####################Interpolation of Tmax for 10 dates.#####################

#This script interpolates station values for the Oregon case study. This program loads the station data from a shapefile

#and perform 8 regressions using the general additive model (GAM). Note that this program:

#1)assumes that the shapefile in the current working 

#2)extract relevant variables from raster images before performing the regressions. 

#This scripts predicts tmas xsing GAM and LST derived from MOD11A1.

#Interactions terms are also included and assessed using the RMSE from validation dataset.

#There are 10 dates used for the GAM interpolation. The dates must be provided as a textfile.

#Script created by Benoit Parmentier on April 4, 2012. 

###Loading r library and packages                                                      # loading the raster package

library(gtools)                                                                        # loading ...

library(mgcv)

library(sp)

library(spdep)

library(rgdal)

###Parameters and arguments

infile1<-"ghcn_or_tmax_b_04142012_OR83M.shp"

path<-"/data/computer/parmentier/Data/IPLANT_project/data_Oregon_stations"

#path<-"H:/Data/IPLANT_project/data_Oregon_stations"

setwd(path) 

infile2<-"dates_interpolation_03052012.txt"                                          #List of 10 dates for the regression

infile2<-"list_365_dates_04212012.txt"

prop<-0.3                                                                            #Proportion of testing retained for validation   

out_prefix<-"_05012012_mod8_LST"

infile3<-"LST_dates_var_names.txt"

infile4<-"models_interpolation_04032012b.txt"

#######START OF THE SCRIPT #############

###Reading the station data and setting up for models' comparison

filename<-sub(".shp","",infile1)              #Removing the extension from file.

ghcn<-readOGR(".", filename)                  #reading shapefile 

proj4string(ghcn) #This retrieves the coordinate system for the SDF

CRS_ghcn<-proj4string(ghcn) #this can be assigned to mean_LST!!!

ghcn = transform(ghcn,Northness = cos(ASPECT)) #Adding a variable to the dataframe

ghcn = transform(ghcn,Eastness = sin(ASPECT))  #adding variable to the dataframe.

ghcn = transform(ghcn,Northness_w = sin(slope)*cos(ASPECT)) #Adding a variable to the dataframe

ghcn = transform(ghcn,Eastness_w = sin(slope)*sin(ASPECT))  #adding variable to the dataframe.

set.seed(100)

dates <-readLines(paste(path,"/",infile2, sep=""))

LST_dates <-readLines(paste(path,"/",infile3, sep=""))

models <-readLines(paste(path,"/",infile4, sep=""))

results <- matrix(1,length(dates),15)            #This is a matrix containing the diagnostic measures from the GAM models.

results_AIC<- matrix(1,length(dates),length(models)+3)  

results_GCV<- matrix(1,length(dates),length(models)+3)

results_DEV<- matrix(1,length(dates),length(models)+3)

results_RMSE<- matrix(1,length(dates),length(models)+3)

cor_LST_LC1<-matrix(1,length(dates),1)      #correlation LST-LC1

cor_LST_LC3<-matrix(1,length(dates),1)      #correlation LST-LC3

cor_LST_tmax<-matrix(1,length(dates),1)    #correlation LST-tmax

#Screening for bad values

ghcn_all<-ghcn

ghcn_test<-subset(ghcn,ghcn$tmax>-150 & ghcn$tmax<400)

ghcn_test2<-subset(ghcn_test,ghcn_test$ELEV_SRTM>0)

ghcn<-ghcn_test2

month_var<-c("mm_01","mm_02","mm_03","mm_04","mm_05","mm_06","mm_07","mm_08","mm_09", "mm_10", "mm_11", "mm_12")

ghcn.subsets <-lapply(dates, function(d) subset(ghcn, date==d)) #this creates a list of 10 subsets data

#note that compare to the previous version date_ column was changed to date

## looping through the dates...

#Change this into  a nested loop, looping through the number of models

for(i in 1:length(dates)){            # start of the for loop #1

  date<-strptime(dates[i], "%Y%m%d")

  month<-strftime(date, "%m")

  LST_month<-paste("mm_",month,sep="")

  ###Regression part 1: Creating a validation dataset by creating training and testing datasets

  mod <-ghcn.subsets[[i]][,match(LST_month, names(ghcn.subsets[[i]]))]

  ghcn.subsets[[i]] = transform(ghcn.subsets[[i]],LST = mod)

