Last updated: 2022-11-06

Checks: 7 0

Knit directory: emlr_obs_analysis/analysis/

This reproducible R Markdown analysis was created with workflowr (version 1.7.0). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.


Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.

Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.

The command set.seed(20210412) was run prior to running the code in the R Markdown file. Setting a seed ensures that any results that rely on randomness, e.g. subsampling or permutations, are reproducible.

Great job! Recording the operating system, R version, and package versions is critical for reproducibility.

Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.

Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.

Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.

The results in this page were generated with repository version 820fff8. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.

Note that you need to be careful to ensure that all relevant files for the analysis have been committed to Git prior to generating the results (you can use wflow_publish or wflow_git_commit). workflowr only checks the R Markdown file, but you know if there are other scripts or data files that it depends on. Below is the status of the Git repository when the results were generated:


Ignored files:
    Ignored:    .Rhistory
    Ignored:    .Rproj.user/
    Ignored:    data/
    Ignored:    output/other/
    Ignored:    output/presentation/
    Ignored:    output/publication/

Untracked files:
    Untracked:  code/visualization_equal_earth.R
    Untracked:  code/visualization_equal_earth_cleaned.R
    Untracked:  code/visualization_equal_earth_cleaned_jdm.R

Unstaged changes:
    Modified:   analysis/_site.yml
    Modified:   code/Workflowr_project_managment.R

Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.


These are the previous versions of the repository in which changes were made to the R Markdown (analysis/CAT_target_budgets.Rmd) and HTML (docs/CAT_target_budgets.html) files. If you’ve configured a remote Git repository (see ?wflow_git_remote), click on the hyperlinks in the table below to view the files as they were in that past version.

File Version Author Date Message
Rmd 820fff8 jens-daniel-mueller 2022-11-06 rebuild website after internal implementing review round 1

version_id_pattern <- "a"
config <- "MLR_basins"

1 Read files

print(version_id_pattern)
[1] "a"
# identify required version IDs

Version_IDs_1 <- list.files(path = "/nfs/kryo/work/jenmueller/emlr_cant/observations",
                            pattern = paste0("v_1", "a"))

Version_IDs_2 <- list.files(path = "/nfs/kryo/work/jenmueller/emlr_cant/observations",
                            pattern = paste0("v_2", "a"))

Version_IDs_3 <- list.files(path = "/nfs/kryo/work/jenmueller/emlr_cant/observations",
                            pattern = paste0("v_3", "a"))

Version_IDs <- c(Version_IDs_1, Version_IDs_2, Version_IDs_3)

# print(Version_IDs)

1.1 Global

for (i_Version_IDs in Version_IDs) {
  # i_Version_IDs <- Version_IDs[1]
  
  # print(i_Version_IDs)
  
  path_version_data     <-
    paste(path_observations,
          i_Version_IDs,
          "/data/",
          sep = "")
  
  # load and join data files
  
  dcant_budget_global <-
    read_csv(paste(path_version_data,
                   "dcant_budget_global.csv",
                   sep = ""))
  
  dcant_budget_global_mod_truth <-
    read_csv(paste(
      path_version_data,
      "dcant_budget_global_mod_truth.csv",
      sep = ""
    ))
  
  dcant_budget_global_bias <-
    read_csv(paste(path_version_data,
                   "dcant_budget_global_bias.csv",
                   sep = ""))
  
  lm_best_predictor_counts <-
    read_csv(paste(path_version_data,
                   "lm_best_predictor_counts.csv",
                   sep = ""))
  
  lm_best_dcant <-
    read_csv(paste(path_version_data,
                   "lm_best_dcant.csv",
                   sep = ""))
  
  dcant_budget_global <- bind_rows(dcant_budget_global,
                                      dcant_budget_global_mod_truth)
  
  dcant_budget_global <- dcant_budget_global %>%
    mutate(Version_ID = i_Version_IDs)
  
  dcant_budget_global_bias <- dcant_budget_global_bias %>%
    mutate(Version_ID = i_Version_IDs)
  
  lm_best_predictor_counts <- lm_best_predictor_counts %>%
    mutate(Version_ID = i_Version_IDs)
  
  lm_best_dcant <- lm_best_dcant %>%
    mutate(Version_ID = i_Version_IDs)

  params_local <-
    read_rds(paste(path_version_data,
                   "params_local.rds",
                   sep = ""))
  
  params_local <- bind_cols(
    Version_ID = i_Version_IDs,
    MLR_basins := str_c(params_local$MLR_basins, collapse = "|"),
    tref1 = params_local$tref1,
    tref2 = params_local$tref2)
  
  tref <- read_csv(paste(path_version_data,
                         "tref.csv",
                         sep = ""))
  
  params_local <- params_local %>%
    mutate(
      median_year_1 = sort(tref$median_year)[1],
      median_year_2 = sort(tref$median_year)[2],
      duration = median_year_2 - median_year_1,
      period = paste(median_year_1, "-", median_year_2)
    )
  
  if (exists("dcant_budget_global_all")) {
    dcant_budget_global_all <-
      bind_rows(dcant_budget_global_all, dcant_budget_global)
  }
  
  if (!exists("dcant_budget_global_all")) {
    dcant_budget_global_all <- dcant_budget_global
  }
  
  if (exists("dcant_budget_global_bias_all")) {
    dcant_budget_global_bias_all <-
      bind_rows(dcant_budget_global_bias_all,
                dcant_budget_global_bias)
  }

  if (!exists("dcant_budget_global_bias_all")) {
    dcant_budget_global_bias_all <- dcant_budget_global_bias
  }
  
    
  if (exists("lm_best_predictor_counts_all")) {
    lm_best_predictor_counts_all <-
      bind_rows(lm_best_predictor_counts_all, lm_best_predictor_counts)
  }
  
  if (!exists("lm_best_predictor_counts_all")) {
    lm_best_predictor_counts_all <- lm_best_predictor_counts
  }
    
  if (exists("lm_best_dcant_all")) {
    lm_best_dcant_all <-
      bind_rows(lm_best_dcant_all, lm_best_dcant)
  }
  
  if (!exists("lm_best_dcant_all")) {
    lm_best_dcant_all <- lm_best_dcant
  }
  
  if (exists("params_local_all")) {
    params_local_all <- bind_rows(params_local_all, params_local)
  }
  
  if (!exists("params_local_all")) {
    params_local_all <- params_local
  }
  
  
}

rm(
  dcant_budget_global,
  dcant_budget_global_bias,
  dcant_budget_global_mod_truth,
  lm_best_predictor_counts,
  lm_best_dcant,
  params_local,
  tref
)

1.2 Basins

# Version_IDs <- Version_IDs[1:length(Version_IDs)-1]

for (i_Version_IDs in Version_IDs) {
  # i_Version_IDs <- Version_IDs[1]
  
  # print(i_Version_IDs)
  
  path_version_data     <-
    paste(path_observations,
          i_Version_IDs,
          "/data/",
          sep = "")
  
