-
Notifications
You must be signed in to change notification settings - Fork 1
/
ghg-processor.qmd
793 lines (650 loc) · 26.5 KB
/
ghg-processor.qmd
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
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
---
title: "GHG processor"
author: "Bond-Lamberty / Wilson"
title-block-banner: true
params:
html_outfile: "ghg-processor.html"
data_folder: "data/"
output_folder: "output/"
metadata_folder: "data/"
MSMT_LENGTH_SEC: 300
DEAD_BAND_SEC: 10
SAVE_RESULTS_FIGURES: true
date: now
date-format: "YYYY-MM-DD HH:mm:ssZ"
format:
html:
toc: true
code-fold: true
editor: visual
---
# Initializing
```{r setup}
#| include: false
library(readr)
library(lubridate)
library(ggplot2)
theme_set(theme_bw())
library(ggrepel)
library(readxl)
library(broom)
library(DT)
library(MASS)
library(tidyr)
# These columns are REQUIRED to be present in the metadata file(s)
# (NB there are can be other columns present; they will be carried through to the results)
METADATA_REQUIRED_FIELDS <- c("Date", "Time_start", "Plot", "Area", "Volume", "Temp")
# Data files must end with ".txt" and may be in sub-folders
data_files <- list.files(params$data_folder,
pattern = "*txt$",
full.names = TRUE, recursive = TRUE)
# Metadata files must end with ".xlsx" and may be in sub-folders
metadata_files <- list.files(params$metadata_folder,
pattern = "*xlsx$",
full.names = TRUE, recursive = TRUE)
```
I see `r length(data_files)` data files to process.
I see `r length(metadata_files)` metadata files to process.
Measurement length is `r params$MSMT_LENGTH_SEC` seconds.
Default dead band time is `r params$DEAD_BAND_SEC` seconds.
Required metafields are: `r paste(METADATA_REQUIRED_FIELDS, collapse = ", ")`.
Writing output to `r normalizePath(params$output_folder)`.
HTML outfile is `r params$html_outfile`.
Working directory is `r getwd()`.
# Data
## Read in GHG data
```{r read-data}
errors <- 0 # error count
# Function to read a data file given by 'fn', filename
# Returns NULL if there's an error reading
read_ghg_data <- function(fn) {
basefn <- basename(fn)
message(Sys.time(), " Processing ", basefn)
dat_raw <- readLines(fn)
# Save the serial number in case we want to look at machine differences later
sn <- trimws(gsub("SN:\t", "", dat_raw[2], fixed = TRUE))
# Parse the timezone from the header and use it to make a TIMESTAMP field
tz <- trimws(gsub("Timezone:\t", "", dat_raw[5], fixed = TRUE))
# These files have five header lines, then the names of the columns in line 6,
# and then the column units in line 7. We only want the names
dat_raw <- dat_raw[-c(1:5, 7)]
# Irritatingly, the units line can repeat in the file (???!?). Remove these
dat_raw <- dat_raw[grep("DATAU", dat_raw, invert = TRUE)]
# Double irritatingly, if there's no remark, the software write \t\t, not
# \t""\t, causing a read error. Replace these instances
dat_raw <- gsub("\t\t", "\tnan\t", dat_raw, fixed = TRUE)
dat <- try({
readr::read_table(I(dat_raw), na = "nan", guess_max = 1e4)
})
# If the try() above succeeded, we haev a data frame and can process it
if(is.data.frame(dat)) {
message("\tRead in ", nrow(dat), " rows of data, ",
min(dat$DATE), " to ", max(dat$DATE))
message("\tInstrument serial number: ", sn)
dat$SN <- sn
message("\tInstrument time zone: ", tz)
# We record the time zone but don't convert the timestamps to it as
# that's not needed right now; metadata time is guaranteed to be same
dat$TIMESTAMP <- lubridate::ymd_hms(paste(dat$DATE, dat$TIME))
dat$TZ <- tz
# Remove unneeded Licor DATE and TIME columns
dat$DATE <- dat$TIME <- NULL
dat$Data_file <- basefn
return(dat)
} else {
warning("File read error for ", basefn)
errors <<- errors + 1
return(NULL)
}
}
dat <- lapply(data_files, read_ghg_data)
dat <- do.call("rbind", dat)
```
Errors: `r errors`
```{r}
#| echo: false
#| output: asis
if(errors) cat("**<span style='color:red'>---> At least one error occurred! <---</span>**")
```
Total observations: `r format(nrow(dat), big.mark = ",")`
## Read in metadata
```{r read-metadata}
errors <- 0
# Function to read a metadata file given by 'fn', filename
# Returns NULL if there's an error reading
read_metadata <- function(fn) {
basefn <- basename(fn)
message(Sys.time(), " Processing ", basefn)
metadat_raw <- try({
read_excel(fn)
})
if(is.data.frame(metadat_raw)) {
if(!all(METADATA_REQUIRED_FIELDS %in% names(metadat_raw))) {
warning("Metadata file ", basefn, " doesn't have required fields!")
