Fighting dynamite fishing with AI

Blast fishing was a serious but unquantified problem in Tanzania. I helped WWF set up a system to monitor this practice using AI to automatically detect blasts in thousands of hours of underwater recordings. With this new data, WWF was able to raise awarness and eventually support communities to fight blast fishing.

The challenge

Blast (or dynamite) fishing is an illegal technique used to kill hundreds of fish in seconds. It can give some fishers a quick gain, but blast fishing is dangerous, wasteful, and destroys the coral reefs that sustain marine life and the livelihoods of thousands of people. WWF was aware of the problem but without any data about its frequency or distribution it was unable to quantify the impact on the ecosystem and therefore it was hard to convince authorities to initiate a campaign against this practice.

The solution

As a consultant, I led a team from multiple organisations to test a system to detect and monitor blast fishing using acoustic data and artificial intelligence. I deployed underwater recorders in strategic locations to record sounds over a couple of months. Then, I used some of the data to train an artificial neural network that automatically detected blasts in thousands of hours of recordings.

The result

I found that blast fishing was more widespread than anticipated. Nearly 20 blasts occurred every day in the vicinity of Dar es Salaam alone. The neural network had a 98% of blasts and proved it was a viable solution to monitor blast fishing in the long term. Using the results I provided, WWF was able to secure additional funding to further investigate the magnitude of the problem and raise awareness among communities and authorities. By 2020, Tanzania finally experienced a reduction in blast fishing.
Skills: Matlab Neural networks Artificial Intelligence R Data visualisation Data science Report writing