  # load and join data files
  
  dcant_budget_basin_AIP <-
    read_csv(paste(path_version_data,
                   "dcant_budget_basin_AIP.csv",
                   sep = ""))
  
  dcant_budget_basin_AIP_mod_truth <-
    read_csv(paste(
      path_version_data,
      "dcant_budget_basin_AIP_mod_truth.csv",
      sep = ""
    ))
  
    
  dcant_budget_basin_AIP <- bind_rows(dcant_budget_basin_AIP,
                                      dcant_budget_basin_AIP_mod_truth)
  
  dcant_budget_basin_AIP_bias <-
    read_csv(paste(path_version_data,
                   "dcant_budget_basin_AIP_bias.csv",
                   sep = ""))
  
  dcant_slab_budget_bias <-
    read_csv(paste0(path_version_data,
                    "dcant_slab_budget_bias.csv"))

  dcant_slab_budget <-
    read_csv(paste0(path_version_data,
                    "dcant_slab_budget.csv"))

  dcant_budget_basin_AIP <- dcant_budget_basin_AIP %>%
    mutate(Version_ID = i_Version_IDs)
  
  dcant_budget_basin_AIP_bias <- dcant_budget_basin_AIP_bias %>%
    mutate(Version_ID = i_Version_IDs)
  
  dcant_slab_budget <- dcant_slab_budget %>%
    mutate(Version_ID = i_Version_IDs)
  
  dcant_slab_budget_bias <- dcant_slab_budget_bias %>%
    mutate(Version_ID = i_Version_IDs)
  
  if (exists("dcant_budget_basin_AIP_all")) {
    dcant_budget_basin_AIP_all <-
      bind_rows(dcant_budget_basin_AIP_all, dcant_budget_basin_AIP)
  }
  
  if (!exists("dcant_budget_basin_AIP_all")) {
    dcant_budget_basin_AIP_all <- dcant_budget_basin_AIP
  }
  
  if (exists("dcant_budget_basin_AIP_bias_all")) {
    dcant_budget_basin_AIP_bias_all <-
      bind_rows(dcant_budget_basin_AIP_bias_all,
                dcant_budget_basin_AIP_bias)
  }
  
  if (!exists("dcant_budget_basin_AIP_bias_all")) {
    dcant_budget_basin_AIP_bias_all <- dcant_budget_basin_AIP_bias
  }
  
  if (exists("dcant_slab_budget_all")) {
    dcant_slab_budget_all <-
      bind_rows(dcant_slab_budget_all, dcant_slab_budget)
  }
  
  if (!exists("dcant_slab_budget_all")) {
    dcant_slab_budget_all <- dcant_slab_budget
  }
  
  if (exists("dcant_slab_budget_bias_all")) {
    dcant_slab_budget_bias_all <-
      bind_rows(dcant_slab_budget_bias_all,
                dcant_slab_budget_bias)
  }
  
  if (!exists("dcant_slab_budget_bias_all")) {
    dcant_slab_budget_bias_all <- dcant_slab_budget_bias
  }
  
}

rm(
  dcant_budget_basin_AIP,
  dcant_budget_basin_AIP_bias,
  dcant_budget_basin_AIP_mod_truth,
  dcant_slab_budget,
  dcant_slab_budget_bias
)

1.3 Basins hemisphere

# Version_IDs <- Version_IDs[1:length(Version_IDs)-1]

for (i_Version_IDs in Version_IDs) {
  # i_Version_IDs <- Version_IDs[1]
  
  # print(i_Version_IDs)
  
  path_version_data     <-
    paste(path_observations,
          i_Version_IDs,
          "/data/",
          sep = "")
  
  # load and join data files
  
  dcant_budget_basin_MLR <-
    read_csv(paste(path_version_data,
                   "dcant_budget_basin_MLR.csv",
                   sep = ""))
  
  dcant_budget_basin_MLR_mod_truth <-
    read_csv(paste(
      path_version_data,
      "dcant_budget_basin_MLR_mod_truth.csv",
      sep = ""
    ))
  
    
  dcant_budget_basin_MLR <- bind_rows(dcant_budget_basin_MLR,
                                      dcant_budget_basin_MLR_mod_truth)
  

  dcant_budget_basin_MLR <- dcant_budget_basin_MLR %>%
    mutate(Version_ID = i_Version_IDs)

  if (exists("dcant_budget_basin_MLR_all")) {
    dcant_budget_basin_MLR_all <-
      bind_rows(dcant_budget_basin_MLR_all, dcant_budget_basin_MLR)
  }
  
  if (!exists("dcant_budget_basin_MLR_all")) {
    dcant_budget_basin_MLR_all <- dcant_budget_basin_MLR
  }

  
}

rm(
  dcant_budget_basin_MLR,
  dcant_budget_basin_MLR_mod_truth
)

1.4 Steady state

for (i_Version_IDs in Version_IDs) {
  # i_Version_IDs <- Version_IDs[1]
  
  # print(i_Version_IDs)
  
  path_version_data     <-
    paste(path_observations,
          i_Version_IDs,
          "/data/",
          sep = "")
  
  # load and join data files
  
  dcant_obs_budget <-
    read_csv(paste0(path_version_data,
                    "anom_dcant_obs_budget.csv"))
  
  dcant_obs_budget <- dcant_obs_budget %>%
    mutate(Version_ID = i_Version_IDs)
  
  if (exists("dcant_obs_budget_all")) {
    dcant_obs_budget_all <-
      bind_rows(dcant_obs_budget_all, dcant_obs_budget)
  }
  
  if (!exists("dcant_obs_budget_all")) {
    dcant_obs_budget_all <- dcant_obs_budget
  }
  
}


rm(dcant_obs_budget)

1.5 Atm CO2

co2_atm <-
  read_csv(paste(path_preprocessing,
                 "co2_atm.csv",
                 sep = ""))
all_predictors <- c("saltempaouoxygenphosphatenitratesilicate")

params_local_all <- params_local_all %>%
  mutate(MLR_predictors = str_remove_all(all_predictors,
                                         MLR_predictors))
dcant_budget_global_all <- dcant_budget_global_all %>%
  filter(estimate == "dcant", 
         method == "total") %>% 
  select(-c(estimate, method)) %>% 
  rename(dcant = value)

dcant_budget_global_all_depth <- dcant_budget_global_all

dcant_budget_global_all <- dcant_budget_global_all %>%
  filter(inv_depth == params_global$inventory_depth_standard)

dcant_budget_global_bias_all <- dcant_budget_global_bias_all %>%
  filter(estimate == "dcant") %>%
  select(-c(estimate))

dcant_budget_global_bias_all_depth <- dcant_budget_global_bias_all

dcant_budget_global_bias_all <- dcant_budget_global_bias_all %>%
  filter(inv_depth == params_global$inventory_depth_standard)
dcant_budget_basin_AIP_all <- dcant_budget_basin_AIP_all %>%
  filter(estimate == "dcant", 
         method == "total") %>% 
  select(-c(estimate, method)) %>% 
  rename(dcant = value)

dcant_budget_basin_AIP_all_depth <- dcant_budget_basin_AIP_all

dcant_budget_basin_AIP_all <- dcant_budget_basin_AIP_all %>%
  filter(inv_depth == params_global$inventory_depth_standard)

dcant_budget_basin_AIP_bias_all <- dcant_budget_basin_AIP_bias_all %>%
  filter(estimate == "dcant") %>% 
  select(-c(estimate))

dcant_budget_basin_AIP_bias_all_depth <- dcant_budget_basin_AIP_bias_all

dcant_budget_basin_AIP_bias_all <- dcant_budget_basin_AIP_bias_all %>%
  filter(inv_depth == params_global$inventory_depth_standard)
dcant_budget_basin_MLR_all <- dcant_budget_basin_MLR_all %>%
  filter(estimate == "dcant", 
         method == "total") %>% 
  select(-c(estimate, method)) %>% 
  rename(dcant = value)

# dcant_budget_basin_MLR_all_depth <- dcant_budget_basin_MLR_all

dcant_budget_basin_MLR_all <- dcant_budget_basin_MLR_all %>%
  filter(inv_depth == params_global$inventory_depth_standard)

# dcant_budget_basin_MLR_bias_all <- dcant_budget_basin_MLR_bias_all %>%
#   filter(estimate == "dcant") %>% 
#   select(-c(estimate))
# 
# dcant_budget_basin_MLR_bias_all_depth <- dcant_budget_basin_MLR_bias_all
# 
# dcant_budget_basin_MLR_bias_all <- dcant_budget_basin_MLR_bias_all %>%
#   filter(inv_depth == params_global$inventory_depth_standard)

2 Bias thresholds

global_bias_rel_max <- 10
global_bias_rel_max
[1] 10
regional_bias_rel_max <- 20
regional_bias_rel_max
[1] 20

3 Individual cases

3.1 Global

3.1.1 Absoulte values

legend_title = expression(atop(Delta * C[ant],
                               (mu * mol ~ kg ^ {
                                 -1
                               })))

dcant_budget_global_all %>%
  ggplot(aes(period, dcant, col = MLR_basins)) +
  geom_jitter(width = 0.05, height = 0) +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(. ~ data_source) +
  ylim(0,NA) +
  theme(axis.text.x = element_text(angle = 45, hjust=1),
        axis.title.x = element_blank())

