Revision 1c15fc49
Added by Adam M. Wilson over 10 years ago
climate/research/oregon/interpolation/kriging_reg.R | ||
---|---|---|
1 |
################## Interpolation of Tmax Using Kriging ####################################### |
|
2 |
########################### Kriging and Cokriging ############################################### |
|
3 |
#This script interpolates station values for the Oregon case study using Kriging and Cokring. # |
|
4 |
#The script uses LST monthly averages as input variables and loads the station data # |
|
5 |
#from a shape file with projection information. # |
|
6 |
#Note that this program: # |
|
7 |
#1)assumes that the shape file is in the current working. # |
|
8 |
#2)relevant variables were extracted from raster images before performing the regressions # |
|
9 |
# and stored shapefile # |
|
10 |
#This scripts predicts tmax using autokrige, gstat and LST derived from MOD11A1. # |
|
11 |
#also included and assessed using the RMSE,MAE,ME and R2 from validation dataset. # |
|
12 |
#TThe dates must be provided as a textfile. # |
|
13 |
#AUTHOR: Benoit Parmentier # |
|
14 |
#DATE: 07/15/2012 # |
|
15 |
#PROJECT: NCEAS INPLANT: Environment and Organisms --TASK#364-- # |
|
16 |
################################################################################################## |
|
17 |
|
|
18 |
###Loading R library and packages |
|
19 |
#library(gtools) # loading some useful tools |
|
20 |
library(mgcv) # GAM package by Wood 2006 (version 2012) |
|
21 |
library(sp) # Spatial pacakge with class definition by Bivand et al. 2008 |
|
22 |
library(spdep) # Spatial pacakge with methods and spatial stat. by Bivand et al. 2012 |
|
23 |
library(rgdal) # GDAL wrapper for R, spatial utilities (Keitt et al. 2012) |
|
24 |
library(gstat) # Kriging and co-kriging by Pebesma et al. 2004 |
|
25 |
library(automap) # Automated Kriging based on gstat module by Hiemstra et al. 2008 |
|
1 |
####################GWR of Tmax for one Date##################### |
|
2 |
#This script generates predicted values from station values for the Oregon case study. This program loads the station data from a shp file |
|
3 |
#and performs Kriging and co-kriging on tmax regression. |
|
4 |
#Script created by Benoit Parmentier on April 17, 2012. |
|
5 |
|
|
6 |
###Loading r library and packages |
|
7 |
library(sp) |
|
8 |
library(spdep) |
|
9 |
library(rgdal) |
|
26 | 10 |
library(spgwr) |
27 | 11 |
library(gpclib) |
28 | 12 |
library(maptools) |
13 |
library(gstat) |
|
29 | 14 |
library(graphics) |
30 |
|
|
31 | 15 |
###Parameters and arguments |
32 | 16 |
|
33 |
infile1<- "ghcn_or_tmax_covariates_06262012_OR83M.shp" #GHCN shapefile containing variables for modeling 2010 |
|
34 |
infile2<-"list_10_dates_04212012.txt" #List of 10 dates for the regression |
|
35 |
#infile2<-"list_365_dates_04212012.txt" |
|
36 |
infile3<-"LST_dates_var_names.txt" #LST dates name |
|
37 |
infile4<-"models_interpolation_05142012.txt" #Interpolation model names |
|
38 |
infile5<-"mean_day244_rescaled.rst" |
|
39 |
inlistf<-"list_files_05032012.txt" #Stack of images containing the Covariates |
|
40 |
|
|
41 |
path<-"/home/parmentier/Data/IPLANT_project/data_Oregon_stations_07152012" #Jupiter LOCATION on Atlas for kriging |
|
42 |
#path<-"H:/Data/IPLANT_project/data_Oregon_stations" #Jupiter Location on XANDERS |
|
43 |
|
|
44 |
setwd(path) |
|
45 |
prop<-0.3 #Proportion of testing retained for validation |
|
46 |
seed_number<- 100 #Seed number for random sampling |
|
47 |
models<-7 #Number of kriging model |
|
48 |
out_prefix<-"_07132012_auto_krig_" #User defined output prefix |
|
17 |
path<- "/data/computer/parmentier/Data/IPLANT_project/data_Oregon_stations/" #Path to all datasets |
|
18 |
setwd(path) |
|
19 |
infile1<-"ghcn_or_tmax_b_04142012_OR83M.shp" #Weather station location in Oregon with input variables |
|
20 |
infile2<-"dates_interpolation_03052012.txt" # list of 10 dates for the regression, more thatn 10 dates may be used |
|
21 |
infile3<-"mean_day244_rescaled.rst" #This image serves as the reference grid for kriging |
|
22 |
infile4<- "orcnty24_OR83M.shp" #Vector file defining the study area: Oregon state and its counties. |
|
23 |
prop<-0.3 #Propotion of weather stations retained for validation/testing |
