Newer
Older
---
output:
pdf_document: default
html_document: default
---
Diurnal Temperature Variation (DTR) Indicator
=============================================
The diurnal temperature variation indicator is a proxy for energy demand. The diurnal temperature indicator is the number of days in a season where the daily temperature variation (tasmax - tasmin) exceeds the vulnerability threshold. Here, the vulnerability threshold is based on the mean daily temperature variation for a reference period plus 5 degrees.
### 1- Load dependencies
This example requires the following system libraries:
- libssl-dev
- libnecdf-dev
- cdo
The **ClimProjDiags R package** should be loaded by running the following lines in R, once it's integrated into CRAN mirror.
library(ClimProjDiags)
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
```
All the other R packages involved can be installed directly from CRAN and loaded as follows:
```r
library(s2dverification)
library(abind)
library(parallel)
```
### 2- Problem, parameters and data definition
To ilustrate the problem, daily maximum or minimum air temperature at 2 m for the reference period (1971 - 2000) and the future scenario rcp2.6 (2006 - 2100) in the northern hemisphere (between -40 - 20 ºE and 25 - 60 ºN) will be created artificially.
A grid of 5 degrees is defined by running the next lines in R:
```r
lat <- seq(25, 60, 5)
lon <- seq(-35, 20 ,5)
```
The synthetic sample of maximum and minimum temperature for the historical periode can be obtained by running the following lines.The maximum temperature data is built by adding random perturbation to a sinusoidal function. The latitudinal behavior of the temperature is considered by subtracting randomly a value proportional to the latitud. The minimum temperature is built by subtracting a random value (with anual cycle included) to the synthetic maximum temperature. Furthermore, attributes of time and dimensions are added to both samples.
```
tmax_historical <- NULL
tmin_historical <- NULL
grid1 <- 293 - 10 * cos(2 * pi / 365 * (1 : 10958)) + rnorm(10958)
grid2 <- 289 - 8 * cos(2 * pi / 365 * (1 : 10958)) - abs(rnorm(10958))
gridlon1 <- NULL
gridlon2 <- NULL
for (i in 1 : 12) {
gridlon1 <- cbind(gridlon1,
grid1 + abs(rnorm(10958, mean = 2, sd = 1) *
cos(2 * pi / 365 * (1 : 10958))))
gridlon2 <- cbind(gridlon2,
grid2 - abs(rnorm(10958, mean = 2, sd = 1) *
cos(2 * pi / 365 * (1 : 10958))))
}
for (j in 1 : 8) {
gridnew1 <- apply(gridlon1, 2, function(x) {x - rnorm(10958, mean = j * 0.001,
sd = 0.1)})
gridnew2 <- apply(gridlon2, 2, function(x) {x -
abs(rnorm(10958, mean = j * 0.75, sd = 0.5))})
tmax_historical <- abind(tmax_historical, gridnew1, along = 3)
tmin_historical <- abind(tmin_historical, gridnew2, along = 3)
}
tmax_historical <- InsertDim(InsertDim(tmax_historical, posdim = 1, lendim = 1),
posdim = 1, lendim = 1)
tmin_historical <- InsertDim(InsertDim(tmin_historical, posdim = 1, lendim = 1),
posdim = 1, lendim = 1)
names(dim(tmax_historical)) <- c("model", "var", "time", "lon", "lat")
names(dim(tmin_historical)) <- c("model", "var", "time", "lon", "lat")
time <- seq(ISOdate(1971, 1, 1), ISOdate(2000, 12, 31), "day")
metadata <- list(time = list(standard_name = 'time', long_name = 'time',
calendar = 'proleptic_gregorian',
units = 'days since 1970-01-01 00:00:00', prec = 'double',
dim = list(list(name = 'time', unlim = FALSE))))
attr(time, "variables") <- metadata
attr(tmax_historical, 'Variables')$dat1$time <- time
attr(tmin_historical, 'Variables')$dat1$time <- time
```
A similar procedure is considered to build the synthetic data for the future projections. However, a trend is added.
