1
|
runGAMCAI <- function(i) { # loop over dates
|
2
|
|
3
|
#ith upates from 10/26/2012
|
4
|
#date<-strptime(dates[i], "%Y%m%d") # interpolation date being processed
|
5
|
date<-strptime(sampling_dat$date[i], "%Y%m%d") # interpolation date being processed, converting the string using specific format
|
6
|
month<-strftime(date, "%m") # current month of the date being processed
|
7
|
LST_month<-paste("mm_",month,sep="") # name of LST month to be matched in the raster stack of covariates and data.frame
|
8
|
|
9
|
#Adding layer LST to the raster stack
|
10
|
|
11
|
pos<-match("LST",layerNames(s_raster)) #Find the position of the layer with name "LST", if not present pos=NA
|
12
|
s_raster<-dropLayer(s_raster,pos) # If it exists drop layer
|
13
|
pos<-match(LST_month,layerNames(s_raster)) #Find column with the current month for instance mm12
|
14
|
r1<-raster(s_raster,layer=pos) #Select layer from stack
|
15
|
layerNames(r1)<-"LST"
|
16
|
#Screen for extreme values" 10/30
|
17
|
min_val<-(-15+273.16) #if values less than -15C then screen out (note the Kelvin units that will need to be changed later in all datasets)
|
18
|
r1[r1 < (min_val)]<-NA
|
19
|
s_raster<-addLayer(s_raster,r1) #Adding current month
|
20
|
|
21
|
###Regression part 1: Creating a validation dataset by creating training and testing datasets
|
22
|
|
23
|
mod_LST <-ghcn.subsets[[i]][,match(LST_month, names(ghcn.subsets[[i]]))] #Match interpolation date and monthly LST average
|
24
|
ghcn.subsets[[i]] <- transform(ghcn.subsets[[i]],LST = mod_LST) #Add the variable LST to the subset dataset
|
25
|
dst$LST<-dst[[LST_month]] #Add also to monthly dataset
|
26
|
|
27
|
#n<-nrow(ghcn.subsets[[i]])
|
28
|
#ns<-n-round(n*prop) #Create a sample from the data frame with 70% of the rows
|
29
|
#nv<-n-ns #create a sample for validation with prop of the rows
|
30
|
#ind.training <- sample(nrow(ghcn.subsets[[i]]), size=ns, replace=FALSE) #This selects the index position for 70% of the rows taken randomly
|
31
|
ind.training<-sampling[[i]]
|
32
|
ind.testing <- setdiff(1:nrow(ghcn.subsets[[i]]), ind.training)
|
33
|
data_s <- ghcn.subsets[[i]][ind.training, ] #Training dataset currently used in the modeling
|
34
|
data_v <- ghcn.subsets[[i]][ind.testing, ] #Testing/validation dataset using input sampling
|
35
|
|
36
|
ns<-nrow(data_s)
|
37
|
nv<-nrow(data_v)
|
38
|
#i=1
|
39
|
date_proc<-sampling_dat$date[i]
|
40
|
date_proc<-strptime(sampling_dat$date[i], "%Y%m%d") # interpolation date being processed
|
41
|
mo<-as.integer(strftime(date_proc, "%m")) # current month of the date being processed
|
42
|
day<-as.integer(strftime(date_proc, "%d"))
|
43
|
year<-as.integer(strftime(date_proc, "%Y"))
|
44
|
|
45
|
datelabel=format(ISOdate(year,mo,day),"%b %d, %Y")
|
46
|
|
47
|
###########
|
48
|
# STEP 1 - LST 10 year monthly averages: THIS IS NOT USED IN CAI method
|
49
|
###########
|
50
|
|
51
|
themolst<-raster(molst,mo) #current month being processed saved in a raster image
|
52
|
min_val<-(-15) #Screening for extreme values
|
53
|
themolst[themolst < (min_val)]<-NA
|
54
|
|
55
|
plot(themolst)
|
56
|
|
57
|
###########
|
58
|
# STEP 2 - Weather station means across same days: Monthly mean calculation
|
59
|
###########
|
60
|
|
61
|
modst=dst[dst$month==mo,] #Subsetting dataset for the relevant month of the date being processed
|
62
|
|
63
|
##########
|
64
|
# STEP 3 - get LST at stations
|
65
|
##########
|
66
|
|
67
|
sta_lola=modst[,c("lon","lat")] #Extracting locations of stations for the current month..