|
24 |
out_prefix<-"_LST_04172012_RMSE" #output name used in the text file result |
|
49 | 25 |
|
50 | 26 |
###STEP 1 DATA PREPARATION AND PROCESSING##### |
51 | 27 |
|
52 |
###Reading the station data and setting up for models' comparison |
|
53 |
filename<-sub(".shp","",infile1) #Removing the extension from file. |
|
54 |
ghcn<-readOGR(".", filename) #reading shapefile |
|
55 |
|
|
56 |
CRS<-proj4string(ghcn) #Storing projection information (ellipsoid, datum,etc.) |
|
57 |
|
|
58 |
mean_LST<- readGDAL(infile5) #Reading the whole raster in memory. This provides a grid for kriging |
|
59 |
proj4string(mean_LST)<-CRS #Assigning coordinate information to prediction grid. |
|
60 |
|
|
61 |
##Extracting the variables values from the raster files |
|
28 |
###Reading the shapefile and raster image from the local directory |
|
62 | 29 |
|
63 |
lines<-read.table(paste(path,"/",inlistf,sep=""), sep=" ") #Column 1 contains the names of raster files |
|
64 |
inlistvar<-lines[,1] |
|
65 |
inlistvar<-paste(path,"/",as.character(inlistvar),sep="") |
|
66 |
covar_names<-as.character(lines[,2]) #Column two contains short names for covaraites |
|
30 |
mean_LST<- readGDAL(infile3) #This reads the whole raster in memory and provide a grid for kriging in a SpatialGridDataFrame object |
|
31 |
filename<-sub(".shp","",infile1) #Removing the extension from file. |
|
32 |
ghcn<-readOGR(".", filename) #Reading station locations from vector file using rgdal and creating a SpatialPointDataFrame |
|
33 |
CRS_ghcn<-proj4string(ghcn) #This retrieves the coordinate system information for the SDF object (PROJ4 format) |
|
34 |
proj4string(mean_LST)<-CRS_ghcn #Assigning coordinates information to SpatialGridDataFrame object |
|
67 | 35 |
|
68 |
s_raster<- stack(inlistvar) #Creating a stack of raster images from the list of variables. |
|
69 |
layerNames(s_raster)<-covar_names #Assigning names to the raster layers |
|
70 |
projection(s_raster)<-CRS |
|
36 |
# Creating state outline from county |
|
71 | 37 |
|
72 |
#stat_val<- extract(s_raster, ghcn3) #Extracting values from the raster stack for every point location in coords data frame. |
|
73 |
pos<-match("ASPECT",layerNames(s_raster)) #Find column with name "value" |
|
74 |
r1<-raster(s_raster,layer=pos) #Select layer from stack |
|
75 |
pos<-match("slope",layerNames(s_raster)) #Find column with name "value" |
|
76 |
r2<-raster(s_raster,layer=pos) #Select layer from stack |
|
77 |
N<-cos(r1*pi/180) |
|
78 |
E<-sin(r1*pi/180) |
|
79 |
Nw<-sin(r2*pi/180)*cos(r1*pi/180) #Adding a variable to the dataframe |
|
80 |
Ew<-sin(r2*pi/180)*sin(r1*pi/180) #Adding variable to the dataframe. |
|
81 |
r<-stack(N,E,Nw,Ew) |
|
82 |
rnames<-c("Northness","Eastness","Northness_w","Eastness_w") |
|
83 |
layerNames(r)<-rnames |
|
84 |
s_raster<-addLayer(s_raster, r) |
|
85 |
s_sgdf<-as(s_raster,"SpatialGridDataFrame") #Conversion to spatial grid data frame |
|
38 |
orcnty<-readOGR(".", "orcnty24_OR83M") |
|
39 |
proj4string(orcnty) #This retrieves the coordinate system for the SDF |
|
40 |
lps <-getSpPPolygonsLabptSlots(orcnty) #Getting centroids county labels |
|
41 |
IDOneBin <- cut(lps[,1], range(lps[,1]), include.lowest=TRUE) #Creating one bin var |
|
42 |
gpclibPermit() #Set the gpclib to True to allow union |
|
43 |
OR_state <- unionSpatialPolygons(orcnty ,IDOneBin) #Dissolve based on bin var |
|
86 | 44 |
|
87 |
### adding var |
|
88 |
ghcn = transform(ghcn,Northness = cos(ASPECT*pi/180)) #Adding a variable to the dataframe |
|
89 |
ghcn = transform(ghcn,Eastness = sin(ASPECT*pi/180)) #adding variable to the dataframe. |
|
90 |
ghcn = transform(ghcn,Northness_w = sin(slope*pi/180)*cos(ASPECT*pi/180)) #Adding a variable to the dataframe |
|
91 |
ghcn = transform(ghcn,Eastness_w = sin(slope*pi/180)*sin(ASPECT*pi/180)) #adding variable to the dataframe. |
|
45 |
# Adding variables for the regressions |
|
92 | 46 |
|
93 |
#Remove NA for LC and CANHEIGHT
|
|
94 |
ghcn$LC1[is.na(ghcn$LC1)]<-0
|
|
95 |
ghcn$LC3[is.na(ghcn$LC3)]<-0
|
|
96 |
ghcn$CANHEIGHT[is.na(ghcn$CANHEIGHT)]<-0
|
|
47 |
ghcn$Northness<- cos(ghcn$ASPECT) #Adding a variable to the dataframe by calculating the cosine of Aspect
|
|
48 |
ghcn$Eastness <- sin(ghcn$ASPECT) #Adding variable to the dataframe.
