Date: 13 March 2024, SHARE Office, Singapore

Interviewer: SHARE, The Singapore-HUJ Alliance for Research and Enterprise.

Interviewee: Prof Matan Gavish, Lead PIs of ISURF at SHARE, Associate Professor at the School of Computer Science and Engineering in the Hebrew University of Jerusalem, Rehovot, Israel.

SHARE had the pleasure of interviewing Professor Matan Gavish, Lead PIs of ISURF at SHARE, Professor at the School of Computer Science and Engineering in the Hebrew University of Jerusalem, Rehovot, Israel. In the interview, Prof. Matan discussed the unique iSURF project initiated with Professor Ng KW from NTU, emphasising advanced sensor integration for real-time transpiration measurement. Prof. Gavish highlighted the importance of high-quality data collection and the use of advanced data science methods like neural networks and AI. He also briefly mentioned the benefits of the cultural gap between Israel and Singapore.

‍

Could you please provide some insight into your background and research interests?

I am a mathematician and data scientist. I completed my PhD in Statistics  and am now on the faculty of School of Computer Science and Engineering at the Hebrew University of Jerusalem. In recent years, I have developed a strong interest in applied data science, particularly in the utilisation of data science and artificial intelligence in agriculture.

Can you give us a brief overview of the objectives of iSURF projects, including the challenges encountered by the team and how they were addressed?

iSURF is a unique open-air tropical plant phenotyping facility. It enables the tracking of plants throughout their growth cycle and provides comprehensive insights into each plant's individual characteristics. We do this by using many different sensors. Some of the sensors are environmental sensors that just hang around. Some of the sensors are inside or under the pot of the plant. And some of the sensors move above the plants and measure the plant’s characteristics without touching it. 

The latter go by “plant remote sensing”. A camera array that is moved by a robotic arm over the plant. The important thing here is that all of that data is gathered at the same time. So, we have a very broad picture that's not invasive. We don't actually cut the plant. I think ISURF is comparable in broad terms to a telescope or a microscope, meaning, the entire facility is a new kind of measurement device. The Hubble Telescope, for example, was used to study a very large variety of scientific phenomena. Similarly, there is a large variety of scientific questions, fundamental scientific questions for plant science and the use of AI in plant science that can be answered using the ISURF facility. All of the data goes to a data centre where it is analysed. And the ISURF team accordingly has a diverse skill set. We have plant physiologists, plant remote sensing experts, data scientists and material scientists. Bringing all of these to the table at the same time is quite unique. 

I'll also mention that the plants do not grow in a clean room. They do not grow in what is known as a growth chamber. They actually grow in an open-air greenhouse. So, there is a roof, but very little else, which means that everything is exposed to changing humidity, temperature, and lighting conditions that are natural. And in doing this, we placed ourselves in a challenging environment. The whole thing would be so much easier to do if we did it in a clean room because we didn't have to care about weather and tropical humidity. We could control the atmosphere and everything else. However, our result would not be directly applicable to the farmer in South East Asia. So, we actually went to great lengths to do this in tropical conditions. And therefore, we believe that the data we are collecting is quite unique. The intensity of the data collection, the variety of the sensors applied, everything in a tropical greenhouse gives us a way to know, with precision, things that farmers in Southeast Asia would very much like to know.

Could you share some details about the collaborative efforts involved in the project?

CREATE projects are, by nature, collaborative between an academic institution in Singapore and a partner academic institution outside of Singapore. I'm at Hebrew University of Jerusalem, whose liaison to CREATE is SHARE  - Singapore & HUJ Alliance for Research and Enterprise. This project came to be in the form of a collaboration between Hebrew University and NTU. 

I am from the Computer Science department, and we have two PIs  from the Faculty of Agriculture at Hebrew University, which is the largest such faculty in Israel. My other co-lead PI is Prof. Ng Kee Woei from the NTU- School of Materials Science and Engineering, along with Prof. He Jie from National Institute of Education, who is a plant physiology expert. We are collaborating with other faculty from NTU and HUJ. While Singapore and Israel are similar in many ways, they also differ significantly in others. I see the brilliance of the CREATE programs, as it uniquely capitalises on this cultural gap, allowing each of us to bring its strong suit into the collaboration. 

Also, interestingly, this is also creating a collaboration inside Hebrew University as well. As I mentioned, this is a major collaboration between Computer Science in Hebrew University and the Faculty of Agriculture in Hebrew University.

What kind of  advanced technologies were implemented during the project? Could you elaborate on those? 

Our approach is very practical. Personally, I found that there is an invisible cognitive bias that took me away from the things that will work towards the things that look very shiny. And in myself, I had to undo this bias and think very critically about what will work. And so, this is part of the core culture of ISURF: we constantly aspire to connect the dots between fundamental scientific questions in plant science and computer science, and what will most significantly “move the needle” for the farmer.

The comprehensive integration of various sensors in ISURF ensures a high level of data quality that will be valuable for years to come. The data being collected and analysed presently will continue to be analysed by ourselves and others in the future, providing enduring insights and benefits.

The other aspect of technology is data science: statistics, artificial intelligence, machine learning is advancing by leaps and bounds. We utilise advanced data science methods like neural networks for vision processing and traditional statistics, combined with innovative statistical techniques, machine vision, and AI technology, to push the boundaries of plant science research

End of interview.

‍