|
68
|
|
69
|
proj_str="+proj=lcc +lat_1=43 +lat_2=45.5 +lat_0=41.75 +lon_0=-120.5 +x_0=400000 +y_0=0 +ellps=GRS80 +units=m +no_defs";
|
70
|
lookup<-function(r,lat,lon) {
|
71
|
xy<-project(cbind(lon,lat),proj_str);
|
72
|
cidx<-cellFromXY(r,xy);
|
73
|
return(r[cidx])
|
74
|
}
|
75
|
sta_tmax_from_lst=lookup(themolst,sta_lola$lat,sta_lola$lon) #Extracted values of LST for the stations
|
76
|
|
77
|
#########
|
78
|
# STEP 4 - bias at stations
|
79
|
#########
|
80
|
|
81
|
sta_bias=sta_tmax_from_lst-modst$TMax; #That is the difference between the monthly LST mean and monthly station mean
|
82
|
#Added by Benoit
|
83
|
modst$LSTD_bias<-sta_bias #Adding bias to data frame modst containning the monthly average for 10 years
|
84
|
|
85
|
bias_xy=project(as.matrix(sta_lola),proj_str)
|
86
|
# png(paste("LST_TMax_scatterplot_",dates[i],out_prefix,".png", sep=""))
|
87
|
# plot(modst$TMax,sta_tmax_from_lst,xlab="Station mo Tmax",ylab="LST mo Tmax",main=paste("LST vs TMax for",datelabel,sep=" "))
|
88
|
# abline(0,1)
|
89
|
# dev.off()
|
90
|
|
91
|
#added by Benoit
|
92
|
#x<-ghcn.subsets[[i]] #Holds both training and testing for instance 161 rows for Jan 1
|
93
|
x<-data_v
|
94
|
d<-data_s
|
95
|
|
96
|
pos<-match("value",names(d)) #Find column with name "value"
|
97
|
#names(d)[pos]<-c("dailyTmax")
|
98
|
names(d)[pos]<-y_var_name
|
99
|
names(x)[pos]<-y_var_name
|
100
|
#names(x)[pos]<-c("dailyTmax")
|
101
|
d$dailyTmax=(as.numeric(d$dailyTmax))/10 #stored as 1/10 degree C to allow integer storage
|
102
|
x$dailyTmax=(as.numeric(x$dailyTmax))/10 #stored as 1/10 degree C to allow integer storage
|
103
|
pos<-match("station",names(d)) #Find column with name "value"
|
104
|
names(d)[pos]<-c("id")
|
105
|
names(x)[pos]<-c("id")
|
106
|
names(modst)[1]<-c("id") #modst contains the average tmax per month for every stations...it has 193 rows
|
107
|
|
108
|
dmoday=merge(modst,d,by="id",suffixes=c("",".y2")) #LOOSING DATA HERE!!! from 113 t0 103
|
109
|
xmoday=merge(modst,x,by="id",suffixes=c("",".y2")) #LOOSING DATA HERE!!! from 48 t0 43
|
110
|
mod_pat<-glob2rx("*.y2")
|
111
|
var_pat<-grep(mod_pat,names(dmoday),value=FALSE) # using grep with "value" extracts the matching names
|
112
|
dmoday<-dmoday[,-var_pat]
|
113
|
mod_pat<-glob2rx("*.y2")
|
114
|
var_pat<-grep(mod_pat,names(xmoday),value=FALSE) # using grep with "value" extracts the matching names
|
115
|
xmoday<-xmoday[,-var_pat] #Removing duplicate columns
|
116
|
|
117
|
#dmoday=merge(modst,d,by="id") #LOOSING DATA HERE!!! from 113 t0 103
|
118
|
#xmoday=merge(modst,x,by="id") #LOOSING DATA HERE!!! from 48 t0 43
|
119
|
#names(dmoday)[4]<-c("lat")
|
120
|
#names(dmoday)[5]<-c("lon") #dmoday contains all the the information: BIAS, monn
|
121
|
#names(xmoday)[4]<-c("lat")
|
122
|
#names(xmoday)[5]<-c("lon") #dmoday contains all the the information: BIAS, monn
|
123
|
|
124
|
data_v<-xmoday
|
125
|
###
|
126
|
|
127
|
#dmoday contains the daily tmax values for training with TMax being the monthly station tmax mean
|
128
|
#xmoday contains the daily tmax values for validation with TMax being the monthly station tmax mean
|
129
|
|
130
|
# windows()
|
131
|
#png(paste("LST_TMax_scatterplot_",dates[i],out_prefix,".png", sep=""))
|
132
|
png(paste("Daily_tmax_monthly_TMax_scatterplot_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],
|
133
|
out_prefix,".png", sep=""))
|
134
|
plot(dailyTmax~TMax,data=dmoday,xlab="Mo Tmax",ylab=paste("Daily for",datelabel),main="across stations in OR")
|
135
|
#savePlot(paste("Daily_tmax_monthly_TMax_scatterplot_",dates[i],out_prefix,".png", sep=""), type="png")
|
136
|
#png(paste("LST_TMax_scatterplot_",dates[i],out_prefix,".png", sep=""))
|
137
|
dev.off()
|
138
|
|
139
|
########
|
140
|
# STEP 5 - interpolate bias/climatology
|
141
|
########
|
142
|
|
143
|
# ?? include covariates like elev, distance to coast, cloud frequency, tree height
|
144
|
#library(fields)
|
145
|
#windows()
|
146
|
#quilt.plot(sta_lola,sta_bias,main="Bias at stations",asp=1)
|
147
|
#US(add=T,col="magenta",lwd=2)
|
148
|
#fitbias<-Tps(bias_xy,sta_bias) #use TPS or krige
|
149
|
|
150
|
#Adding options to use only training stations: 07/11/2012
|
151
|
bias_xy<-project(as.matrix(sta_lola),proj_str)
|
152
|
clim_xy<-project(as.matrix(sta_lola),proj_str) #This is the coordinates of monthly station location (193)
|
153
|
#bias_xy2=project(as.matrix(c(dmoday$lon,dmoday$lat),proj_str)
|
154
|
if(bias_val==1){
|
155
|
sta_bias<-dmoday$LSTD_bias
|
156
|
bias_xy<-cbind(dmoday$x_OR83M,dmoday$y_OR83M) #This will use only stations from training daily samples for climatology step if bias_val=1
|
157
|
}
|
158
|
|
159
|
