Yuri-Shendryk

Yuri Shendryk  

 

 

[Image of a split circle appears with photos in each half of the circle flashing through of various CSIRO activities and the circle then morphs into the CSIRO logo]

 

[Image changes to show a digital image of the Great Barrier Reef, and Yuri Shendryk can be seen inset on the top right talking, and text appears: Helping sugarcane farmers protect the Great Barrier Reef, Yuri Shendryk, March 2021]

 

Yuri Shendryk: So, my name is Yuri Shendryk and until recently I was a post-doctoral fellow at CSIRO. So today, I would like to present the summary of my work as part of the Digiscape Future Science Platform, where we aimed at helping sugarcane farmers to protect the Great Barrier Reef.

 

[Image changes to show a new slide showing four photos of the Great Barrier Reef, an Australian map, and a close view map of the Great Barrier Reef area, and Yuri can be seen inset talking in the top right, and text appears: Threats to the Reef, Rising sea temperatures and agricultural runoff threaten the Great Barrier Reef]

 

So, most of you are probably aware that the Great Barrier Reef is currently in a very poor condition. This is mainly due to rising sea temperatures and agricultural runoff causing the Reef to bleach and die. While rising sea temperatures linked to climate change is probably the main reason the health of the Great Barrier Reef is deteriorating, in Digiscape we focused on resolving agricultural problems and that’s why we aimed at addressing the second major threat to the Reef, which is reducing agricultural pollutants that come from the sugarcane fields. So on this slide you can see that sugarcane farms in Australia are stretched all along the Great Barrier Reef coast and the problem is that farmers usually apply high rates of fertiliser to their crop, which leads to excess fertiliser run-off and poor water quality on the Reef.

 

[Image changes to show a new slide showing a diagram showing drone, Landsat, and weather information feeding into the Cloud (Senaps), and then a person can be seen next to a Smartphone, and Yuri can be seen inset talking in the top right, and text appears: Hypothesis, Provision of timely information on crop yields will help sugarcane farmers make better decisions about their fertiliser use]

 

So to address this problem, we decided to test a hypothesis that provision of timely information on sugarcane yields will help farmers make better decisions about their fertiliser use. So we live in the era of digitisation with easy access to big data and the ability to analyse it in real time. So we decided that we can use large volumes of remote and proximal sensing, so for example, satellite and drone imagery, CSIRO’s computational platforms and data analytics to, first of all, build the predictive model of sugarcane yield, and secondly, deliver information on how management practices affect yield to the farmers through an application or a web portal. So at this stage, you might ask, what your prediction has to do with water quality in the Great Barrier Reef, and my answer to that is by knowing yield predictions early in the growing season, farmers can not only do their harvest and irrigation planning, but also their fertilisation planning.

 

[Image changes to show a new slide showing a series of photos of sugar cane fields, a LiDAR sensor, a drone, cut sugar cane, a Landsat satellite, a sugar cane harvester, and a sensor, and Yuri can be seen inset talking in the top right, and text appears: Fine-scale yield prediction, Large-scale yield prediction]

 

Now, to develop models for predicting sugarcane yields, we relied on remote sensing technology. So we decided to develop for yield from the different scales allowing farmers to inform their management decisions at within-field and their whole-of-farm scales. So for the fine-scale yield prediction we tested a combination of drone-based multispectral imaging, and LiDAR scanning technologies, which actually hasn't been done before. While for large-scale yield prediction, we tested different satellite sensors and that included multispectral and radar that have revisit rates of up to one day at resolution as low as 3m.

 

[Image changes to show a new slide showing a digital image of a sugar cane crop on the right, a drone, a LIDAR scanner, and a farmer looking at cut sugar cane on the left, and Yuri can be seen inset talking in the top right, and text appears: Fine-scale yield prediction]

 

Right, so for the finer scale, for the fine-scale yield prediction we used drone-based multispectral LiDAR technologies in combination with extensive field measurements of sugarcane yield. So, in these photos, you can actually see the drone system equipped with LiDAR and multispectral sensor that we used for this work. So the interesting part is that the LiDAR system that we used has been developed at CSIRO and now is available commercially through a spin-out company called Emesent. On this slide you can also see the actual process of yield measurements in the field which involved the use of differential GPS, and cutting and weighing of sugarcane stocks.

 

[Image changes to show a black screen and then the previous slide returns]

 

So our drone system allowed this performance from [03:27] and generate predictive models of yield at different stages of sugarcane development and basically the earliest stage at which reliable fine-scale prediction is possible was as soon as at ten weeks from planting. So, essentially the earliest stage at which you can predict fine-scale sugarcane yield was at ten weeks.

