enriching the sample size experiment

Set Up

library(ggplot2)
library(dplyr)
library(knitr)
BASE_DIR <- "/Users/dorh012/projects/2021/cleanup_stitch/stitches"

pangeo_table <- read.csv(file.path(BASE_DIR, "stitches", "data", "pangeo_table.csv"), stringsAsFactors = FALSE)

req_exps <- c("ssp126", "ssp245", "ssp585", "ssp370")
pangeo_table %>% 
  filter(variable == "tas" & domain == "Amon") %>% 
  filter(experiment %in% req_exps) -> 
  tas_exps
tas_exps %>% 
  group_by(variable, model) %>% 
  summarise(experiment = n_distinct(experiment)) %>% 
  filter(experiment == length(req_exps)) %>% 
  ungroup ->
  full_experiments 

`summarise()` has grouped output by 'variable'. You can override using the `.groups` argument.

How many models meet these conditions?

How many models meet the full 4 experiment criteria?

length(full_experiments$model)

[1] 23

Of these models waht does the ensemble count look like?

tas_exps %>% 
  filter(model %in% full_experiments$model) %>% 
  group_by(variable, model, experiment) %>%
  summarise(ensemble = n_distinct(ensemble)) -> 
  ensemble_counts

`summarise()` has grouped output by 'variable', 'model'. You can override using the `.groups` argument.

Of the models with all the experiments how many meet the criteria of ~10 to 8 per ensemble?

Unfortunately it looks like there is only one model that meets these requirements. Let’s do the experiment with the various tolerances & take a look at the results.

Target ssp245

read.csv(file.path(BASE_DIR, "MPI-ESM1-2-LR_ssp245results.csv"), stringsAsFactors = FALSE) %>% 
  mutate(tol = as.character(tol)) %>% 
  mutate(stitching_id = paste0(stitching_id, '~', tol)) -> 
  ssp245_data

ssp245_data %>%
  mutate(model = "MPI-ESM1-2-LR") %>% 
  group_by(model, tol) %>% 
  summarise(count = n_distinct(stitching_id)) -> 
  stitching_count 

`summarise()` has grouped output by 'model'. You can override using the `.groups` argument.

stitching_count %>% 
  kable()

model	tol	count
MPI-ESM1-2-LR	0.07	11
MPI-ESM1-2-LR	0.1	15
MPI-ESM1-2-LR	0.13	20

list.files(file.path(BASE_DIR, "stitches", "data", "tas-data"), "MPI-ESM1-2-LR", full.names = TRUE) %>% 
  read.csv(stringsAsFactors = FALSE) %>% 
  filter(grepl(x = experiment, pattern = "ssp")) %>%  
  group_by(experiment) %>%  
  summarise(n = n_distinct(ensemble))

list.files(file.path(BASE_DIR, "stitches", "data", "tas-data"), "MPI-ESM1-2-LR", full.names = TRUE) %>% 
  read.csv(stringsAsFactors = FALSE) %>% 
  filter(experiment ==  "ssp245") -> 
  og_data
ssp245_data %>% 
  left_join(stitching_count) %>% 
  mutate(tol_label = paste0(tol, " (", count,")")) %>% 
  ggplot() + 
  geom_line(aes(year, value, color = tol, groupby = stitching_id)) + 
  geom_line(data = og_data, aes(year, value, groupby = ensemble, color = "comp. data"), alpha = 0.4) + 
  facet_wrap("tol_label", ncol = 1) + 
  theme_bw() + 
  labs(y = "Deg C") + 
  scale_color_manual(values = c("0.1" = "#E69F00","0.13" = "#56B4E9", "0.07" = "#009E73", "comp. data" = "black"))

Joining, by = "tol"
Ignoring unknown aesthetics: groupbyIgnoring unknown aesthetics: groupby

Target ssp370

read.csv(file.path(BASE_DIR, "MPI-ESM1-2-LR_ssp370results.csv"), stringsAsFactors = FALSE) %>% 
  mutate(tol = as.character(tol)) %>% 
  mutate(stitching_id = paste0(stitching_id, '~', tol)) -> 
  ssp370_data

ssp370_data %>%
  mutate(model = "MPI-ESM1-2-LR") %>% 
  group_by(model, tol) %>% 
  summarise(count = n_distinct(stitching_id)) -> 
  stitching_count 

`summarise()` has grouped output by 'model'. You can override using the `.groups` argument.

stitching_count %>% 
  kable()

model	tol	count
MPI-ESM1-2-LR	0.07	11
MPI-ESM1-2-LR	0.1	14
MPI-ESM1-2-LR	0.13	18

list.files(file.path(BASE_DIR, "stitches", "data", "tas-data"), "MPI-ESM1-2-LR", full.names = TRUE) %>% 
  read.csv(stringsAsFactors = FALSE) %>% 
  filter(experiment ==  "ssp370") -> 
  og_data
ssp370_data %>% 
  left_join(stitching_count) %>% 
  mutate(tol_label = paste0(tol, " (", count,")")) %>% 
  ggplot() + 
  geom_line(aes(year, value, color = tol, groupby = stitching_id)) + 
  geom_line(data = og_data, aes(year, value, groupby = ensemble, color = "comp. data"), alpha = 0.4) + 
  facet_wrap("tol_label", ncol = 1) + 
  theme_bw() + 
  labs(y = "Deg C") + 
  scale_color_manual(values = c("0.1" = "#E69F00","0.13" = "#56B4E9", "0.07" = "#009E73", "comp. data" = "black"))

Joining, by = "tol"
Ignoring unknown aesthetics: groupbyIgnoring unknown aesthetics: groupby