sta_clim<-modst$TMax #This contains the monthly climatology...used in the prediction of the monthly surface
|
160
|
clim_covar<-data_month$ELEV_SRTM
|
161
|
#fitbias<-Krig(bias_xy,sta_bias,theta=1e5) #use TPS or krige
|
162
|
fitclim<-Krig(clim_xy,sta_clim,theta=1e5)
|
163
|
|
164
|
theta_val<-100000
|
165
|
kf<-exp(-rdist(clim_xy/theta_val))
|
166
|
image(kf)
|
167
|
kf_fun<-function(dist,theta=1,C=NA){
|
168
|
exp(-rdist(dist/theta))
|
169
|
}
|
170
|
plot(sort(kf[1,],decreasing=T),type="l")
|
171
|
|
172
|
fitclim2<-Krig(x=clim_xy,Y=sta_clim,Z=clim_covar,theta=1e5)
|
173
|
fitclim2<-Krig(clim_xy,sta_clim,theta=theta_val)
|
174
|
fitclim2<-Krig(clim_xy,sta_clim,cov.function="kf_fun",theta=1e5)
|
175
|
|
176
|
#The output is a krig object using fields: modif 10/30
|
177
|
#mod9a<-fitbias
|
178
|
mod_krtmp1<-fitclim
|
179
|
model_name<-paste("mod_kr","month",sep="_")
|
180
|
assign(model_name,mod_krtmp1)
|
181
|
|
182
|
# Creating plot of bias surface and saving it
|
183
|
#X11()
|
184
|
png(paste("Climtology_surface_LST_TMax_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],
|
185
|
out_prefix,".png", sep="")) #Create file to write a plot
|
186
|
datelabel2=format(ISOdate(year,mo,day),"%B ") #added by Benoit, label
|
187
|
surface(fitclim,col=rev(terrain.colors(100)),asp=1,main=paste("Interpolated clim for",datelabel2,sep=" ")) #Plot to file
|
188
|
#savePlot(paste("Bias_surface_LST_TMax_",dates[i],out_prefix,".png", sep=""), type="png")
|
189
|
dev.off() #Release the hold to the file
|
190
|
|
191
|
#US(add=T,col="magenta",lwd=2)
|
192
|
|
193
|
##########
|
194
|
# STEP 7 - interpolate delta across space: this is the daily deviation from the monthly average
|
195
|
##########
|
196
|
|
197
|
daily_sta_lola=dmoday[,c("lon","lat")] #could be same as before but why assume merge does this - assume not
|
198
|
daily_sta_xy=project(as.matrix(daily_sta_lola),proj_str)
|
199
|
daily_delta=dmoday$dailyTmax-dmoday$TMax
|
200
|
|
201
|
daily_deltaclim<-dmoday$dailyTmax-dmoday$TMax #For daily surface interpolation...
|
202
|
daily_deltaclim_v<-data_v$dailyTmax-data_v$TMax #For validation...
|
203
|
#dmoday$daily_deltaclim <-daily_deltaclim
|
204
|
#fitdelta<-Tps(daily_sta_xy,daily_delta) #use TPS or krige
|
205
|
fitdelta<-Krig(daily_sta_xy,daily_delta,theta=1e5) #use TPS or krige
|
206
|
fitdeltaclim<-Krig(daily_sta_xy,daily_deltaclim,theta=1e5) #use TPS or krige
|
207
|
|
208
|
#Kriging using fields package: modif 10/30
|
209
|
#mod9b<-fitdelta
|
210
|
mod_krtmp2<-fitdeltaclim
|
211
|
model_name<-paste("mod_kr","day",sep="_")
|
212
|
assign(model_name,mod_krtmp2)
|
213
|
|
214
|
# Creating plot of bias surface and saving it
|
215
|
#X11()
|
216
|
png(paste("Deltaclim_surface_TMax_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],
|
217
|
out_prefix,".png", sep=""))
|
218
|
surface(fitdeltaclim,col=rev(terrain.colors(100)),asp=1,main=paste("Interpolated deltaclim for",datelabel,sep=" "))
|
219
|
#savePlot(paste("Delta_surface_LST_TMax_",dates[i],out_prefix,".png", sep=""), type="png")
|
220
|
dev.off()
|
221
|
#US(add=T,col="magenta",lwd=2)
|
222
|
#
|
223
|
|
224
|
#### Added by Benoit on 06/19
|
225
|
data_s<-dmoday #put the
|
226
|
#data_s$daily_delta<-daily_delta
|
227
|
data_s$daily_deltaclim<-daily_deltaclim
|
228
|
data_v$daily_deltaclim<-daily_deltaclim_v
|
229
|
#data_s$y_var<-daily_delta #y_var is the variable currently being modeled, may be better with BIAS!!
|
230
|
#data_s$y_var<-data_s$LSTD_bias
|
231
|
#### Added by Benoit ends
|
232
|
|
233
|
#########
|
234
|
# STEP 8 - assemble final answer - T= LST-Bias(interpolated)+delta(interpolated) (This is for fusion not implemented in this script...)
|
235
|
# T= clim(interpolated) + deltaclim(interpolated) (This is for CAI)
|
236
|
#########
|
237
|
|
238
|
#bias_rast=interpolate(themolst,fitbias) #interpolation using function from raster package
|
239
|
clim_rast=interpolate(themolst,fitclim) #interpolation using function from raster package
|
240
|
#themolst is raster layer, fitbias is "Krig" object from bias surface
|
241
|
clim_rast2=interpolate(themolst,fitclim2) #interpolation using function from raster package
|
242
|
|
243
|
clim_rast2=interpolate(ELEV_SRTM,fitclim2,xyOnly=FALSE) #interpolation using function from raster package
|
244
|
|
245
|
#Saving kriged surface in raster images
|
246
|
data_name<-paste("clim_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],sep="")
|
247
|
raster_name<-paste("CAI_",data_name,out_prefix,".rst", sep="")
|
248
|
writeRaster(clim_rast, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
249
|
|
250
|
#daily_delta_rast=interpolate(themolst,fitdelta) #Interpolation of the bias surface...
