1 |
0163d0e2
|
Benoit Parmentier
|
#Function to be used with GAM_fusion_analysis_raster_prediction_mutlisampling.R
|
2 |
|
|
#runClimFusion<-function(r_stack,data_training,data_testing,data_training){
|
3 |
|
|
|
4 |
|
|
####
|
5 |
|
|
#TODO:
|
6 |
|
|
#Add log file and calculate time and sizes for processes-outputs
|
7 |
|
|
|
8 |
3e1b1ed4
|
Benoit Parmentier
|
runClim_KGFusion<-function(j){
|
9 |
0163d0e2
|
Benoit Parmentier
|
#Make this a function with multiple argument that can be used by mcmapply??
|
10 |
|
|
#This creates clim fusion layers...
|
11 |
|
|
|
12 |
|
|
#Functions used in the script
|
13 |
|
|
predict_raster_model<-function(in_models,r_stack,out_filename){
|
14 |
|
|
#This functions performs predictions on a raster grid given input models.
|
15 |
|
|
#Arguments: list of fitted models, raster stack of covariates
|
16 |
|
|
#Output: spatial grid data frame of the subset of tiles
|
17 |
|
|
#s_sgdf<-as(r_stack,"SpatialGridDataFrame") #Conversion to spatial grid data frame
|
18 |
|
|
list_rast_pred<-vector("list",length(in_models))
|
19 |
|
|
for (i in 1:length(in_models)){
|
20 |
|
|
mod <-in_models[[i]] #accessing GAM model ojbect "j"
|
21 |
|
|
raster_name<-out_filename[[i]]
|
22 |
|
|
if (inherits(mod,"gam")) {
|
23 |
|
|
#rpred<- predict(mod, newdata=s_sgdf, se.fit = TRUE) #Using the coeff to predict new values.
|
24 |
|
|
#rast_pred2<- predict(object=s_raster,model=mod,na.rm=TRUE) #Using the coeff to predict new values.
|
25 |
|
|
raster_pred<- predict(object=s_raster,model=mod,na.rm=FALSE) #Using the coeff to predict new values.
|
26 |
|
|
names(raster_pred)<-"y_pred"
|
27 |
|
|
writeRaster(raster_pred, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
28 |
|
|
print(paste("Interpolation:","mod", j ,sep=" "))
|
29 |
|
|
list_rast_pred[[i]]<-raster_name
|
30 |
|
|
}
|
31 |
|
|
}
|
32 |
|
|
if (inherits(mod,"try-error")) {
|
33 |
|
|
print(paste("no gam model fitted:",mod[1],sep=" "))
|
34 |
|
|
}
|
35 |
|
|
return(list_rast_pred)
|
36 |
|
|
}
|
37 |
|
|
|
38 |
|
|
fit_models<-function(list_formulas,data_training){
|
39 |
|
|
#This functions several models and returns model objects.
|
40 |
|
|
#Arguments: - list of formulas for GAM models
|
41 |
|
|
# - fitting data in a data.frame or SpatialPointDataFrame
|
42 |
|
|
#Output: list of model objects
|
43 |
|
|
list_fitted_models<-vector("list",length(list_formulas))
|
44 |
|
|
for (k in 1:length(list_formulas)){
|
45 |
|
|
formula<-list_formulas[[k]]
|
46 |
|
|
mod<- try(gam(formula, data=data_training))
|
47 |
|
|
model_name<-paste("mod",k,sep="")
|
48 |
|
|
assign(model_name,mod)
|
49 |
|
|
list_fitted_models[[k]]<-mod
|
50 |
|
|
}
|
51 |
|
|
return(list_fitted_models)
|
52 |
|
|
}
|
53 |
|
|
#Model and response variable can be changed without affecting the script
|
54 |
|
|
prop_month<-0 #proportion retained for validation
|
55 |
|
|
run_samp<-1
|
56 |
|
|
list_formulas<-vector("list",nmodels)
|
57 |
|
|
|
58 |
|
|
list_formulas[[1]] <- as.formula("y_var ~ s(elev_1)", env=.GlobalEnv)
|
59 |
|
|
