This community is in archive. Visit community.xprize.org for the current XPRIZE Community.
Digital transformation in agriculture
Caroline
Posts: 47 XPRIZE
A recent Forbes article highlighted three of the most promising AgTech developments and some limitation:
1) Sensors and Data
Farms and farmers will be connected and more capable of sharing data and increasing efficiencies due to AI-driven sensors. Such sensors will be distributed in different parts of the farm and will monitor the quality of plants, soil, animal health, and weather.
2) Research and Development
Applying AI algorithms to more quickly determine which new hybrid plants would have the best chances of growth under real-life planting conditions will save product development time and cost.
3) Image Recognition
Training AI to recognize thousands of plant and animal species to improve drone abilities of detecting pest disease and crop damage is the next big thing.
Limitations
1) Sensors and Data
Farms and farmers will be connected and more capable of sharing data and increasing efficiencies due to AI-driven sensors. Such sensors will be distributed in different parts of the farm and will monitor the quality of plants, soil, animal health, and weather.
2) Research and Development
Applying AI algorithms to more quickly determine which new hybrid plants would have the best chances of growth under real-life planting conditions will save product development time and cost.
3) Image Recognition
Training AI to recognize thousands of plant and animal species to improve drone abilities of detecting pest disease and crop damage is the next big thing.
Limitations
- Data Availability: in farming data is produced and collected over long periods of time.
- Talent Availability: tech-savvy industries are competing over a limited number of young AI talent.
2
Comments
In the case of the Canadian Prairies where zero-till agriculture is widely adopted, new technologies need to be adopted to work well zero-till. In the case of the world, zero-till is a practice which should be adopted widely.
on the first point, are there specific emerging platforms that have the potential to help farmers make sense of all of this? it sounds like they have tons of data streams but no consolidated management system. is the need to have a centralized place for all the data, an integrated system that can make sense of it all and adjust different production factors in real time, or both?
Some areas will use tillage because there is a perception that tillage helps manage moisture surpluses. If a grower doesn't have access to herbicide tolerant crops or access to herbicide in general, they will need some form of tillage to manage weeds. Row crops like corn and cotton 'need' some sort of tillage, though soybeans are up for debate. Also, in areas with lots of cheap labour, the time spent tilling might not be as costly.
I was able to find some older stats (2008/2009) about zero-till adoption worldwide by continent: South America had 46.8%, North America was at 37.8%, Australia/NZ 11.5%, Asia 2.3%, Europe 1.1%, Africa 0.3%.
I was at a soil conservation conference a few years ago and one presenter mentioned how they were gaining traction with zero till adoption in an area of conflict because it meant less time on the tractor, where there was risk of being shot at. There was some discussion of little handheld one-row direct seeders as well, so the philosophy is adoptable to smallholders.
In my opinion, the #1 technology that all farmers could benefit from is accurate long term weather forecasting.
The AI push to classify plant and animal species, iNAT, appears to use images that are from a typical, oblique viewing perspective that facilitates photo interpretation. However, most drone imaging of crops have imaging sensors aimed perpendicularly (i.e. nadir) to the ground surface and these images are much more difficult for even humans to interpret.
In addition, the scale of any sensor measurement will impact AI accuracy. Multispectral cameras used for drone image analysis of crops have very low spatial resolution so the nadir images often don't have much spatial detail. Even high spatial resolution imagers need to be flown very low, often so low that flights and data processing take substantially longer (which make it harder to achieve a return on investment), in order to come close to the level of spatial detail on the iNAT images.
Finally, timing is a big issue. Many crops, especially some dicots, have very spectrally and morphologically distinct development stages. For instance, lavender or canola look very different when comparing pre-bloom versus during bloom. In addition, several economically important plants have different shape characteristics before and after stem elongation. This isn't as simple as identifying a cat, which mostly just grows after it is born. Plants undergo much more anatomical development throughout their lifecycles, which makes their appearance a moving target and therefore more difficult to allocate image regions to classification groups.
notice that i didn't use the term AI, and not just because of your (correct) admonition that it's being overused as a buzzword (i suppose we can thank branding/marketing types who are searching around for a contemporary term to explain somewhat related concepts). there are a lot of different technologies that interplay in the examples we've discussed - drones, AI-driven video processing and analysis, sensing data, data on plant lifecycles and morphology, etc.
i'm curious, where do you see these trends converging? how far are we in your estimation from a point at which drone analysis can be done from a distance with low resolution video but with high accuracy?