|
251
|
daily_deltaclim_rast=interpolate(themolst,fitdeltaclim) #Interpolation of the bias surface...
|
252
|
|
253
|
#plot(daily_delta_rast,main="Raster Daily Delta")
|
254
|
|
255
|
#Saving kriged surface in raster images
|
256
|
data_name<-paste("deltaclim_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],sep="")
|
257
|
raster_name<-paste("CAI_",data_name,out_prefix,".rst", sep="")
|
258
|
writeRaster(daily_deltaclim_rast, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
259
|
|
260
|
#tmax_predicted=themolst+daily_delta_rast-bias_rast #Final surface as a raster layer...eqt ok
|
261
|
tmax_predicted<-daily_deltaclim_rast + clim_rast #Final surface as a raster layer...
|
262
|
#tmp6<-data_s$daily_deltaclim +data_s$TMax
|
263
|
#tmp7<-extract(tmax_predicted,data_s)
|
264
|
#plot(tmax_predicted,main="Predicted daily")
|
265
|
|
266
|
#Saving kriged surface in raster images
|
267
|
data_name<-paste("tmax_predicted_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],sep="")
|
268
|
raster_name<-paste("CAI_",data_name,out_prefix,".rst", sep="")
|
269
|
writeRaster(tmax_predicted, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
270
|
|
271
|
########
|
272
|
# check: assessment of results: validation
|
273
|
########
|
274
|
RMSE<-function(x,y) {return(mean((x-y)^2)^0.5)}
|
275
|
MAE_fun<-function(x,y) {return(mean(abs(x-y)))}
|
276
|
#ME_fun<-function(x,y){return(mean(abs(y)))}
|
277
|
#FIT ASSESSMENT
|
278
|
sta_pred_data_s=lookup(tmax_predicted,data_s$lat,data_s$lon)
|
279
|
rmse_fit=RMSE(sta_pred_data_s,data_s$dailyTmax)
|
280
|
mae_fit=MAE_fun(sta_pred_data_s,data_s$dailyTmax)
|
281
|
|
282
|
sta_pred=lookup(tmax_predicted,data_v$lat,data_v$lon)
|
283
|
#sta_pred=lookup(tmax_predicted,daily_sta_lola$lat,daily_sta_lola$lon)
|
284
|
#rmse=RMSE(sta_pred,dmoday$dailyTmax)
|
285
|
#pos<-match("value",names(data_v)) #Find column with name "value"
|
286
|
#names(data_v)[pos]<-c("dailyTmax")
|
287
|
tmax<-data_v$dailyTmax
|
288
|
#data_v$dailyTmax<-tmax
|
289
|
rmse=RMSE(sta_pred,tmax)
|
290
|
mae<-MAE_fun(sta_pred,tmax)
|
291
|
r2<-cor(sta_pred,tmax)^2 #R2, coef. of var
|
292
|
me<-mean(sta_pred-tmax)
|
293
|
|
294
|
#plot(sta_pred~dmoday$dailyTmax,xlab=paste("Actual daily for",datelabel),ylab="Pred daily",main=paste("RMSE=",rmse))
|
295
|
|
296
|
png(paste("Predicted_tmax_versus_observed_scatterplot_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",
|
297
|
sampling_dat$run_samp[i],out_prefix,".png", sep=""))
|
298
|
plot(sta_pred~tmax,xlab=paste("Actual daily for",datelabel),ylab="Pred daily",main=paste("RMSE=",rmse))
|
299
|
abline(0,1)
|
300
|
#savePlot(paste("Predicted_tmax_versus_observed_scatterplot_",dates[i],out_prefix,".png", sep=""), type="png")
|
301
|
dev.off()
|
302
|
#resid=sta_pred-dmoday$dailyTmax
|
303
|
resid=sta_pred-tmax
|
304
|
#quilt.plot(daily_sta_lola,resid)
|
305
|
|
306
|
|
307
|
###BEFORE GAM prediction the data object must be transformed to SDF
|
308
|
|
309
|
coords<- data_v[,c('x_OR83M','y_OR83M')]
|
310
|
coordinates(data_v)<-coords
|
311
|
proj4string(data_v)<-CRS #Need to assign coordinates...
|
312
|
coords<- data_s[,c('x_OR83M','y_OR83M')]
|
313
|
coordinates(data_s)<-coords
|
314
|
proj4string(data_s)<-CRS #Need to assign coordinates..
|
315
|
coords<- modst[,c('x_OR83M','y_OR83M')]
|
316
|
coordinates(modst)<-coords
|
317
|
proj4string(modst)<-CRS #Need to assign coordinates..
|
318
|
|
319
|
ns<-nrow(data_s) #This is added to because some loss of data might have happened because of the averaging...
|
320
|
nv<-nrow(data_v)
|
321
|
|
322
|
###GAM PREDICTION
|
323
|
|
324
|
#data_s$y_var<-data_s$dailyTmax #This shoudl be changed for any variable!!!
|
325
|
#data_v$y_var<-data_v$dailyTmax
|
326
|
#data_v$y_var<-data_v$daily_deltaclim
|
327
|
data_s$y_var<-data_s$daily_deltaclim
|
328
|
data_v$y_var<-data_v$daily_deltaclim
|
329
|
|
330
|
if (climgam==1){ #This is an option to use covariates in the daily surface...