return(NULL)
}
metadat_raw$Metadata_file <- basefn
# Note that it's guaranteed that metadata time is instrument time,
# so we don't worry about any time zones right now
metadat <- metadat_raw
# Metadata files can, optionally, have "Dead_band" and "Msmt_length" columns
# If not present, insert those columns with NA values
if(!"Dead_band" %in% names(metadat)) {
metadat$Dead_band <- NA_real_
} else {
message("\tCustom Dead_band column exists")
}
if(!"Msmt_stop" %in% names(metadat)) {
metadat$Msmt_stop <- NA_real_
} else {
message("\tCustom Msmt_stop column exists")
}
} else {
warning("File read error for ", basefn)
errors <<- errors + 1
return(NULL)
}
metadat
}
metadat <- lapply(metadata_files, read_metadata)
metadat <- do.call("rbind", metadat)
if(!nrow(metadat)) stop("No metadata read!")
metadat$Obs <- seq_len(nrow(metadat))
```
Errors: `r errors`
Total metadata rows: `r nrow(metadat)`
## Metadata-data matching
```{r join-data}
# For each row of metadata, find corresponding observational data
zero_matches <- data.frame()
match_info <- data.frame()
matched_dat <- list()
for(i in seq_len(nrow(metadat))) {
ts <- metadat$Time_start[i]
start_time <- metadat$Date[i] + hour(ts) * 60 * 60 + minute(ts) * 60 + second(ts)
end_time <- start_time + params$MSMT_LENGTH_SEC
# Subset data following timestamp in metadata and store
x <- subset(dat, TIMESTAMP >= start_time & TIMESTAMP < end_time)
# message("Metadata row ", i, " ", metadat$Plot[i], " start = ", start_time, " matched ", nrow(x), " data rows")
info <- data.frame(Row = i,
Plot = metadat$Plot[i],
start_time = start_time,
Rows_matched = nrow(x),
Min_timestamp = NA,
Max_timestamp = NA)
if(nrow(x)) {
# We have matched data!
x$Obs <- i
# Assign dead band: value given in Quarto params, but can be overridden in metadata
if(is.na(metadat$Dead_band[i])) {
x$DEAD_BAND_SEC <- params$DEAD_BAND_SEC
} else {
x$DEAD_BAND_SEC <- metadat$Dead_band[i]
}
# Assign stop time (in seconds, not including dead band): value given in
# Quarto params, but can be overridden in metadata
if(is.na(metadat$Msmt_stop[i])) {
x$MSMT_STOP_SEC <- params$MSMT_LENGTH_SEC
} else {
x$MSMT_STOP_SEC <- metadat$Msmt_stop[i]
}
# Calculate ELAPSED_SECS, which is relative time (in s) since start of measurement
x$ELAPSED_SECS <- time_length(interval(min(x$TIMESTAMP), x$TIMESTAMP), "seconds")
info$Min_timestamp <- min(x$TIMESTAMP)
info$Max_timestamp <- max(x$TIMESTAMP)
} else {
# No data matched the date and time given in this row of the metadata
message("\tWARNING; file: ", metadat$File[i])
zero_matches <- rbind(zero_matches, data.frame(File = metadat$File[i],
Row = i,
Date = metadat$Date[i],
Time_start = metadat$Time_start[i]))
}
match_info <- rbind(match_info, info)
matched_dat[[i]] <- x
}
# Combine all subsetted data together and merge with metadata
combined_dat <- do.call("rbind", matched_dat)
combined_dat <- merge(metadat, combined_dat, by = "Obs")
# Compute a few basic stats about the combined data
n <- nrow(combined_dat)
neg_CO2 <- sum(combined_dat$CO2 < 0, na.rm = TRUE)
na_CO2 <- sum(is.na(combined_dat$CO2), na.rm = TRUE)
na_CH4 <- sum(is.na(combined_dat$CH4), na.rm = TRUE)
datatable(match_info)
```
Checking for metadata rows with no matching data...
