-
Notifications
You must be signed in to change notification settings - Fork 1
/
script_modelingL8.R
280 lines (237 loc) · 11.5 KB
/
script_modelingL8.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
## Example: http://uc-r.github.io/gbm_regression
## 1. Load libraries
library(gbm) # basic implementation
library(caret) # an aggregator package for performing many machine learning models
library(pdp) # model visualization
library(ggplot2) # model visualization
library(lime) # model visualization
library(dplyr)
library(raster)
## 2. Set working directory and load data
setwd("C:/Users/neliq/Documents/NIF/Rothamsted/")
dat1 <- read.csv('11.models/Landsat/datLandsat2013.csv', sep=',')
head(dat1)
## 3. Create data partitions
set.seed(125)
trainIndex <- createDataPartition(dat1$TC_perc, p = .75,
list = FALSE,
times = 1)
dat_train = dat1[ trainIndex,]
dat_test = dat1[-trainIndex,]
X_train <- dat_train[,-1]
Y_train <- dat_train[,1]
X_test <- dat_test[,-1]
Y_test <- dat_test[,1]
## 4. Train model and save
gbm.fitGNSVI <- gbm(formula = TC_perc ~ .-GNDVI,
data = dat_train,
n.trees = 1000,
interaction.depth = 1,
shrinkage = 0.01,
cv.folds = 10,
n.cores = 4, # will use all cores by default
verbose = FALSE)
saveRDS(gbm.fit, "finalGBM_modelL82013.rds")
par(mar = c(5, 8, 1, 1))
summary(gbm.fit,
cBars = 14,
method = relative.influence,
las = 2)
# 4.1. Check RMSE and plot
print(paste("RMSE: ", round(sqrt(min(gbm.fit$cv.error)),2), "%"))
gbm.perf(gbm.fit, method = "cv")
# 4.2. predict values for raining and test data
predTrainDefault <- predict(gbm.fit, n.trees = gbm.fit$n.trees, X_train)
predDefault <- predict(gbm.fit, n.trees = gbm.fit$n.trees, X_test)
# 4.3. Print metrics and plot
RSQUARE <- function(y_actual,y_predict){
cor(y_actual,y_predict)^2
}
print(paste("RMSE training data: ",
round(caret::RMSE(predTrainDefault, Y_train),2), "%"))
print(paste("RMSE test data: ",
round(caret::RMSE(predDefault, Y_test),2), "%"))
print(paste("Rsquare training data: ",
round(RSQUARE(Y_train, predTrainDefault),2)))
print(paste("Rsquare training data: ",
round(RSQUARE(Y_test, predDefault),2)))
par(mfrow=c(1,2))
plot(Y_train, PRED, ylim = c(0,10), xlim = c(0,10))
abline(0,1)
plot(Y_test, PREDS, ylim = c(0,10), xlim = c(0,10))
abline(0,1)
## 5. Load raster and predict
# 2013
list.files()
covs <- stack("5.satellites_images/Landsat/5m/median/L8MedianDEMTerrain2013.tif")
names(covs) <- c("B2", "B3", "B4", "B5", "B6", "B7", 'B10',
"NDVI", "GNDVI", "NDMI", "EVI", "BSI","DEM",
"ASP", "CI", "LS", "PC", "RSP", "SLP", "TWI")
plot(covs)
covs.df <- as.data.frame(covs, xy = TRUE)
predMapOpt <- predict(gbm.fit, newdata = covs.df)
preds <- cbind(covs.df[,1:2], predMapOpt)
colnames(preds) <- c('X', 'Y', 'CTotal')
preds_map <- rasterFromXYZ(preds)
pal <- colorRampPalette(c("#d7191c","#fdae61","#ffffc0","#a6d96a","#1a9641"))
plot(preds_map,col = pal(100))
area_field <- shapefile("C:/Users/neliq/Documents/NIF/Rothamsted/2.shapefiles/area_field.shp")
plot(area_field, add = TRUE)
crs(preds_map) <- '+proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +units=m +no_defs'
writeRaster(preds_map, '6.results/CTotalL8_2013.tif')
## Predict to raster of other years
# 2016
terrainStack2016 <- stack("5.satellites_images/Landsat/5m/L8DEMTerrain2016.tif")
names(terrainStack2016) <- c("B2", "B3", "B4", "B5", "B6", "B7", 'B10',
