-
Notifications
You must be signed in to change notification settings - Fork 1
/
Collinearity_ENMEval.R
212 lines (185 loc) · 8.75 KB
/
Collinearity_ENMEval.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
#### SET-UP ####
# This script is part of Courtney Stuart's first MSc chapter in the lab of
# Dr. Stephanie Green at the University of Alberta (2019-2021). This script
# assesses correlation among a series of spatial predictors, including:
# benthic habitat classifications, bathymetric and topographic surfaces,
# and bottom water conditions. These data are specific to southern Florida
# and were examined using Pearson correlation coefficients and variance
# inflation factors (VIF) before being used to model habitat suitability for
# two critical reef fish in the Florida Keys - gray snapper (Lutjanus griseus
# (lg)) and bluestriped grunt (Haemulon sciurus (hs)). This script also
# includes functions used to optimize settings for the MaxEnt machine
# learning algorithm.
#### TO USE THIS FILE ####
# This is script 2 of 4 in Courtney's data analysis pipeline. To prepare
# the data required to run this script, please first run the
# Full_Data_Prep.R file (unless data are provided by directly by Courtney).
#### CONTACT ####
# Courtney Stuart (cestuart@ualberta.ca)
# libraries
library(easypackages)
libraries("raster", "sf", "tmap", "dplyr", "usdm", "rgdal", "usdm",
"sdmpredictors", "PNWColors", "corrplot", "Cairo")
# working directories
setwd("Z:/Courtney/Stuart_MSc_Ch1/") # main project folder
# data directories
temp_wd = "Z:/Courtney/Stuart_MSc_Ch1/Temporary/" # temporary files
csv_wd = "Z:/Courtney/Stuart_MSc_Ch1/GitHub/FL_Habitat_Suitability/Data/" # for writing tabular data
fish_wd = "Z:/Courtney/Stuart_MSc_Ch1/Species_Occurrence/" # for fish data
spatial_wd = "Z:/Courtney/Stuart_MSc_Ch1/Spatial_Predictors/" # for spatial predictor rasters
plots_wd = "Z:/Courtney/Stuart_MSc_Ch1/GitHub/FL_Habitat_Suitability/Figures/"
# save PROJ.4 string for standard projection (ESPG:26958 NAD 83/Florida East) and source LAT/LON data (EPSG:4326 WGS 84/World Geodetic System 1984)
my_crs = CRS("+init=epsg:26958")
gcs = CRS("+init=epsg:4326")
# load environmental rasters
habitat = raster(paste0(spatial_wd, "Habitat.asc"))
mg_dist = raster(paste0(spatial_wd, "Mangrove_Dist.asc"))
depth = raster(paste0(spatial_wd, "Depth.asc"))
sd_depth = raster(paste0(spatial_wd, "StDev_Depth.asc"))
slope = raster(paste0(spatial_wd, "Slope.asc"))
curvature = raster(paste0(spatial_wd, "Curvature.asc"))
plan_curve = raster(paste0(spatial_wd, "Plan_Curve.asc"))
bpi_fine = raster(paste0(spatial_wd, "BPI_Fine.asc"))
bpi_broad = raster(paste0(spatial_wd, "BPI_Broad.asc"))
rugosity = raster(paste0(spatial_wd, "Rugosity.asc"))
sum_temp = raster(paste0(spatial_wd, "Mean_Sum_Temp.asc"))
sum_do = raster(paste0(spatial_wd, "Mean_Sum_DO.asc"))
sum_sal = raster(paste0(spatial_wd, "Mean_Sum_Sal.asc"))
win_temp = raster(paste0(spatial_wd, "Mean_Win_Temp.asc"))
win_do = raster(paste0(spatial_wd, "Mean_Win_DO.asc"))
win_sal = raster(paste0(spatial_wd, "Mean_Win_Sal.asc"))
# define crs
crs(habitat) = my_crs
crs(mg_dist) = my_crs
crs(depth) = my_crs
crs(sd_depth) = my_crs
crs(slope) = my_crs
crs(curvature) = my_crs
crs(plan_curve) = my_crs
crs(bpi_fine) = my_crs
crs(bpi_broad) = my_crs
crs(rugosity) = my_crs
crs(sum_temp) = my_crs
crs(sum_do) = my_crs
crs(sum_sal) = my_crs
crs(win_temp) = my_crs
crs(win_do) = my_crs
crs(win_sal) = my_crs
# create raster stack
env = stack(x = c(habitat, mg_dist, depth, sd_depth, slope, curvature,
plan_curve, bpi_fine, bpi_broad, rugosity, sum_temp, sum_do,
sum_sal, win_temp, win_do, win_sal))
#### CORRELATION AND VIF ####
set.seed(42)
no_cores = parallel::detectCores() - 2
cl = snow::makeCluster(no_cores)
# full pearson correlation matrix on all spatial predictors
pearson = pearson_correlation_matrix(env)
head(pearson)
snow::stopCluster(cl)
# plot full correlation matrix
palette = pnw_palette("Shuksan2", 200, type = "continuous")
par(mar=c(0,0,0,0))
corrplot(pearson, method = "color", col = palette, type = "upper",
order = "original", addCoef.col = "black", number.cex = 0.55,
number.digits = 2, tl.col = "black", tl.srt = 40, tl.cex = 0.8)
# save plot as png
Cairo(file=paste0(plots_wd, "Correlation_Full_Predictor_Set.png"),
bg="white",