```{r}
#| echo: false
#| output: asis
if(nrow(zero_matches)) {
cat("**<span style='color:red'>---> At least one metadata row had no matching data! <---</span>**\n\n")
}
```
```{r}
# Table of zero-match data, if any
if(nrow(zero_matches)) {
knitr::kable(zero_matches)
}
```
Checking for unused data files...
```{r}
#| echo: false
#| output: asis
unused_files <- setdiff(unique(dat$Data_file), unique(combined_dat$Data_file))
if(length(unused_files)) {
cat("**<span style='color:red'>---> At least one data file was unused! <---</span>**\n\n")
}
```
```{r}
# Table of unused files, if any
if(length(unused_files)) {
knitr::kable(data.frame(Data_file = unused_files))
}
```
Total rows of combined data: `r format(nrow(combined_dat), big.mark = ",")`
## CO2 and CH4 measurement data
Observations with negative CO2: `r neg_CO2` (`r round(neg_CO2 / n * 100, 0)`%)
Observations with missing CO2: `r na_CO2` (`r round(na_CO2 / n * 100, 0)`%)
Distribution of CO2 values:
```{r}
# Print distribution and (below) plot random subsample of CO2 data
OVERALL_PLOT_N <- 300
summary(combined_dat$CO2)
```
Plot of `r OVERALL_PLOT_N` random CO2 values (y axis 5%-95%):
```{r plot-obs}
#| fig-width: 8
#| fig-height: 4
ylim_co2 <- quantile(combined_dat$CO2, probs = c(0.05, 0.95), na.rm = TRUE)
dat_small <- combined_dat[sample.int(n, OVERALL_PLOT_N, replace = TRUE),]
ggplot(dat_small, aes(TIMESTAMP, CO2, color = Plot)) +
geom_point(na.rm = TRUE) +
coord_cartesian(ylim = ylim_co2)
```
Observations with missing CH4: `r na_CH4` (`r round(na_CH4 / n * 100, 0)`%)
Distribution of CH4 values:
```{r}
# Print distribution and (below) plot random subsample of CH4 data
summary(combined_dat$CH4)
```
Plot of `r OVERALL_PLOT_N` random CH4 values (y axis 5%-95%):
```{r plot-ch4}
#| fig-width: 8
#| fig-height: 4
ylim_ch4 <- quantile(combined_dat$CH4, probs = c(0.05, 0.95), na.rm = TRUE)
dat_small <- combined_dat[sample.int(n, OVERALL_PLOT_N, replace = TRUE),]
ggplot(dat_small, aes(TIMESTAMP, CH4, color = Plot)) +
geom_point(na.rm = TRUE) +
coord_cartesian(ylim = ylim_ch4)
```
## Observations by plot and day
```{r plot-co2}
#| fig-width: 8
ggplot(combined_dat, aes(ELAPSED_SECS, CO2, group = Obs)) +
geom_point(na.rm = TRUE, color = "purple") +
facet_wrap(~Plot, scales = "free") +
# Plot both linear and curvilinear fits
#geom_smooth(method = lm, formula = 'y ~ poly(x, 2)', na.rm = TRUE, linetype = 2, color = "blue") +
geom_smooth(method = lm, formula = 'y ~ x', na.rm = TRUE, color = "black", linewidth = 0.5) +
geom_vline(aes(xintercept = DEAD_BAND_SEC), linetype = 2, color = "green") +
geom_vline(aes(xintercept = MSMT_STOP_SEC), linetype = 2, color = "red")
ggplot(combined_dat, aes(ELAPSED_SECS, CH4, group = Obs)) +
geom_point(na.rm = TRUE, color = "blue") +
facet_wrap(~Plot, scales = "free") +
#geom_smooth(method = lm, formula = 'y ~ poly(x, 2)', na.rm = TRUE, linetype = 2, color = "blue") +
geom_smooth(method = lm, formula = 'y ~ x', na.rm = TRUE, color = "black", linewidth = 0.5) +
geom_vline(aes(xintercept = DEAD_BAND_SEC), linetype = 2, color = "green") +
geom_vline(aes(xintercept = MSMT_STOP_SEC), linetype = 2, color = "red")
```
# Fluxes
## Model fitting
```{r model-fitting}
model_fit_error <- FALSE
# Fit a model to units of gas per day
fit_model <- function(df, depvar) {
# Metadata
info <- data.frame(Obs = df$Obs[1],