"NDVI", "GNDVI", "NDMI", "EVI", "BSI","DEM",
"ASP", "CI", "LS", "PC", "RSP", "SLP", "TWI")
covs2016.df <- as.data.frame(terrainStack2016, xy = TRUE)
predMapOpt2016 <- predict(gbm.fit, newdata = covs2016.df)
preds2016 <- cbind(covs.df[,1:2], predMapOpt2016)
colnames(preds2016) <- c('X', 'Y', 'CTotal')
mapCT2016 <- rasterFromXYZ(preds2016)
pal <- colorRampPalette(c("#d7191c","#fdae61","#ffffc0","#a6d96a","#1a9641"))
plot(mapCT2016,col = pal(100))
plot(area_field, add = TRUE)
crs(mapCT2016) <- '+proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +units=m +no_defs'
writeRaster(mapCT2016, '6.results/CTotalL8_2016.tif', overwrite = TRUE)
# 2018
terrainStack2018 <- stack("5.satellites_images/Landsat/5m/L8DEMTerrain2018.tif")
names(terrainStack2018) <- c("B2", "B3", "B4", "B5", "B6", "B7", 'B10',
"NDVI", "GNDVI", "NDMI", "EVI", "BSI","DEM",
"ASP", "CI", "LS", "PC", "RSP", "SLP", "TWI")
covs2018.df <- as.data.frame(terrainStack2018, xy = TRUE)
predMapOpt2018 <- predict(gbm.fit, newdata = covs2018.df)
preds2018 <- cbind(covs2018.df[,1:2], predMapOpt2018)
colnames(preds2018) <- c('X', 'Y', 'CTotal')
mapCT2018 <- rasterFromXYZ(preds2018)
pal <- colorRampPalette(c("#d7191c","#fdae61","#ffffc0","#a6d96a","#1a9641"))
plot(mapCT2018,col = pal(100))
plot(area_field, add = TRUE)
crs(mapCT2018) <- '+proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +units=m +no_defs'
writeRaster(mapCT2018, '6.results/CTotalL8_2018.tif', overwrite = TRUE)
# 2019
terrainStack2019 <- stack("5.satellites_images/Landsat/5m/L8DEMTerrain2019.tif")
names(terrainStack2019) <- c("B2", "B3", "B4", "B5", "B6", "B7", 'B10',
"NDVI", "GNDVI", "NDMI", "EVI", "BSI","DEM",
"ASP", "CI", "LS", "PC", "RSP", "SLP", "TWI")
covs2019.df <- as.data.frame(terrainStack2019, xy = TRUE)
predMapOpt2019 <- predict(gbm.fit, newdata = covs2019.df)
preds2019 <- cbind(covs2019.df[,1:2], predMapOpt2019)
colnames(preds2019) <- c('X', 'Y', 'CTotal')
mapCT2019 <- rasterFromXYZ(preds2019)
pal <- colorRampPalette(c("#d7191c","#fdae61","#ffffc0","#a6d96a","#1a9641"))
plot(mapCT2019, col = pal(100))
plot(area_field, add = TRUE)
crs(mapCT2019) <- '+proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +units=m +no_defs'
writeRaster(mapCT2019, '6.results/CTotalL8_2019.tif', overwrite = TRUE)
# 2020
terrainStack2020 <- stack("5.satellites_images/Landsat/5m/L8DEMTerrain2020.tif")
names(terrainStack2020) <- c("B2", "B3", "B4", "B5", "B6", "B7", 'B10',
"NDVI", "GNDVI", "NDMI", "EVI", "BSI","DEM",
"ASP", "CI", "LS", "PC", "RSP", "SLP", "TWI")
covs2020.df <- as.data.frame(terrainStack2020, xy = TRUE)
predMapOpt2020 <- predict(gbm.fit, newdata = covs2020.df)
preds2020 <- cbind(covs2020.df[,1:2], predMapOpt2020)
colnames(preds2020) <- c('X', 'Y', 'CTotal')
mapCT2020 <- rasterFromXYZ(preds2020)
pal <- colorRampPalette(c("#d7191c","#fdae61","#ffffc0","#a6d96a","#1a9641"))
plot(mapCT2020, col = pal(100))
plot(area_field, add = TRUE)
crs(mapCT2020) <- '+proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +units=m +no_defs'
writeRaster(mapCT2020, '6.results/CTotalL8_2020.tif', overwrite = TRUE)
# 2021
terrainStack2021 <- stack("5.satellites_images/Landsat/5m/L8DEMTerrain2021.tif")
names(terrainStack2021) <- c("B2", "B3", "B4", "B5", "B6", "B7", 'B10',
"NDVI", "GNDVI", "NDMI", "EVI", "BSI","DEM",
