TITLE: Using `any()` with `ifelse()` and `group_by()`
DATE: 2022-08-14
AUTHOR: John L. Godlee
====================================================================


I am doing some more work to update the SEOSAW database. When we 
measure the growth and mortality of trees to monitor biomass 
dynamics, it is important to have a consistent way of recording 
mortality. For trees, especially trees with multiple stems, the 
concept of "death" is not as simple as for animals. We generally 
classify a tree as dead if all the above-ground tissue on the tree 
appears to be dead, i.e. has no living leaves or buds, and no sap 
below the bark. Sometimes though, a single stem on a multi-stemmed 
tree will die, but the rest of the tree continues to live. 
Sometimes a stem or tree may appear dead, but comes back to life 
later. In savanna systems especially, it's fairly common for trees 
to partially die, or to resurrect after a disturbance event like a 
drought or a fire.

  [SEOSAW database]: https://seosaw.github.io

In SEOSAW we previously classified stems as either "alive", 
"topkilled", "resprouting" or "dead". I first heard the term 
topkill used in a paper by Bill Hoffmann (Hoffmann et al. 2009), 
where they define it as "the complete death of aerial biomass". In 
SEOSAW we tried to tighten up the term a bit by defining it as a 
stem where there is no living tissue above the point of DBH 
measurement. A resprouting stem is defined as a topkilled stem 
where there is new material sprouting from the stem. The difficulty 
with this old system was that it was very subjective to decide 
whether a stem was topkilled or dead, and whether a stem was 
resprouting or top-killed. How do you define whether living tissue 
on a stem is the result of resprouting? Does resprouting tissue 
have to be growing directly from the above-ground parts of the 
topkilled stem to be classed as resprouting?

  [Hoffmann et al. 2009]: https://doi.org/10.1890/08-0741.1

We recently came up with a revised mortality classification. This 
classification only has "alive", "resprouting", and "dead". A stem 
is classified as resprouting if it doesn't appear to have any 
living tissue above the DBH measurement, but there are signs of 
life elsewhere on the tree. Other signs of life can come from other 
stems on the tree, from resprouting material at the base of the 
tree, or from resprouting material on the stem of interest below 
the DBH measurement. A stem can only be classified as dead if all 
stems on the tree are dead and there is no sign of life anywhere 
else on the above-ground parts of the organism. It follows then 
that on a multi-stemmed tree, a stem cannot be classified as dead 
unless all other stems on the tree are dead. Additionally however, 
even if all stems on a tree are lacking living tissue this still 
doesn't mean that the stems will be classified as dead if there is 
living material somewhere else on the tree, i.e. from the base of 
the stool on a heavily coppiced individual.

In light of this new classification, I had to adjust the mortality 
data in some old censuses. I wrote a check that would flag stems 
classified as dead when other stems on the tree were classified as 
alive or resprouting.

    # Split data by plot ID, tree ID, and census date
    x_split <- split(x, list(x$plot_id, x$tree_id, x$census_date), 
        drop = TRUE)

    # Only check trees with more than one stem
    x_split_fil <- x_split[unlist(lapply(x_split, nrow)) > 1]

    # If there are some trees with multiple stems, check each one
    if (length(x_split_fil) > 0) { 
      # For each tree
      x_zomb <- unlist(lapply(x_split_fil, function(i) {
        # Flag if any stems are dead and any other stems are alive 
or resprouting
        status <- i$stem_status[!is.na(i$stem_status)]
        any(status == "d") & any(status %in% c("a", "r"))
      }))
      names(x_zomb[x_zomb])
    }

I also concocted a little snippet using {dplyr} that I could use to 
adjust the values of stems that had been classified as dead to 
resprouting.

    x %>% 
      group_by(plot_id, tree_id, census_date) %>%
      mutate(
        stem_status = case_when(
          # If any stems on the tree are alive or resprouting 
          # change dead stems to resprouting
          any(stem_status %in% c("a", "r")) & stem_status == "d" ~ 
"r",
          TRUE ~ stem_status)
      )