TIMESTAMP = mean(df$TIMESTAMP),
DEAD_BAND_SEC = unique(df$DEAD_BAND_SEC),
MSMT_STOP_SEC = unique(df$MSMT_STOP_SEC),
TZ = df$TZ[1],
SN = df$SN[1],
Data_file = df$Data_file[1],
Gas = depvar,
Volume = df$Volume[1],
Temp = df$Temp[1],
Area = df$Area[1]
)
# Remove the dead band data
df <- df[df$ELAPSED_SECS > df$DEAD_BAND_SEC,]
# Remove the after-end data
df <- df[df$ELAPSED_SECS <= df$MSMT_STOP_SEC,]
# Fit a linear model
try(mod <- lm(df[,depvar] ~ df$ELAPSED_SECS))
if(!exists("mod")) {
message("ERROR fitting model for obs ", df$Obs[1], " ", df$Plot[1], " ", depvar)
model_fit_error <- model_fit_error & FALSE
}
# Model statistics
model_stats <- glance(mod)
# Slope and intercept statistics
slope_stats <- tidy(mod)[2,-1]
names(slope_stats) <- paste("slope", names(slope_stats), sep = "_")
intercept_stats <- tidy(mod)[1,-1]
names(intercept_stats) <- paste("int", names(intercept_stats), sep = "_")
# Add robust regression slope as a QA/QC check
robust <- MASS::rlm(df[,depvar] ~ df$ELAPSED_SECS)
slope_stats$slope_estimate_robust <- coefficients(robust)[2]
# Add polynomial regression R2 as a QA/QC check
poly <- lm(df[,depvar] ~ poly(df$ELAPSED_SECS, 3))
slope_stats$r.squared_poly <- summary(poly)$r.squared
res <- cbind(info, model_stats, slope_stats, intercept_stats)
# Round to a sensible number of digits and return
numerics <- sapply(res, is.numeric)
res[numerics] <- round(res[numerics], 3)
res
}
results_ch4 <- lapply(split(combined_dat, combined_dat$Obs), fit_model, "CH4")
results_ch4 <- do.call("rbind", results_ch4)
results_co2 <- lapply(split(combined_dat, combined_dat$Obs), fit_model, "CO2")
results_co2 <- do.call("rbind", results_co2)
```
```{r}
#| echo: false
#| output: asis
if(model_fit_error) cat("**<span style='color:red'>---> At least one model didn't fit! <---</span>**")
```
## Unit conversion
```{r unit-conversion}
required_fields <- c("Volume", "Temp")
if(!all(required_fields %in% names(combined_dat))) {
stop("We need volume ('Volume') and temperature ('Temp')")
}
# The slopes and their errors, from model-fitting above, are in ppb/s (CH4)
# and ppm/s (CO2). Convert them to µmol/m2/day and mmol/m2/day respectively.
# CH4
# Use known volume of chamber to convert ppb/s to liters/s:
# ppb CH4 * volume of chamber (m3) * 1000 L per m3
CH4_slope_L <- with(results_ch4, (slope_estimate / 1e9) * Volume * 1000)
CH4_slope_err_L <- with(results_ch4, (slope_std.error / 1e9) * Volume * 1000)
# Use ideal gas law to calculate µmol of CH4 per m2 ground:
# (atm pressure * L CH4) / (R [in L*atm / T[K] * mol] * T[K]) * 10^6 µmol/mol
TEMP_K <- with(results_ch4, Temp + 273.15)
AREA <- results_ch4$Area
S_PER_DAY <- 60 * 60 * 24
results_ch4$flux_estimate <- (1 * CH4_slope_L) / (0.08206 * TEMP_K) * 1e6 / AREA * S_PER_DAY
results_ch4$flux_units <- "µmol/m2/day"
results_ch4$flux_std.error <- (1 * CH4_slope_err_L) / (0.08206 * TEMP_K) * 1e6 / AREA * S_PER_DAY
# CO2
# Use known volume of chamber to convert ppm/s to liters/s:
# ppm CO2 * volume of chamber (m3) * 1000 L per m3
CO2_slope_L <- with(results_co2, (slope_estimate / 1e9) * Volume * 1000)
CO2_slope_err_L <- with(results_co2, (slope_std.error / 1e9) * Volume * 1000)
# Use ideal gas law to calculate mmol of CO2 per m2 ground:
# (atm pressure * L CO2) / (R [in L*atm / T[K] * mol] * T[K]) * 10^3 mmol/mol
#combined_dat$`CO2 mmol/m2` <- (1 * CO2_L) / (0.08206 * TEMP_K) * 1e3 / AREA