|
331
|
data_s$y_var<-data_s$TMax
|
332
|
data_v$y_var<-data_v$TMax
|
333
|
data_month<-modst
|
334
|
data_month$y_var<-modst$TMax
|
335
|
}
|
336
|
|
337
|
#Model and response variable can be changed without affecting the script
|
338
|
#list_formulas<-vector("list",nmodels)
|
339
|
|
340
|
#list_formulas[[1]] <- as.formula("y_var ~ s(lat) + s(lon) + s(ELEV_SRTM)", env=.GlobalEnv)
|
341
|
#list_formulas[[2]] <- as.formula("y_var~ s(lat,lon)+ s(ELEV_SRTM)", env=.GlobalEnv)
|
342
|
#list_formulas[[3]] <- as.formula("y_var~ s(lat) + s (lon) + s (ELEV_SRTM) + s (Northness)+ s (Eastness) + s(DISTOC)", env=.GlobalEnv)
|
343
|
#list_formulas[[4]] <- as.formula("y_var~ s(lat) + s (lon) + s(ELEV_SRTM) + s(Northness) + s (Eastness) + s(DISTOC) + s(LST)", env=.GlobalEnv)
|
344
|
#list_formulas[[5]] <- as.formula("y_var~ s(lat,lon) +s(ELEV_SRTM) + s(Northness,Eastness) + s(DISTOC) + s(LST)", env=.GlobalEnv)
|
345
|
#list_formulas[[6]] <- as.formula("y_var~ s(lat,lon) +s(ELEV_SRTM) + s(Northness,Eastness) + s(DISTOC) + s(LST)+s(LC1)", env=.GlobalEnv)
|
346
|
#list_formulas[[7]] <- as.formula("y_var~ s(lat,lon) +s(ELEV_SRTM) + s(Northness,Eastness) + s(DISTOC) + s(LST)+s(LC3)", env=.GlobalEnv)
|
347
|
#list_formulas[[8]] <- as.formula("y_var~ s(lat,lon) +s(ELEV_SRTM) + s(Northness,Eastness) + s(DISTOC) + s(LST) + s(LC1,LC3)", env=.GlobalEnv)
|
348
|
|
349
|
|
350
|
#This can be entered as textfile or option later...ok for running now on 10/30/2012
|
351
|
#list_formulas[[1]] <- as.formula("y_var~ s(ELEV_SRTM)", env=.GlobalEnv)
|
352
|
#list_formulas[[2]] <- as.formula("y_var~ s(LST)", env=.GlobalEnv)
|
353
|
#list_formulas[[3]] <- as.formula("y_var~ s(LST) + s(ELEV_SRTM)", env=.GlobalEnv)
|
354
|
#list_formulas[[4]] <- as.formula("y_var~ s(LST,ELEV_SRTM)", env=.GlobalEnv)
|
355
|
#list_formulas[[5]] <- as.formula("y_var~ s(lat,lon,ELEV_SRTM)", env=.GlobalEnv)
|
356
|
|
357
|
if (climgam==1){ #This will automatically use monthly station data in the second step
|
358
|
|
359
|
for (j in 1:nmodels){
|
360
|
formula<-list_formulas[[j]]
|
361
|
mod<- try(gam(formula, data=data_month))
|
362
|
model_name<-paste("mod",j,sep="")
|
363
|
assign(model_name,mod)
|
364
|
}
|
365
|
|
366
|
} else if (climgam==0){ #This will use daily delta in the second step
|
367
|
|
368
|
for (j in 1:nmodels){
|
369
|
formula<-list_formulas[[j]]
|
370
|
mod<- try(gam(formula, data=data_s))
|
371
|
model_name<-paste("mod",j,sep="")
|
372
|
assign(model_name,mod)
|
373
|
}
|
374
|
|
375
|
}
|
376
|
|
377
|
### Added by benoit
|
378
|
#Store results using TPS
|
379
|
j=nmodels+1
|
380
|
results_RMSE[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
381
|
results_RMSE[2]<- ns #number of stations used in the training stage
|
382
|
results_RMSE[3]<- "RMSE"
|
383
|
|
384
|
results_RMSE[j+3]<- rmse #Storing RMSE for the model j
|
385
|
|
386
|
results_RMSE_f[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
387
|
results_RMSE_f[2]<- ns #number of stations used in the training stage
|
388
|
results_RMSE_f[3]<- "RMSE_f"
|
389
|
results_RMSE_f[j+3]<- rmse_fit #Storing RMSE for the model j
|
390
|
|
391
|
results_MAE_f[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
392
|
results_MAE_f[2]<- ns #number of stations used in the training stage
|
393
|
results_MAE_f[3]<- "RMSE_f"
|
394
|
results_MAE_f[j+3]<- mae_fit #Storing RMSE for the model j
|
395
|
|
396
|
results_MAE[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
397
|
results_MAE[2]<- ns #number of stations used in the training stage
|
398
|
results_MAE[3]<- "MAE"
|
399
|
results_MAE[j+3]<- mae #Storing RMSE for the model j
|
400
|
|
401
|
results_ME[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
402
|
results_ME[2]<- ns #number of stations used in the training stage
|
403
|
results_ME[3]<- "ME"
|
404
|
results_ME[j+3]<- me #Storing RMSE for the model j
|
405
|
|
406
|
results_R2[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
407
|
results_R2[2]<- ns #number of stations used in the training stage
|
408
|
results_R2[3]<- "R2"
|
409
|
results_R2[j+3]<- r2 #Storing RMSE for the model j
|
410
|
|
411
|
pred_mod<-paste("pred_mod",j,sep="")
|
412
|
#Adding the results back into the original dataframes.
|
413
|
data_s[[pred_mod]]<-sta_pred_data_s
|
414
|
data_v[[pred_mod]]<-sta_pred
|
415
|
|
416
|
#Model assessment: RMSE and then krig the residuals....!