|
|
49 |
ghcn$Northness_w <- sin(ghcn$slope)*cos(ghcn$ASPECT) #Adding a variable to the dataframe
|
|
50 |
ghcn$Eastness_w <- sin(ghcn$slope)*sin(ghcn$ASPECT) #Adding variable to the dataframe.
|
|
97 | 51 |
|
98 |
set.seed(seed_number) #Using a seed number allow results based on random number to be compared...
|
|
52 |
set.seed(100) #This set a seed number for the random sampling to make results reproducible.
|
|
99 | 53 |
|
100 |
dates <-readLines(paste(path,"/",infile2, sep="")) |
|
101 |
LST_dates <-readLines(paste(path,"/",infile3, sep=""))
|
|
102 |
#models <-readLines(paste(path,"/",infile4, sep=""))
|
|
54 |
dates <-readLines(paste(path,"/",infile2, sep="")) #Reading dates in a list from the textile.
|
|
55 |
results <- matrix(1,length(dates),4) #This is a matrix containing the diagnostic measures from the GAM models.
|
|
56 |
results_mod_n<-matrix(1,length(dates),3)
|
|
103 | 57 |
|
104 |
#models<-5 |
|
105 |
#Model assessment: specific diagnostic/metrics for GAM |
|
106 |
results_AIC<- matrix(1,length(dates),models+3) |
|
107 |
results_GCV<- matrix(1,length(dates),models+3) |
|
58 |
#Screening for bad values and setting the valid range |
|
108 | 59 |
|
109 |
#Model assessment: general diagnostic/metrics |
|
110 |
results_RMSE <- matrix(1,length(dates),models+3) |
|
111 |
results_MAE <- matrix(1,length(dates),models+3) |
|
112 |
results_ME <- matrix(1,length(dates),models+3) |
|
113 |
results_R2 <- matrix(1,length(dates),models+3) #Coef. of determination for the validation dataset |
|
114 |
results_RMSE_f<- matrix(1,length(dates),models+3) |
|
115 |
|
|
116 |
|
|
117 |
#Screening for bad values: value is tmax in this case |
|
118 |
#ghcn$value<-as.numeric(ghcn$value) |
|
119 |
ghcn_all<-ghcn |
|
120 |
ghcn_test<-subset(ghcn,ghcn$value>-150 & ghcn$value<400) |
|
121 |
ghcn_test2<-subset(ghcn_test,ghcn_test$ELEV_SRTM>0) |
|
60 |
ghcn_test<-subset(ghcn,ghcn$tmax>-150 & ghcn$tmax<400) #Values are in tenth of degrees C |
|
61 |
ghcn_test2<-subset(ghcn_test,ghcn_test$ELEV_SRTM>0) #No elevation below sea leve is allowed. |
|
122 | 62 |
ghcn<-ghcn_test2 |
123 |
#coords<- ghcn[,c('x_OR83M','y_OR83M')] |
|
124 |
|
|
125 |
|
|
126 | 63 |
|
127 | 64 |
###CREATING SUBSETS BY INPUT DATES AND SAMPLING |
128 | 65 |
ghcn.subsets <-lapply(dates, function(d) subset(ghcn, ghcn$date==as.numeric(d))) #Producing a list of data frame, one data frame per date. |
129 | 66 |
|
130 | 67 |
for(i in 1:length(dates)){ # start of the for loop #1 |
131 | 68 |
#i<-3 #Date 10 is used to test kriging |
132 |
|
|
133 |
#This allows to change only one name of the |
|
134 |
|
|
135 | 69 |
date<-strptime(dates[i], "%Y%m%d") |
136 | 70 |
month<-strftime(date, "%m") |
137 | 71 |
LST_month<-paste("mm_",month,sep="") |
138 |
#adding to SpatialGridDataFrame |
|
139 |
#t<-s_sgdf[,match(LST_month, names(s_sgdf))] |
|
140 |
#s_sgdf$LST<-s_sgdf[c(LST_month)] |
|
141 | 72 |
mod <-ghcn.subsets[[i]][,match(LST_month, names(ghcn.subsets[[i]]))] |
142 | 73 |
ghcn.subsets[[i]]$LST <-mod[[1]] |
143 | 74 |
|
... | ... | |
149 | 80 |
data_s <- ghcn.subsets[[i]][ind.training, ] |
150 | 81 |
data_v <- ghcn.subsets[[i]][ind.testing, ] |
151 | 82 |
|
83 |
###STEP 2 KRIGING### |
|
84 |
|
|
85 |
#Kriging tmax |
|
152 | 86 |
|
153 |
###BEFORE Kringing the data object must be transformed to SDF |
|
87 |
hscat(tmax~1,data_s,(0:9)*20000) # 9 lag classes with 20,000m width |
|
88 |