TEMP_K <- with(results_co2, Temp + 273.15)
AREA <- results_co2$Area
results_co2$flux_estimate <- (1 * CO2_slope_L) / (0.08206 * TEMP_K) * 1e6 / AREA * S_PER_DAY
results_co2$flux_units <- "mmol/m2/day"
results_co2$flux_std.error <- (1 * CO2_slope_err_L) / (0.08206 * TEMP_K) * 1e6 / AREA * S_PER_DAY
# Combine CH4 and CO2 results and re-merge with metadata
results <- rbind(results_co2, results_ch4)
# Drop columns we don't need because they'll come in with the merge
results$Area <- results$Temp <- results$Volume <- NULL
# Re-merge with metadata
results <- as_tibble(merge(results, metadat, by = "Obs"))
```
Total rows of results: `r format(nrow(results), big.mark = ",")`
## Flux summary
```{r flux-summary}
#| fig-width: 8
# Flux table and visualization
formatRound(datatable(results),
which(sapply(results, is.numeric)), digits = 3)
ggplot(results, aes(x = TIMESTAMP, y = flux_estimate, color = Plot)) +
geom_point() +
geom_errorbar(aes(ymin = flux_estimate - flux_std.error,
ymax = flux_estimate + flux_std.error)) +
geom_line(aes(group = Plot)) +
facet_wrap(~paste(Gas, flux_units) + Site, scales = "free")
```
# QA/QC
```{r qaqc-helpers}
#| include: false
# General function: plot distribution of a single column,
# identify outliers
onecol_plot_and_flag <- function(results, col, probs = c(0.05, 0.95),
left = TRUE, right = TRUE) {
p <- ggplot(results, aes(.data[[col]])) +
geom_histogram(bins = 30) +
facet_wrap(Gas ~., scales = "free", ncol = 1)
print(p)
# Find which ones are outside of quantiles and flag
q_flags <- rep(FALSE, nrow(results))
co2 <- results$Gas == "CO2"
x <- as.data.frame(results)
q_co2 <- quantile(x[co2, col], probs = probs)
q_ch4 <- quantile(x[!co2, col], probs = probs)
if(left) q_flags[co2] <- q_flags[co2] | x[co2, col] < min(q_co2)
if(right) q_flags[co2] <- q_flags[co2] | x[co2, col] > max(q_co2)
if(left) q_flags[!co2] <- q_flags[!co2] | x[!co2, col] < min(q_ch4)
if(right) q_flags[!co2] <- q_flags[!co2] | x[!co2, col] > max(q_ch4)
# Create plot labels for all data, then NA the ones inside the cutoff
results$lab <- x$Obs
results$lab[!q_flags] <- NA_character_
p <- ggplot(results, aes(.data[[col]], flux_estimate)) +
geom_point() +
facet_wrap(~Gas, scales = "free") +
geom_rug(sides = "b", alpha = 0.3) +
geom_label_repel(aes(label = lab), size = 2.5, na.rm = TRUE)
print(p)
return(q_flags)
}
# General function: plot two columns against each other, coloring by
# standardized residual. Return vector of T/F flags if s.resid > cutoff
twocol_plot_and_flag <- function(results, col1, col2,
add_1to1_line = FALSE,
rstandard_cutoff = 2) {
results <- as.data.frame(results)
# Identify standardized residuals outside of our cutoff
m <- lm(results[,col2] ~ results[,col1])
results$rstandard <- round(abs(rstandard(m)), 2)
resid_flags <- results$rstandard > rstandard_cutoff
# Create plot labels for all data, then NA the ones inside the cutoff
results$lab <- with(results, paste(paste0("#", Obs),
Date, "\n",
Plot,
"\nsresid =", rstandard))
results$lab[!resid_flags] <- NA_character_
p <- ggplot(results, aes(.data[[col1]], .data[[col2]])) +
geom_point()
if(add_1to1_line) p <- p + geom_abline()
p <- p +
geom_label_repel(aes(label = lab), size = 2.5, na.rm = TRUE) +
facet_wrap(Gas ~., scales = "free", ncol = 1)
