Rushikesh Chavan is the Director of The Habitats Belief (THT), a nonprofit organisation working to guard India’s pure habitats and indigenous species of wildlife, with a concentrate on the conservation of lesser-known species and habitats.
Working throughout India, THT works with over 120 companions protecting an space of over six lakh hectares. It has developed cutting-edge applied sciences to resolve complicated conservation challenges.
A post-graduate in atmosphere and biodiversity from the College of Mumbai, Rushikesh has beforehand labored with the Wildlife Conservation Belief and the Bombay Pure Historical past Society.
With over 25 years of expertise throughout on-ground implementation and coverage design, he has additionally pioneered India’s first coverage assume tank on Conservation Behaviour Science, bringing collectively conservation biology, psychology, and information analytics to tell decision-making.
Rushikesh spoke to indianexpress.com on the challenges of utilizing tech for conservation, the influence of their grants, the brand new tech they’re engaged on, and the way AI will influence the conservation tech house. Edited excerpts:
Venkatesh Kannaiah: Are you able to give us an outline of the work of The Habitats Belief?
Rushikesh Chavan: Based in 2018 by Roshni Nadar Malhotra of HCL Applied sciences, THT is targeted on challenges arising from biodiversity loss, the local weather disaster and world water shortage.
Story continues beneath this advert
We work in two methods. We offer grants to companions starting from Rs 3 lakh to Rs 1 crore to help their efforts. Past grants, we additionally establish particular themes the place we fund companions and are a part of the execution course of.
Once we discover that no organisation is engaged on an essential problem in a geography, we execute the challenge ourselves.
We additionally work with numerous authorities organisations. One amongst them is MITRA, the coverage assume tank in Maharashtra that works on the strains of NITI Aayog, integrating ecological priorities into their regional planning efforts, such because the Vidarbha plan.
Venkatesh Kannaiah: Inform us about a few of your grantees/companions and the way they’re utilizing tech to sort out problems with conservation and biodiversity?
Story continues beneath this advert
Rushikesh Chavan: Hoolock Gibbons are elusive primates, and the one ape present in rainforest habitats of Northeast India. Given the issue in direct sightings, scientists are utilizing its vocalisation habits to estimate its inhabitants.
THT, in collaboration with Conservation Initiatives, is utilizing acoustic recorders to detect these gibbons. We’re additionally creating AI and machine-learning-based fashions to course of acoustic information collected by these recorders and establish gibbon calls and their numbers in a given locality.
The Indian wolf is a species below menace, and we’re working with The Grasslands Belief to watch wolf packs and their behaviour utilizing satellite tv for pc radio collars, thermal drones and digicam trapping. Whereas radio collars assist observe the each day motion of wolves, thermal cameras assist in detecting their behaviour at evening.
Monitoring land use and land cowl utilizing satellite tv for pc imagery is crucial for conserving wildlife and biodiversity. Open Pure Ecosystems (ONEs) are one of many largest varieties of habitat kinds in India that embrace grasslands, scrublands and open savannah. These habitats have been incorrectly categorised as wastelands in authorities atlases. We’re working with pioneering researchers like MD Madhusudan and Pradeep Koulgi to precisely map these habitats and thereby prohibit their overexploitation.
Story continues beneath this advert
Venkatesh Kannaiah: Inform us about your Tech for Conservation programme and its highlights.
Rushikesh Chavan: At The Habitats Belief, we see tech as a device and never as an answer in itself. In India, expertise solely works if the encircling ecosystem helps it. So in our Tech for Conservation programme, we concentrate on three themes.
First, as soon as we establish a crucial conservation hole, we develop tech the place crucial. For instance, our work in bioacoustics has been very efficient. We’ve additionally developed AI-based digicam entice fashions to assist kind and course of massive volumes of wildlife information. We’ve constructed our personal ROVs (Remotely Operated Autos) to check marine and ocean ecosystems, sending them right down to depths of almost 200 metres. This sort of work is uncommon in India.
Second, we assess whether or not the ecosystem exists for that expertise to be efficient. We associate with researchers, conservationists, and subject organisations to deploy the tech.
Story continues beneath this advert
Conservation tech is pricey. Since we’re a philanthropic organisation, we put money into constructing the tech after which make it open entry wherever doable.
Venkatesh Kannaiah: How are you utilizing AI and machine studying in your conservation efforts?
Rushikesh Chavan: First, at a techniques degree. For instance, in Maharashtra, we’re working with NITI Aayog, utilizing AI, GIS, and large-scale information modelling to design development pathways that combine ecology and the economic system. When you don’t keep the ecological integrity of a panorama, there isn’t any level focusing solely on a single species.
