The nature of ecological information

Having access to the right information, collected at the right place and time, is key for any successful ecological research. The current type, format, and amount of information managed by ecologists have dramatically changed over the last decade. However, the analysis of such information hasn’t changed much when compared to other industries and scientific fields.

Ecologists have traditionally used statistical methods developed by researchers from other fields. These methods (e.g. hypothesis testing and p-values) were often designed for information collected under “controlled environments”. Unfortunately, the natural world is far from “controlled”. This conundrum is likely to have limited the scope and impact of some ecological research.

Ecology often relies on observational data. The complex, unstructured and not normally distributed nature of this data represent a great analytical challenge. To make the most of this messy information, and consequently increase the efficiency of research and management efforts, ecologists need versatile, robust and accurate statistical tools.

Artificial intelligence vs machine learning

There are endless discussions about the “correct” description of artificial intelligence and machine learning. The following definitions are my best attempt at keeping them simple and relevant to this post.

• Artificial Intelligence (AI) is the theory and development of computer systems able to perform tasks normally requiring human intelligence.
• Machine learning (ML) is the field of computer science that aims to teach computers how to learn and act without being explicitly programmed. There are >100 different ML algortihms, including convolutional neural networks, boosted regression trees and random forests.

Why ecologists need to know about AI and ML?

The complex interactions that take place in nature makes it difficult for ecologists to determine causality. Nowadays, gathering large amounts of environmental data over multiple sites and at different points in time is easier than ever. This data, if sampled and analysed with the right tools, has the potential to enable ecologists to gain a holistic understanding of natural ecosystems.

ML algorithms are well-suited to analyse both large and small data sets. ML can outperform the predictive power and inference of common statistical tests used in ecology (e.g. Chi-square test, t-test, or ANOVA models). However, the use of ML in ecology is low compared to other scientific fields.

There are seeds and seedlings of change though. The number of conservation projects using AI increases as more ecologists understand the benefits of ML and NGOs, governments and private companies support its use for conservation. Some examples include ML algorithms applied to model species distribution, explore human-nature interactions, and recognise species from photos, video, or audio data.

The use of AI tools in ecology is far from reaching its peak. Further collaborations between ecologists, statisticians and computer scientists will elucidate the true potential of AI for wildlife conservation, ecology, and management of natural resources.