print(p)
return(resid_flags)
}
# General function: given a results subset, reshape to long form,
# merge with concentration data, and plot for closer inspection
# Typically code will call twocol_plot_and_flag above to identify potential
# issues, and then this function to plot data in detail
plot_group <- function(results, combined_dat) {
plotdata <- merge(results[c("Obs", "adj.r.squared")], combined_dat, by = "Obs")
plotdata_long <- tidyr::pivot_longer(plotdata, c("CO2", "CH4"),
names_to = "Gas")
model_data <- subset(plotdata_long,
ELAPSED_SECS > DEAD_BAND_SEC & ELAPSED_SECS <= MSMT_STOP_SEC)
ggplot(plotdata_long, aes(ELAPSED_SECS, value, group = Obs, color = Gas)) +
scale_color_manual(values = c("CH4" = "blue", "CO2" = "purple")) +
geom_point(na.rm = TRUE) +
geom_vline(aes(xintercept = DEAD_BAND_SEC), linetype = 2, color = "green") +
geom_vline(aes(xintercept = MSMT_STOP_SEC), linetype = 2, color = "red") +
facet_wrap(~paste(paste0("#", Obs), date(TIMESTAMP)) + paste(Plot, Gas),
scales = "free") +
theme(strip.text = element_text(size = 8)) +
geom_smooth(data = model_data,
method = lm, formula = 'y ~ x', na.rm = TRUE, color = "black")
}
```
## R2 distribution
```{r qaqc-r2}
#| fig-width: 8
#| fig-height: 6
results$FLAG_R2 <- onecol_plot_and_flag(results,
"adj.r.squared",
right = FALSE)
plot_group(subset(results, FLAG_R2), combined_dat)
```
## y-intercept distribution
```{r qaqc-yint}
#| fig-width: 8
#| fig-height: 6
# The intercept is quite variable, so use {0.01, 0.99}
results$FLAG_INTERCEPT <- onecol_plot_and_flag(results, "int_estimate",
probs = c(0.01, 0.99))
plot_group(subset(results, FLAG_INTERCEPT), combined_dat)
```
## Robust regression divergence
This can indicate outlier problems or ebullition events.
```{r qaqc-robust}
#| fig-width: 8
#| fig-height: 6
results$FLAG_ROBUST_DIV <- twocol_plot_and_flag(results,
"slope_estimate",
"slope_estimate_robust",
add_1to1_line = TRUE)
plot_group(subset(results, FLAG_ROBUST_DIV), combined_dat)
```
## Polynomial regression divergence
This can indicate curvature of the gas concentrations time series due to saturation, etc.
```{r qaqc-poly}
#| fig-width: 8
#| fig-height: 6
results$FLAG_POLY_DIV <- twocol_plot_and_flag(results,
"adj.r.squared",
"r.squared_poly",
add_1to1_line = TRUE)
plot_group(subset(results, FLAG_POLY_DIV), combined_dat)
```
## QA/QC summary
```{r qaqc-summary}
res1 <- results["Obs"]
res2 <- results[grepl("^FLAG_", names(results))]
res <- pivot_longer(cbind(res1, res2), starts_with("FLAG"), names_to = "Flag")
res <- subset(res, value)
f <- function(x) paste(unique(x), collapse = ", ")
knitr::kable(aggregate(Obs ~ Flag, data = res, f))
```
```{r}
#| include: false
# -----------------------------------------------------------------------------
is_outlier <- function(x, devs = 5.2) {
# See: Davies, P.L. and Gather, U. (1993).
# "The identification of multiple outliers" (with discussion)
# J. Amer. Statist. Assoc., 88, 782-801.
x <- na.omit(x)
lims <- median(x) + c(-1, 1) * devs * mad(x, constant = 1)
return(x < lims[1] | x > lims[2])
}
```
# Output
## Data
```{r write-output}
now_string <- function() format(Sys.time(), "%Y-%m-%d")
mkfn <- function(name, extension) { # "make filename" helper function
file.path(params$output_folder,
paste(paste(name, now_string(), sep = "_"), extension, sep = "."))