3.1.2 Biases

dcant_budget_global_bias_all %>%
  ggplot(aes(period, dcant_bias, col=MLR_basins)) +
  geom_hline(yintercept = 0) +
  scale_color_brewer(palette = "Dark2") +
  labs(y = expression(atop(Delta * C[ant] ~ bias,
                               (PgC)))) +
  geom_point() +
  theme(axis.text.x = element_blank(),
        axis.title.x = element_blank())

p_global_bias <-
  dcant_budget_global_bias_all %>%
  ggplot() +
  geom_hline(yintercept = global_bias_rel_max * c(-1, 1),
             linetype = 2) +
  geom_hline(yintercept = 0) +
  scale_color_brewer(palette = "Dark2") +
  labs(y = expression(Delta * C[ant] ~ bias ~ ("%")),
       title = "Model-based assesment") +
  theme(axis.title.x = element_blank()) +
  geom_point(aes(period, dcant_bias_rel, col = MLR_basins),
  alpha = 0.7) +
  theme(axis.text.x = element_text(angle = 45, hjust = 1),
        axis.title.x = element_blank())

p_global_bias

dcant_budget_global_bias_all %>%
  group_by(period) %>%
  summarise(
    dcant_bias_sd = sd(dcant_bias),
    dcant_bias = mean(dcant_bias),
    dcant_bias_rel_sd = sd(dcant_bias_rel),
    dcant_bias_rel = mean(dcant_bias_rel)
  ) %>%
  ungroup() %>%
  kable() %>%
  kable_styling() %>%
  scroll_box(height = "300px")
period dcant_bias_sd dcant_bias dcant_bias_rel_sd dcant_bias_rel
1994 - 2004 1.513148 1.450167 8.577937 8.220899
1994 - 2014 1.577394 4.547167 4.100536 11.820647
2004 - 2014 1.561493 3.179500 7.497084 15.265508

3.2 Basins

3.2.1 Absoulte values

dcant_budget_basin_AIP_all %>%
  ggplot(aes(period, dcant, col = MLR_basins)) +
  geom_jitter(width = 0.05, height = 0) +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(basin_AIP ~ data_source) +
  ylim(0,NA) +
  theme(axis.text.x = element_text(angle = 45, hjust=1),
        axis.title.x = element_blank())

3.2.2 Biases

dcant_budget_basin_AIP_bias_all %>%
  ggplot(aes(period, dcant_bias, col=MLR_basins)) +
  geom_hline(yintercept = 0) +
  geom_point() +
  facet_grid(basin_AIP ~ .)

dcant_budget_basin_AIP_bias_all %>%
  ggplot() +
  geom_tile(aes(y = 0, height = regional_bias_rel_max * 2,
                x = "2004 - 2014", width = Inf,
                fill = "bias\nthreshold"), alpha = 0.5) +
  geom_hline(yintercept = 0) +
  scale_fill_manual(values = "grey70", name = "") +
  scale_color_brewer(palette = "Dark2") +
  labs(y = expression(Delta ~ C[ant] ~ bias)) +
  theme(axis.title.x = element_blank()) +
  geom_jitter(aes(period, dcant_bias_rel, col = MLR_basins),
              width = 0.05, height = 0) +
  facet_grid(. ~ basin_AIP)

p_regional_bias <- 
  dcant_budget_basin_AIP_bias_all %>%
  ggplot() +
  geom_hline(yintercept = regional_bias_rel_max * c(-1,1),
             linetype = 2) +
  geom_hline(yintercept = 0) +
  scale_color_brewer(palette = "Dark2") +
  labs(y = expression(Delta * C[ant] ~ bias ~ ("%")),
       title = "Model-based assesment") +
  theme(axis.title.x = element_blank()) +
  geom_point(aes(period, dcant_bias_rel, col = MLR_basins),
             alpha = 0.7) +
    theme(axis.text.x = element_text(angle = 45, hjust=1),
        axis.title.x = element_blank()) +
  facet_grid(. ~ basin_AIP) +
  theme(
  strip.background = element_blank(),
  strip.text.x = element_blank()
)

p_regional_bias

3.3 Slab budgets

3.3.1 Absolute values

dcant_slab_budget_all %>%
  filter(data_source == "obs",
         period != "1994 - 2014") %>% 
  ggplot(aes(MLR_basins, dcant, fill = gamma_slab)) +
  geom_hline(yintercept = 0, col = "red") +
  geom_col() +
  scale_fill_scico_d(direction = -1) +
  facet_grid(basin_AIP ~ period)

dcant_slab_budget_all %>%
  filter(data_source == "obs",
         period != "1994 - 2014") %>%
  group_by(basin_AIP) %>%
  group_split() %>%
  map(
    ~ ggplot(data = .x,
             aes(MLR_basins, dcant, fill = gamma_slab)) +
      geom_hline(yintercept = 0) +
      geom_col() +
      scale_fill_scico_d(direction = -1) +
      labs(title = paste("data_source:", unique(.x$basin_AIP))) +
      facet_grid(gamma_slab ~ period)
  )
[[1]]


[[2]]


[[3]]

3.3.2 Bias

dcant_slab_budget_bias_all %>%
  filter(period != "1994 - 2014") %>%
  group_by(basin_AIP) %>%
  group_split() %>%
  # head(1) %>% 
  map(
    ~ ggplot(data = .x,
             aes(gamma_slab, dcant_bias, fill = gamma_slab)) +
      geom_col() +
      coord_flip() +
      scale_x_discrete(limits = rev) +
      scale_fill_scico_d(direction = -1) +
      facet_grid(period ~ MLR_basins) +
      labs(title = paste("data_source:", unique(.x$basin_AIP)))
    )
[[1]]
Warning: Removed 40 rows containing missing values (position_stack).


[[2]]
Warning: Removed 24 rows containing missing values (position_stack).


[[3]]
Warning: Removed 108 rows containing missing values (position_stack).

3.3.3 Spread

dcant_slab_budget_all %>%
  filter(period != "1994 - 2014",
         data_source != "mod_truth") %>%
  group_by(data_source, basin_AIP, gamma_slab, period) %>%
  summarise(dcant_range = max(dcant) - min(dcant)) %>%
  ungroup() %>%
  group_split(basin_AIP) %>%
  # head(1) %>%
  map(
    ~ ggplot(data = .x,
             aes(gamma_slab, dcant_range, fill = gamma_slab)) +
      geom_col() +
      coord_flip() +
      scale_x_discrete(limits = rev) +
      scale_fill_scico_d(direction = -1) +
      facet_grid(period ~ data_source) +
      labs(title = paste("data_source:", unique(.x$basin_AIP)))
  )
`summarise()` has grouped output by 'data_source', 'basin_AIP', 'gamma_slab'.
You can override using the `.groups` argument.
[[1]]


[[2]]


[[3]]

3.4 Basins hemisphere

3.4.1 Absoulte values

dcant_budget_basin_MLR_all %>%
  ggplot(aes(period, dcant, col = MLR_basins)) +
  geom_jitter(width = 0.05, height = 0) +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(basin ~ data_source) +
  ylim(0,NA) +
  theme(axis.text.x = element_text(angle = 45, hjust=1),
        axis.title.x = element_blank())

3.4.2 Biases

dcant_budget_basin_MLR_bias_all <-
  dcant_budget_basin_MLR_all %>%
  filter(data_source %in% c("mod", "mod_truth")) %>%
  pivot_wider(names_from = data_source,
              values_from = dcant) %>%
  mutate(dcant_bias = mod - mod_truth,
         dcant_bias_rel = 100*(mod - mod_truth)/mod_truth)
  
dcant_budget_basin_MLR_bias_all %>%   
  ggplot(aes(period, dcant_bias, col=MLR_basins)) +
  geom_hline(yintercept = 0) +
  geom_point() +
  facet_grid(basin ~ .)

