Colin - weeks 3 and 4

[bpbond]

1. Identify soil temperature columns
2. Go through them one by one. Extract the depth and put it into its own column, and then put the calculated Q10 into its own column

main_results <- list()
for( ... a loop through the datasets ...) {

results <- data.frame(columns = temperature_columns, depth = NA, q10 = NA, n = NA)
for(tcol in temperature_columns) {
     results$depth <- ...
     results$n <- nrow(dat)
     results$q10 <- calc_q10(dat[tcol], dat$flux)
}

main_results[[dataset]] <- results
}

main_results <- bind_rows(main_results, .id = "Dataset")

[bpbond]

This is all inside a larger loop going through the datasets themselves.

In the above code I'm assuming that temperature_columns is a character vector holding the names of CSR_Txxx columns.

[bpbond]

There is CSR_TAIR and/or CSR_TCHAMBER. Be careful with these - you will want to set depth to ...? Let's say -1 for now.

[bpbond]

Here's a better grep: tcols <- grep("^CSR_T(AIR|[0-9]+)", names(dsd))

[bpbond]

Next steps after 2020-07-23 phone call with @jinshijian and @10aDing :

• Colin to send or post summary of existing output - i.e. Q10 for entire datasets
• Calculating by week - here is where the dplyr group_by syntax is really handy. You can use the lubridate package to do the week calculation. So for example

  df %>%
  mutate(woy = week(CSR_TIMESTAMP_BEGIN)) %>%
  group_by(week) %>%
  summarize(...)

• Calculate Rh and Rs Q10s separately, for day and night (using information in ports)
• We'd like to calculate sunset and sunrise (and thus day and night) dynamically. See https://www.esrl.noaa.gov/gmd/grad/solcalc/solareqns.PDF . I would suggest that this is its own function, with a bit of test code like we did for calc_q10.

[bpbond]

So the output data frame from the above processing should look something like this

• Dataset
• Year
• Week of year
• Measurement variable (Rs or Rh)
• Day or night
• Temperature depth
• Q10
• Various statistics (R2, N, p-value of regression/slope)
• Mean SM, mean T, mean flux

[bpbond]

@jinshijian noted good idea of start with just one dataset, work all the way through developing pipeline, and after robust you can expand to all

[bpbond]

Notes from call with @10aDing 2020-07-29:

• Let's get the two open PRs merged (note #11 has a conflict currently)
• At that point I will clone, open a feedback branch and then a PR with some suggestions
• Colin to work on adding day/night calculation - if sunriseset or whatever only takes single values, you can use vapply to apply to a whole column
• Colin to work on adding visualization and summaries of Q10 - histograms, etc.

[bpbond]

Let's stick with just calculating for Rh for now.