  #Screening LST values

  #ghcn.subsets[[i]]<-subset(ghcn.subsets[[i]],ghcn.subsets[[i]]$LST> 258 & ghcn.subsets[[i]]$LST<313)

  n<-nrow(ghcn.subsets[[i]])

  ns<-n-round(n*prop)  #Create a sample from the data frame with 70% of the rows

  nv<-n-ns             #create a sample for validation with prop of the rows

  #ns<-n-round(n*prop)  #Create a sample from the data frame with 70% of the rows

  ind.training <- sample(nrow(ghcn.subsets[[i]]), size=ns, replace=FALSE) #This selects the index position for 70% of the rows taken randomly

  ind.testing <- setdiff(1:nrow(ghcn.subsets[[i]]), ind.training)

  data_s <- ghcn.subsets[[i]][ind.training, ]

  data_v <- ghcn.subsets[[i]][ind.testing, ]

  #mod <-data_s[,match(LST_dates[i], names(data_s))]

  #data_s = transform(data_s,LST = mod)

  #data_v = transform(data_v,LST = mod)

  ####Regression part 2: GAM models

  mod1<- gam(tmax~ s(lat) + s (lon) + s (ELEV_SRTM), data=data_s)

  #mod2<- gam(tmax~ s(lat,lon,ELEV_SRTM), data=data_s)

  mod2<- gam(tmax~ s(lat,lon) + s(ELEV_SRTM), data=data_s)

  mod3<- gam(tmax~ s(lat) + s (lon) + s (ELEV_SRTM) +  s (Northness)+ s (Eastness) + s(DISTOC), data=data_s)

  mod4<- gam(tmax~ s(lat) + s (lon) + s(ELEV_SRTM) + s(Northness) + s (Eastness) + s(DISTOC) + s(LST), data=data_s)

  mod5<- gam(tmax~ s(lat,lon) +s(ELEV_SRTM) + s(Northness,Eastness) + s(DISTOC) + s(LST), data=data_s)

  mod6<- gam(tmax~ s(lat,lon) +s(ELEV_SRTM) + s(Northness,Eastness) + s(DISTOC) + s(LST,LC1), data=data_s)

  mod7<- gam(tmax~ s(lat,lon) +s(ELEV_SRTM) + s(Northness,Eastness) + s(DISTOC) + s(LST,LC3), data=data_s)

  mod8<- gam(tmax~ s(lat,lon) +s(ELEV_SRTM) + s(Northness,Eastness) + s(DISTOC) + s(LST) + s(LC1), data=data_s)

  ####Regression part 3: Calculating and storing diagnostic measures

  results_AIC[i,1]<- dates[i]  #storing the interpolation dates in the first column

  results_AIC[i,2]<- ns        #number of stations used in the training stage

  results_AIC[i,3]<- AIC (mod1)

  results_AIC[i,4]<- AIC (mod2)

  results_AIC[i,5]<- AIC (mod3)

  results_AIC[i,6]<- AIC (mod4)

  results_AIC[i,7]<- AIC (mod5)

  results_AIC[i,8]<- AIC (mod6)

  results_AIC[i,9]<- AIC (mod7)

  results_AIC[i,10]<- AIC (mod8)

  results_GCV[i,1]<- dates[i]  #storing the interpolation dates in the first column

  results_GCV[i,2]<- ns        #number of stations used in the training stage

  results_GCV[i,3]<- mod1$gcv.ubre

  results_GCV[i,4]<- mod2$gcv.ubre

  results_GCV[i,5]<- mod3$gcv.ubre

  results_GCV[i,6]<- mod4$gcv.ubre

  results_GCV[i,7]<- mod5$gcv.ubre

  results_GCV[i,8]<- mod6$gcv.ubre

  results_GCV[i,9]<- mod7$gcv.ubre

  results_GCV[i,10]<- mod7$gcv.ubre

  results_DEV[i,1]<- dates[i]  #storing the interpolation dates in the first column

  results_DEV[i,2]<- ns        #number of stations used in the training stage

  results_DEV[i,3]<- mod1$deviance

  results_DEV[i,4]<- mod2$deviance

  results_DEV[i,5]<- mod3$deviance

  results_DEV[i,6]<- mod4$deviance

  results_DEV[i,7]<- mod5$deviance

  results_DEV[i,8]<- mod6$deviance

  results_DEV[i,9]<- mod7$deviance

  results_DEV[i,10]<- mod8$deviance

  #####VALIDATION: Prediction checking the results using the testing data########

  y_mod1<- predict(mod1, newdata=data_v, se.fit = TRUE) #Using the coeff to predict new values.

