1
|
runKriging <- function(i) { # loop over dates
|
2
|
|
3
|
date<-strptime(dates[i], "%Y%m%d") # interpolation date being processed
|
4
|
month<-strftime(date, "%m") # current month of the date being processed
|
5
|
LST_month<-paste("mm_",month,sep="") # name of LST month to be matched
|
6
|
|
7
|
#i=1
|
8
|
date_proc<-dates[i]
|
9
|
date_proc<-strptime(dates[i], "%Y%m%d") # interpolation date being processed
|
10
|
mo<-as.integer(strftime(date_proc, "%m")) # current month of the date being processed
|
11
|
day<-as.integer(strftime(date_proc, "%d"))
|
12
|
year<-as.integer(strftime(date_proc, "%Y"))
|
13
|
|
14
|
|
15
|
#Adding layer LST to the raster stack
|
16
|
|
17
|
pos<-match(LST_month,layerNames(s_raster)) #Find column with the current month for instance mm12
|
18
|
r1<-raster(s_raster,layer=pos) #Select layer from stack
|
19
|
layerNames(r1)<-"LST"
|
20
|
s_raster<-addLayer(s_raster,r1) #Adding current month
|
21
|
s_sgdf<-as(s_raster,"SpatialGridDataFrame") #Conversion to spatial grid data frame
|
22
|
|
23
|
###Regression part 1: Creating a validation dataset by creating training and testing datasets
|
24
|
|
25
|
mod_LST <-ghcn.subsets[[i]][,match(LST_month, names(ghcn.subsets[[i]]))] #Match interpolation date and monthly LST average
|
26
|
ghcn.subsets[[i]] = transform(ghcn.subsets[[i]],LST = mod_LST) #Add the variable LST to the subset dataset
|
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
|
|
39
|
###BEFORE model prediction the data object must be transformed to SDF
|
40
|
|
41
|
coords<- data_v[,c('x_OR83M','y_OR83M')]
|
42
|
coordinates(data_v)<-coords
|
43
|
proj4string(data_v)<-CRS #Need to assign coordinates...
|
44
|
coords<- data_s[,c('x_OR83M','y_OR83M')]
|
45
|
coordinates(data_s)<-coords
|
46
|
proj4string(data_s)<-CRS #Need to assign coordinates..
|
47
|
|
48
|
ns<-nrow(data_s) #This is added to because some loss of data might have happened because of the averaging...
|
49
|
nv<-nrow(data_v)
|
50
|
|
51
|
### PREDICTION/ Interpolation
|
52
|
|
53
|
pos<-match("value",names(data_s)) #Find column with name "value"
|
54
|
names(data_s)[pos]<-y_var_name
|
55
|
pos<-match("value",names(data_v)) #Find column with name "value"
|
56
|
names(data_v)[pos]<-y_var_name
|
57
|
|
58
|
#if y_var_name=="dailyTmax"
|
59
|
data_v$y_var<-data_v[[y_var_name]]/10 #Note that values are divided by 10 because the var is temp
|
60
|
data_s$y_var<-data_s[[y_var_name]]/10
|
61
|
|
62
|
#Model and response variable can be changed without affecting the script
|
63
|
|
64
|
formula1 <- as.formula("y_var ~1", env=.GlobalEnv)
|
65
|
formula2 <- as.formula("y_var~ x_OR83M+y_OR83M", env=.GlobalEnv)
|
66
|
formula3 <- as.formula("y_var~ x_OR83M+y_OR83M+ELEV_SRTM", env=.GlobalEnv)
|
67
|
formula4 <- as.formula("y_var~ x_OR83M+y_OR83M+DISTOC", env=.GlobalEnv)
|
68
|
formula5 <- as.formula("y_var~ x_OR83M+y_OR83M+ELEV_SRTM+DISTOC", env=.GlobalEnv)
|
69
|
formula6 <- as.formula("y_var~ x_OR83M+y_OR83M+Northness+Eastness", env=.GlobalEnv)
|
70
|
formula7 <- as.formula("y_var~ LST", env=.GlobalEnv)
|
71
|
formula8 <- as.formula("y_var~ x_OR83M+y_OR83M+LST", env=.GlobalEnv)
|
72
|
formula9 <- as.formula("y_var~ x_OR83M+y_OR83M+ELEV_SRTM+LST", env=.GlobalEnv)
|
73
|
|
74
|
mod1<- try(autoKrige(formula1, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
75
|
mod2<- try(autoKrige(formula2, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
76
|
mod3<- try(autoKrige(formula3, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
77
|
mod4<- try(autoKrige(formula4, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