list_formulas[[2]] <- as.formula("y_var ~ s(LST)", env=.GlobalEnv)
|
60 |
|
|
list_formulas[[3]] <- as.formula("y_var ~ s(elev_1,LST)", env=.GlobalEnv)
|
61 |
|
|
list_formulas[[4]] <- as.formula("y_var ~ s(lat) + s(lon)+ s(elev_1)", env=.GlobalEnv)
|
62 |
|
|
list_formulas[[5]] <- as.formula("y_var ~ s(lat,lon,elev_1)", env=.GlobalEnv)
|
63 |
|
|
list_formulas[[6]] <- as.formula("y_var ~ s(lat,lon) + s(elev_1) + s(N_w,E_w) + s(LST)", env=.GlobalEnv)
|
64 |
|
|
list_formulas[[7]] <- as.formula("y_var ~ s(lat,lon) + s(elev_1) + s(N_w,E_w) + s(LST) + s(LC2)", env=.GlobalEnv)
|
65 |
|
|
list_formulas[[8]] <- as.formula("y_var ~ s(lat,lon) + s(elev_1) + s(N_w,E_w) + s(LST) + s(LC6)", env=.GlobalEnv)
|
66 |
|
|
list_formulas[[9]] <- as.formula("y_var ~ s(lat,lon) + s(elev_1) + s(N_w,E_w) + s(LST) + s(DISTOC)", env=.GlobalEnv)
|
67 |
|
|
lst_avg<-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")
|
68 |
|
|
|
69 |
|
|
data_month<-dst[dst$month==j,] #Subsetting dataset for the relevant month of the date being processed
|
70 |
|
|
LST_name<-lst_avg[j] # name of LST month to be matched
|
71 |
|
|
data_month$LST<-data_month[[LST_name]]
|
72 |
|
|
|
73 |
|
|
#LST bias to model...
|
74 |
|
|
data_month$LSTD_bias<-data_month$LST-data_month$TMax
|
75 |
|
|
data_month$y_var<-data_month$LSTD_bias #Adding bias as the variable modeled
|
76 |
|
|
mod_list<-fit_models(list_formulas,data_month) #only gam at this stage
|
77 |
3e1b1ed4
|
Benoit Parmentier
|
cname<-paste("mod",1:length(mod_list),sep="") #change to more meaningful name?
|
78 |
0163d0e2
|
Benoit Parmentier
|
names(mod_list)<-cname
|
79 |
|
|
#Adding layer LST to the raster stack
|
80 |
3e1b1ed4
|
Benoit Parmentier
|
pos<-match("elev",names(s_raster))
|
81 |
0163d0e2
|
Benoit Parmentier
|
layerNames(s_raster)[pos]<-"elev_1"
|
82 |
|
|
|
83 |
|
|
pos<-match("LST",names(s_raster)) #Find the position of the layer with name "LST", if not present pos=NA
|
84 |
|
|
s_raster<-dropLayer(s_raster,pos) # If it exists drop layer
|
85 |
|
|
LST<-subset(s_raster,LST_name)
|
86 |
|
|
names(LST)<-"LST"
|
87 |
|
|
#Screen for extreme values": this needs more thought, min and max val vary with regions
|
88 |
|
|
#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)
|
89 |
|
|
#r1[r1 < (min_val)]<-NA
|
90 |
|
|
s_raster<-addLayer(s_raster,LST) #Adding current month
|
91 |
|
|
|
92 |
|
|
#Now generate file names for the predictions...
|
93 |
3e1b1ed4
|
Benoit Parmentier
|
list_out_filename<-vector("list",length(mod_list))
|
94 |
|
|
names(list_out_filename)<-cname
|
95 |
|
|
|
96 |
0163d0e2
|
Benoit Parmentier
|
for (k in 1:length(list_out_filename)){
|
97 |
3e1b1ed4
|
Benoit Parmentier
|
#j indicate which month is predicted
|
98 |
|
|
data_name<-paste("bias_LST_month_",j,"_",cname[k],"_",prop_month,
|
99 |
0163d0e2
|
Benoit Parmentier
|
"_",run_samp,sep="")
|
100 |
|
|
raster_name<-paste("fusion_",data_name,out_prefix,".tif", sep="")
|
101 |
|
|
list_out_filename[[k]]<-raster_name
|
102 |
|
|
}
|
103 |
|
|
|
104 |
|
|
#now predict values for raster image...