dcant_budget_basin_MLR_bias_all %>%
  ggplot() +
  geom_tile(aes(y = 0, height = regional_bias_rel_max * 2,
                x = "2004 - 2014", width = Inf,
                fill = "bias\nthreshold"), alpha = 0.5) +
  geom_hline(yintercept = 0) +
  scale_fill_manual(values = "grey70", name = "") +
  scale_color_brewer(palette = "Dark2") +
  labs(y = expression(Delta ~ C[ant] ~ bias)) +
  theme(axis.title.x = element_blank()) +
  geom_jitter(aes(period, dcant_bias_rel, col = MLR_basins),
              width = 0.05, height = 0) +
  facet_grid(. ~ basin)

p_regional_bias <- 
  dcant_budget_basin_MLR_bias_all %>%
  ggplot() +
  geom_hline(yintercept = regional_bias_rel_max * c(-1,1),
             linetype = 2) +
  geom_hline(yintercept = 0) +
  scale_color_brewer(palette = "Dark2") +
  labs(y = expression(Delta * C[ant] ~ bias ~ ("%")),
       title = "Model-based assesment") +
  theme(axis.title.x = element_blank()) +
  geom_point(aes(period, dcant_bias_rel, col = MLR_basins),
             alpha = 0.7) +
    theme(axis.text.x = element_text(angle = 45, hjust=1),
        axis.title.x = element_blank()) +
  facet_grid(. ~ basin) +
  theme(
  strip.background = element_blank(),
  strip.text.x = element_blank()
)

p_regional_bias

dcant_budget_basin_MLR_bias_all %>%
  group_by(period, basin) %>%
  summarise(
    dcant_bias_sd = sd(dcant_bias),
    dcant_bias = mean(dcant_bias),
    dcant_bias_rel_sd = sd(dcant_bias_rel),
    dcant_bias_rel = mean(dcant_bias_rel)
  ) %>%
  ungroup() %>%
  kable() %>%
  kable_styling() %>%
  scroll_box(height = "300px")
`summarise()` has grouped output by 'period'. You can override using the
`.groups` argument.
period basin dcant_bias_sd dcant_bias dcant_bias_rel_sd dcant_bias_rel
1994 - 2004 Indian 0.8287145 -0.0811667 17.189681 -1.683606
1994 - 2004 N_Atlantic 0.2931037 0.3101667 15.023254 15.897830
1994 - 2004 N_Pacific 0.7540646 0.3498333 27.763791 12.880461
1994 - 2004 S_Atlantic 0.3277698 0.1913333 13.110793 7.653333
1994 - 2004 S_Pacific 0.7238496 0.6803333 12.806964 12.037037
1994 - 2014 Indian 1.2922363 1.0248333 12.368264 9.808895
1994 - 2014 N_Atlantic 0.4612550 0.2641667 10.736848 6.149131
1994 - 2014 N_Pacific 0.6201006 1.0050000 10.621799 17.214800
1994 - 2014 S_Atlantic 0.5469557 0.8838333 10.341383 16.710783
1994 - 2014 S_Pacific 1.3460879 1.3695000 10.685782 10.871636
2004 - 2014 Indian 0.7856189 1.0045000 13.964076 17.854604
2004 - 2014 N_Atlantic 0.2189168 0.0581667 9.335471 2.480455
2004 - 2014 N_Pacific 0.3402215 0.6805000 10.897551 21.796925
2004 - 2014 S_Atlantic 0.5565156 0.7550000 19.953947 27.070635
2004 - 2014 S_Pacific 0.9599571 0.6820000 13.822276 9.820014

4 Ensemble

4.1 Global

dcant_budget_global_all_in <- dcant_budget_global_all %>% 
  filter(data_source %in% c("mod", "obs"))

dcant_budget_global_ensemble <- dcant_budget_global_all_in %>% 
  group_by(data_source, period, tref2) %>% 
  summarise(dcant_mean = mean(dcant),
            dcant_sd = sd(dcant),
            dcant_range = max(dcant)- min(dcant)) %>% 
  ungroup()
`summarise()` has grouped output by 'data_source', 'period'. You can override
using the `.groups` argument.

4.1.1 Mean

legend_title = expression(Delta * C[ant]~(PgC))

ggplot() +
  geom_col(data = dcant_budget_global_ensemble,
           aes(x = period,
               y = dcant_mean),
           fill = "darkgrey") +
  geom_errorbar(
    data = dcant_budget_global_ensemble,
    aes(
      x = period,
      y = dcant_mean,
      ymax = dcant_mean + dcant_sd,
      ymin = dcant_mean - dcant_sd
    ),
    width = 0.1
  ) +
  geom_point(
    data = dcant_budget_global_all,
    aes(period, dcant, col = MLR_basins),
    alpha = 0.7,
    position = position_jitter(width = 0.2, height = 0)
  ) +
  scale_y_continuous(limits = c(0,70), expand = c(0,0)) +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(. ~ data_source) +
  labs(y = legend_title) +
  theme(axis.text.x = element_text(angle = 45, hjust=1),
        axis.title.x = element_blank())

p_global_dcant <- ggplot() +
  geom_col(data = dcant_budget_global_ensemble %>% 
             filter(data_source == "obs"),
           aes(x = period,
               y = dcant_mean),
           fill = "darkgrey") +
    geom_point(
    data = dcant_budget_global_all %>% 
             filter(data_source == "obs"),
    aes(period, dcant, col = MLR_basins),
    alpha = 0.7,
    position = position_jitter(width = 0.1, height = 0)
  ) +
  geom_errorbar(
    data = dcant_budget_global_ensemble %>% 
             filter(data_source == "obs"),
    aes(
      x = period,
      y = dcant_mean,
      ymax = dcant_mean + dcant_sd,
      ymin = dcant_mean - dcant_sd
    ),
    width = 0.1
  ) +
  scale_y_continuous(limits = c(0,70), expand = c(0,0)) +
  scale_color_brewer(palette = "Dark2") +
  labs(y = legend_title,
       title = "Observation-based results") +
  theme(axis.text.x = element_blank(),
        axis.title.x = element_blank())

p_global_dcant_bias <-
p_global_dcant / p_global_bias +
  plot_layout(guides = 'collect',
              heights = c(2,1))

p_global_dcant_bias

# ggsave(plot = p_global_dcant_bias,
#        path = here::here("output/publication"),
#        filename = "Fig_global_dcant_budget.png",
#        height = 5,
#        width = 5)

rm(p_global_bias, p_global_dcant, p_global_dcant_bias)

4.1.2 Mean vs atm CO2

dcant_ensemble <- dcant_budget_global_ensemble %>% 
  filter(data_source == "obs",
         period != "1994 - 2014") %>% 
  select(year = tref2, dcant_mean, dcant_sd)

tcant_S04 <- bind_cols(year = 1994, dcant_mean = 118, dcant_sd = 19)

tcant_ensemble <- full_join(dcant_ensemble, tcant_S04)
Joining, by = c("year", "dcant_mean", "dcant_sd")
tcant_ensemble <- left_join(tcant_ensemble, co2_atm)
Joining, by = "year"
co2_atm_pi <- bind_cols(pCO2 = 280, dcant_mean = 0, year = 1750, dcant_sd = 0)

tcant_ensemble <- full_join(tcant_ensemble, co2_atm_pi)
Joining, by = c("year", "dcant_mean", "dcant_sd", "pCO2")
tcant_ensemble <- tcant_ensemble %>% 
  arrange(year) %>% 
  mutate(tcant = cumsum(dcant_mean),
         tcant_sd = cumsum(dcant_sd))

tcant_ensemble %>% 
  ggplot(aes(pCO2, tcant, ymin = tcant - tcant_sd, ymax = tcant + tcant_sd)) +
  geom_ribbon(fill = "grey80") +
  geom_point() +
  geom_line() +
  scale_x_continuous(breaks = seq(280, 400, 30),
                     sec.axis = dup_axis(labels =  c(1750, 1940, 1980, 2000, 2015),
                                         name = "Year")) +
  geom_text(aes(label = year), nudge_x = -5, nudge_y = 5) +
  labs(x = expression(Atmospheric~pCO[2]~(µatm)),
       y = expression(Total~oceanic~C[ant]~(PgC)))

