The question that began this is easy to ask. Walking a familiar trail, wanting a list of what grows along it — where would that list come from?
Kīpuka Puaulu is a good place to try. It is a small kīpuka — an island of older forest surrounded by younger lava flows — on the flank of Mauna Loa, in Hawaiʻi Volcanoes National Park. It has long been known for its native trees, and for a hau kuahiwi, Hibiscadelphus giffardianus, first described from this very forest. If any Hawaiian forest has been thoroughly written down, it is this one.
The writing-down lives in herbaria. This chapter gathers those records for a two-kilometre circle around the trail and turns them into a checklist. The steps are few, and the result looks like an answer. Whether it is one is the subject of the rest of the book.
## --- Standard packages ---
## data handling and graphics (dplyr, readr, ggplot2, ...)
library(tidyverse)
## formatted tables
library(gt)
## --- Package from github/kimbridges ---
## install once with: install_github("kimbridges/checklistr")
library(checklistr)
## --- Options ---
## suppress read_csv() column-type messages
options(readr.show_col_types = FALSE)A herbarium record is a pressed plant with a label. The label carries a name, the person who collected it, the date, the place, and the institution that keeps the sheet. Multiply that by a century of botanists and you have the raw material for a flora — scattered across the world’s collections, and now catalogued in databases a laptop can reach.
Those databases are pooled by GBIF — the Global Biodiversity Information Facility — an international, open aggregator that gathers digitized specimen and observation records from thousands of institutions into a single index anyone can query. What checklistr retrieves are its preserved-specimen records: the herbarium sheets, now reachable from a laptop.
That reach is worth pausing on. Before these records were digitized — and long before GBIF consolidated them — seeing a specimen meant one of two things: requesting a loan and waiting for the sheet to arrive by mail, or travelling to the herbarium yourself. I once had to visit the Central National Herbarium near Kolkata to see specimens I needed, and what should have been a simple errand became, that day, genuinely dangerous — for reasons that are not important here. During the COVID-19 closures I tried to reach several of the large US herbaria and found them, understandably, off limits. For me, a digital record is more than a convenience. The aggregation is not tidiness; it is what makes the specimens reachable at all.
checklistr makes that reach a single call. It asks GBIF for every vascular-plant specimen collected within two kilometres of a point.
The records that come back are saved for this document, so the pages render without waiting on the network. Here is what a few of them look like — each a plant, a person, a year, a collection.
| species | recordedBy | year | institutionCode |
|---|---|---|---|
| Solanum americanum | O Degener|I Degener | 1976 | AK |
| Coprosma ernodeoides | O Degener|I Degener | 1977 | AK |
| Dubautia ciliolata | A R Jamieson | 1991 | AK |
| Melicope cinerea | L M Cranwell|O H Selling|C Skottsberg | 1938 | AK |
| Morella faya | W R Sykes | 1989 | AK |
| Pipturus albidus | W R Sykes | 1989 | AK |
Every row is a real event: someone stood in this forest, on a known day, and gathered that plant. The checklist we are about to build is assembled entirely from events like these.
Going from a pile of records to a list of species takes a few steps. Names have to be resolved to accepted species, because a plant may have been filed under several names over the years. Duplicate records — the same specimen entered more than once — have to be removed, so the counts mean something. And to make the list useful, rather than merely correct, each species is given a growth form, so trees can be told from herbs.
checklistr does this as one pipeline.
## resolve names, deduplicate, flag determination conflicts,
## and assemble the accepted-species checklist
checklist <- records |>
add_primary_collector() |>
build_checklist()
## add growth form (trees, shrubs, herbs, graminoids, ferns)
growthform <- gift_growthform(cache = "gift_growthform.rds")
checklist <- add_lifeform(checklist, growthform)The saved output of that pipeline is the checklist itself. It sorts the forest into a familiar shape.
| lifeform | taxa |
|---|---|
| tree | 43 |
| herb | 37 |
| fern | 33 |
| graminoid | 24 |
| shrub | 23 |
The pipeline turns a century of records into a checklist of 160 taxa. A sample of the native trees — the frame the forest is built on — shows how ordinary the result looks.
| accepted_name | family | n_records |
|---|---|---|
| Charpentiera obovata | Amaranthaceae | 1 |
| Ilex anomala | Aquifoliaceae | 2 |
| Cheirodendron trigynum | Araliaceae | 1 |
| Perrottetia sandwicensis | Dipentodontaceae | 1 |
| Acacia koa | Fabaceae | 4 |
| Sophora chrysophylla | Fabaceae | 3 |
| Hibiscadelphus giffardianus | Malvaceae | 5 |
| Hibiscadelphus hualalaiensis | Malvaceae | 3 |
Read that table quickly and it is exactly what was asked for: the plants that grow here, ready to jog a memory on the trail. Acacia koa. Metrosideros polymorpha. The names a botanist walking the loop would want at hand.
But look again at how it was built. Every name was resolved against a shifting taxonomy. Every count depended on a judgment about which records were duplicates. Every row rests on someone having collected the plant, at some time, and on that specimen surviving and being catalogued — and on no one having collected the plants that are missing. The list is not a fact about the forest. It is a reconstruction, assembled from evidence, and every entry is a claim.
The chapters that follow take those claims apart, one kind at a time — who made the record, and when; what the map of the records reveals about scale; how complete the reconstruction is; where it contradicts itself; and where the experts who built it disagree. A checklist, it turns out, is an argument. This one is a good place to start reading it.