|
105 |
3e1b1ed4
|
Benoit Parmentier
|
rast_bias_list<-predict_raster_model(mod_list,s_raster,list_out_filename)
|
106 |
|
|
names(rast_bias_list)<-cname
|
107 |
|
|
#Some modles will not be predicted...remove them
|
108 |
|
|
rast_bias_list<-rast_bias_list[!sapply(rast_bias_list,is.null)] #remove NULL elements in list
|
109 |
|
|
|
110 |
|
|
mod_rast<-stack(rast_bias_list) #stack of bias raster images from models
|
111 |
0163d0e2
|
Benoit Parmentier
|
rast_clim_list<-vector("list",nlayers(mod_rast))
|
112 |
3e1b1ed4
|
Benoit Parmentier
|
names(rast_clim_list)<-names(rast_bias_list)
|
113 |
0163d0e2
|
Benoit Parmentier
|
for (k in 1:nlayers(mod_rast)){
|
114 |
|
|
clim_fus_rast<-LST-subset(mod_rast,k)
|
115 |
3e1b1ed4
|
Benoit Parmentier
|
data_name<-paste("clim_LST_month_",j,"_",names(rast_clim_list)[k],"_",prop_month,
|
116 |
0163d0e2
|
Benoit Parmentier
|
"_",run_samp,sep="")
|
117 |
|
|
raster_name<-paste("fusion_",data_name,out_prefix,".tif", sep="")
|
118 |
|
|
rast_clim_list[[k]]<-raster_name
|
119 |
|
|
writeRaster(clim_fus_rast, filename=raster_name,overwrite=TRUE) #Wri
|
120 |
|
|
}
|
121 |
3e1b1ed4
|
Benoit Parmentier
|
|
122 |
|
|
#Adding Kriging for Climatology options
|
123 |
|
|
|
124 |
|
|
bias_xy<-coordinates(data_month)
|
125 |
|
|
fitbias<-Krig(bias_xy,data_month$LSTD_bias,theta=1e5) #use TPS or krige
|
126 |
|
|
mod_krtmp1<-fitbias
|
127 |
|
|
model_name<-"mod_kr"
|
128 |
|
|
|
129 |
|
|
bias_rast<-interpolate(LST,fitbias) #interpolation using function from raster package
|
130 |
|
|
#Saving kriged surface in raster images
|
131 |
|
|
data_name<-paste("bias_LST_month_",j,"_",model_name,"_",prop_month,
|
132 |
|
|
"_",run_samp,sep="")
|
133 |
|
|
raster_name_bias<-paste("fusion_",data_name,out_prefix,".tif", sep="")
|
134 |
|
|
writeRaster(bias_rast, filename=raster_name_bias,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
135 |
|
|
|
136 |
|
|
#now climatology layer
|
137 |
|
|
clim_rast<-LST-bias_rast
|
138 |
|
|
data_name<-paste("clim_LST_month_",j,"_",model_name,"_",prop_month,
|
139 |
|
|
"_",run_samp,sep="")
|
140 |
|
|
raster_name_clim<-paste("fusion_",data_name,out_prefix,".tif", sep="")
|
141 |
|
|
writeRaster(clim_rast, filename=raster_name_clim,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
142 |
|
|
|
143 |
|
|
#Adding to current objects
|
144 |
|
|
mod_list[[model_name]]<-mod_krtmp1
|
145 |
|
|
rast_bias_list[[model_name]]<-raster_name_bias
|
146 |
|
|
rast_clim_list[[model_name]]<-raster_name_clim
|
147 |
|
|
|
148 |
|
|
#Prepare object to return
|
149 |
|
|
clim_obj<-list(rast_bias_list,rast_clim_list,data_month,mod_list,list_formulas)
|
150 |
0163d0e2
|
Benoit Parmentier
|
names(clim_obj)<-c("bias","clim","data_month","mod","formulas")
|
151 |
3e1b1ed4
|
Benoit Parmentier
|
|
152 |
|
|
save(clim_obj,file= paste("clim_obj_month_",j,"_",out_prefix,".RData",sep=""))
|
153 |
|
|
|
154 |
0163d0e2
|
Benoit Parmentier
|
return(clim_obj)