v<-variogram(tmax~1, data_s) # This plots a sample varigram for date 10 fir the testing dataset |
|
89 |
plot(v) |
|
90 |
v.fit<-fit.variogram(v,vgm(2000,"Sph", 150000,1000)) #Model variogram: sill is 2000, spherical, range 15000 and nugget 1000 |
|
91 |
plot(v, v.fit) #Compare model and sample variogram via a graphical plot |
|
92 |
tmax_krige<-krige(tmax~1, data_s,mean_LST, v.fit) #mean_LST provides the data grid/raster image for the kriging locations to be predicted. |
|
154 | 93 |
|
155 |
coords<- data_v[,c('x_OR83M','y_OR83M')] |
|
156 |
coordinates(data_v)<-coords |
|
157 |
proj4string(data_v)<-CRS #Need to assign coordinates... |
|
158 |
coords<- data_s[,c('x_OR83M','y_OR83M')] |
|
159 |
coordinates(data_s)<-coords |
|
160 |
proj4string(data_s)<-CRS #Need to assign coordinates.. |
|
94 |
#Cokriging tmax |
|
95 |
g<-gstat(NULL,"tmax", tmax~1, data_s) #This creates a gstat object "g" that acts as container for kriging specifications. |
|
96 |
g<-gstat(g, "SRTM_elev",ELEV_SRTM~1,data_s) #Adding variables to gstat object g |
|
97 |
g<-gstat(g, "LST", LST~1,data_s) |
|
161 | 98 |
|
162 |
#This allows to change only one name of the data.frame |
|
163 |
pos<-match("value",names(data_s)) #Find column with name "value" |
|
164 |
names(data_s)[pos]<-c("tmax") |
|
165 |
data_s$tmax<-data_s$tmax/10 #TMax is the average max temp for months |
|
166 |
pos<-match("value",names(data_v)) #Find column with name "value" |
|
167 |
names(data_v)[pos]<-c("tmax") |
|
168 |
data_v$tmax<-data_v$tmax/10 |
|
169 |
#dstjan=dst[dst$month==9,] #dst contains the monthly averages for tmax for every station over 2000-2010 |
|
170 |
############## |
|
171 |
###STEP 2 KRIGING### |
|
99 |
vm_g<-variogram(g) #Visualizing multivariate sample variogram. |
|
100 |
vm_g.fit<-fit.lmc(vm_g,g,vgm(2000,"Sph", 100000,1000)) #Fitting variogram for all variables at once. |
|
101 |
plot(vm_g,vm_g.fit) #Visualizing variogram fit and sample |
|
102 |
vm_g.fit$set <-list(nocheck=1) #Avoid checking and allow for different range in variogram |
|
103 |
co_kriged_surf<-predict(vm_g.fit,mean_LST) #Prediction using co-kriging with grid location defined from input raster image. |
|
104 |
#co_kriged_surf$tmax.pred #Results stored in SpatialGridDataFrame with tmax prediction accessible in dataframe. |
|
172 | 105 |
|
173 |
#Kriging tmax |
|
174 | 106 |
|
175 |
# hscat(tmax~1,data_s,(0:9)*20000) # 9 lag classes with 20,000m width |
|
176 |
# v<-variogram(tmax~1, data_s) # This plots a sample varigram for date 10 fir the testing dataset |
|
177 |
# plot(v) |
|
178 |
# v.fit<-fit.variogram(v,vgm(2000,"Sph", 150000,1000)) #Model variogram: sill is 2000, spherical, range 15000 and nugget 1000 |
|
179 |
# plot(v, v.fit) #Compare model and sample variogram via a graphical plot |
|
180 |
# tmax_krige<-krige(tmax~1, data_s,mean_LST, v.fit) #mean_LST provides the data grid/raster image for the kriging locations to be predicted. |
|
107 |
#spplot.vcov(co_kriged_surf) #Visualizing the covariance structure |
|
181 | 108 |
|
182 |
krmod1<-autoKrige(tmax~1, data_s,s_sgdf,data_s) #Use autoKrige instead of krige: with data_s for fitting on a grid |
|
183 |
krmod2<-autoKrige(tmax~x_OR83M+y_OR83M,input_data=data_s,new_data=s_sgdf,data_variogram=data_s) |
|
184 |
krmod3<-autoKrige(tmax~x_OR83M+y_OR83M+ELEV_SRTM,input_data=data_s,new_data=s_sgdf,data_variogram=data_s) |
|
185 |