 

[Image changes to show a new slide showing a Field level graph, a Mill level graph, and an aerial satellite map with insets of fields along the Queensland coast, and Yuri can be seen inset talking in the top right, and text appears: Large-scale yield prediction]

 

Now for large-scale yield prediction, we used yield information from the harvesters provided by the farmers in combination with remote-sensing and climatic products. So this included multispectral and radar satellite imagery, elevation, rainfall, and temperature information. Using machine-learning we achieved moderate accuracies of yield prediction at the field level, as you can see on the left-hand side scatter plot. But I think the more important finding was first of all that we could generate such predictions as early as four months before the harvest season started. And secondly, that accuracies were much higher once we started looking at regional yields by aggregating yield from all farms within a mill region so you can see this fact on the right-hand side scatter plot.

 

[Image changes to show a new slide showing a photo of colleagues around a table looking at a presentation, a group of colleagues posing for a photo, the web portal digital image of yield prediction, and Yuri can be seen inset talking in the top right, and text appears: App development]

 

So once we developed predictive models of sugarcane yield at multiples scales, we had to find a way how to deliver this information to the farmers, so they can inform their management practices. And the obvious choice at the time was to develop a web portal. So, unfortunately there was limited funding within Digiscape to address software engineering needs of our project, so we had to improvise and we ended up employing 12 students from the University of Queensland to build the prototype of our application using Google Earth’s engine. Here, you can see our students that actually made it happen and a snapshot of our web portal for yield prediction.

 

[Image changes to show a new slide showing two males seated at a table looking at a computer and notebooks with speech bubbles surrounding them, and Yuri can be seen inset talking in the top right, and text heading and text appears in the speech bubbles: User experience, I trust cane yield predictions, but not sugar yield predictions, How to export yield predictions and import them into my fertilizer application equipment?, Elevation information is very useful, Yield predictions make a lot of sense, even small patterns are picked up very well, Yield predictions are very useful… need to evaluate them over multiple growing seasons before using them to inform my farming practices, I don’t mind other farmers knowing my fertilization rates and yields, I’d use yield predictions to track my nitrogen trials, Mill would love these yield predictions]

 

Finally, equipped with our yield-prediction models and a prototype of our web portal to visualise our predictions, we approached multiple sugarcane farmers to get feedback on our product. So generally, feedback was very positive, with all the farmers indicating that our yield prediction made a lot of sense to them, however they also indicated that they would need to evaluate those predictions over multiple growing seasons before using them to inform their farming practices. So that basically means that the adoption of yield predictions as a tool to decide on how much to fertilise a sugar cane field might take multiple years to implement.

 

[Image changes to show a new slide showing a photo of a tropical fish, and the Great Barrier Reef Foundation and Dendra Systems logos, and Yuri can be seen inset in the top right talking, and text appears: Conclusions, We haven’t saved the Great Barrier Reef yet, Sugarcane yield can be accurately predicted, Continuation of the project, App development, Extensive user testing, Landed a job at Dendra Systems]

 

So this brings us to the conclusions, and first of all I’m going to state the obvious saying that we haven’t saved the Great Barrier Reef yet. But I think we’re on the right track when it comes to at least resolving water quality issues in the Great Barrier Reef. Another point is that we demonstrated that sugarcane yield can be accurately predicted at both fine and large scales using remote-sensing and most importantly yield could be predicted long before the harvest.

 

There’s also a need to develop this project further, specifically a proper platform for distributing yield prediction in near real-time is needed. Likely just before finishing my post-doctoral fellowship, we submitted a proposal to the Great Barrier Reef Foundation to do this, and as far as I know our proposal has been shortlisted at this stage. So fingers crossed for the longevity of this project. Finally while I’m unable to lead this project at CSIRO any more, this project probably helped me to land an exciting job at Dendra Systems, where I currently lead the development of carbon prediction models for reforestation projects using remote sensing.

 

[Image changes to show a new slide showing inset two journal papers from the International Journal of Applied Earth Observation and Geoinformation, and one inset paper from the Field Crops Research, and Yuri can be seen inset talking in the top right and text appears: Papers]

 

And yeah if you missed any of the information from today’s presentation or if you would like to learn more about my research in sugarcane yield prediction you are welcome to read these papers that we managed to publish on this topic.

 

[Image changes to show a new slide showing a digital image of the Great Barrier Reef, and text appears: Helping sugarcane farmers protect the Great Barrier Reef, Yuri Shendryk, March 2021, yuri.shendryk@gmail.com]

 

And finally, thank you for listening and yeah I’m happy to take any questions.

 

[Image changes to show a white screen and the CSIRO logo and text appears: CSIRO, Australia’s National Science Agency]