|
155 |
|
|
}
|
156 |
|
|
|
157 |
|
|
## Run function for kriging...?
|
158 |
|
|
|
159 |
4719fdd7
|
Benoit Parmentier
|
runGAMFusion <- function(i) { # loop over dates
|
160 |
3e1b1ed4
|
Benoit Parmentier
|
#Change this to allow explicitly arguments...
|
161 |
|
|
#Arguments:
|
162 |
|
|
#1)list of climatology files for all models...(12*nb of models)
|
163 |
|
|
#2)data_s:training
|
164 |
|
|
#3)data_v:testing
|
165 |
|
|
#4)list of dates??
|
166 |
|
|
#5)stack of covariates: not needed at this this stage
|
167 |
|
|
#6)dst: data at the monthly time scale
|
168 |
4719fdd7
|
Benoit Parmentier
|
|
169 |
0163d0e2
|
Benoit Parmentier
|
#Function used in the script
|
170 |
|
|
|
171 |
4719fdd7
|
Benoit Parmentier
|
date<-strptime(sampling_dat$date[i], "%Y%m%d") # interpolation date being processed
|
172 |
|
|
month<-strftime(date, "%m") # current month of the date being processed
|
173 |
|
|
LST_month<-paste("mm_",month,sep="") # name of LST month to be matched
|
174 |
0f4426cb
|
Benoit Parmentier
|
proj_str<-proj4string(dst)
|
175 |
d93bea39
|
Benoit Parmentier
|
|
176 |
4719fdd7
|
Benoit Parmentier
|
###Regression part 1: Creating a validation dataset by creating training and testing datasets
|
177 |
0f4426cb
|
Benoit Parmentier
|
data_day<-ghcn.subsets[[i]]
|
178 |
|
|
mod_LST <- ghcn.subsets[[i]][,match(LST_month, names(ghcn.subsets[[i]]))] #Match interpolation date and monthly LST average
|
179 |
|
|
data_day$LST <- as.data.frame(mod_LST)[,1] #Add the variable LST to the dataset
|
180 |
|
|
dst$LST<-dst[[LST_month]] #Add the variable LST to the monthly dataset
|
181 |
|
|
|
182 |
4719fdd7
|
Benoit Parmentier
|
ind.training<-sampling[[i]]
|
183 |
0f4426cb
|
Benoit Parmentier
|
ind.testing <- setdiff(1:nrow(data_day), ind.training)
|
184 |
|
|
data_s <- data_day[ind.training, ] #Training dataset currently used in the modeling
|
185 |
|
|
data_v <- data_day[ind.testing, ] #Testing/validation dataset using input sampling
|
186 |
4719fdd7
|
Benoit Parmentier
|
|
187 |
|
|
ns<-nrow(data_s)
|
188 |
|
|
nv<-nrow(data_v)
|
189 |
|
|
#i=1
|
190 |
|
|
date_proc<-sampling_dat$date[i]
|
191 |
|
|
date_proc<-strptime(sampling_dat$date[i], "%Y%m%d") # interpolation date being processed
|
192 |
|
|
mo<-as.integer(strftime(date_proc, "%m")) # current month of the date being processed
|
193 |
|
|
day<-as.integer(strftime(date_proc, "%d"))
|
194 |
|
|
year<-as.integer(strftime(date_proc, "%Y"))
|
195 |
|
|
|
196 |
55056785
|
Benoit Parmentier
|
modst<-dst[dst$month==mo,] #Subsetting dataset for the relevant month of the date being processed
|
197 |
3e1b1ed4
|
Benoit Parmentier
|
#Change to y_var...could be TMin
|
198 |
|
|
#modst$LSTD_bias <- modst$LST-modst$y_var
|
199 |
|
|
modst$LSTD_bias <- modst$LST-modst$TMax; #That is the difference between the monthly LST mean and monthly station mean
|
200 |
0163d0e2
|
Benoit Parmentier
|
|
201 |
0f4426cb
|
Benoit Parmentier
|
x<-as.data.frame(data_v)
|
202 |
|
|
