The accelerating rate of biodiversity loss is one of the most critical threats facing our society today. Addressing this threat is currently a top priority for policy makers, NGO’s, and most recently, business leaders. However, current efforts in conservation, management and restoration will only be as effective at curtailing the loss of biodiversity as the information used to guide and evaluate them. Although we might have a notion of what a sustainable landscape might look like, or what nature positivity might sound like on a farm, we are often at a loss when tasked with quantifying the impact of sustainability efforts and regenerative practices for biodiversity. How do we best measure biodiversity? What sources of data are most informative? Can we use the same metric across landscapes and continents? How much will this cost?
Artificial intelligence has been the present and future of our society for over 75 years. However, it is more recent advances in Machine Learning, and even more recently, Deep Learning, that have been driving the current explosion of AI applications worldwide, from facial recognition software to self-driving cars. A challenge as big as how to measure life is best met by leveraging these technological advances to develop an approach for quantifying biodiversity that is robust, scalable, and computationally efficient. Dr. Carla Gomes and her research group at the Institute for Computational Sustainability at Cornell University have developed a computational model that leverages the latest advancements in AI to quantify biodiversity in a framework that is a reliable and scalable.
The Global Biodiversity Information Facility (GBIF) houses the world’s most comprehensive dataset on biodiversity. Approximately half of the records in GBIF come from one single source – eBird. To date, eBird boasts over 1.52 Billion observations on birds submitted by over 910k citizen scientists worldwide, and growing at about 40% per year. These data are not just impressive in scale and volume, but also in the highly curated nature that they are collected and stored. The eBird data base boasts thousands of built-in filters, and a network of thousands of reviewers worldwide, that allows scientists and decision-makers to use eBird as a source of information for applications that range from population assessments for listing under the Endangered Species Act, to informing regulatory processes for the sighting of wind-energy developments. The built-in processes for quality-control in eBird allows for the community that contributes data to be highly diverse, ranging from community members and local tour guides, to trained survey technicians and professional birdwatchers. This results in an inclusive process for quantifying biodiversity that is not just accurate and scalable, but it is also cost-effective.
The Biodiversity Progress Index
In 2019, Nespresso’s AAA Sustainable Quality Program, the Cornell Lab of Ornithology, and the Institute for Computational Sustainability partnered to jointly develop the Biodiversity Progress Index (BPI), with the support of ECOM Sustainable Management Services and the INCAE business school. The BPI results from the application of a deep-neural network model to eBird data to quantify avian richness as a proxy for biodiversity. We are able to estimate species richness at high-resolution, for an entire country or region, for every month of the year.
One challenge with measuring biodiversity is that species richness varies due to many factors, including evolution. This results in patterns such as geographical differences in richness that are not related to investments in conservation or restoration. In addition, it complicates comparisons of biodiversity across regions within a country, or among countries. To address this, the BPI takes the measure of species richness and translates it into community completeness– a metric that tells us how many species are present given how many species should be present in that habitat. Community completeness is the most accurate metric for assessing how biodiversity is affected by environmental restoration and conservation efforts because it tells us the proportion of species – and the identity of those species – that are missing from a given location. To the left, you can species community completeness for Costa Rica. Values closer to 1 indicate areas with more complete avian communities, and vice versa.
Metrics that make up the BPI include:
- Species richness
- How many species are on my farm?
- How many species are benefiting from our sustainability project?
- Community completeness
- How complete is the bird community on my farm?
- Which species are present and which are missing?
- How is community completeness changing over time?
- What factors are driving species loss?
- Landscape-level performance:
- How does my sustainability certification program compare to the landscape around it?
- How much biodiversity are National Parks protecting relative to the surrounding landscape or the entire country?
The BPI leverages the practice of using birds as indicators of biodiversity and environmental health. Birds use a wide range of natural and human-managed environments, and many species respond quickly to changes in their surroundings, making them highly reliable and scalable indicators of environmental health. In addition, birdwatching has been shown to be beneficial for human well-being and local economies. The hemispherical migrations of birds also helps connect people and actions across national and international borders. Because of these factors, birds are now the most well-studied taxonomic group on the planet, providing the information we need to better understand patterns and drivers of global biodiversity loss.