# ggsave(path = "output/publication",
#        filename = "Fig_global_dcant_budget_vs_atm_pCO2.png",
#        height = 4,
#        width = 7)

4.1.3 Sum decades

dcant_budget_global_all_in_sum <-
  dcant_budget_global_all_in %>%
  filter(period != "1994 - 2014") %>%
  arrange(tref1) %>%
  group_by(data_source, MLR_basins) %>%
  mutate(dcant = dcant + lag(dcant)) %>% 
  ungroup() %>%
  drop_na() %>% 
  mutate(estimate = "sum")

dcant_budget_global_all_in_sum <-
  bind_rows(
    dcant_budget_global_all_in_sum,
    dcant_budget_global_all_in %>%
      filter(period == "1994 - 2014") %>%
      mutate(estimate = "direct")
  )

ggplot() +
  geom_point(
    data = dcant_budget_global_all_in_sum,
    aes(estimate, dcant, col = MLR_basins),
    alpha = 0.7,
    position = position_jitter(width = 0, height = 0)
  ) +
  scale_y_continuous(limits = c(0,70), expand = c(0,0)) +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(. ~ data_source) +
  theme(axis.text.x = element_text(angle = 45, hjust=1),
        axis.title.x = element_blank())

4.1.4 Mean bias

dcant_budget_global_ensemble_bias <- full_join(
  dcant_budget_global_ensemble %>%
    filter(data_source == "mod") %>% 
    select(period, dcant_mean, dcant_sd),
  dcant_budget_global_all %>%
    filter(data_source == "mod_truth",
           MLR_basins == unique(dcant_budget_global_all$MLR_basins)[1]) %>% 
    select(period, dcant)
)
Joining, by = "period"
dcant_budget_global_ensemble_bias <- dcant_budget_global_ensemble_bias %>% 
  mutate(dcant_mean_bias = dcant_mean - dcant,
         dcant_mean_bias_rel = 100 * dcant_mean_bias / dcant)

dcant_budget_global_ensemble_bias %>%
  ggplot(aes(period, dcant_mean_bias)) +
  geom_hline(yintercept = 0) +
  geom_point()

dcant_budget_global_ensemble_bias %>%
  ggplot(aes(period, dcant_mean_bias_rel)) +
  geom_hline(yintercept = 0) +
  geom_point()

4.1.5 Vertical patterns

4.1.5.1 Absoulte values

dcant_budget_global_all_depth %>%
  filter(data_source != "mod_truth") %>% 
  group_by(data_source) %>%
  group_split() %>%
  # head(1) %>%
  map(
    ~  ggplot(data = .x,
              aes(dcant, MLR_basins, fill=period)) +
      geom_vline(xintercept = 0) +
      geom_col(position = "dodge") +
      scale_fill_brewer(palette = "Dark2") +
      facet_grid(inv_depth ~ .) +
      labs(title = paste("data_source:", unique(.x$data_source)))
  )
[[1]]


[[2]]

4.1.5.2 Biases

dcant_budget_global_bias_all_depth %>%
  ggplot(aes(dcant_bias, MLR_basins, fill = period)) +
  geom_vline(xintercept = 0) +
  geom_col(position = "dodge") +
  scale_fill_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ .)

dcant_budget_global_bias_all_depth %>%
  ggplot(aes(dcant_bias_rel, MLR_basins, fill = period)) +
  geom_vline(xintercept = 0) +
  geom_col(position = "dodge") +
  scale_fill_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ .)

rm(dcant_budget_global_all,
   dcant_budget_global_all_depth,
   dcant_budget_global_bias_all,
   dcant_budget_global_bias_all_depth,
   dcant_budget_global_ensemble,
   dcant_budget_global_ensemble_bias)

4.2 Basins

dcant_budget_basin_AIP_ensemble <- dcant_budget_basin_AIP_all %>% 
  filter(data_source %in% c("mod", "obs")) %>% 
  group_by(basin_AIP, data_source, period) %>% 
  summarise(dcant_mean = mean(dcant),
            dcant_sd = sd(dcant),
            dcant_range = max(dcant)- min(dcant)) %>% 
  ungroup()
`summarise()` has grouped output by 'basin_AIP', 'data_source'. You can override
using the `.groups` argument.

4.2.1 Mean

dcant_budget_basin_AIP_ensemble %>%
  ggplot(aes(period, dcant_mean, col=basin_AIP)) +
  geom_pointrange(aes(ymax = dcant_mean + dcant_sd,
                      ymin = dcant_mean - dcant_sd),
                  shape = 21) +
  facet_grid(. ~ data_source)

p_regional_dcant <- ggplot() +
  geom_col(
    data = dcant_budget_basin_AIP_ensemble %>%
      filter(data_source == "obs"),
    aes(x = period,
        y = dcant_mean),
    fill = "darkgrey"
  ) +
  geom_point(
    data = dcant_budget_basin_AIP_all %>%
      filter(data_source == "obs"),
    aes(period, dcant, col = MLR_basins),
    position = position_jitter(width = 0.1, height = 0),
    alpha = 0.7
  ) +
  geom_errorbar(
    data = dcant_budget_basin_AIP_ensemble %>%
      filter(data_source == "obs"),
    aes(
      x = period,
      y = dcant_mean,
      ymax = dcant_mean + dcant_sd,
      ymin = dcant_mean - dcant_sd
    ),
    width = 0.1
  ) +
  scale_y_continuous(limits = c(0, 35), expand = c(0, 0)) +
  scale_color_brewer(palette = "Dark2") +
  labs(y = legend_title,
       title = "Observation-based results") +
  theme(axis.text.x = element_blank(),
        axis.title.x = element_blank()) +
  facet_grid(. ~ basin_AIP)

p_regional_dcant_bias <-
p_regional_dcant / p_regional_bias +
  plot_layout(guides = 'collect',
              heights = c(2,1))

p_regional_dcant_bias

# ggsave(plot = p_regional_dcant_bias,
#        path = "output/publication",
#        filename = "Fig_regional_dcant_budget.png",
#        height = 5,
#        width = 10)

rm(p_regional_bias, p_regional_dcant, p_regional_dcant_bias)

4.2.2 Mean bias

dcant_budget_basin_AIP_ensemble_bias <- full_join(
  dcant_budget_basin_AIP_ensemble %>%
    filter(data_source == "mod") %>% 
    select(basin_AIP, period, dcant_mean, dcant_sd),
  dcant_budget_basin_AIP_all %>%
    filter(data_source == "mod_truth",
           MLR_basins == unique(dcant_budget_basin_AIP_all$MLR_basins)[1]) %>% 
    select(basin_AIP, period, dcant)
)
Joining, by = c("basin_AIP", "period")
dcant_budget_basin_AIP_ensemble_bias <- dcant_budget_basin_AIP_ensemble_bias %>% 
  mutate(dcant_mean_bias = dcant_mean - dcant,
         dcant_mean_bias_rel = 100 * dcant_mean_bias / dcant)


dcant_budget_basin_AIP_ensemble_bias %>%
  ggplot(aes(period, dcant_mean_bias, col = basin_AIP)) +
  geom_hline(yintercept = 0) +
  geom_point()

dcant_budget_basin_AIP_ensemble_bias %>%
  ggplot(aes(period, dcant_mean_bias_rel, col = basin_AIP)) +
  geom_hline(yintercept = 0) +
  geom_point()

4.2.3 Vertical patterns

4.2.3.1 Absoulte values

dcant_budget_basin_AIP_all_depth %>%
  filter(data_source != "mod_truth") %>%
  group_by(data_source) %>%
  group_split() %>%
  # head(1) %>%
  map(
    ~  ggplot(data = .x,
              aes(dcant, MLR_basins, fill = basin_AIP)) +
      geom_vline(xintercept = 0) +
      geom_col() +
      scale_fill_brewer(palette = "Dark2") +
      facet_grid(inv_depth ~ period) +
      labs(title = paste("data_source:", unique(.x$data_source)))
  )
[[1]]


