
install.packages("rgdal")  
install.packages("rgeos")  
install.packages("dplyr") 
  
# load those packages
library(maps)
library(mapdata)
library(maptools)  
library(sp)
library(rgdal)
library(rgeos)
library(dplyr)  


# set your working directory
# setwd("your wd")

# load a map (shapefiles) with a given administrative division from the working directory. 
# in our case, we used France (administrative division 2, i.e. Departments) obtained from: https://gadm.org/download_country_v3.html
FRANCE2=readOGR("FRA_adm2.shp")    

# Reproject France shapefiles to UTM 28N 
FRANCE2_28N=spTransform(FRANCE2, CRS("+init=epsg:2192"))


################################################################

# MAKE THE JOIN OF SHAPEFILE AND DATABASE

# load the csv file with the data on the number of herds and animals for each of your geographical units
# for the sake of the exercise, we set a fixed 500 herds and 5000 cattle per Department
census=read.csv("CENSUSDATA.csv", sep=";",header=T)

# transform ID2 in shapefile to integer for joining with vaccination data
FRANCE2_28N@data$ID_2 = as.numeric(as.character(FRANCE2_28N@data$ID_2))

# Join data from cvs file (with census data) to the shapefile using the code for the administrative divisions
FRANCE2_28N@data = left_join(FRANCE2_28N@data, census, by=c("ID_2" = "MAPNUM"))



################################################################

# SIMULATE THE RANDOM DISTRIBUTION OF ALL CATTLE HERDS IN FRANCE

# create a list to store the (randomly distributted) farms in the different departments
farms=list()

# get the number of geographical units in your shapefile
ndep=length(FRANCE2_28N$NAME_2)

# create matrices to store size and average size of cattle herds in each Department
Z1=matrix(ncol=ndep,nrow=1,byrow=T)
Z2=matrix(ncol=ndep,nrow=1,byrow=T)

# loop to simulate random distribution of herds
for(i in 1: ndep){ 
  
  # generate random farms within a given department (i)
  farms[[i]]=spsample(FRANCE2_28N[i,],FRANCE2_28N$CATHERDS[i],type="random")
  
  # number of herds in department i
  Z1[i]=FRANCE2_28N$CATHERDS[i]
  
  # average number of bovines per farm in department i
  Z2[i]=FRANCE2_28N$CATHEADS[i]/FRANCE2_28N$CATHERDS[i]
  
  # vector with average census for a given department
  bov=rep(Z2[i], Z1[i])
  
  # vector with department to which the herds belong
  depto=as.character(FRANCE2_28N$Dep[i])
  
  # generate elements of SpatialPointsDataFrame: df with coordinates and df1 with dataframe
  df=data.frame(x = farms[[i]]$x, y=farms[[i]]$y)
  df1=data.frame(bov,depto)
  mat=as.matrix(df) 
  
  # generate elements of SpatialPoints
  sp1=SpatialPoints(coords = mat)
  
  # generate elements of SpatialPointsDataFrame
  farms[[i]]=SpatialPointsDataFrame(sp1, data = df1)
  
}


# check cattle herds (in one of the 96 departments)
i=43
plot(FRANCE2_28N[i,])
plot(farms[[i]],col="red",cex=0.5,pch=16,add=T)

# create a SpatialPointsDataFrame object with all the farms in France by combining all the randomly distributted farms in the different departments
FARMS=do.call("rbind", farms)

# check their distribution
plot(FRANCE2_28N)
plot(FARMS,col="red",cex=0.01,add=T)



################################################################

# LOOP TO SIMULATE DISEASE SPREAD 

# get number of farms in France
nfarms=length(FARMS)

# select weekly speed of spread of disease, in our case 7.3 Km per week
speed=7.3

# list in which each element is a dataframe with all the farms (within a corresponding radius) and their census, given an initial herd and a time for disease spread
farmlist=list()

# list in which each element is a SpatialPointsDataFrame with all the farms (within a corresponding radius), given an initial herd and a time for disease spread
farmlist2=list()

# timesteps in disease spread (columns in dataframe => i)
# trials or number of farms randomly sampled (rows in dataframe => j)
timesteps=5  # i
trials=500  # j

# matrices of j (trials or number of farms randomly sampled, i.e. rows in dataframe) by i (timesteps in disease spread, i.e. columns in dataframe) dimensions for storing a) number of farms and b) number of animals for the different buffers, respectively.
matrix.farms=matrix(ncol=timesteps,nrow=trials,byrow=T)
matrix.animals=matrix(ncol=timesteps,nrow=trials,byrow=T)

# matrix to store randomly selected farm plus the Department to which it belongs
matrix.herd=matrix(ncol=2,nrow=trials,byrow=T)

# random farms selected
rndF=matrix(ncol=trials,nrow=1,byrow=T)

# loop
for(j in 1: trials){ 
  
  # select 1 herd randomly
  rndF[j]=sample(1:nfarms, 1, replace=T)
  
  # store value of random herd
  matrix.herd[j,1]=rndF[j]
  matrix.herd[j,2]=as.character(FARMS$depto[rndF[j]])
  
  for(i in 1: timesteps){   
    
    # generate buffer around random farm (rndF[i]) for time i
    BUFF1=gBuffer(FARMS[rndF[j],],width=speed*i*1000)
    
    # get a list of the farms within buffers
    k=((j-1)*timesteps)+i
    
    farmlist[[k]]=over(BUFF1,FARMS,returnList=T)
    farmlist2[[k]]=FARMS[!is.na(over(FARMS,BUFF1)),]
    
    # number of farms
    z1=length(farmlist[[k]]$buffer[ ,1])
    matrix.farms[j,i]=z1
    
    # mean number of animals
    z2=(farmlist[[k]]$buffer[1,1])
    z3=z1*z2
    matrix.animals[j,i]=round(z3)
  }
}



# plot results of a random simulation (e.g. 44) with all the farms within a given radius
x=44
plot(FRANCE2_28N)
plot(farmlist2[[x]],col="red",cex=0.05,add=T)

# get a dataframe with all the herds from that simulation
farmlist[[x]]

# transform matrix.herd from matrix to dataframe and name columns
matrix.herd=as.data.frame(matrix.herd)
colnames(matrix.herd)[1]="Initial herd"
colnames(matrix.herd)[2]="Initial Department"

# save results for herds and for animals
RES.HERDS=data.frame(matrix.herd,matrix.farms)
RES.ANIMALS=data.frame(matrix.herd,matrix.animals)


################################################################