krmod4<-autoKrige(tmax~x_OR83M+y_OR83M+DISTOC,input_data=data_s,new_data=s_sgdf,data_variogram=data_s) |
|
186 |
krmod5<-autoKrige(tmax~x_OR83M+y_OR83M+ELEV_SRTM+DISTOC,input_data=data_s,new_data=s_sgdf,data_variogram=data_s) |
|
187 |
krmod6<-autoKrige(tmax~x_OR83M+y_OR83M+Northness+Eastness,input_data=data_s,new_data=s_sgdf,data_variogram=data_s) |
|
188 |
krmod7<-autoKrige(tmax~x_OR83M+y_OR83M+Northness+Eastness,input_data=data_s,new_data=s_sgdf,data_variogram=data_s) |
|
189 |
#krmod8<-autoKrige(tmax~LST,input_data=data_s,new_data=s_sgdf,data_variogram=data_s) |
|
190 |
#krmod9<-autoKrige(tmax~x_OR83M+y_OR83M+LST,input_data=data_s,new_data=s_sgdf,data_variogram=data_s) |
|
109 |
tmax_krig1_s <- overlay(tmax_krige,data_s) #This overlays the kriged surface tmax and the location of weather stations |
|
110 |
tmax_cokrig1_s<- overlay(co_kriged_surf,data_s) #This overalys the cokriged surface tmax and the location of weather stations |
|
111 |
tmax_krig1_v <- overlay(tmax_krige,data_v) #This overlays the kriged surface tmax and the location of weather stations |
|
112 |
tmax_cokrig1_v<- overlay(co_kriged_surf,data_v) |
|
191 | 113 |
|
192 |
krig1<-krmod1$krige_output #Extracting Spatial Grid Data frame |
|
193 |
krig2<-krmod2$krige_output |
|
194 |
krig3<-krmod3$krige_outpu |
|
195 |
krig4<-krmod4$krige_output |
|
196 |
krig5<-krmod5$krige_output |
|
197 |
krig6<-krmod6$krige_output #Extracting Spatial Grid Data frame |
|
198 |
krig7<-krmod7$krige_output |
|
199 |
#krig8<-krmod8$krige_outpu |
|
200 |
#krig9<-krmod9$krige_output |
|
114 |
data_s$tmax_kr<-tmax_krig1_s$var1.pred #Adding the results back into the original dataframes. |
|
115 |
data_v$tmax_kr<-tmax_krig1_v$var1.pred |
|
116 |
data_s$tmax_cokr<-tmax_cokrig1_s$tmax.pred |
|
117 |
data_v$tmax_cokr<-tmax_cokrig1_v$tmax.pred |
|
201 | 118 |
|
202 |
#tmax_krig1_s <- overlay(krige,data_s) #This overlays the kriged surface tmax and the location of weather stations |
|
203 |
#tmax_krig1_v <- overlay(krige,data_v) |
|
204 |
# |
|
205 |
# #Cokriging tmax |
|
206 |
# g<-gstat(NULL,"tmax", tmax~1, data_s) #This creates a gstat object "g" that acts as container for kriging specifications. |
|
207 |
# g<-gstat(g, "SRTM_elev",ELEV_SRTM~1,data_s) #Adding variables to gstat object g |
|
208 |
# g<-gstat(g, "LST", LST~1,data_s) |
|
119 |
#Co-kriging only on the validation sites for faster computing |
|
209 | 120 |
|
210 |
# vm_g<-variogram(g) #Visualizing multivariate sample variogram. |
|
211 |
# vm_g.fit<-fit.lmc(vm_g,g,vgm(2000,"Sph", 100000,1000)) #Fitting variogram for all variables at once. |
|
212 |
# plot(vm_g,vm_g.fit) #Visualizing variogram fit and sample |
|
213 |
# vm_g.fit$set <-list(nocheck=1) #Avoid checking and allow for different range in variogram |
|
214 |
# co_kriged_surf<-predict(vm_g.fit,mean_LST) #Prediction using co-kriging with grid location defined from input raster image. |
|
215 |
# #co_kriged_surf$tmax.pred #Results stored in SpatialGridDataFrame with tmax prediction accessible in dataframe. |
|
121 |
cokrig1_dv<-predict(vm_g.fit,data_v) |
|
122 |
cokrig1_ds<-predict(vm_g.fit,data_s) |
|
123 |
data_s$tmax_cokr<-cokrig1_ds$tmax.pred |
|
124 |
data_v$tmax_cokr<-cokrig1_dv$tmax.pred |
|
216 | 125 |
|
217 |
#spplot.vcov(co_kriged_surf) #Visualizing the covariance structure |
|
218 |
|
|
219 |
# tmax_cokrig1_s<- overlay(co_kriged_surf,data_s) #This overalys the cokriged surface tmax and the location of weather stations |
|