[[2]]

4.2.3.2 Biases

dcant_budget_basin_AIP_bias_all_depth %>%
  ggplot(aes(dcant_bias, MLR_basins, fill = basin_AIP)) +
  geom_vline(xintercept = 0) +
  geom_col() +
  scale_fill_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ period)

dcant_budget_basin_AIP_bias_all_depth %>%
  ggplot(aes(dcant_bias_rel, MLR_basins, fill = basin_AIP)) +
  geom_vline(xintercept = 0) +
  geom_col(position = "dodge") +
  scale_fill_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ period)

5 Steady state

dcant_obs_budget_all %>%
  group_by(inv_depth) %>%
  group_split() %>%
  # head(1) %>% 
  map(
    ~ ggplot(data = .x,
             aes(estimate, dcant_pos, fill = basin_AIP)) +
      scale_fill_brewer(palette = "Dark2") +
      geom_col() +
      facet_grid(MLR_basins ~ period) +
      labs(title = paste("inventory depth:",unique(.x$inv_depth)))
  )
[[1]]


[[2]]


[[3]]


[[4]]


[[5]]

6 Predictor analysis

lm_best_predictor_counts_all <-
  full_join(lm_best_predictor_counts_all,
            params_local_all)
Joining, by = "Version_ID"
lm_best_predictor_counts_all <- lm_best_predictor_counts_all %>% 
  mutate(n_predictors_total = rowSums(across(aou:temp), na.rm = TRUE)/10)

lm_best_predictor_counts_all %>%
  ggplot(aes(x = MLR_basins, y = n_predictors_total)) +
  # ggdist::stat_halfeye(
  #   adjust = .5,
  #   width = .6,
  #   .width = 0,
  #   justification = -.2,
  #   point_colour = NA
  # ) +
  geom_boxplot(width = 0.5,
               outlier.shape = NA) +
  gghalves::geom_half_point(
    side = "l",
    range_scale = .4,
    alpha = .5,
    aes(col = gamma_slab)
  ) +
  scale_color_viridis_d() +
  facet_grid(basin ~ data_source)

lm_best_predictor_counts_all %>%
  pivot_longer(aou:temp,
               names_to = "predictor",
               values_to = "count") %>%
  group_split(predictor) %>%
  # head(1) %>%
  map(
    ~ ggplot(data = .x,
             aes(MLR_basins, count, color = gamma_slab)) +
      geom_jitter(alpha = 0.5) +
      scale_color_viridis_d() +
      labs(title = paste0("predictor:", unique(.x$predictor))) +
      coord_cartesian(ylim = c(0, 10)) +
      facet_grid(basin ~ data_source)
  )
[[1]]


[[2]]
Warning: Removed 42 rows containing missing values (geom_point).


[[3]]
Warning: Removed 7 rows containing missing values (geom_point).


[[4]]
Warning: Removed 8 rows containing missing values (geom_point).


[[5]]
Warning: Removed 1 rows containing missing values (geom_point).


[[6]]
Warning: Removed 2 rows containing missing values (geom_point).


[[7]]

lm_best_dcant_all <-
  full_join(lm_best_dcant_all,
            params_local_all)
Joining, by = "Version_ID"
lm_best_dcant_all %>%
  count(basin, data_source, gamma_slab, MLR_basins, period) %>%
  ggplot(aes(MLR_basins, n)) +
  geom_jitter(height = 0, alpha = 0.3) +
  facet_grid(basin ~ data_source)

7 Drift and bias

dcant_budget_global_all_dissic %>%
  filter(estimate == "dcant") %>%
  ggplot(aes(inv_depth, value, col = !!sym(config))) +
  geom_hline(yintercept = 0) +
  scale_color_brewer(palette = "Dark2") +
  geom_point() +
  geom_path() +
  labs(y = "DIC change (PgC)") +
  facet_grid(data_source ~ period, scales = "free_y")

dcant_budget_global_bias_all_decomposition <-
  dcant_budget_global_bias_all_decomposition %>%
  filter(estimate == "dcant") %>%
  select(inv_depth, dcant_bias, contribution, !!sym(config), period) %>%
  pivot_wider(names_from = contribution,
              values_from = dcant_bias)

dcant_budget_global_bias_all_decomposition <-
  full_join(
    dcant_budget_global_bias_all_decomposition,
    dcant_budget_global_bias_all_depth %>%
      select(inv_depth, !!sym(config), period, mod_truth)
  )
Joining, by = c("inv_depth", "MLR_basins", "period")
dcant_budget_global_bias_all_decomposition %>%
  ggplot(aes(`dcant offset`, `delta C* - mod_truth`, col = !!sym(config))) +
  geom_vline(xintercept = 0, col = "grey50") +
  geom_hline(yintercept = 0, col = "grey50") +
  geom_abline(intercept = 0, slope = 1) +
  geom_point() +
  coord_fixed() +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ period)

dcant_budget_global_bias_all_decomposition %>%
  ggplot(aes(`dcant offset`, `C* prediction error`, col = !!sym(config))) +
  geom_vline(xintercept = 0, col = "grey50") +
  geom_hline(yintercept = 0, col = "grey50") +
  geom_abline(intercept = 0, slope = 1) +
  geom_point() +
  coord_fixed() +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ period)

dcant_budget_global_bias_all_decomposition %>%
  ggplot(aes(
    `dcant offset`,
    `C* prediction error` + `delta C* - mod_truth`,
    col = !!sym(config)
  )) +
  geom_vline(xintercept = 0, col = "grey50") +
  geom_hline(yintercept = 0, col = "grey50") +
  geom_abline(intercept = 0, slope = 1) +
  geom_point() +
  coord_fixed() +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ period)

dcant_budget_global_bias_all_decomposition %>%
  ggplot(aes(`dcant offset`, `C* drift`, col = !!sym(config))) +
  geom_vline(xintercept = 0, col = "grey50") +
  geom_hline(yintercept = 0, col = "grey50") +
  geom_abline(intercept = 0, slope = 1) +
  geom_point() +
  coord_fixed() +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ period)

dcant_budget_global_bias_all_decomposition %>%
  ggplot(aes(
    `dcant offset` - `C* drift`,
    `C* prediction error`,
    col = !!sym(config)
  )) +
  geom_vline(xintercept = 0, col = "grey50") +
  geom_hline(yintercept = 0, col = "grey50") +
  geom_abline(intercept = 0, slope = 1) +
  geom_point() +
  coord_fixed() +
  scale_color_brewer(palette = "Dark2") +
  facet_grid(inv_depth ~ period)

dcant_budget_global_bias_all_decomposition %>%
  ggplot(aes(
    x = period,
    fill = !!sym(config),
    col = !!sym(config)
  )) +
  geom_hline(yintercept = 0) +
  geom_point(
    aes(y = `dcant offset`, shape = "dcant offset"),
    position = position_nudge(x = -0.05),
    alpha = 0.5
  ) +
  geom_point(
    aes(y = `dcant offset` - `C* drift`, shape = "dcant offset - C* drift"),
    position = position_nudge(x = 0.05),
    alpha = 0.5
  ) +
  scale_color_brewer(palette = "Dark2") +
  scale_fill_brewer(palette = "Dark2") +
  scale_shape_manual(values = c(21,23)) +
  facet_grid(inv_depth ~ .)

dcant_budget_global_bias_all_decomposition <-
  dcant_budget_global_bias_all_decomposition %>%
  mutate(
    `dcant offset rel` = 100 * `dcant offset` / mod_truth,
    `dcant offset rel corr` = 100 * (`dcant offset` - `C* drift`) / mod_truth,
    `C* prediction error rel` = 100 * (`C* prediction error`) / mod_truth
  )

dcant_budget_global_bias_all_decomposition %>%
  ggplot(aes(
    x = period,
    fill = !!sym(config),
    col = !!sym(config)
  )) +
  geom_hline(yintercept = 0) +
  geom_point(
    aes(y = `dcant offset rel`, shape = "dcant offset"),
    position = position_nudge(x = -0.05),
    alpha = 0.5
  ) +
  geom_point(
    aes(y = `dcant offset rel corr`, shape = "dcant offset - C* drift"),
    position = position_nudge(x = 0.05),
    alpha = 0.5
  ) +
  scale_color_brewer(palette = "Dark2") +
  scale_fill_brewer(palette = "Dark2") +
  scale_shape_manual(values = c(21,23)) +
  facet_grid(inv_depth ~ .)

