Skip to main content
Special Feature

Tech gone wild

On the beaches of Goa, Tech for Wildlife uses GIS data and aerial imagery to examine nesting patterns of olive ridley turtles.

Researchers are aiding wildlife and forest conservation efforts with tech-enabled tools that examine a host of jungle sights and sounds.

Bathed in the silver gleam of a full moon night, forest officials at Gujarat's Gir National Park & Wildlife Sanctuary have been setting out every month on a number-crunching exercise. Their aim is to count the lions in the forest, aided by the brightness of the moon.

But two summers ago, the in-house exercise, popularly known as Poonam Avlokan, took a new turn: it segued into the 15th Asiatic Lion Population Estimation exercise.

Lion censuses require large human participation, and are carried out every five years with the help of partner organisations and volunteers. But the exercise came to a standstill because of the constraints posed by the COVID-19 pandemic. So, the scope of Poonam Avlokan was expanded, and the counting by the forest officials in June 2020 was taken as the actual census.

It found 674 Asiatic lions (up from 523 in 2015) inhabiting a nearly 30,000-sq-km area. Data such as the time of spotting a lion, GPS location and individual identification marks were noted. The lions were all observed visually and physically. Some were fitted with radio collars, as well, their movements tracked by a high-tech unit, which followed the radio signals.

Technology is being used to aid and enhance wildlife and forest conservation. Take the Gir Hi-Tec Monitoring Unit, for instance. Set up in 2019, it has not just been tracking the movement of radio-collared lions but has also introduced a variety of tech-enabled tools to help the sanctuary in its conservation and management efforts. Earlier this year, an AI-enabled, computer-based photo identification tool called SIMBA (Software with Intelligent Marking Based identification of Asiatic lions) was deployed for identification.

SIMBA leverages machine learning to distinguish between lions, and automates the process of individual identification. A range of data, such as scars on the face, notches on the ears and whisker spot patterns, is used to identify the lions. Whisker spots, in particular, are useful in identification because they do not change over time and differ from one lion to another. Each whisker spot is a small and dark area surrounding a single whisker. SIMBA has been enabled to distinguish between lions using a dataset of nearly 1,000 lion images. With this, forest officials are now able to create a database of all the local lions, each with a unique identification number and information on other attributes.

The foresters hope to use the data for conservation and management. Currently, there is one source population of lions in Gir and some satellite populations, or smaller groups of lions living slightly away from the source group. SIMBA will help track the "movement between source and satellite population and (detect) if any lion is exploring a new area," says Ram Mohan, Deputy Conservator of Forests at the Gir National Park & Wildlife Sanctuary. With SIMBA, it will also be possible to figure out the original location of a lion in a new area. "This will help us get rid of 'duplication' in our data, making it more 'accurate'," Mohan says.

Unlike the Gir exercise, tiger censuses in India use camera trap images. Cameras installed at many locations in forests detect movements of animals and capture images. From these, the images of tigers are picked out and used for the four-yearly tiger census.

Since camera traps capture any activity in the vicinity, a huge repository of different kinds of images is collected. During the 2018 tiger census, a repository of nearly 35 million images was created. For the census, tiger images had to be selected from this collection. As it was near impossible to manually sieve out tiger images from such a vast collection, researchers developed a module called CaTRAT (Camera Trap data Repository and Analysis Tool) that uses a deep neural network for image classification.

"It takes in any image as an input and tells you the species of the animal in the image," says Saket Anand, Associate Professor at the Indraprastha Institute of Information Technology (IIIT), Delhi. Anand had led the development of the module.

CONTINUE READING

Get unlimited digital access on any device.

Get the print magazine delivered at home.

Subscribe

Already a subscriber? Login

PAST ISSUES - Free to Read

share-alt-square
Volume 01 Issue 04 Jul-Aug 2022
Read This Issue
share-alt-square
Volume 01 Edition 03 Sep-Oct 2021
Read This Issue
share-alt-square
Volume 01 Edition 02 Jul-Aug 2021
Read This Issue
Search by Keywords, Topic or Author

© 2021 IIT MADRAS - All rights reserved

Powered by RAGE