220 |
# tmax_cokrig1_v<- overlay(co_kriged_surf,data_v) |
|
126 |
#Calculate RMSE and then krig the residuals....! |
|
221 | 127 |
|
222 |
for (j in 1:models){ |
|
223 |
|
|
224 |
mod<-paste("krig",j,sep="") |
|
225 |
krmod<-get(mod) |
|
226 |
krig_val_s <- overlay(krmod,data_s) #This overlays the kriged surface tmax and the location of weather stations |
|
227 |
krig_val_v <- overlay(krmod,data_v) #This overlays the kriged surface tmax and the location of weather stations |
|
228 |
|
|
229 |
pred_krmod<-paste("pred_krmod",j,sep="") |
|
230 |
#Adding the results back into the original dataframes. |
|
231 |
data_s[[pred_krmod]]<-krig_val_s$var1.pred |
|
232 |
data_v[[pred_krmod]]<-krig_val_v$var1.pred |
|
233 |
|
|
234 |
#Model assessment: RMSE and then krig the residuals....! |
|
235 |
|
|
236 |
res_mod_kr_s<- data_s$tmax - data_s[[pred_krmod]] #Residuals from kriging training |
|
237 |
res_mod_kr_v<- data_v$tmax - data_v[[pred_krmod]] #Residuals from kriging validation |
|
238 |
|
|
239 |
RMSE_mod_kr_s <- sqrt(sum(res_mod_kr_s^2,na.rm=TRUE)/(nv-sum(is.na(res_mod_kr_s)))) #RMSE from kriged surface training |
|
240 |
RMSE_mod_kr_v <- sqrt(sum(res_mod_kr_v^2,na.rm=TRUE)/(nv-sum(is.na(res_mod_kr_v)))) #RMSE from kriged surface validation |
|
241 |
MAE_mod_kr_s<- sum(abs(res_mod_kr_s),na.rm=TRUE)/(nv-sum(is.na(res_mod_kr_s))) #MAE from kriged surface training #MAE, Mean abs. Error FOR REGRESSION STEP 1: GAM |
|
242 |
MAE_mod_kr_v<- sum(abs(res_mod_kr_v),na.rm=TRUE)/(nv-sum(is.na(res_mod_kr_v))) #MAE from kriged surface validation |
|
243 |
ME_mod_kr_s<- sum(res_mod_kr_s,na.rm=TRUE)/(nv-sum(is.na(res_mod_kr_s))) #ME, Mean Error or bias FOR REGRESSION STEP 1: GAM |
|
244 |
ME_mod_kr_v<- sum(res_mod_kr_v,na.rm=TRUE)/(nv-sum(is.na(res_mod_kr_v))) #ME, Mean Error or bias FOR REGRESSION STEP 1: GAM |
|
245 |
R2_mod_kr_s<- cor(data_s$tmax,data_s[[pred_krmod]],use="complete.obs")^2 #R2, coef. of determination FOR REGRESSION STEP 1: GAM |
|
246 |
R2_mod_kr_v<- cor(data_v$tmax,data_v[[pred_krmod]],use="complete.obs")^2 #R2, coef. of determinationFOR REGRESSION STEP 1: GAM |
|
247 |
#(nv-sum(is.na(res_mod2))) |
|
248 |
#Writing out results |
|
249 |
|
|
250 |
results_RMSE[i,1]<- dates[i] #storing the interpolation dates in the first column |
|
251 |
results_RMSE[i,2]<- ns #number of stations used in the training stage |
|
252 |
results_RMSE[i,3]<- "RMSE" |
|
253 |
results_RMSE[i,j+3]<- RMSE_mod_kr_v |
|
254 |
#results_RMSE_kr[i,3]<- res_mod_kr_v |
|
255 |
|
|
256 |
results_MAE[i,1]<- dates[i] #storing the interpolation dates in the first column |
|
257 |
results_MAE[i,2]<- ns #number of stations used in the training stage |
|
258 |
results_MAE[i,3]<- "MAE" |
|
259 |
results_MAE[i,j+3]<- MAE_mod_kr_v |
|
260 |
#results_RMSE_kr[i,3]<- res_mod_kr_v |
|
261 |
|
|
262 |
results_ME[i,1]<- dates[i] #storing the interpolation dates in the first column |
|
263 |
results_ME[i,2]<- ns #number of stations used in the training stage |
|
264 |
results_ME[i,3]<- "ME" |
|
265 |
results_ME[i,j+3]<- ME_mod_kr_v |
|
266 |
#results_RMSE_kr[i,3]<- res_mod_kr_v |
|
267 |
|
|
268 |
results_R2[i,1]<- dates[i] #storing the interpolation dates in the first column |
|
269 |
results_R2[i,2]<- ns #number of stations used in the training stage |
|
270 |
results_R2[i,3]<- "R2" |
|
271 |
results_R2[i,j+3]<- R2_mod_kr_v |
|
272 |
#results_RMSE_kr[i,3]<- res_mod_kr_v |
|
273 |
|
|
274 |