|
417
|
|
418
|
res_mod_s<- data_s$dailyTmax - data_s[[pred_mod]] #Residuals from kriging training
|
419
|
res_mod_v<- data_v$dailyTmax - data_v[[pred_mod]] #Residuals from kriging validation
|
420
|
|
421
|
name2<-paste("res_mod",j,sep="")
|
422
|
data_v[[name2]]<-as.numeric(res_mod_v)
|
423
|
data_s[[name2]]<-as.numeric(res_mod_s)
|
424
|
|
425
|
#ns<-nrow(data_s) #This is added to because some loss of data might have happened because of the averaging...
|
426
|
#nv<-nrow(data_v)
|
427
|
#browser()
|
428
|
|
429
|
mod_obj<-vector("list",nmodels+2) #This will contain the model objects fitting: 10/30
|
430
|
mod_obj[[nmodels+1]]<-mod_kr_month #Storing climatology object
|
431
|
mod_obj[[nmodels+2]]<-mod_kr_day #Storing delta object
|
432
|
|
433
|
for (j in 1:nmodels){
|
434
|
|
435
|
##Model assessment: specific diagnostic/metrics for GAM
|
436
|
|
437
|
name<-paste("mod",j,sep="") #modj is the name of The "j" model (mod1 if j=1)
|
438
|
mod<-get(name) #accessing GAM model ojbect "j"
|
439
|
mod_obj[[j]]<-mod #storing current model object
|
440
|
|
441
|
#If mod "j" is not a model object
|
442
|
if (inherits(mod,"try-error")) {
|
443
|
results_m1[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
444
|
results_m1[1,2]<- ns #number of stations used in the training stage
|
445
|
results_m1[1,3]<- "AIC"
|
446
|
results_m1[1,j+3]<- NA
|
447
|
|
448
|
results_m2[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
449
|
results_m2[1,2]<- ns #number of stations used in the training
|
450
|
results_m2[1,3]<- "GCV"
|
451
|
results_m2[1,j+3]<- NA
|
452
|
|
453
|
results_m3[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
454
|
results_m3[1,2]<- ns #number of stations used in the training stage
|
455
|
results_m3[1,3]<- "DEV"
|
456
|
results_m3[1,j+3]<- NA
|
457
|
|
458
|
results_RMSE_f[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
459
|
results_RMSE_f[1,2]<- ns #number of stations used in the training stage
|
460
|
results_RMSE_f[1,3]<- "RSME_f"
|
461
|
results_RMSE_f[1,j+3]<- NA
|
462
|
|
463
|
results_MAE_f[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
464
|
results_MAE_f[1,2]<- ns #number of stations used in the training stage
|
465
|
results_MAE_f[1,3]<- "MAE_f"
|
466
|
results_MAE_f[1,j+3]<-NA
|
467
|
|
468
|
results_R2_f[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
469
|
results_R2_f[1,2]<- ns #number of stations used in the training stage
|
470
|
results_R2_f[1,3]<- "R2_f"
|
471
|
results_R2_f[1,j+3]<- NA #Storing R2 for the model j
|
472
|
|
473
|
|
474
|
results_RMSE[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
475
|
results_RMSE[1,2]<- ns #number of stations used in the training stage
|
476
|
results_RMSE[1,3]<- "RMSE"
|
477
|
results_RMSE[1,j+3]<- NA #Storing RMSE for the model j
|
478
|
results_MAE[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
479
|
results_MAE[1,2]<- ns #number of stations used in the training stage
|
480
|
results_MAE[1,3]<- "MAE"
|
481
|
results_MAE[1,j+3]<- NA #Storing MAE for the model j
|
482
|
results_ME[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
483
|
results_ME[1,2]<- ns #number of stations used in the training stage
|
484
|
results_ME[1,3]<- "ME"
|
485
|
results_ME[1,j+3]<- NA #Storing ME for the model j
|
486
|
results_R2[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
487
|
results_R2[1,2]<- ns #number of stations used in the training stage
|
488
|
results_R2[1,3]<- "R2"
|
489
|
results_R2[1,j+3]<- NA #Storing R2 for the model j
|
490
|
|
491
|
}
|
492
|
|
493
|
#If mod is a modelobject
|
494
|
|
495
|
#If mod "j" is not a model object
|
496
|
if (inherits(mod,"gam")) {
|
497
|
|
498
|
# model specific metrics
|
499
|
results_m1[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
500
|
results_m1[1,2]<- ns #number of stations used in the training stage
|
501
|
results_m1[1,3]<- "AIC"
|
502
|
results_m1[1,j+3]<- AIC (mod)
|
503
|
|
504
|
results_m2[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
505
|
results_m2[1,2]<- ns #number of stations used in the training
|
506
|
results_m2[1,3]<- "GCV"
|
507
|
results_m2[1,j+3]<- mod$gcv.ubre
|
508
|
|
509
|
results_m3[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
510
|
results_m3[1,2]<- ns #number of stations used in the training stage
|
511
|
results_m3[1,3]<- "DEV"
|
512
|
results_m3[1,j+3]<- mod$deviance
|
513
|
|
514
|
##Model assessment: general diagnostic/metrics
|
515
|
##validation: using the testing data
|
516
|
if (predval==1) {
|
517
|
|
518
|
##Model assessment: specific diagnostic/metrics for GAM
|
519
|
|
520
|
name<-paste("mod",j,sep="") #modj is the name of The "j" model (mod1 if j=1)
|
521
|
mod<-get(name) #accessing GAM model ojbect "j"
|
522
|
|
523
|
s_sgdf<-as(s_raster,"SpatialGridDataFrame") #Conversion to spatial grid data frame
|
524
|
|
525
|
rpred<- predict(mod, newdata=s_sgdf, se.fit = TRUE) #Using the coeff to predict new values.