  y_mod2<- predict(mod2, newdata=data_v, se.fit = TRUE)            

  y_mod3<- predict(mod3, newdata=data_v, se.fit = TRUE) 

  y_mod4<- predict(mod4, newdata=data_v, se.fit = TRUE) 

  y_mod5<- predict(mod5, newdata=data_v, se.fit = TRUE) 

  y_mod6<- predict(mod6, newdata=data_v, se.fit = TRUE)

  y_mod7<- predict(mod7, newdata=data_v, se.fit = TRUE)

  y_mod8<- predict(mod8, newdata=data_v, se.fit = TRUE)

  res_mod1<- data_v$tmax - y_mod1$fit #Residuals for GMA model that resembles the ANUSPLIN interpolation

  res_mod2<- data_v$tmax - y_mod2$fit   #Residuals for GAM model that resembles the PRISM interpolation                               

  res_mod3<- data_v$tmax - y_mod3$fit  

  res_mod4<- data_v$tmax - y_mod4$fit

  res_mod5<- data_v$tmax - y_mod5$fit

  res_mod6<- data_v$tmax - y_mod6$fit

  res_mod7<- data_v$tmax - y_mod7$fit

  res_mod8<- data_v$tmax - y_mod8$fit

  RMSE_mod1 <- sqrt(sum(res_mod1^2)/nv)          

  RMSE_mod2 <- sqrt(sum(res_mod2^2)/nv)

  RMSE_mod3 <- sqrt(sum(res_mod3^2)/nv)

  RMSE_mod4 <- sqrt(sum(res_mod4^2)/nv)

  RMSE_mod5 <- sqrt(sum(res_mod5^2)/nv)

  RMSE_mod6 <- sqrt(sum(res_mod6^2)/nv)

  RMSE_mod7 <- sqrt(sum(res_mod7^2)/nv)

  RMSE_mod8 <- sqrt(sum(res_mod8^2)/nv)

  results_RMSE[i,1]<- dates[i]  #storing the interpolation dates in the first column

  results_RMSE[i,2]<- ns        #number of stations used in the training stage

  results_RMSE[i,3]<- RMSE_mod1

  results_RMSE[i,4]<- RMSE_mod2

  results_RMSE[i,5]<- RMSE_mod3

  results_RMSE[i,6]<- RMSE_mod4

  results_RMSE[i,7]<- RMSE_mod5

  results_RMSE[i,8]<- RMSE_mod6

  results_RMSE[i,9]<- RMSE_mod7

  results_RMSE[i,10]<- RMSE_mod8

  #Saving dataset in dataframes

  data_name<-paste("ghcn_v_",dates[[i]],sep="")

  assign(data_name,data_v)

  data_name<-paste("ghcn_s_",dates[[i]],sep="")

  assign(data_name,data_s)

  #ghcn_v<-ls(pattern="ghcn_v_")

  cor_LST_LC1[i]<-cor(ghcn.subsets[[i]]$LST,ghcn.subsets[[i]]$LC1)

  cor_LST_LC3[i]<-cor(ghcn.subsets[[i]]$LST,ghcn.subsets[[i]]$LC3)

  #end of the for loop #1

  }

## Plotting and saving diagnostic measures

results_RMSEnum <-results_RMSE

results_AICnum <-results_AIC

mode(results_RMSEnum)<- "numeric"

mode(results_AICnum)<- "numeric"

# Make it numeric first

# Now turn it into a data.frame...

results_table_RMSE<-as.data.frame(results_RMSEnum)

results_table_AIC<-as.data.frame(results_AICnum)

colnames(results_table_RMSE)<-c("dates","ns","mod1", "mod2","mod3", "mod4", "mod5", "mod6", "mod7", "mod8")

colnames(results_table_AIC)<-c("dates","ns","mod1", "mod2","mod3", "mod4", "mod5", "mod6", "mod7", "mod8")

#results_table_RMSE

write.table(results_table_RMSE, file= paste(path,"/","results_GAM_Assessment",out_prefix,".txt",sep=""), sep=",")

write.table(results_table_AIC, file= paste(path,"/","results_GAM_Assessment",out_prefix,".txt",sep=""),sep=",", append=TRUE)

###Analysing the results from the 365 days run: Summarize by month

for(i in 1:nrow(results_table_RMSE)){

  date<-results_table_RMSE$dates[i]

  date<-strptime(date, "%Y%m%d")

  results_table_RMSE$month[i]<-as.integer(strftime(date, "%m"))

}

average<-aggregate(cbind(mod1,mod2,mod3,mod4,mod5,mod6,mod7,mod8)~month,data=results_table_RMSE,mean, na.rm=TRUE)

average<-aggregate(cbind(mod1,mod2,mod3,mod4,mod5,mod6,mod7,mod8)~month,data=results_table_RMSE, FUN=mean)