78
|
mod5<- try(autoKrige(formula5, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
79
|
mod6<- try(autoKrige(formula6, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
80
|
mod7<- try(autoKrige(formula7, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
81
|
mod8<- try(autoKrige(formula8, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
82
|
mod9<- try(autoKrige(formula9, input_data=data_s,new_data=s_sgdf,data_variogram=data_s))
|
83
|
|
84
|
#tmax_predicted=themolst+daily_delta_rast-bias_rast #Final surface?? but daily_rst
|
85
|
|
86
|
### Model assessment
|
87
|
|
88
|
for (j in 1:nmodels){
|
89
|
|
90
|
##Model assessment: specific diagnostic/metrics for GAM
|
91
|
|
92
|
name<-paste("mod",j,sep="") #modj is the name of The "j" model (mod1 if j=1)
|
93
|
mod<-get(name) #accessing GAM model ojbect "j"
|
94
|
#krmod_auto<-get(mod)
|
95
|
|
96
|
#If mod "j" is not a model object
|
97
|
if (inherits(mod,"try-error")) {
|
98
|
|
99
|
results_m1[1,1]<- dates[i] #storing the interpolation dates in the first column
|
100
|
results_m1[1,2]<- ns #number of stations used in the training stage
|
101
|
results_m1[1,3]<- "SSERR"
|
102
|
results_m1[1,j+3]<- NA
|
103
|
|
104
|
results_m2[1,1]<- dates[i] #storing the interpolation dates in the first column
|
105
|
results_m2[1,2]<- ns #number of stations used in the training
|
106
|
results_m2[1,3]<- "GCV"
|
107
|
results_m2[1,j+3]<- NA
|
108
|
|
109
|
results_m3[1,1]<- dates[i] #storing the interpolation dates in the first column
|
110
|
results_m3[1,2]<- ns #number of stations used in the training stage
|
111
|
results_m3[1,3]<- "DEV"
|
112
|
results_m3[1,j+3]<- NA
|
113
|
|
114
|
results_RMSE_f[1,1]<- dates[i] #storing the interpolation dates in the first column
|
115
|
results_RMSE_f[1,2]<- ns #number of stations used in the training stage
|
116
|
results_RMSE_f[1,3]<- "RSME_f"
|
117
|
results_RMSE_f[1,j+3]<- NA
|
118
|
|
119
|
results_MAE_f[1,1]<- dates[i] #storing the interpolation dates in the first column
|
120
|
results_MAE_f[1,2]<- ns #number of stations used in the training stage
|
121
|
results_MAE_f[1,3]<- "MAE_f"
|
122
|
results_MAE_f[1,j+3]<-NA
|
123
|
|
124
|
results_RMSE[1,1]<- dates[i] #storing the interpolation dates in the first column
|
125
|
results_RMSE[1,2]<- ns #number of stations used in the training stage
|
126
|
results_RMSE[1,3]<- "RMSE"
|
127
|
results_RMSE[1,j+3]<- NA #Storing RMSE for the model j
|
128
|
results_MAE[1,1]<- dates[i] #storing the interpolation dates in the first column
|
129
|
results_MAE[1,2]<- ns #number of stations used in the training stage
|
130
|
results_MAE[1,3]<- "MAE"
|
131
|
results_MAE[1,j+3]<- NA #Storing MAE for the model j
|
132
|
results_ME[1,1]<- dates[i] #storing the interpolation dates in the first column
|
133
|
results_ME[1,2]<- ns #number of stations used in the training stage
|
134
|
results_ME[1,3]<- "ME"
|
135
|
results_ME[1,j+3]<- NA #Storing ME for the model j
|
136
|
results_R2[1,1]<- dates[i] #storing the interpolation dates in the first column
|
137
|
results_R2[1,2]<- ns #number of stations used in the training stage
|
138
|
results_R2[1,3]<- "R2"
|
139
|
results_R2[1,j+3]<- NA #Storing R2 for the model j
|
140
|
|
141
|
}
|
142
|
|
143
|
#If mod is a modelobject
|
144
|
|
145
|
#If mod "j" is not a model object
|
146
|
if (inherits(mod,"autoKrige")) {
|
147
|
|
148
|
rpred<-mod$krige_output #Extracting the SptialGriDataFrame from the autokrige object
|
149
|
|
150
|
#rpred<- predict(mod, newdata=s_sgdf, se.fit = TRUE) #Using the coeff to predict new values.