type="png",
units="in",
width=6,
height=5,
pointsize=12,
dpi=600)
par(mar=c(0,0,0,0))
corrplot(pearson, method = "color", col = palette, type = "upper",
order = "original", addCoef.col = "black", number.cex = 0.55,
number.digits = 2, tl.col = "black", tl.srt = 40, tl.cex = 0.8)
dev.off()
# save as CSV
write.csv(pearson, paste0(csv_wd, "Correlation_Full_Predictor_Set.csv"))
## examine correlation coefficients and variance inflation factors (VIF)
#install.packages('usdm')
library(usdm)
cl = snow::makeCluster(no_cores)
x = sampleRandom(env, 10000, na.rm = TRUE)
snow::stopCluster(cl)
vif = vif(as.data.frame(x))
vif
write.csv(vif, paste0(csv_wd, "VIF_Full_Predictor_Set.csv"), row.names = F)
# remove predictors that exceeded thresholds
# removing DO for both seasons, SD depth, and plan curvature
env2 = stack(x = c(habitat, mg_dist, depth, slope, curvature, bpi_fine,
bpi_broad, rugosity, sum_temp, sum_sal, win_temp, win_sal))
# reduced correlation matrix
cl = snow::makeCluster(no_cores)
pearson2 = (pearson_correlation_matrix(env2))
head(pearson2)
snow::stopCluster(cl)
# plot and save the reduced correlation matrix
corrplot(pearson2, method = "color", col = palette, type = "upper",
order = "original", addCoef.col = "black", number.cex = 0.55,
number.digits = 2, tl.col = "black", tl.srt = 40, tl.cex = 0.8)
Cairo(file=paste0(plots_wd, "Correlation_Reduced_Predictor_Set.png"),
bg="white",
type="png",
units="in",
width=6,
height=5,
pointsize=12,
dpi=300)
par(mar=c(0,0,0,0))
corrplot(pearson2, method = "color", col = palette, type = "upper",
order = "original", addCoef.col = "black", number.cex = 0.55,
number.digits = 2, tl.col = "black", tl.srt = 40, tl.cex = 0.8)
dev.off()
# save csv
write.csv(pearson2, paste0(csv_wd, "Correlation_Reduced_Predictor_Set.csv"))
# explore correlation coefficients and VIFs
cl = snow::makeCluster(no_cores)
x2 = sampleRandom(env2, 10000, na.rm = TRUE)
snow::stopCluster(cl)
vif2 = vif(as.data.frame(x2))
vif2
write.csv(vif2, paste0(csv_wd, "VIF_Reduced_Predictor_Set.csv"), row.names = F)
#### ENM EVALUTATE ####
#install.packages("rJava", dependencies = T)
#install.packages("ENMeval", dependecies = T)
#Sys.setenv(JAVA_HOME="C:/Program Files/Java/jdk-15.0.2/")
library(rJava)
library(ENMeval)
# need rJava first, find the dismo directory and move the maxent.jar file there (manually)
system.file("java", package = "dismo")
# Evaluate what the best settings are for MaxEnt
# need to read in the bias file created based on the distribution of survey effort,
# this will guide the selection of background points by MaxEnt
bias = raster(paste0(fish_wd, "Presence_Only/Bias.asc"))
crs(bias) = my_crs
# study domain is very large, so select 10,000 background points
# (the base settings for MaxEnt)
bg = as.data.frame(xyFromCell(bias, sample(which(!is.na(values(subset(env2, 1)))), 10000,
prob = values(bias)[!is.na(values(subset(env2, 1)))])))
bg = bg %>% dplyr::rename(LON_M = x, LAT_M = y)
# save to our temporary folder just in case
write.csv(bg, paste0(temp_wd, "Background_Points.csv"), row.names = FALSE)
# run evaluation using 10-fold cross-validation & background points selected based
# on bias file. Remember to specify that variable 1 in the raster stack (habitat)
# is categorical and use only the presence data !
enm_wd = "Z:/Courtney/Stuart_MSc_Ch1/ENMevaluate/"
lg_pres_only = read.csv(paste0(fish_wd, "Presence_Only/Subadult_Gray_Snapper_PO_Train.csv"))[,-1]
lg_enm_eval = ENMevaluate(lg_pres_only, env2, method = "randomkfold", kfolds = 10,
categoricals = 1, algorithm = 'maxent.jar', bg.coords = bg,
RMvalues = c(0.25, 0.50, 1.0, 2.0, 5.0),
fc = c("L", "LQ", "LQH", "LQHP"),
parallel = TRUE, numCores = no_cores, progbar = TRUE,
updateProgress = TRUE)
write.csv(lg_enm_eval@results, paste(enm_wd, "Subadult_Gray_Snapper_ENMeval.csv"))
hs_pres_only = read.csv(paste0(fish_wd, "Presence_Only/Subadult_Bluestriped_Grunt_PO_Train.csv"))[,-1]
hs_enm_eval = ENMevaluate(hs_pres_only, env2, method = "randomkfold", kfolds = 10,
categoricals = 1, algorithm = 'maxent.jar', bg.coords = bg,
RMvalues = c(0.25, 0.50, 1.0, 2.0, 5.0),
fc = c("L", "LQ", "LQH", "LQHP"),
parallel = TRUE, numCores = no_cores, progbar = TRUE,
updateProgress = TRUE)
write.csv(hs_enm_eval@results, paste(enm_wd, "Subadult_Bluestriped_Grunt_ENMeval.csv"))