name3<-paste("res_kr_mod",j,sep="") |
|
275 |
#as.numeric(res_mod) |
|
276 |
#data_s[[name3]]<-res_mod_kr_s |
|
277 |
data_s[[name3]]<-as.numeric(res_mod_kr_s) |
|
278 |
#data_v[[name3]]<-res_mod_kr_v |
|
279 |
data_v[[name3]]<-as.numeric(res_mod_kr_v) |
|
280 |
#Writing residuals from kriging |
|
281 |
|
|
282 |
#Saving kriged surface in raster images |
|
283 |
data_name<-paste("mod",j,"_",dates[[i]],sep="") |
|
284 |
krig_raster_name<-paste("krmod_",data_name,out_prefix,".tif", sep="") |
|
285 |
writeGDAL(krmod,fname=krig_raster_name, driver="GTiff", type="Float32",options ="INTERLEAVE=PIXEL") |
|
286 |
krig_raster_name<-paste("krmod_",data_name,out_prefix,".rst", sep="") |
|
287 |
writeRaster(raster(krmod), filename=krig_raster_name) #Writing the data in a raster file format...(IDRISI) |
|
288 |
|
|
289 |
#krig_raster_name<-paste("Kriged_tmax_",data_name,out_prefix,".tif", sep="") |
|
290 |
#writeGDAL(tmax_krige,fname=krig_raster_name, driver="GTiff", type="Float32",options ="INTERLEAVE=PIXEL") |
|
291 |
#X11() |
|
292 |
#plot(raster(co_kriged_surf)) |
|
293 |
#title(paste("Tmax cokriging for date ",dates[[i]],sep="")) |
|
294 |
#savePlot(paste("Cokriged_tmax",data_name,out_prefix,".png", sep=""), type="png") |
|
295 |
#dev.off() |
|
296 |
#X11() |
|
297 |
#plot(raster(tmax_krige)) |
|
298 |
#title(paste("Tmax Kriging for date ",dates[[i]],sep="")) |
|
299 |
#savePlot(paste("Kriged_res_",data_name,out_prefix,".png", sep=""), type="png") |
|
300 |
#dev.off() |
|
301 |
# |
|
302 |
|
|
303 |
} |
|
128 |
res_mod1<- data_v$tmax - data_v$tmax_kr #Residuals from kriging. |
|
129 |
res_mod2<- data_v$tmax - data_v$tmax_cokr #Residuals from cokriging. |
|
304 | 130 |
|
305 |
# #Co-kriging only on the validation sites for faster computing |
|
306 |
# |
|
307 |
# cokrig1_dv<-predict(vm_g.fit,data_v) |
|
308 |
# cokrig1_ds<-predict(vm_g.fit,data_s) |
|
309 |
# # data_s$tmax_cokr<-cokrig1_ds$tmax.pred |
|
310 |
# # data_v$tmax_cokr<-cokrig1_dv$tmax.pred |
|
311 |
# |
|
312 |
# #Calculate RMSE and then krig the residuals....! |
|
313 |
# |
|
314 |
# res_mod1<- data_v$tmax - data_v$tmax_kr #Residuals from kriging. |
|
315 |
# res_mod2<- data_v$tmax - data_v$tmax_cokr #Residuals from cokriging. |
|
316 |
# |
|
317 |
# RMSE_mod1 <- sqrt(sum(res_mod1^2,na.rm=TRUE)/(nv-sum(is.na(res_mod1)))) #RMSE from kriged surface. |
|
318 |
# RMSE_mod2 <- sqrt(sum(res_mod2^2,na.rm=TRUE)/(nv-sum(is.na(res_mod2)))) #RMSE from co-kriged surface. |
|
319 |
# #(nv-sum(is.na(res_mod2))) |
|
131 |
RMSE_mod1 <- sqrt(sum(res_mod1^2,na.rm=TRUE)/(nv-sum(is.na(res_mod1)))) #RMSE from kriged surface. |
|
132 |
RMSE_mod2 <- sqrt(sum(res_mod2^2,na.rm=TRUE)/(nv-sum(is.na(res_mod2)))) #RMSE from co-kriged surface. |
|
133 |
#(nv-sum(is.na(res_mod2))) |
|
320 | 134 |
|
321 | 135 |
#Saving the subset in a dataframe |
322 | 136 |
data_name<-paste("ghcn_v_",dates[[i]],sep="") |
323 | 137 |
assign(data_name,data_v) |
324 | 138 |
data_name<-paste("ghcn_s_",dates[[i]],sep="") |
325 | 139 |
assign(data_name,data_s) |
326 |
|
|
327 |
# results[i,1]<- dates[i] #storing the interpolation dates in the first column |
|
328 |
# results[i,2]<- ns #number of stations in training |
|
329 |
# results[i,3]<- RMSE_mod1 |
|
330 |
# results[i,4]<- RMSE_mod2 |
|
331 |
# |
|
332 |
# results_mod_n[i,1]<-dates[i] |
|
333 |
# results_mod_n[i,2]<-(nv-sum(is.na(res_mod1))) |
|
334 |
# results_mod_n[i,3]<-(nv-sum(is.na(res_mod2))) |
|