|
151
|
y_pred<-rpred$var1.pred #is the order the same?
|
152
|
#y_prederr<-rpred$var1.var
|
153
|
raster_pred<-r1
|
154
|
layerNames(raster_pred)<-"y_pred"
|
155
|
clearValues(raster_pred) #Clear values in memory, just in case...
|
156
|
values(raster_pred)<-as.numeric(y_pred) #Assign values to every pixels
|
157
|
|
158
|
data_name<-paste("predicted_mod",j,"_",dates[[i]],sep="")
|
159
|
raster_name<-paste("Kriging_",data_name,out_prefix,".rst", sep="")
|
160
|
writeRaster(raster_pred, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
161
|
#writeRaster(r2, filename=raster_name,overwrite=TRUE) #Writing the data in a raster file format...(IDRISI)
|
162
|
|
163
|
#Save png plot here...
|
164
|
data_name<-paste("predicted_mod",j,"_",dates[[i]],sep="")
|
165
|
png_name<-paste("Kriging_plot_",data_name,out_prefix,".png", sep="")
|
166
|
png(png_name) #Create file to write a plot
|
167
|
#datelabel2=format(ISOdate(year,mo,day),"%B ") #Plot label
|
168
|
plot(mod) #Plot to file the autokrige object
|
169
|
#savePlot(paste("Bias_surface_LST_TMax_",dates[i],out_prefix,".png", sep=""), type="png")
|
170
|
dev.off() #Release the hold to the file
|
171
|
|
172
|
pred_sgdf<-as(raster_pred,"SpatialGridDataFrame" ) #Conversion to spatial grid data frame
|
173
|
#rpred_val_s <- overlay(raster_pred,data_s) #This overlays the kriged surface tmax and the location of weather stations
|
174
|
|
175
|
rpred_val_s <- overlay(pred_sgdf,data_s) #This overlays the kriged surface tmax and the location of weather stations
|
176
|
rpred_val_v <- overlay(pred_sgdf,data_v) #This overlays the kriged surface tmax and the location of weather stations
|
177
|
|
178
|
pred_mod<-paste("pred_mod",j,sep="")
|
179
|
#Adding the results back into the original dataframes.
|
180
|
data_s[[pred_mod]]<-rpred_val_s$y_pred
|
181
|
data_v[[pred_mod]]<-rpred_val_v$y_pred
|
182
|
|
183
|
#Model assessment: RMSE and then krig the residuals....!