|
526
|
y_pred<-rpred$fit
|
527
|
raster_pred<-r1
|
528
|
layerNames(raster_pred)<-"y_pred"
|
529
|
values(raster_pred)<-as.numeric(y_pred)
|
530
|
data_name<-paste("predicted_mod",j,"_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],sep="")
|
531
|
raster_name<-paste("GAMCAI_",data_name,out_prefix,".rst", sep="")
|
532
|
writeRaster(raster_pred, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
533
|
#writeRaster(r2, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
534
|
|
535
|
tmax_predicted_CAI<-raster_pred + clim_rast #Final surface as a raster layer...taht is if daily prediction with GAM
|
536
|
if (climgam==1){
|
537
|
tmax_predicted_CAI<-raster_pred + daily_deltaclim_rast #Final surface as a raster layer...
|
538
|
}
|
539
|
|
540
|
layerNames(tmax_predicted_CAI)<-"y_pred"
|
541
|
data_name<-paste("predicted_mod",j,"_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],sep="")
|
542
|
raster_name<-paste("GAMCAI_tmax_predicted_",data_name,out_prefix,".rst", sep="")
|
543
|
writeRaster(tmax_predicted_CAI, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
544
|
#writeRaster(r2, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
545
|
|
546
|
pred_sgdf<-as(tmax_predicted_CAI,"SpatialGridDataFrame") #Conversion to spatial grid data frame
|
547
|
#rpred_val_s <- overlay(raster_pred,data_s) #This overlays the kriged surface tmax and the location of weather stations
|
548
|
|
549
|
rpred_val_s <- overlay(pred_sgdf,data_s) #This overlays the kriged surface tmax and the location of weather stations
|
550
|
rpred_val_v <- overlay(pred_sgdf,data_v) #This overlays the kriged surface tmax and the location of weather stations
|
551
|
|
552
|
pred_mod<-paste("pred_mod",j,sep="")
|
553
|
#Adding the results back into the original dataframes.
|
554
|
data_s[[pred_mod]]<-rpred_val_s$y_pred
|
555
|
data_v[[pred_mod]]<-rpred_val_v$y_pred
|
556
|
|
557
|
#Model assessment: RMSE and then krig the residuals....!
|
558
|
|
559
|
res_mod_s<- data_s$dailyTmax - data_s[[pred_mod]] #Residuals from kriging training
|
560
|
res_mod_v<- data_v$dailyTmax - data_v[[pred_mod]] #Residuals from kriging validation
|
561
|
|
562
|
}
|
563
|
|
564
|
if (predval==0) {
|
565
|
|
566
|
y_mod<- predict(mod, newdata=data_v, se.fit = TRUE) #Using the coeff to predict new values.
|
567
|
|
568
|
pred_mod<-paste("pred_mod",j,sep="")
|
569
|
#Adding the results back into the original dataframes.
|
570
|
data_s[[pred_mod]]<-as.numeric(mod$fit)
|
571
|
data_v[[pred_mod]]<-as.numeric(y_mod$fit)
|
572
|
|
573
|
#Model assessment: RMSE and then krig the residuals....!
|
574
|
#y_var_name<-"dailyTmax"
|
575
|
res_mod_s<- data_s$dailyTmax - data_s[[pred_mod]] #Residuals from kriging training
|
576
|
res_mod_v<- data_v$dailyTmax - data_v[[pred_mod]] #Residuals from kriging validation
|
577
|
}
|
578
|
|
579
|
#y_var_fit= mod$fit #move it
|
580
|
#Use res_mod_s so the R2 is based on daily station training
|
581
|
R2_mod_f<- cor(data_s$dailyTmax,res_mod_s, use="complete")^2
|
582
|
RMSE_mod_f<- sqrt(mean(res_mod_s^2,na.rm=TRUE))
|
583
|
|
584
|
results_RMSE_f[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
585
|
results_RMSE_f[1,2]<- ns #number of stations used in the training stage
|
586
|
results_RMSE_f[1,3]<- "RSME_f"
|
587
|
#results_RMSE_f[1,j+3]<-sqrt(mean(mod$residuals^2,na.rm=TRUE))
|
588
|
results_RMSE_f[1,j+3]<-sqrt(mean(res_mod_s^2,na.rm=TRUE))
|
589
|
|
590
|
results_MAE_f[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
591
|
results_MAE_f[1,2]<- ns #number of stations used in the training stage
|
592
|
results_MAE_f[1,3]<- "MAE_f"
|
593
|
#results_MAE_f[j+3]<-sum(abs(y_var_fit-data_s$y_var))/ns
|
594
|
results_MAE_f[1,j+3]<-mean(abs(res_mod_s),na.rm=TRUE)
|
595
|
|
596
|
results_R2_f[1,1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
597
|
results_R2_f[1,2]<- ns #number of stations used in the training stage
|
598
|
results_R2_f[1,3]<- "R2_f"
|
599
|
results_R2_f[1,j+3]<- R2_mod_f #Storing R2 for the model j
|
600
|
|
601
|
#### Now calculate validation metrics
|
602
|
res_mod<-res_mod_v
|
603
|
|
604
|
#RMSE_mod <- sqrt(sum(res_mod^2)/nv) #RMSE FOR REGRESSION STEP 1: GAM
|
605
|
RMSE_mod<- sqrt(mean(res_mod^2,na.rm=TRUE))
|
606
|
#MAE_mod<- sum(abs(res_mod),na.rm=TRUE)/(nv-sum(is.na(res_mod))) #MAE from kriged surface validation