dcant_budget_global_bias_all_decomposition <-
  dcant_budget_global_bias_all_decomposition %>%
  pivot_longer(-c(inv_depth:period),
               names_to = "estimate",
               values_to = "value")


dcant_budget_global_bias_all_decomposition %>%
  group_by(inv_depth, estimate) %>%
  summarise(mean = mean(value),
            sd = sd(value)) %>%
  ungroup() %>%
  kable() %>%
  kable_styling() %>%
  scroll_box(height = "300px")
`summarise()` has grouped output by 'inv_depth'. You can override using the
`.groups` argument.
inv_depth estimate mean sd
100 C* drift 0.3106667 0.1130247
100 C* prediction error -0.5596111 0.3693567
100 C* prediction error rel -11.4837057 7.0494179
100 dcant offset 0.3001111 0.3080336
100 dcant offset rel 6.5058791 6.7734518
100 dcant offset rel corr -0.0481478 6.6575298
100 delta C* - mod_truth -0.2620000 0.1826685
100 mod_truth 4.7563333 1.7425592
500 C* drift 1.8180000 0.6614652
500 C* prediction error -1.9073333 1.2302951
500 C* prediction error rel -11.1553704 5.9071511
500 dcant offset 0.6138333 1.0800516
500 dcant offset rel 3.8144418 6.5430753
500 dcant offset rel corr -6.9517799 6.2585468
500 delta C* - mod_truth 0.9033333 0.3296997
500 mod_truth 16.9600000 6.2242462
1000 C* drift 3.6083333 1.3124168
1000 C* prediction error -3.3052778 1.9886341
1000 C* prediction error rel -14.5604856 7.3706540
1000 dcant offset 1.0151111 1.5928394
1000 dcant offset rel 4.5222476 7.1492004
1000 dcant offset rel corr -11.4103612 7.0990382
1000 delta C* - mod_truth 2.7026667 0.9845265
1000 mod_truth 22.7516667 8.3557301
3000 C* drift 3.3400000 1.2153477
3000 C* prediction error -1.2218333 1.7257355
3000 C* prediction error rel -5.0224865 7.9414601
3000 dcant offset 3.0589444 1.9551632
3000 dcant offset rel 11.7690183 7.2025806
3000 dcant offset rel corr -1.3232266 7.6893151
3000 delta C* - mod_truth 2.6516667 0.9668653
3000 mod_truth 25.6453333 9.4254881
10000 C* drift 3.6306667 1.3214391
10000 C* prediction error 0.2030556 2.3973145
10000 C* prediction error rel 0.4089855 10.2625458
10000 dcant offset 3.8608333 2.7882101
10000 dcant offset rel 14.5526389 9.8470769
10000 dcant offset rel corr 0.5324328 10.5514540
10000 delta C* - mod_truth 2.0286667 0.7386383
10000 mod_truth 26.0420000 9.5725748
dcant_budget_global_bias_all_decomposition %>%
  group_by(inv_depth, estimate, period) %>%
  summarise(mean = mean(value),
            sd = sd(value)) %>%
  ungroup() %>%
  kable() %>%
  kable_styling() %>%
  scroll_box(height = "300px")
`summarise()` has grouped output by 'inv_depth', 'estimate'. You can override
using the `.groups` argument.
inv_depth estimate period mean sd
100 C* drift 1994 - 2004 0.2340000 0.0000000
100 C* drift 1994 - 2014 0.4660000 0.0000000
100 C* drift 2004 - 2014 0.2320000 0.0000000
100 C* prediction error 1994 - 2004 -0.2018333 0.2192299
100 C* prediction error 1994 - 2014 -0.8480000 0.3110183
100 C* prediction error 2004 - 2014 -0.6290000 0.2446696
100 C* prediction error rel 1994 - 2004 -6.0811489 6.6053010
100 C* prediction error rel 1994 - 2014 -11.8867396 4.3596626
100 C* prediction error rel 2004 - 2014 -16.4832285 6.4116764
100 dcant offset 1994 - 2004 0.3061667 0.0514254
100 dcant offset 1994 - 2014 0.4330000 0.3225226
100 dcant offset 2004 - 2014 0.1611667 0.4141746
100 dcant offset rel 1994 - 2004 9.2246661 1.5494230
100 dcant offset rel 1994 - 2014 6.0695262 4.5209217
100 dcant offset rel 2004 - 2014 4.2234451 10.8536310
100 dcant offset rel corr 1994 - 2004 2.1743497 1.5494230
100 dcant offset rel corr 1994 - 2014 -0.4625736 4.5209217
100 dcant offset rel corr 2004 - 2014 -1.8562194 10.8536310
100 delta C* - mod_truth 1994 - 2004 -0.3830000 0.0000000
100 delta C* - mod_truth 1994 - 2014 -0.3920000 0.0000000
100 delta C* - mod_truth 2004 - 2014 -0.0110000 0.0000000
100 mod_truth 1994 - 2004 3.3190000 0.0000000
100 mod_truth 1994 - 2014 7.1340000 0.0000000
100 mod_truth 2004 - 2014 3.8160000 0.0000000
500 C* drift 1994 - 2004 1.3750000 0.0000000
500 C* drift 1994 - 2014 2.7270000 0.0000000
500 C* drift 2004 - 2014 1.3520000 0.0000000
500 C* prediction error 1994 - 2004 -1.1121667 0.8811119
500 C* prediction error 1994 - 2014 -2.8666667 1.4164305
500 C* prediction error 2004 - 2014 -1.7431667 0.6842670
500 C* prediction error rel 1994 - 2004 -9.4692777 7.5020170
500 C* prediction error rel 1994 - 2014 -11.2683438 5.5677301
500 C* prediction error rel 2004 - 2014 -12.7284897 4.9964734
500 dcant offset 1994 - 2004 0.8475000 0.7334597
500 dcant offset 1994 - 2014 0.8990000 1.4163956
500 dcant offset 2004 - 2014 0.0950000 0.9674035
500 dcant offset rel 1994 - 2004 7.2158365 6.2448674
500 dcant offset rel 1994 - 2014 3.5338050 5.5675926
500 dcant offset rel 2004 - 2014 0.6936838 7.0639177
500 dcant offset rel corr 1994 - 2004 -4.4912729 6.2448674
500 dcant offset rel corr 1994 - 2014 -7.1855346 5.5675926
500 dcant offset rel corr 2004 - 2014 -9.1785323 7.0639177
500 delta C* - mod_truth 1994 - 2004 0.7090000 0.0000000
500 delta C* - mod_truth 1994 - 2014 1.3550000 0.0000000
500 delta C* - mod_truth 2004 - 2014 0.6460000 0.0000000
500 mod_truth 1994 - 2004 11.7450000 0.0000000
500 mod_truth 1994 - 2014 25.4400000 0.0000000
500 mod_truth 2004 - 2014 13.6950000 0.0000000
1000 C* drift 1994 - 2004 2.7210000 0.0000000
1000 C* drift 1994 - 2014 5.4120000 0.0000000
1000 C* drift 2004 - 2014 2.6920000 0.0000000
1000 C* prediction error 1994 - 2004 -2.3566667 1.7889628
1000 C* prediction error 1994 - 2014 -4.9470000 2.2246272
1000 C* prediction error 2004 - 2014 -2.6121667 0.6370926
1000 C* prediction error rel 1994 - 2004 -15.0106157 11.3946675
1000 C* prediction error rel 1994 - 2014 -14.4958537 6.5186719
1000 C* prediction error rel 2004 - 2014 -14.1749873 3.4571988
1000 dcant offset 1994 - 2004 0.8808333 1.6476046
1000 dcant offset 1994 - 2014 1.5180000 2.1753928
1000 dcant offset 2004 - 2014 0.6465000 0.8313663
1000 dcant offset rel 1994 - 2004 5.6104034 10.4942969
1000 dcant offset rel 1994 - 2014 4.4480910 6.3744040
1000 dcant offset rel 2004 - 2014 3.5082483 4.5114298
1000 dcant offset rel corr 1994 - 2004 -11.7208068 10.4942969
1000 dcant offset rel corr 1994 - 2014 -11.4103203 6.3744040
1000 dcant offset rel corr 2004 - 2014 -11.0999566 4.5114298
1000 delta C* - mod_truth 1994 - 2004 1.9870000 0.0000000
1000 delta C* - mod_truth 1994 - 2014 4.0550000 0.0000000
1000 delta C* - mod_truth 2004 - 2014 2.0660000 0.0000000
1000 mod_truth 1994 - 2004 15.7000000 0.0000000
1000 mod_truth 1994 - 2014 34.1270000 0.0000000
1000 mod_truth 2004 - 2014 18.4280000 0.0000000
3000 C* drift 1994 - 2004 2.5340000 0.0000000
3000 C* drift 1994 - 2014 5.0100000 0.0000000
3000 C* drift 2004 - 2014 2.4760000 0.0000000
3000 C* prediction error 1994 - 2004 -1.7530000 1.5980787
3000 C* prediction error 1994 - 2014 -1.8406667 1.6154944
3000 C* prediction error 2004 - 2014 -0.0718333 1.6052929
3000 C* prediction error rel 1994 - 2004 -9.9376417 9.0594032
3000 C* prediction error rel 1994 - 2014 -4.7849295 4.1995800
3000 C* prediction error rel 2004 - 2014 -0.3448883 7.7073791
3000 dcant offset 1994 - 2004 1.4501667 1.5131480
3000 dcant offset 1994 - 2014 4.5471667 1.5773942
3000 dcant offset 2004 - 2014 3.1795000 1.5614927
3000 dcant offset rel 1994 - 2004 8.2208995 8.5779367
3000 dcant offset rel 1994 - 2014 11.8206475 4.1005361
3000 dcant offset rel 2004 - 2014 15.2655080 7.4970843
3000 dcant offset rel corr 1994 - 2004 -6.1441799 8.5779367
3000 dcant offset rel corr 1994 - 2014 -1.2031645 4.1005361
3000 dcant offset rel corr 2004 - 2014 3.3776647 7.4970843
3000 delta C* - mod_truth 1994 - 2004 1.9180000 0.0000000
3000 delta C* - mod_truth 1994 - 2014 3.9780000 0.0000000
3000 delta C* - mod_truth 2004 - 2014 2.0590000 0.0000000
3000 mod_truth 1994 - 2004 17.6400000 0.0000000
3000 mod_truth 1994 - 2014 38.4680000 0.0000000
3000 mod_truth 2004 - 2014 20.8280000 0.0000000
10000 C* drift 1994 - 2004 2.7700000 0.0000000
10000 C* drift 1994 - 2014 5.4460000 0.0000000
10000 C* drift 2004 - 2014 2.6760000 0.0000000
10000 C* prediction error 1994 - 2004 -1.3288333 1.5591230
10000 C* prediction error 1994 - 2014 0.2348333 2.7613722
10000 C* prediction error 2004 - 2014 1.7031667 1.9916137
10000 C* prediction error rel 1994 - 2004 -7.4228205 8.7092113
10000 C* prediction error rel 1994 - 2014 0.6011656 7.0690222
10000 C* prediction error rel 2004 - 2014 8.0486114 9.4117181
10000 dcant offset 1994 - 2004 1.4425000 1.6810007
10000 dcant offset 1994 - 2014 5.6878333 2.8839545
10000 dcant offset 2004 - 2014 4.4521667 1.9610988
10000 dcant offset rel 1994 - 2004 8.0577589 9.3900164
10000 dcant offset rel 1994 - 2014 14.5606670 7.3828289
10000 dcant offset rel 2004 - 2014 21.0394909 9.2675148
10000 dcant offset rel corr 1994 - 2004 -7.4153726 9.3900164
10000 dcant offset rel corr 1994 - 2014 0.6190854 7.3828289
10000 dcant offset rel corr 2004 - 2014 8.3935857 9.2675148
10000 delta C* - mod_truth 1994 - 2004 1.4860000 0.0000000
10000 delta C* - mod_truth 1994 - 2014 3.0430000 0.0000000
10000 delta C* - mod_truth 2004 - 2014 1.5570000 0.0000000
10000 mod_truth 1994 - 2004 17.9020000 0.0000000
10000 mod_truth 1994 - 2014 39.0630000 0.0000000
10000 mod_truth 2004 - 2014 21.1610000 0.0000000

