I was digging into the issue with the labels and from what I can tell, I actually over-engineered my timezone thing 😄 It appears that NWIS always reports things in the standard time of whatever timezone, not the daylight savings time. I was adjusting for the need to convert from daylight time (DT) to standard time (ST) for some times of the year by subtracting an hour. What this ended up doing was moving those dates that fall into the "daylight time" part of the year back an hour even though they were already in standard time. I've added a note in the code and changed how it was adjusting.

I don't think it is worth regenerating the historic data at this point since it is a simple hour shift. Though, I can't tell how I missed that issue in my tests and checks before. Don't really have time to think about that, but am re-running the current dates with this fix now. Code and updated data to come soon.

Gah, but sadly not every site is reported in standard so now I end up with some sites on Oct 1, 2022 seems impossible to win here. I might just do a filter here and call it. Alternatively, I could make an in file that represents these sites that go against the norm and use that to do something different and we can add to it in the future if we need to. @ceenell let me know what you think is best.

library(tidyverse)
n_vals_per_date <- read_csv("1_fetch/out/gw_data.csv") %>%
  group_by(Date) %>%
  tally() 
head(n_vals_per_date, 3)
tail(n_vals_per_date, 3)

> head(n_vals_per_date, 3)
# A tibble: 3 x 2
  Date           n
  <date>     <int>
1 2021-10-01  2130
2 2021-10-02  2133
3 2021-10-03  2130
> tail(n_vals_per_date, 3)
# A tibble: 3 x 2
  Date           n
  <date>     <int>
1 2022-09-29  1521
2 2022-09-30  1518
3 2022-10-01    76  # <--- GAH! Now there some bleeding into the next year.

There are 76 sites across 4 states that have this issue.

library(tidyverse)
sites_in_fy23 <- read_csv("1_fetch/out/gw_data.csv") %>% 
  filter(Date > as.Date('2022-09-30'))

gw_sites_sf <- scipiper::scmake('gw_sites_sf')
gw_sites_sf_fy23 <- gw_sites_sf %>% 
  filter(site_no %in% sites_in_fy23$site_no) %>% 
  mutate(state_abbr = dataRetrieval::stateCdLookup(state_cd))
gw_sites_sf_fy23 %>% 
  st_drop_geometry() %>% 
  group_by(state_abbr) %>% 
  tally()

# A tibble: 4 x 2
  state_abbr     n
  <chr>      <int>
1 CA             4
2 IL            45
3 IN            10
4 MI            17

Document this in an issue as a future improvement and then filter it. You were thinking to chop the last day off for the 76 sites, right? I wouldn't want to drop the sites altogether, and the time lag is going to be insignificant in the animation.

Sounds good. And yes, going to chop off just the last day not dump those sites.

See #78