#average on all the data.frame

averaget<-aggregate(results_table_RMSE, by=list(results_table_RMSE$month),FUN=mean, na.rm=TRUE)

#mediant<-aggregate(results_table_RMSE, by=list(results_table_RMSE$month),FUN=median, na.rm=TRUE)

#average_lowt<-aggregate(results_table_RMSE, by=list(results_table_RMSE$month), FUN=function(v) t.test(v)$conf.int[1])

#average_up<-aggregate(cbind(mod1,mod2,mod3,mod4,mod5,mod6,mod7,mod8)~month,data=results_table_RMSE, function(v) t.test(v)$conf.int[2])

median<-aggregate(cbind(mod1,mod2,mod3,mod4,mod5,mod6,mod7,mod8)~month,data=results_table_RMSE, median, na.rm=TRUE)

average_low<-aggregate(cbind(mod1,mod2,mod3,mod4,mod5,mod6,mod7,mod8)~month,data=results_table_RMSE, function(v) t.test(v)$conf.int[1])

average_up<-aggregate(cbind(mod1,mod2,mod3,mod4,mod5,mod6,mod7,mod8)~month,data=results_table_RMSE, function(v) t.test(v)$conf.int[2])

mod<-names(averaget)

mod<-mod[4:11]

#Saving graphic plots

for(i in 1:length(mod)){

  X11(width=14,height=10)

  name<-mod[i]

  barplot2(average[[name]],plot.ci=TRUE, ci.l=average_low[[name]], ci.u=average_up[[name]],main="Mean RMSE per month", names.arg=c("Jan", "Feb", "Mar", "Apr", "May", "Jun","Jul", "Aug", "Sep","Oct", "Nov", "Dec"),ylim=c(20,30),ylab="RMSE in tenth deg C",xlab=name)

  #title(paste("Sampling RMSE for mod",i,sep=""))

  savePlot(paste("barplot_results_RMSE_month_",name,out_prefix,".png", sep=""), type="png")

  dev.off() 

}

for(i in 1:length(mod)){

  X11(width=14,height=10)

  name<-mod[i]

  barplot2(average[[name]],plot.ci=TRUE, ci.l=average_low[[name]], ci.u=average_up[[name]],main=paste(" Mean RMSE per month ",name, sep=""), names.arg=c("Jan", "Feb", "Mar", "Apr", "May", "Jun","Jul", "Aug", "Sep","Oct", "Nov", "Dec"),ylim=c(20,30),ylab="RMSE in tenth deg C",xlab=name)

  #title(paste("Sampling RMSE for mod",i,sep=""))

  savePlot(paste("barplot_results_RMSE_month_",name,out_prefix,".png", sep=""), type="png")

  dev.off() 

  X11(width=14,height=10)

  name<-mod[i]

  hist(results_table_RMSE[[name]],breaks=15, main=paste(" Histogram RMSE_",name, sep=""),xlab=paste("RMSE ",name, sep=""))

  savePlot(paste("Hist_results_RMSE_365_",name,out_prefix,".png", sep=""), type="png")

  dev.off()

}

for(i in 1:length(mod)){

  X11(width=14,height=10)

  name<-mod[i]

  hist(results_table_RMSE[[name]],breaks=15, main=paste(" Histogram RMSE_",name, sep=""),xlab=paste("RMSE ",name, sep=""))

  savePlot(paste("Hist_results_RMSE_365_",name,out_prefix,".png", sep=""), type="png")

  dev.off()

}

r<-(results_table_RMSE[,3:10]) #selecting only the columns related to models...

mean_r<-mean(r)

median_r<-sapply(r, median)

sd_r<-sapply(r, sd)

barplot(mean_r,ylim=c(23,26),ylab="RMSE in tenth deg C")

barplot(median_r,ylim=c(23,26),ylab="RMSE in tenth deg C",add=TRUE,inside=FALSE,beside=TRUE) # put both on the same plot

barplot(sd_r,ylim=c(6,8),ylab="RMSE in tenth deg C") # put both on the same plot

height<-rbind(mean_r,median_r)

barplot(height,ylim=c(23,26),ylab="RMSE in tenth deg C",beside=TRUE,legend=rownames(height))

barplot(height,ylim=c(23,26),ylab="RMSE in tenth deg C",beside=TRUE, col=c("darkblue","red"),legend=rownames(height)) # put both on the same plot

barplot2(mean_r,median_r,ylim=c(23,26),ylab="RMSE in tenth deg C") # put both on the same plot

#Collect var explained and p values for each var...

# End of script##########

names(results_table_RMSE)