|
607
|
MAE_mod<- mean(abs(res_mod), na.rm=TRUE)
|
608
|
#ME_mod<- sum(res_mod,na.rm=TRUE)/(nv-sum(is.na(res_mod))) #ME, Mean Error or bias FOR REGRESSION STEP 1: GAM
|
609
|
ME_mod<- mean(res_mod,na.rm=TRUE) #ME, Mean Error or bias FOR REGRESSION STEP 1: GAM
|
610
|
#R2_mod<- cor(data_v$y_var,data_v[[pred_mod]])^2 #R2, coef. of var FOR REGRESSION STEP 1: GAM
|
611
|
pred_mod<-paste("pred_mod",j,sep="")
|
612
|
R2_mod<- cor(data_v$dailyTmax,data_v[[pred_mod]], use="complete")^2
|
613
|
results_RMSE[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
614
|
results_RMSE[2]<- ns #number of stations used in the training stage
|
615
|
results_RMSE[3]<- "RMSE"
|
616
|
results_RMSE[j+3]<- RMSE_mod #Storing RMSE for the model j
|
617
|
results_MAE[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
618
|
results_MAE[2]<- ns #number of stations used in the training stage
|
619
|
results_MAE[3]<- "MAE"
|
620
|
results_MAE[j+3]<- MAE_mod #Storing MAE for the model j
|
621
|
results_ME[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
622
|
results_ME[2]<- ns #number of stations used in the training stage
|
623
|
results_ME[3]<- "ME"
|
624
|
results_ME[j+3]<- ME_mod #Storing ME for the model j
|
625
|
results_R2[1]<- sampling_dat$date[i] #storing the interpolation dates in the first column
|
626
|
results_R2[2]<- ns #number of stations used in the training stage
|
627
|
results_R2[3]<- "R2"
|
628
|
results_R2[j+3]<- R2_mod #Storing R2 for the model j
|
629
|
|
630
|
#Saving residuals and prediction in the dataframes: tmax predicted from GAM
|
631
|
|
632
|
name2<-paste("res_mod",j,sep="")
|
633
|
data_v[[name2]]<-as.numeric(res_mod_v)
|
634
|
data_s[[name2]]<-as.numeric(res_mod_s)
|
635
|
#end of loop calculating RMSE
|
636
|
}
|
637
|
}
|
638
|
|
639
|
#if (i==length(dates)){
|
640
|
|
641
|
|
642
|
#Specific diagnostic measures related to the testing datasets
|
643
|
|
644
|
results_table_RMSE<-as.data.frame(results_RMSE)
|
645
|
results_table_MAE<-as.data.frame(results_MAE)
|
646
|
results_table_ME<-as.data.frame(results_ME)
|
647
|
results_table_R2<-as.data.frame(results_R2)
|
648
|
results_table_RMSE_f<-as.data.frame(results_RMSE_f)
|
649
|
results_table_MAE_f<-as.data.frame(results_MAE_f)
|
650
|
results_table_R2_f<-as.data.frame(results_R2_f)
|
651
|
|
652
|
results_table_m1<-as.data.frame(results_m1)
|
653
|
results_table_m2<-as.data.frame(results_m2)
|
654
|
results_table_m3<-as.data.frame(results_m3)
|
655
|
|
656
|
tb_metrics1<-rbind(results_table_RMSE,results_table_MAE, results_table_ME,
|
657
|
results_table_R2,results_table_RMSE_f,results_table_MAE_f,results_table_R2_f) #
|
658
|
tb_metrics2<-rbind(results_table_m1,results_table_m2, results_table_m3)
|
659
|
|
660
|
#Preparing labels
|
661
|
mod_labels<-rep("mod",nmodels+1)
|
662
|
index<-as.character(1:(nmodels+1))
|
663
|
mod_labels<-paste(mod_labels,index,sep="")
|
664
|
cname<-c("dates","ns","metric", mod_labels)
|
665
|
#cname<-c("dates","ns","metric","mod1", "mod2","mod3", "mod4", "mod5", "mod6", "mod7","mod8","mod9")
|
666
|
colnames(tb_metrics1)<-cname
|
667
|
#cname<-c("dates","ns","metric","mod1", "mod2","mod3", "mod4", "mod5", "mod6", "mod7","mod8")
|
668
|
colnames(tb_metrics2)<-cname[1:(nmodels+3)]
|
669
|
|
670
|
print(paste(sampling_dat$date[i],"processed"))
|
671
|
# Kriging object may need to be modified...because it contains the full image of prediction!!
|
672
|
##loop through model objects data frame and set field to zero...
|
673
|
|
674
|
#mod_obj<-list(mod1,mod2,mod3,mod4,mod5,mod6,mod7,mod8,mod9a,mod9b)
|
675
|
#names(mod_obj)<-c("mod1","mod2","mod3","mod4","mod5","mod6","mod7","mod8","mod9a","mod9b") #generate names automatically??
|
676
|
mod_labels_kr<-c("mod_kr_month", "mod_kr_day")
|
677
|
names(mod_obj)<-c(mod_labels[1:nmodels],mod_labels_kr)
|
678
|
results_list<-list(data_s,data_v,tb_metrics1,tb_metrics2,mod_obj,data_month,list_formulas)
|
679
|
names(results_list)<-c("data_s","data_v","tb_metrics1","tb_metrics2","mod_obj","data_month","formulas")
|
680
|
save(results_list,file= paste(path,"/","results_list_metrics_objects_",sampling_dat$date[i],"_",sampling_dat$prop[i],"_",sampling_dat$run_samp[i],
|
681
|
out_prefix,".RData",sep=""))
|
682
|
return(results_list)
|
683
|
#return(tb_diagnostic1)
|
684
|
}
|