sessionInfo()
R version 4.1.2 (2021-11-01)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: openSUSE Leap 15.3

Matrix products: default
BLAS:   /usr/local/R-4.1.2/lib64/R/lib/libRblas.so
LAPACK: /usr/local/R-4.1.2/lib64/R/lib/libRlapack.so

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] kableExtra_1.3.4   geomtextpath_0.1.0 colorspace_2.0-2   marelac_2.1.10    
 [5] shape_1.4.6        ggforce_0.3.3      metR_0.11.0        scico_1.3.0       
 [9] patchwork_1.1.1    collapse_1.7.0     forcats_0.5.1      stringr_1.4.0     
[13] dplyr_1.0.7        purrr_0.3.4        readr_2.1.1        tidyr_1.1.4       
[17] tibble_3.1.6       ggplot2_3.3.5      tidyverse_1.3.1    workflowr_1.7.0   

loaded via a namespace (and not attached):
 [1] fs_1.5.2           gghalves_0.1.1     bit64_4.0.5        lubridate_1.8.0   
 [5] gsw_1.0-6          RColorBrewer_1.1-2 webshot_0.5.2      httr_1.4.2        
 [9] rprojroot_2.0.2    tools_4.1.2        backports_1.4.1    bslib_0.3.1       
[13] utf8_1.2.2         R6_2.5.1           DBI_1.1.2          withr_2.4.3       
[17] tidyselect_1.1.1   processx_3.5.2     bit_4.0.4          compiler_4.1.2    
[21] git2r_0.29.0       textshaping_0.3.6  cli_3.1.1          rvest_1.0.2       
[25] xml2_1.3.3         labeling_0.4.2     sass_0.4.0         scales_1.1.1      
[29] checkmate_2.0.0    SolveSAPHE_2.1.0   callr_3.7.0        systemfonts_1.0.3 
[33] digest_0.6.29      svglite_2.0.0      rmarkdown_2.11     oce_1.5-0         
[37] pkgconfig_2.0.3    htmltools_0.5.2    highr_0.9          dbplyr_2.1.1      
[41] fastmap_1.1.0      rlang_1.0.2        readxl_1.3.1       rstudioapi_0.13   
[45] jquerylib_0.1.4    generics_0.1.1     farver_2.1.0       jsonlite_1.7.3    
[49] vroom_1.5.7        magrittr_2.0.1     Rcpp_1.0.8         munsell_0.5.0     
[53] fansi_1.0.2        lifecycle_1.0.1    stringi_1.7.6      whisker_0.4       
[57] yaml_2.2.1         MASS_7.3-55        grid_4.1.2         parallel_4.1.2    
[61] promises_1.2.0.1   crayon_1.4.2       haven_2.4.3        hms_1.1.1         
[65] seacarb_3.3.0      knitr_1.37         ps_1.6.0           pillar_1.6.4      
[69] reprex_2.0.1       glue_1.6.0         evaluate_0.14      getPass_0.2-2     
[73] data.table_1.14.2  modelr_0.1.8       vctrs_0.3.8        tzdb_0.2.0        
[77] tweenr_1.0.2       httpuv_1.6.5       cellranger_1.1.0   gtable_0.3.0      
[81] polyclip_1.10-0    assertthat_0.2.1   xfun_0.29          broom_0.7.11      
[85] later_1.3.0        viridisLite_0.4.0  ellipsis_0.3.2     here_1.0.1