d<-as.data.frame(data_s)
|
203 |
|
|
#x[x$value==-999.9]<-NA
|
204 |
|
|
for (j in 1:nrow(x)){
|
205 |
|
|
if (x$value[j]== -999.9){
|
206 |
|
|
x$value[j]<-NA
|
207 |
|
|
}
|
208 |
|
|
}
|
209 |
|
|
for (j in 1:nrow(d)){
|
210 |
|
|
if (d$value[j]== -999.9){
|
211 |
|
|
d$value[j]<-NA
|
212 |
|
|
}
|
213 |
|
|
}
|
214 |
|
|
#x[x$value==-999.9]<-NA
|
215 |
|
|
#d[d$value==-999.9]<-NA
|
216 |
4719fdd7
|
Benoit Parmentier
|
pos<-match("value",names(d)) #Find column with name "value"
|
217 |
|
|
#names(d)[pos]<-c("dailyTmax")
|
218 |
|
|
names(d)[pos]<-y_var_name
|
219 |
|
|
names(x)[pos]<-y_var_name
|
220 |
|
|
#names(x)[pos]<-c("dailyTmax")
|
221 |
|
|
pos<-match("station",names(d)) #Find column with name "value"
|
222 |
|
|
names(d)[pos]<-c("id")
|
223 |
|
|
names(x)[pos]<-c("id")
|
224 |
|
|
names(modst)[1]<-c("id") #modst contains the average tmax per month for every stations...
|
225 |
55056785
|
Benoit Parmentier
|
|
226 |
3e1b1ed4
|
Benoit Parmentier
|
dmoday <-merge(modst,d,by="id",suffixes=c("",".y2")) #LOOSING DATA HERE!!! from 113 t0 103
|
227 |
|
|
xmoday <-merge(modst,x,by="id",suffixes=c("",".y2")) #LOOSING DATA HERE!!! from 48 t0 43
|
228 |
55056785
|
Benoit Parmentier
|
mod_pat<-glob2rx("*.y2")
|
229 |
|
|
var_pat<-grep(mod_pat,names(dmoday),value=FALSE) # using grep with "value" extracts the matching names
|
230 |
|
|
dmoday<-dmoday[,-var_pat]
|
231 |
|
|
mod_pat<-glob2rx("*.y2")
|
232 |
|
|
var_pat<-grep(mod_pat,names(xmoday),value=FALSE) # using grep with "value" extracts the matching names
|
233 |
|
|
xmoday<-xmoday[,-var_pat] #Removing duplicate columns
|
234 |
4719fdd7
|
Benoit Parmentier
|
|
235 |
|
|
data_v<-xmoday
|
236 |
|
|
|
237 |
|
|
#dmoday contains the daily tmax values for training with TMax being the monthly station tmax mean
|
238 |
|
|
#xmoday contains the daily tmax values for validation with TMax being the monthly station tmax mean
|
239 |
|
|
|
240 |
|
|
##########
|
241 |
|
|
# STEP 7 - interpolate delta across space
|
242 |
|
|
##########
|
243 |
|
|
|
244 |
3e1b1ed4
|
Benoit Parmentier
|
daily_delta<-dmoday$dailyTmax-dmoday$TMax
|
245 |
|
|
daily_delta_xy<-as.matrix(cbind(dmoday$x,dmoday$y))
|
246 |
|
|
fitdelta<-Krig(daily_delta_xy,daily_delta,theta=1e5) #use TPS or krige
|
247 |
d93bea39
|
Benoit Parmentier
|
mod_krtmp2<-fitdelta
|
248 |
|
|
model_name<-paste("mod_kr","day",sep="_")
|
249 |
4719fdd7
|
Benoit Parmentier
|
data_s<-dmoday #put the
|
250 |
|
|
data_s$daily_delta<-daily_delta
|
251 |
|
|
|
252 |
|
|
#########
|
253 |
|
|
# STEP 8 - assemble final answer - T=LST+Bias(interpolated)+delta(interpolated)
|
254 |
|
|
#########
|
255 |
|
|
|
256 |
3e1b1ed4
|
Benoit Parmentier
|
rast_clim_list<-rast_clim_yearlist[[mo]] #select relevant month
|
257 |
|
|
rast_clim_month<-raster(rast_clim_list[[1]])
|
258 |
4719fdd7
|
Benoit Parmentier
|
|
259 |
3e1b1ed4
|
Benoit Parmentier
|
daily_delta_rast<-interpolate(rast_clim_month,fitdelta) #Interpolation of the bias surface...
