Measuring biodiversity — how machine learning can help achieve impact goals
3 ways in which AI is being used to measure biodiversity loss and help conservation efforts towards restoration and rehabilitation
A few decades ago, a small community of scientists and students would measure biodiversity by demarcating a representative plot of land in a sensitive or vulnerable area, counting the number and abundance of species, and using statistical methods to extrapolate conclusions and predictions from the data. This worked well enough at a local or regional level.
Since then, the food and resource needs of a constantly increasing human population has led to widespread land degradation and deforestation, decimating our planet’s biodiversity at an unprecedented global scale. The entire ecosystem on earth has become vulnerable and at risk.
The United Nations has included several targets relating to biodiversity in the Sustainable Development Goals (SDGs) that humankind is striving to achieve by 2030. Governments, NGOs, companies, communities and individuals have launched thousands of initiatives to try and achieve these targets. The conundrum lies in how to measure the impact of these initiatives in order to assess whether we are achieving the SDGs.
The ongoing collection and analysis of large amounts of data requires greater computing power than is possible for the small group of scientists who are involved in these studies. This is where artificial intelligence can provide solutions. Machine learning models can process complex information and can also be trained to predict outcomes and trends, enabling us to proactively manage risks that may arise in the future.
Here are three ways in which innovations in technology are helping to achieve the ambitious targets contained in the SDGs …
1. Botanist in your pocket
SDG15 (relating to life on land) includes the conservation and restoration of ecosystems, and halting the loss of biodiversity. A highly biodiverse area like the Cape Floristic Region has over 9000 plant species within a million hectares. Even if we had several teams of botanists working around the clock, it would be difficult to be able to correctly identify, count and monitor all species.
AI-based platforms and applications enable anyone with a mobile phone to upload photos of plants together with GPS information, and to get feedback from other users regarding the identification of the species and its inclusion in a database. Advances in deep learning also mean that models can be trained to identify species from photos without constant human interaction.
Plant data can therefore be collected on a massive scale by farmers, hikers, home owners, tourists, students and other members of communities. Machine learning enables scientists to analyse these large amounts of data, make predictions about their sustainability and put measures in place to mitigate their extinction.
2. Game ranger on patrol 24/7
The SDGs include targets to protect and prevent the extinction of threatened species, and to take urgent action to end poaching and trafficking of protected species. Conservation parks struggle to prevent the poaching of rhinos, elephants, and other protected species. Poaching is often discovered after the fact, when the deaths or removal of animals can only be recorded and not prevented.
AI-based security camera systems are able to detect movement within conservation areas and use image classification to identify whether the motion is caused by an animal or person. Poachers can therefore be detected before they attack an animal, and a message sent to rangers in real-time so that they can respond immediately.
Images of poachers are captured so that gangs operating in a particular area can be identified and a record kept of the individuals involved. Patterns of behaviour can also be discerned and predictions made about poaching activity.
3. Teach a man to fish … sustainably
Decades of overfishing have resulted in depleted fish populations in our oceans and rivers. It is estimated that almost a quarter of fishing globally is illegal and unregulated. SDG14 (relating to life below water) includes the target of restoring fish stocks by regulating sustainable fishing practices and management plans.
Using technology to manage fish stocks is currently one of the most innovative areas of sustainability. Machine learning is increasingly being used to supplement sonar and electronic monitoring systems that were already in place, to provide end-to-end traceability.
Sonar technology is already used to monitor fish populations before they are caught, i.e. while they are still under water. Experiments are being conducted to use machine learning to identify and classify the different species. The system tracks fish distribution and is able to make recommendations to fishermen about where to find the most profitable and sustainable fishing areas.
Electronic monitoring of fishing vessels together with GPS technology can also be used to monitor their coordinates, detect when they are fishing in restricted or vulnerable areas, and plot the density of vessels in a particular area to be alerted of potential overfishing.
Cameras attached to fishing vessels monitor what is being caught through video footage. Machine learning classification models can be used to identify what species are being caught and transmit images to a central database.
One of the consequences of overfishing is the loss of millions of tonnes of bycatch (other marine creatures that are unintentionally captured during the fishing process). Standardised acoustic deterrents like pingers have proven to be ineffective in preventing this because they actually attract animals like seals which have come to associate the sound with large amounts of fish. New models are being tested which detect marine animals through sonar technology, and identify them by species using machine learning classification. Acoustic deterrents are then activated to emit specific frequencies or no sound, depending on which animals are close to the nets. Continuous monitoring and training of the model is expected to result in greater efficiency in reducing loss of marine life bycatch.
Conclusion
Traditional ways to conserve biodiversity are not adequate to enable us to achieve the SDG targets by 2030. Innovations in technology, especially in artificial intelligence, are helping to automate processes to facilitate conservation efforts towards restoration and rehabilitation of biodiversity.