|
184
|
|
185
|
res_mod_s<- data_s$y_var - data_s[[pred_mod]] #Residuals from kriging training
|
186
|
res_mod_v<- data_v$y_var - data_v[[pred_mod]] #Residuals from kriging validation
|
187
|
|
188
|
####ADDED ON JULY 20th
|
189
|
res_mod<-res_mod_v
|
190
|
|
191
|
#RMSE_mod <- sqrt(sum(res_mod^2)/nv) #RMSE FOR REGRESSION STEP 1: GAM
|
192
|
RMSE_mod<- sqrt(mean(res_mod^2,na.rm=TRUE))
|
193
|
#MAE_mod<- sum(abs(res_mod),na.rm=TRUE)/(nv-sum(is.na(res_mod))) #MAE from kriged surface validation
|
194
|
MAE_mod<- mean(abs(res_mod), na.rm=TRUE)
|
195
|
#ME_mod<- sum(res_mod,na.rm=TRUE)/(nv-sum(is.na(res_mod))) #ME, Mean Error or bias FOR REGRESSION STEP 1: GAM
|
196
|
ME_mod<- mean(res_mod,na.rm=TRUE) #ME, Mean Error or bias FOR REGRESSION STEP 1: GAM
|
197
|
#R2_mod<- cor(data_v$y_var,data_v[[pred_mod]])^2 #R2, coef. of var FOR REGRESSION STEP 1: GAM
|
198
|
R2_mod<- cor(data_v$y_var,data_v[[pred_mod]], use="complete")^2
|
199
|
|
200
|
R2_mod_f<- cor(data_s$y_var,data_s[[pred_mod]], use="complete")^2
|
201
|
RMSE_mod_f<- sqrt(mean(res_mod_s^2,na.rm=TRUE))
|
202
|
#MAE_mod<- sum(abs(res_mod),na.rm=TRUE)/(nv-sum(is.na(res_mod))) #MAE from kriged surface validation
|
203
|
MAE_mod_f<- mean(abs(res_mod_s), na.rm=TRUE)
|
204
|
|
205
|
results_m1[1,1]<- dates[i] #storing the interpolation dates in the first column
|
206
|
results_m1[1,2]<- ns #number of stations used in the training stage
|
207
|
results_m1[1,3]<- "SSERR"
|
208
|
results_m1[1,j+3]<- mod$sserr
|
209
|
|
210
|
results_m2[1,1]<- dates[i] #storing the interpolation dates in the first column
|
211
|
results_m2[1,2]<- ns #number of stations used in the training
|
212
|
results_m2[1,3]<- "GCV"
|
213
|
results_m2[1,j+3]<- NA
|
214
|
|
215
|
results_m3[1,1]<- dates[i] #storing the interpolation dates in the first column
|
216
|
results_m3[1,2]<- ns #number of stations used in the training stage
|
217
|
results_m3[1,3]<- "DEV"
|
218
|
results_m3[1,j+3]<- NA
|
219
|
|
220
|
results_RMSE_f[1,1]<- dates[i] #storing the interpolation dates in the first column
|
221
|
results_RMSE_f[1,2]<- ns #number of stations used in the training stage
|
222
|
results_RMSE_f[1,3]<- "RSME_f"
|
223
|
results_RMSE_f[1,j+3]<-RMSE_mod_f
|
224
|
|
225
|
results_MAE_f[1,1]<- dates[i] #storing the interpolation dates in the first column
|
226
|
results_MAE_f[1,2]<- ns #number of stations used in the training stage
|
227
|
results_MAE_f[1,3]<- "MAE_f"
|
228
|
results_MAE_f[1,j+3]<-MAE_mod_f
|
229
|
|
230
|
results_R2_f[1,1]<- dates[i] #storing the interpolation dates in the first column
|
231
|
results_R2_f[1,2]<- ns #number of stations used in the training stage
|
232
|
results_R2_f[1,3]<- "R2_f"
|
233
|
results_R2_f[1,j+3]<- R2_mod_f #Storing R2 for the model j
|
234
|
|
235
|
results_RMSE[1,1]<- dates[i] #storing the interpolation dates in the first column
|
236
|
results_RMSE[1,2]<- ns #number of stations used in the training stage
|
237
|
results_RMSE[1,3]<- "RMSE"
|
238
|
results_RMSE[1,j+3]<- RMSE_mod #Storing RMSE for the model j
|
239
|
results_MAE[1,1]<- dates[i] #storing the interpolation dates in the first column
|
240
|
results_MAE[1,2]<- ns #number of stations used in the training stage
|
241
|
results_MAE[1,3]<- "MAE"
|
242
|
results_MAE[1,j+3]<- MAE_mod #Storing MAE for the model j
|
243
|
results_ME[1,1]<- dates[i] #storing the interpolation dates in the first column
|
244
|