Second, we use AI and machine studying primarily for producing proof. We use proof actively to tell choices, form coverage, and design interventions.
Story continues beneath this advert
For instance, all our digicam entice information processing makes use of AI/ML fashions. Our bioacoustic work depends closely on machine studying to separate species calls from ambient noise. Video and sonar evaluation from marine explorations additionally makes use of ML fashions. Right this moment, in case you are not utilizing AI and ML, you might be dropping out.
We’ve additionally explored massive language fashions (LLMs). However that house is evolving very quickly with numerous fashions, so we’re continually rethinking our technique there.
So broadly, AI, ML, and LLMs are utilized by us for large-scale information evaluation, for sample recognition in ecological datasets, techniques modelling, situation constructing and decision-support frameworks.
The place we’ve not absolutely cracked the issue but is implementation. As an example, think about a drone with infrared flying over a forest. AI can detect hotspots or anomalies, say, early-stage forest fires. That offers us info. However how do you change that into one thing a forest officer can instantly act upon, cost-effectively? That’s nonetheless a spot.
Story continues beneath this advert
In India, price is a severe constraint. Superior AI techniques could be costly to deploy at scale. AI in India nonetheless has to cross that cost-effectiveness barrier for large-scale, on-ground operational deployment.
AI is extraordinarily highly effective for information technology, proof constructing, and technique design. However relating to direct intervention, changing or automating ground-level work, we’re not absolutely there but.
Venkatesh Kannaiah: What instruments have you ever constructed for serving to conservation efforts?
Rushikesh Chavan: We’ve developed an underwater ROV to discover deeper marine ecosystems. It permits us to seize video, imagery, and side-scan sonar information at depths of as much as 200 metres; areas which are in any other case very tough to check in India.
Story continues beneath this advert
The second innovation is a multi-array microphone system. Usually, when you place a microphone in a forest, it captures sound from a single route. We developed a multi-directional microphone array that may not solely file sounds but in addition triangulate the place precisely the sound is coming from. That spatial accuracy is crucial for ecological research.
Now, coming to software program options. One main device we’ve developed is a digicam entice image-sorting platform. India generates thousands and thousands of wildlife pictures via digicam traps. We constructed a base AI mannequin that enables customers to add their digicam entice information and robotically kind pictures.
On high of that, there are plugins for species classifiers that may be added. For instance, if a forest division needs to isolate tiger pictures, a plugin can extract all tiger pictures from the dataset. We’ve supplied this device without spending a dime to locations like Kaziranga Tiger Reserve, particularly throughout massive workouts just like the All India Tiger Estimation, the place they generate lakhs of pictures with restricted capability to course of them.
The subsequent layer of software program pertains to analysing video and sonar information, particularly from the rover. Digital camera traps produce nonetheless pictures, that are simpler to course of. However underwater video and side-scan sonar information are much more complicated. So we’re growing fashions that may analyse video streams scientifically, not simply create a species checklist, however apply statistical fashions to generate sturdy ecological insights.
Then there’s our bioacoustics work. Utilizing the multi-array microphones, we seize massive volumes of forest sound information. We talked earlier concerning the Hoolock Gibbon challenge. The subsequent stage is figuring out people via vocal signatures. Simply as you and I sound completely different, animals have distinctive vocal patterns. There’s a idea in acoustic science that will enable differentiation between people. If that works for gibbons, and when mixed with directional information from the microphone arrays, we are able to apply statistical fashions to reliably estimate the variety of animals or birds that exist in a given panorama. That may be a breakthrough.
We’ve developed round 5 to 6 main expertise options, a few of that are within the public area and actively getting used, and others are at numerous levels of improvement. As and when the fashions are validated, we launch them brazenly.
Venkatesh Kannaiah: Inform us about 5 improvements or applied sciences which are altering the face of conservation.
Rushikesh Chavan: There’s the Acoustic Microphone Array tech for monitoring species which are extra heard than seen, which we spoke about.
There’s the ROV tech. People can safely dive and examine species solely at depths of round 30-50 metres, however these underwater robots, unoccupied and piloted from the floor, develop into essential for deepwater duties, scientific analysis, and surveillance.
There’s the metabarcoding and genetic evaluation tech. DNA evaluation solely allowed us to know if a species was current at a spot, provided that tissue, hair or excreta is obtainable. Now, superior methods of DNA evaluation can inform us if a species is experiencing a inhabitants decline, affected by illnesses or susceptible to vary.