"ASP", "CI", "LS", "PC", "RSP", "SLP", "TWI")
covs2021.df <- as.data.frame(terrainStack2021, xy = TRUE)
predMapOpt2021 <- predict(gbm.fit, newdata = covs2021.df)
preds2021 <- cbind(covs2021.df[,1:2], predMapOpt2021)
colnames(preds2021) <- c('X', 'Y', 'CTotal')
mapCT2021 <- rasterFromXYZ(preds2021)
pal <- colorRampPalette(c("#d7191c","#fdae61","#ffffc0","#a6d96a","#1a9641"))
plot(mapCT2021, col = pal(100))
plot(area_field, add = TRUE)
crs(mapCT2021) <- '+proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +units=m +no_defs'
writeRaster(mapCT2021, '6.results/CTotalL8_2021.tif', overwrite = TRUE)
## Take mean by fields
all_maps <- stack(preds_map, mapCT2016, mapCT2018, mapCT2019, mapCT2020, mapCT2021)
names(all_maps) <- c("Pred2012", "Pred2016", "Pred2018", "Pred2019", "Pred2020", "Pred2021")
plot(all_maps)
mapCTFields <- cbind(area_field$Field_Name,
raster::extract(all_maps, area_field, fun = mean, df = TRUE))
write.csv(mapCTFields, "6.results/PredsbyFieldLandsat/preds_by_fieldL8.csv")
{## Hyperparameter tuning ##############################################
# 6.1. create hyperparameter grid
hyper_grid <- expand.grid(shrinkage = c(.01, .05, .1),
interaction.depth = c(1, 3, 5),
n_trees = c(100, 500, 700, 100),
n.minobsinnode = c(5, 10, 15),
bag.fraction = c(.65, .8, 1),
optimal_trees = 0, # a place to dump results
min_RMSE = 0) # a place to dump results
nrow(hyper_grid)
# 6.2. Grid search
for(i in 1:nrow(hyper_grid)) {
# reproducibility
set.seed(123)
# train model
gbm.tune <- gbm(formula = TC_perc ~ .,
data = dat_train,
distribution = "gaussian",
n.trees = hyper_grid$n_trees[i],
interaction.depth = hyper_grid$interaction.depth[i],
shrinkage = hyper_grid$shrinkage[i],
n.minobsinnode = hyper_grid$n.minobsinnode[i],
bag.fraction = hyper_grid$bag.fraction[i],
train.fraction = .75,
cv.folds = 5,
n.cores = 3,
verbose = TRUE)
# add min training error and trees to grid
hyper_grid$optimal_trees[i] <- which.min(gbm.tune$valid.error)
hyper_grid$min_RMSE[i] <- sqrt(min(gbm.tune$valid.error))
}
hyper_grid %>%
dplyr::arrange(min_RMSE) %>%
head(10)
## 7. Train tunned GBM model
gbm.fit.final <- gbm(formula = TC_perc ~ .,
data = dat_train,
distribution = "gaussian",
n.trees = 700,
interaction.depth = 2,
shrinkage = 0.01,
n.minobsinnode = 20,
bag.fraction = .65,
train.fraction = 1,
n.cores = 2,
verbose = FALSE)
saveRDS(gbm.fit.final, "finalGBM_model.rds")
par(mar = c(5, 8, 1, 1))
summary(gbm.fit,
cBars = 18,
method = relative.influence,
las = 2)
gbm.fit %>%
partial(pred.var = "EVI", n.trees = gbm.fit$n.trees, grid.resolution = 100) %>%
autoplot(rug = TRUE, train = dat_train)
# 7.1. predict values for training and test data
predTrainOpt <- predict(gbm.fit.final, n.trees = gbm.fit.final$n.trees, X_train)
predOpt <- predict(gbm.fit.final, n.trees = gbm.fit.final$n.trees, X_test)
# 7.3. Print metrics and plot
print(paste("RMSE training data: ",
round(caret::RMSE(predTrainOpt, Y_train),2), "%"))
print(paste("RMSE test data: ",
round(caret::RMSE(predOpt, Y_test),2), "%"))
print(paste("Rsquare training data: ",
round(RSQUARE(Y_train, predTrainOpt),2)))
print(paste("Rsquare training data: ",
round(RSQUARE(Y_test, predOpt),2)))
par(mfrow=c(1,2))
plot(Y_train, predTrainOpt, ylim = c(0,10), xlim = c(0,10))
abline(0,1)
plot(Y_test, predOpt, ylim = c(0,10), xlim = c(0,10))
abline(0,1)
}