What are the most recognized metrics so far and how are they computed?

[cboettig]

Biodiversity metrics are more often defined in the abstract than in operational definitions (i.e. software implementations), and less still are they computed in 'real time' from available streaming data sources. There's lots riding on these metrics and lots of scope for progress in how the metrics are defined.

In this exploratory repository, we're seeking to get a better sense of what metrics already exist, and what data (and assumptions) feed them. This is possibly a stepping stone to making that process more automated and more modular, so that each step in a pipeline can start to be built on the best available data and methods, instead of the most conveniently at hand.

A few examples to get us rolling.

WWF Living Planet Index

The WWF's Living Planet Index has been around since 2007, with it's most recent report published last year (2022). Like most influential indices, there is plenty of controversy about what it is based on and how it is computed (e.g. see this recent Nature piece and linked replies).

GEO BON EBVs

GEO BON's Essential Biodiversity Variables ("EBVs") are another highly visible effort to define metrics, currently listing eight biodiversity indicators. Most of the technical definitions of these metrics are things we can't actually measure with available data, but may be able to approximate in various ways.

For instance, the 'biodiversity intactness index, BII' is defined as the average abundance of a species in an area relative to its reference or pristine abundance, averaged over wide number of representative species. In practice it is being calculated based on the PREDICTS database using various rough proxies, drawing data from studies not designed to measure anything like this. Other indicators may be a bit more grounded in observational data.

Wrangling indicators

Despite their challenges and weaknesses, these numbers already matter, and are becoming increasing cited by to multi-lateral agreements and corporate 'nature positive' goals. The Convention on Biodiversity (CBD) in 2010 set the 20 Aichi Biodiversity Targets for 2020, though according to CBD's own report none of the 20 were fully achieved, many for lack of measurable definitions in the first place, e.g. see 1, 2. Several of the EBV indicators are considered to be metrics of Aichi Targets.

The Biodiversity Indicators Partnership seeks to wrangle together a list of many of the more common indicators, including those mentioned above. (As you can see, one can do a lot of clicking without seeing an actual indicator, let alone the data behind it).

Exploratory Data Analysis

A good first step will be in tracking down the data sources behind these indicators and doing a bit of exploratory data analysis.

1. What is the spatial, temporal, and taxonomic coverage and resolution of each indicator?
2. What are the underlying data sources, and how are they assembled or what bias is there in the data selection?
3. To what extent are these indicators based on the same data, and showing the same patterns?
4. What are some of the emerging potential data sources (e.g. in remote sensing, camera traps, eDNA, citizen science efforts, etc) that might provide richer data sources for addressing some of these gaps?
5. What are the challenges in being able to assimilate those sources?

[gizarp]

This is super useful @cboettig, thanks!

I took the plunge and started to play with the LPI dataset while trying to reproduce the paper you mentioned. Some code is provided to run the paper's BHM model, but there doesn't seem to be much on data preparation, visualization, etc. so this might be a good and interesting exercise.

My hope is to use this as a starting point to maybe develop more comprehensive general-purpose tools to manipulate the LPI data (and potentially other similar data), and learn from the process!

Will update on more!