|
260 |
4719fdd7
|
Benoit Parmentier
|
|
261 |
|
|
#Saving kriged surface in raster images
|
262 |
|
|
data_name<-paste("daily_delta_",sampling_dat$date[i],"_",sampling_dat$prop[i],
|
263 |
|
|
"_",sampling_dat$run_samp[i],sep="")
|
264 |
3e1b1ed4
|
Benoit Parmentier
|
raster_name_delta<-paste("fusion_",data_name,out_prefix,".tif", sep="")
|
265 |
|
|
writeRaster(daily_delta_rast, filename=raster_name_delta,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
266 |
|
|
|
267 |
|
|
#Now predict daily after having selected the relevant month
|
268 |
|
|
temp_list<-vector("list",length(rast_clim_list))
|
269 |
|
|
for (j in 1:length(rast_clim_list)){
|
270 |
|
|
rast_clim_month<-raster(rast_clim_list[[j]])
|
271 |
|
|
temp_predicted<-rast_clim_month+daily_delta_rast
|
272 |
4719fdd7
|
Benoit Parmentier
|
|
273 |
3e1b1ed4
|
Benoit Parmentier
|
data_name<-paste(y_var_name,"_predicted_",names(rast_clim_list)[j],"_",
|
274 |
|
|
sampling_dat$date[i],"_",sampling_dat$prop[i],
|
275 |
|
|
"_",sampling_dat$run_samp[i],sep="")
|
276 |
|
|
raster_name<-paste("fusion_",data_name,out_prefix,".tif", sep="")
|
277 |
|
|
writeRaster(temp_predicted, filename=raster_name,overwrite=TRUE)
|
278 |
|
|
temp_list[[j]]<-raster_name
|
279 |
4719fdd7
|
Benoit Parmentier
|
}
|
280 |
|
|
|
281 |
3e1b1ed4
|
Benoit Parmentier
|
mod_krtmp2<-fitdelta
|
282 |
|
|
model_name<-paste("mod_kr","day",sep="_")
|
283 |
|
|
names(tmp_list)<-names(rast_clim_list)
|
284 |
|
|
coordinates(data_s)<-cbind(data_s$x,data_s$y)
|
285 |
|
|
proj4string(data_s)<-proj_str
|
286 |
|
|
coordinates(data_v)<-cbind(data_v$x,data_v$y)
|
287 |
|
|
proj4string(data_v)<-proj_str
|
288 |
|
|
|
289 |
|
|
delta_obj<-list(temp_list,rast_clim_list,raster_name_delta,data_s,
|
290 |
|
|
data_v,sampling_dat[i,],mod_krtmp2)
|
291 |
|
|
|
292 |
|
|
obj_names<-c(y_var_name,"clim","delta","data_s","data_v",
|
293 |
|
|
"sampling_dat",model_name)
|
294 |
|
|
names(delta_obj)<-obj_names
|
295 |
|
|
save(delta_obj,file= paste("delta_obj_",sampling_dat$date[i],"_",sampling_dat$prop[i],
|
296 |
4719fdd7
|
Benoit Parmentier
|
"_",sampling_dat$run_samp[i],out_prefix,".RData",sep=""))
|
297 |
3e1b1ed4
|
Benoit Parmentier
|
return(delta_obj)
|
298 |
|
|
|
299 |
|
|
}
|
300 |
|
|
|