140 |
|
|
141 |
krig_raster_name<-paste("coKriged_tmax_",data_name,out_prefix,".tif", sep="") |
|
142 |
writeGDAL(co_kriged_surf,fname=krig_raster_name, driver="GTiff", type="Float32",options ="INTERLEAVE=PIXEL") |
|
143 |
krig_raster_name<-paste("Kriged_tmax_",data_name,out_prefix,".tif", sep="") |
|
144 |
writeGDAL(tmax_krige,fname=krig_raster_name, driver="GTiff", type="Float32",options ="INTERLEAVE=PIXEL") |
|
145 |
X11() |
|
146 |
plot(raster(co_kriged_surf)) |
|
147 |
title(paste("Tmax cokriging for date ",dates[[i]],sep="")) |
|
148 |
savePlot(paste("Cokriged_tmax",data_name,out_prefix,".png", sep=""), type="png") |
|
149 |
dev.off() |
|
150 |
X11() |
|
151 |
plot(raster(tmax_krige)) |
|
152 |
title(paste("Tmax Kriging for date ",dates[[i]],sep="")) |
|
153 |
savePlot(paste("Kriged_res_",data_name,out_prefix,".png", sep=""), type="png") |
|
154 |
dev.off() |
|
155 |
|
|
156 |
results[i,1]<- dates[i] #storing the interpolation dates in the first column |
|
157 |
results[i,2]<- ns #number of stations in training |
|
158 |
results[i,3]<- RMSE_mod1 |
|
159 |
results[i,4]<- RMSE_mod2 |
|
160 |
|
|
161 |
results_mod_n[i,1]<-dates[i] |
|
162 |
results_mod_n[i,2]<-(nv-sum(is.na(res_mod1))) |
|
163 |
results_mod_n[i,3]<-(nv-sum(is.na(res_mod2))) |
|
335 | 164 |
} |
336 | 165 |
|
337 | 166 |
## Plotting and saving diagnostic measures |
338 |
results_table_RMSE<-as.data.frame(results_RMSE) |
|
339 |
results_table_MAE<-as.data.frame(results_MAE) |
|
340 |
results_table_ME<-as.data.frame(results_ME) |
|
341 |
results_table_R2<-as.data.frame(results_R2) |
|
342 |
|
|
343 |
cname<-c("dates","ns","metric","krmod1", "krmod2","krmod3", "krmod4", "mkrod5") |
|
344 |
colnames(results_table_RMSE)<-cname |
|
345 |
colnames(results_table_MAE)<-cname |
|
346 |
colnames(results_table_ME)<-cname |
|
347 |
colnames(results_table_R2)<-cname |
|
348 |
|
|
349 |
|
|
350 |
#Summary of diagnostic measures are stored in a data frame |
|
351 |
tb_diagnostic1<-rbind(results_table_RMSE,results_table_MAE, results_table_ME, results_table_R2) # |
|
352 |
#tb_diagnostic1_kr<-rbind(results_table_RMSE_kr,results_table_MAE_kr, results_table_ME_kr, results_table_R2_kr) |
|
353 |
#tb_diagnostic2<-rbind(results_table_AIC,results_table_GCV, results_table_DEV,results_table_RMSE_f) |
|
354 |
|
|
355 |
write.table(tb_diagnostic1, file= paste(path,"/","results_GAM_Assessment_measure1",out_prefix,".txt",sep=""), sep=",") |
|
356 |
#write.table(tb_diagnostic1_kr, file= paste(path,"/","results_GAM_Assessment_measure1_kr_",out_prefix,".txt",sep=""), sep=",") |
|
357 |
#write.table(tb_diagnostic2, file= paste(path,"/","results_GAM_Assessment_measure2_",out_prefix,".txt",sep=""), sep=",") |
|
167 |
results_num <-results |
|
168 |
mode(results_num)<- "numeric" |
|
169 |
# Make it numeric first |
|
170 |
# Now turn it into a data.frame... |
|
358 | 171 |
|
172 |
results_table<-as.data.frame(results_num) |
|
173 |
colnames(results_table)<-c("dates","ns","RMSE") |
|
359 | 174 |
|
360 |
#### END OF SCRIPT ##### |
|
175 |
write.csv(results_table, file= paste(path,"/","results_Kriging_Assessment",out_prefix,".txt",sep="")) |
Also available in: Unified diff
Revert "Merge branch 'ag/interp' of code.nceas.ucsb.edu:environmental-layers into aw/precip"
This reverts commit f9c712987ba814b83b2c2cc058c6c1f9b07933c1, reversing
changes made to f0375becc9fb9e13e55c5b7a48be5c5f98064f87.