}
flux_fn <- mkfn("results", "csv")
message("Writing final flux dataset ", flux_fn, "...")
write_csv(results, flux_fn)
obs_fn <- mkfn("observations", "csv")
message("Writing compiled observations dataset ", obs_fn, "...")
write_csv(combined_dat, obs_fn)
if(params$SAVE_RESULTS_FIGURES) {
message("Writing plot outputs...", appendLF = FALSE)
# Loop through each line, generating plot and filename and saving
for(i in seq_len(nrow(results))) {
p <- plot_group(results[i,], combined_dat)
fn <- mkfn(paste("obs", results$Obs[i], results$Date[i],
results$Plot[i], sep = "_"), "pdf")
ggsave(fn, plot = p, width = 8, height = 5)
}
message("\t", i, " written")
}
```
## Metadata
Note this doesn't include any 'extra' fields in the metadata file.
```{r write-metadata}
results_fields <- c(
"adj.r.squared" = "Linear flux model adjusted R2",
"AIC" = "Linear flux model AIC",
"Area" = "Area entry from metadata file, m2",
"BIC" = "Linear flux model BIC",
"Data_file" = "Data file name",
"Date" = "Date entry from metadata file",
"DEAD_BAND_SEC" = "Dead band used for flux calculation, s",
"Dead_band" = "Dead band entry from metadata file, if present, s",
"deviance" = "Linear flux model deviance",
"df.residual" = "Linear flux model degrees of freedom",
"df" = "Linear flux model degrees of freedom",
"FLAG_INTERCEPT" = "Flag: linear model intercept flagged as unusual",
"FLAG_POLY_DIV" = "Flag: linear model R2 diverges from polynomial model",
"FLAG_R2" = "Flag: linear model R2 flagged as unusual",
"FLAG_ROBUST_DIV" = "Flag: linear model flux diverges from robust model",
"flux_estimate" = "Gas flux computed from slope of linear flux model",
"flux_std.error" = "Standard error of flux_estimate",
"flux_units" = "Units of flux_estimate and flux_std.error",
"Gas" = "Gas (CO2 or CH4)",
"int_estimate" = "Intercept of linear flux model, ppb (CH4) or ppm (CO2)",
"int_p.value" = "P-value of intercept of linear flux model",
"int_statistic" = "Statistic (t-value) of intercept of linear flux model",
"int_std.error" = "Standard error of intercept of linear flux model",
"logLik" = "Linear flux model log-likelihood",
"Metadata_file" = "Metadata file name",
"MSMT_STOP_SEC" = "Measurement stop used for flux calculation, s",
"Msmt_stop" = "Measurement stop entry from metadata file, if present, s",
"nobs" = "Linear flux model number of observations",
"Obs" = "Observation number, i.e. order in metadata file(s)",
"p.value" = "Linear flux model p-value",
"Plot" = "Plot entry from metadata file",
"r.squared_poly" = "Polynomial flux model R2",
"r.squared" = "Linear flux model R2",
"sigma" = "Linear flux model sigma",
"slope_estimate_robust" = "Slope of robust flux model, ppb (CH4) or ppm (CO2) /s",
"slope_estimate" = "Slope of linear flux model, ppb (CH4) or ppm (CO2) /s",
"slope_p.value" = "P-value of slope of linear flux model",
"slope_statistic" = "Statistic (t-value) of slope of linear flux model",
"slope_std.error" = "Standard error of slope of linear flux model",
"SN" = "Serial number from analyzer",
"statistic" = "Linear flux model F-statistic",
"Temp" = "Temperature entry from metadata file, degC",
"Time_start" = "Time_start entry from metadata file",
"TIMESTAMP" = "Original timestamp from analyzer",
"TZ" = "Time zone from analyzer",
"Volume" = "Volume entry from metadata file, m3"
)
# Which metadata names above don't occur in results?
not_there <- names(results_fields)[!names(results_fields) %in% names(results)]
if(length(not_there)) {
message("Field metadata descriptions NOT in results: ", paste(not_there, collapse = ", "))
}
results_meta <- data.frame(Field_name = names(results))
results_meta$Description <- results_fields[results_meta$Field_name]
# Mark any metadata fields that aren't required - they're extra, from user
from_metadata <- is.na(results_meta$Description) &
results_meta$Field_name %in% names(metadat)
results_meta$Description[from_metadata] <- "<from metadata file>"
datatable(results_meta)
metadata_fn <- mkfn("results_metadata", "csv")
message("Writing metadata ", metadata_fn, "...")
write_csv(results_meta, metadata_fn)
```
All done!
# Reproducibility
```{r}
sessionInfo()
```