```
tmax_projection <- NULL
tmin_projection <- NULL
grid1 <- 293 - 10 * cos(2 * pi / 365 * (1 : 34698)) + rnorm(34698) +
(1 : 34698) * rnorm(1, mean = 4) / 34698
grid2 <- 289 - 8 * cos(2 * pi / 365 * (1 : 34698)) - abs(rnorm(34698)) +
(1 : 34698) * rnorm(1, mean = 4) / 40000
gridlon1 <- NULL
gridlon2 <- NULL
for (i in 1 : 12) {
gridlon1 <- cbind(gridlon1,
grid1 + abs(rnorm(34698, mean = 2, sd = 1) *
cos(2 * pi / 365 * (1 : 34698))))
gridlon2 <- cbind(gridlon2,
grid2 - abs(rnorm(34698, mean = 2, sd = 1) *
cos(2 * pi / 365 * (1 : 34698))))
}
for (j in 1 : 8) {
gridnew1 <- apply(gridlon1, 2, function(x) {x - rnorm(34698, mean = j * 0.01,
sd = 0.1)})
gridnew2 <- apply(gridlon2, 2, function(x) {x -
abs(rnorm(34698, mean = j * 0.75, sd = 0.5))})
tmax_projection <- abind(tmax_projection, gridnew1, along = 3)
tmin_projection <- abind(tmin_projection, gridnew2, along = 3)
}
tmax_projection <- InsertDim(InsertDim(tmax_projection, posdim = 1, lendim = 1),
posdim = 1, lendim = 1)
tmin_projection <- InsertDim(InsertDim(tmin_projection, posdim = 1, lendim = 1),
posdim = 1, lendim = 1)
names(dim(tmax_projection)) <- c("model", "var", "time", "lon", "lat")
names(dim(tmin_projection)) <- c("model", "var", "time", "lon", "lat")
time <- seq(ISOdate(2006, 1, 1), ISOdate(2100, 12, 31), "day")
metadata <- list(time = list(standard_name = 'time', long_name = 'time',
calendar = 'proleptic_gregorian',
units = 'days since 1970-01-01 00:00:00', prec = 'double',
dim = list(list(name = 'time', unlim = FALSE))))
attr(time, "variables") <- metadata
attr(tmax_projection, 'Variables')$dat1$time <- time
attr(tmin_projection, 'Variables')$dat1$time <- time
```
### 3- Reference diurnal temperature variation
The function `DTRRef`, from the **ClimProjDiags package**, computes the mean between maximum and minimum temperature for each season during the selected reference period:
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
```r
dtr_reference <- DTRRef(tmax = tmax_historical, tmin = tmin_historical, by.seasons = TRUE, ncores = NULL)
```
The label `dtr_reference` contains the output for each season and gridpoint:
```r
> str(dtr_reference)
List of 2
$ dtr.ref: num [1:4, 1, 1, 1:12, 1:8] 5.09 2.9 5.04 2.91 5.09 ...
$ season : chr [1:4] "DJF" "MAM" "JJA" "SON"
```
### 4- Computing the diurnal temperature variation indicator
The diurnal temperature variation indicator is computed with the function `DTRIndicator` indicating the maximum and minimum temperature for the future projectionn and the reference temperature variation:
```r
dtr_indicator <- DTRIndicator(tmax = tmax_projection, tmin = tmin_projection,
ref = dtr_reference,
by.seasons = TRUE, ncores = NULL)
```
The function returns a list of three elements, being the label `indicator` the desired output:
```r
> str(dtr_indicator)
List of 3
$ indicator: int [1:96, 1:4, 1, 1, 1:12, 1:8] 0 1 0 0 0 2 0 1 1 0 ...
$ year : chr [1:96] "2006" "2007" "2008" "2009" ...
$ season : chr [1:4] "DJF" "JJA" "MAM" "SON"
```
*Note: the total number of years in `dtr_indicator$year` (96) is greater than the period examined (in this case 95 years of future projection). It's due to the last december of the time series, which belongs to the next winter (in this example 2101 winter).*
### 5- Visualizing the diurnal temperature variation indicator
A four panel plot can be generated to visualize the seasonal indicator running the following lines:
```r
dtr_rcp <- array(dim = c(length(lon), length(lat), 4))
for (j in 1:4){
dtr_rcp[,,j] <- Mean1Dim(dtr_indicator$indicator[,j,,,,], 1)
}
breaks <- 0 : (max(dtr_rcp) + 1)
PlotLayout(PlotEquiMap, c(1, 2), lon = lon, lat = lat, var = dtr_rcp,
titles = c('DJF', 'MAM', 'JJA', 'SON'),
toptitle = "DTR", filled.continents = FALSE,
units = "Days", title_scale = 0.5, axelab = FALSE,
draw_separators = TRUE, subsampleg = 1,
brks = breaks, color_fun = clim.palette("yellowred"),
bar_extra_labels = c(2, 0, 0, 0),
fileout = "SpatialDTR.png")
```
Furthermore, the future diurnal temperature variation can be compared with the observed during the reference period. So, the diurnal temperature variation indicator is computed for the reference period by running the next lines:
```r
dtr_indicator_reference <- DTRIndicator(tmax = tmax_historical,
tmin = tmin_historical,
ref = dtr_reference, by.seasons = TRUE,
ncores = NULL)
```
By ordinary subtraction, the comparison between the reference and the future projection diurnal temperature variation indicator can be computed. To visualize the result, `PlotLayout` function will be applied again. The resulting plot, will be saved in the working directory.
```r
dtr_diff <- array(dim=c(length(lat), length(lon), 4))
for (i in 1:4){
dtr_diff[,,i] <- Mean1Dim(dtr_indicator$indicator[,i,1,1,,], 1) -
Mean1Dim(dtr_indicator_reference$indicator[,i,1,1,,], 1)
}
PlotLayout(PlotEquiMap, c(1, 2), lon = lon, lat = lat, var = dtr_diff,
titles = c('DJF', 'MAM', 'JJA', 'SON'),
toptitle = "Change in DTR indicator",
filled.continents = FALSE, units = "Days",
axelab = FALSE, draw_separators = TRUE, subsampleg = 1,
brks = -2 : 2, bar_extra_labels = c(2, 0, 0, 0),
fileout = "SpatialComparisonDTR.png")
```
*Note: the outputs of this example, including the figures, are artificial.*