There’s thermal imaging to trace and monitor elusive nocturnal species. Nocturnal behaviour monitoring has been a problem in conditions of human-wildlife battle. Now, thermal-imaging-equipped drones are used for surveillance of websites the place a battle has occurred and have usually helped managers to securely find and rescue animals and alert individuals.
There’s auto-classification and GSM-enabled digicam entice tech. Digital camera traps are deployed to file wildlife, and now developments in fashions are serving to us establish species shortly and ship that info on to our telephones and techniques. It helps safety businesses detect poachers.
Venkatesh Kannaiah: Inform us about world out-of-the-box startups working in conservation tech and the issues they’re fixing.
Rushikesh Chavan: There’s Conservation X Labs, which makes use of AI, superior sensors and open innovation to assist conservationists reply to threats comparable to poaching, invasive species and habitat loss.
There’s Nightjar Tech, which develops AI-powered wildlife monitoring {hardware} and an clever camera-alert system that detects animals, individuals and threats in actual time.
Sara Beery Lab works on laptop imaginative and prescient and machine studying for ecological monitoring, creating AI strategies to automate wildlife picture and sensor information evaluation for conservation science.RoboticsCats is a Hong Kong-based machine imaginative and prescient and AI startup that builds environmental monitoring and early wildfire detection techniques utilizing laptop imaginative and prescient to detect smoke, hearth and biodiversity cues from surveillance cameras.
There’s additionally EarthRanger, a data-driven software program platform broadly utilized by conservation groups to combine subject information, sensor feeds and AI insights into real-time decision-making instruments that enhance wildlife monitoring, battle mitigation and habitat safety globally.
Venkatesh Kannaiah: What do you assume are one or two low-hanging fruit, the place AI can intervene on this tech conservation house?
Rushikesh Chavan: One low-hanging fruit for AI and ML is air pollution monitoring and remediation.
Utilizing satellite tv for pc imagery and distant sensing, AI fashions can estimate numerous parameters for pH or contamination ranges. After getting that intelligence layer, the subsequent step is automation.
For instance, invasive species like water hyacinth choke many Indian water our bodies. Eradicating them manually at scale is extraordinarily tough. AI-driven monitoring techniques mixed with small, automated, solar-powered elimination items may establish hotspots and deploy focused mechanical motion. That may be a house the place AI plus automation could make a distinction.
If we take into consideration one thing like cleansing the Ganga, a purely human-driven effort at that scale is sort of unimaginable. However AI-assisted monitoring and mechanised interventions may considerably enhance effectivity.
A second alternative is decision-support techniques for managers. AI-powered dashboards can combine digicam traps, satellite tv for pc imagery, hearth alerts, and floor stories right into a single interface.
A 3rd promising house is ecological restoration. India has immense biodiversity, and restoration requires native ecological information — what species to plant, by which soil, below what rainfall regime. That experience at present sits with a handful of specialists. AI fashions can combine biodiversity information, soil profiles, rainfall information, and historic vegetation patterns to suggest species mixes.
Venkatesh Kannaiah: How is the tech ecosystem round conservation and biodiversity in India?
Rushikesh Chavan: One main problem is price. Conservation tech is pricey, and in contrast to sectors like finance or e-commerce, conservation doesn’t have deep-pocketed prospects.
The second problem is scale and sustainability for startups. Many startups construct glorious proof-of-concept instruments however battle to make the enterprise mannequin viable.
One other is dependency on massive expertise platforms. Conservation tech builders are sometimes adapting present instruments — satellite tv for pc imagery platforms, cloud infrastructure, mapping instruments — relatively than constructing all the things from scratch. If an organization modifications its pricing mannequin, API entry guidelines, or licensing phrases, that disruption immediately impacts conservation functions.
Venkatesh Kannaiah: Inform us concerning the two huge tech challenges you need to remedy.
Rushikesh Chavan: The problem is easy methods to design expertise that’s usable in real-world Indian situations. Instruments which are light-weight, low-bandwidth, device-agnostic, and straightforward to deploy.
The second problem is native adaptation. What works in Chennai might not work in Bengaluru or Delhi. Ecological techniques differ. Soundscapes differ. Landscapes differ. Even governance techniques differ. A restoration mannequin constructed for Himalayan ecosystems gained’t robotically apply to dry grasslands.
In any other case, you find yourself constructing a brand new expertise for each single use case, which is inefficient. Growth takes time, effort, and cash. And by the point you end constructing one thing, the broader expertise panorama might have already shifted, making elements of your system out of date.