results_ME[1,2]<- ns #number of stations used in the training stage
|
245
|
results_ME[1,3]<- "ME"
|
246
|
results_ME[1,j+3]<- ME_mod #Storing ME for the model j
|
247
|
results_R2[1,1]<- dates[i] #storing the interpolation dates in the first column
|
248
|
results_R2[1,2]<- ns #number of stations used in the training stage
|
249
|
results_R2[1,3]<- "R2"
|
250
|
results_R2[1,j+3]<- R2_mod #Storing R2 for the model j
|
251
|
|
252
|
#Saving residuals and prediction in the dataframes: tmax predicted from GAM
|
253
|
|
254
|
name2<-paste("res_mod",j,sep="")
|
255
|
data_v[[name2]]<-as.numeric(res_mod_v)
|
256
|
data_s[[name2]]<-as.numeric(res_mod_s)
|
257
|
#end of loop calculating RMSE
|
258
|
}
|
259
|
}
|
260
|
|
261
|
#if (i==length(dates)){
|
262
|
|
263
|
#Specific diagnostic measures related to the testing datasets
|
264
|
|
265
|
results_table_RMSE<-as.data.frame(results_RMSE)
|
266
|
results_table_MAE<-as.data.frame(results_MAE)
|
267
|
results_table_ME<-as.data.frame(results_ME)
|
268
|
results_table_R2<-as.data.frame(results_R2)
|
269
|
results_table_RMSE_f<-as.data.frame(results_RMSE_f)
|
270
|
results_table_MAE_f<-as.data.frame(results_MAE_f)
|
271
|
results_table_R2_f<-as.data.frame(results_R2_f)
|
272
|
|
273
|
results_table_m1<-as.data.frame(results_m1)
|
274
|
results_table_m2<-as.data.frame(results_m2)
|
275
|
results_table_m3<-as.data.frame(results_m3)
|
276
|
|
277
|
tb_metrics1<-rbind(results_table_RMSE,results_table_MAE, results_table_ME,
|
278
|
results_table_R2,results_table_RMSE_f,results_table_MAE_f,results_table_R2_f) #
|
279
|
tb_metrics2<-rbind(results_table_m1,results_table_m2, results_table_m3)
|
280
|
cname<-c("dates","ns","metric","mod1", "mod2","mod3", "mod4", "mod5", "mod6", "mod7","mod8","mod9")
|
281
|
colnames(tb_metrics1)<-cname
|
282
|
cname<-c("dates","ns","metric","mod1", "mod2","mod3", "mod4", "mod5", "mod6", "mod7","mod8","mod9")
|
283
|
colnames(tb_metrics2)<-cname
|
284
|
#colnames(results_table_RMSE)<-cname
|
285
|
#colnames(results_table_RMSE_f)<-cname
|
286
|
#tb_diagnostic1<-results_table_RMSE #measures of validation
|
287
|
#tb_diagnostic2<-results_table_RMSE_f #measures of fit
|
288
|
|
289
|
#write.table(tb_diagnostic1, file= paste(path,"/","results_fusion_Assessment_measure1",out_prefix,".txt",sep=""), sep=",")
|
290
|
|
291
|
#}
|
292
|
print(paste(dates[i],"processed"))
|
293
|
# Kriging object may need to be modified...because it contains the full image of prediction!!
|
294
|
##loop through model objects data frame and set field to zero...
|
295
|
|
296
|
mod_obj<-list(mod1,mod2,mod3,mod4,mod5,mod6,mod7,mod8,mod9)
|
297
|
names(mod_obj)<-c("mod1","mod2","mod3","mod4","mod5","mod6","mod7","mod8","mod9") #generate names automatically??
|
298
|
#results_list<-list(data_s,data_v,tb_metrics1,tb_metrics2)
|
299
|
#save(mod_obj,file= paste(path,"/","results_list_mod_objects_",dates[i],out_prefix,".RData",sep=""))
|
300
|
|
301
|
for (j in 1:nmodels){
|
302
|
if (inherits(mod_obj[[j]],"autoKrige")){
|
303
|
mod_obj[[j]]$krige_output<-NULL
|
304
|
}
|
305
|
}
|
306
|
results_list<-list(data_s,data_v,tb_metrics1,tb_metrics2,mod_obj)
|
307
|
names(results_list)<-c("data_s","data_v","tb_metrics1","tb_metrics2","mod_obj")
|
308
|
save(results_list,file= paste(path,"/","results_list_metrics_objects_",dates[i],out_prefix,".RData",sep=""))
|
309
|
return(results_list)
|
310
|
#return(tb_diagnostic1)
|
311
|
}
|