BIEN Growth Forms Module

Contents

I. Overview
II. Methods
III. Usage
IV. Version info

I. Overview

Code for inferring growth forms of species within the BIEN database. This is done in two general steps: (1) compile all growth form attributions for each species, and (2) infer a single consensus growth form for each species. Family- and genus-level attributions are used to impute growth forms to species for which species-level attributions are not available.

For each consensus growth form assigned to each species, a confidence score (0-1), categorical confidence flag ('very high', 'high', 'med', 'low', 'very low') and attribution method ('direct from species', 'imputed from family', 'imputed from genus') is also provided.

This early 'alpha' release is a series of SQL only, with commands intended to be pasted into the postgres SQL interface (see III. Usage for sequence). Also note that the initial import steps require manual inspection and correction of taxonomy and growth form vocabulary. See instructions in individual scripts.

The individual scripts in this workflow will be redeveloped as semi-automated pipeline at a later date.

III. Methods

All growth form attributions at the species, genus or family level were compiled from the BIEN traits table (agg_traits) and the BIEN observations table (view_full_occurrence_individual).

Taxonomy

Although taxon names had previously been standardized with the TNRS, names were checked to ensure that each species belonged to exactly one family. A small number of names (33) had homonym issues and were corrected by manual inspection.

Vocabulary.

All growth form names and codes were standardized to conform to the following simplified growth form schema:

Growth form	Notes
herb	includes grasses, epiphytic herbs and non-woody parasites
bambusoid	woody tree-like grasses
shrub	includes shrubby hemi-parasites)
tree
vina	herbaceous climbers, originating on the ground
liana	herbaceous climbers, originating on the ground
hemiepiphyte	climbers, germinating in the forest canopy as an epipthye and secondarily sending roots to the ground.

Consensus growth form (gf_cons)

For each species, the growth form with the highest proportion of attributions (observations, or obs) was selected as the consensus growth form.

Consensus growth form confidence score (gf_conf)

A confidence score from 0-1 was calculated for each consensus growth form attributed to each species, using the following formula:

For species, gf_conf is calculated as follows:

gf_conf = GREATEST( gf_cons_prop - low_n_penalty, 0 ) (1)

Where,

gf_cons_prop = gf_cons_obs / total_obs (2)

and

low_n_penalty = (1-(LEAST(total_obs, pthresh)::numeric/100))/ pdenom (3)

For genera and families, gf_conf is calculated as follows:

gf_conf = GREATEST( gf_cons_prop - low_n_penalty - low_species_penalty, 0 ) (4)

where

low_species_penalty = ( 1 - spp_gf_prop ) * pmax_spp_gf (5)

In equation (2) gf_cons_obs is the total growth form attribution for the consensus growth form and total_obs is the total growth form attributions. In equation (3), pthresh is the minimum sample size threshold, below which the penalty is non-zero; pdenom is the "probability denominator", a scaling parameter which sets the maximum penalty subtracted at total_obs=0. At pthresh=100, pdenom=5 sets a maximum penalty of 0.2. That is, no species-level observations (total_obs=0) result in an additional 20% being subtracted from gf_cons_prop.

In equation (5), spp_gf_prop is the proportion of species in the genus or family having a growth form attribution. pmax_spp_gf set the maximum penalty; a pmax_spp_gf=0.2 caps the maximum penalty at 0.2 (20%).

Parameter settings used were pthresh=100, pdenom=5 and pmax_spp_gf. Using these setting, the maximum amount by which gf_cons_prop can be reduced is 0.4. Use of the Postgres function GREATEST in equations (1) and (4) prevents gf_conf from dropping below zero.

Consensus growth form confidence flag (gf_conf_flag)

gf_conf_flag converts gf_conf to confidence categories, as follows:

Value	gf_conf range
very high	>=0.90
high	<0.90 AND>=0.75
med	<0.75 AND>=0.5
low	<0.5 AND>=0.25
very low	<0.25

Growth form imputation

For a new species list, consensus growth forms are assigned by first joining tp species, then th genera, and finally, growth form is assigned to the remining species by joining to family. For each determination, the method used to determine growth form, gf_method, is recorded as follows:

gf_method	meaning
direct from species	gf assigned by joining on species
imputed from genus	gf assigned by joining on genus
imputed from family	gf assigned by joining on family

III. Usage

The scripts must be run in the following order:

1. gf_traits.sql
2. gf_obs.sql
3. gf_consensus_species.sql
4. gf_consensus_genus.sql
5. gf_consensus_family.sql
6. gf_manual_corrections.sql
7. gf_conf
8. gf_impute
9. gf_correct_syn_species.sql (optional; see script)

The individual scripts in this workflow will be redeveloped as semi-automated pipeline at a later date.

IV. Version info

0.1

• First working beta release. Scripts run separately, in sequence shown in II. Usage.