@oliversanchez I'm wondering if you have any insights here? btw, thanks for joining the community! as you can see we've got a lively discussion going on very much in line with your current work.
I think that manned aircraft are often more appropriate than drones for some ag remote sensing jobs, especially covering large contiguous tracts of land.
I would need more context to address the accuracy part of your question.
Click here to read the whole story.
India’s agriculture budget doubled over five years to Rs 57,600 crore in 2018-’19. As a proportion of the total budget, however, the allocations for agriculture were largely static, at an average of 2% over the past four years, rising from an average of 1.3%.
How India compares with the world
In 2017, 18% of Chinese were employed in agriculture compared to 43% of Indians, according to World Bank data. Over 26 years till 2017, the proportion fell 37 percentage points in China, while in India it fell 21 percentage points, keeping India above the world average of 26.4% of people employed in agriculture.
Although agricultural income is not taxed in India, farmers have been “‘implicitly taxed’ through restrictive marketing and trade policies that have an in-built consumer bias of controlling agri-prices.
Technologizing Agriculture
Agriculture is ripe for the adoption of new technologies to help monitor and manage assets on a granular level, and everything from the Internet of Things (IoT) sensors, robots, and drones are being used by farms around the globe.
Big data also holds enormous promise for urban farmers — people who are turning rooftops and abandoned lots into small farms.
Robotic use of multiple cameras to collect color images and three-dimensional (3D) depth information on growing plants, robotic harvesting on indoor farms etc. are progressive trends toward incorporating precision agriculture techniques to improve yields and become more sustainable agriculture.
Executable Needs:
Rented Agricultural Machines equipped with computer vision and machine learning capabilities to eliminate use of herbicide, pesticides volumes.
Subsidies on sensor based machinery that check incremental changes in the soil's composition and to test soil moisture levels to identify flooding, overwatering, or ground freezing.
Awareness of IoT-enabled water and fertilizer delivery valves that can remotely be monitored and managed.
Drone technology, using unmanned aerial vehicles (UAVs) equipped with a package of high-definition cameras, IR sensors, and image-recognition capabilities to monitor crops, which can provide significant increases in efficiency
What do you see as the most important technological developments in the farming and food sector? Which technologies show the most promise? What are the limitations -- for farmers, distributors and consumers?
In my view, one of the more promising approaches, lesser-known in the tech communities, is permaculture. While most people who support permaculture take a decidedly low-tech approach (based on what I can only characterize as a misguidedly anti-scientific worldview), it could be effectively married to high-tech innovations with the promise of multiple levels of extraordinary synergy.
As I define it, the essence of permaculture is treating food production as a system of plants, animals and microbial life that are curated by intelligence to support each other. In this manner, certain plants protect other plants by repelling invaders, while all of the plants can actually exchange water and nutrients underground, via a system of mycelial filaments, thereby balancing supplies. (Think of the movie Avatar, but for real! I've read that 95% of Earth's flora evolved to live synergistically with underground mycelium networks. These networks are ignored in modern agriculture, especially monoculture. Yet one can run home experiments and easily see their value.)
Such systems can be largely self-sustaining in terms of fertilizer, water production, and other necessities of plant life. While they have tended to be labor-intensive, in my view the introduction of certain forms of advanced technology could make such systems largely labor-free. Some permaculturists have demonstrated, on a pilot basis, the potential of creating new micro-climates. (This has enabled, for example, growing fruit trees at elevations previously considered impossible.)
Further, permaculture could be used to take essentially useless desert land--itself the result of human mismanagement in most cases--and turn it lush once again. This further opens up the possibility of new villages and even cities and city-states arising on uninhabited territory, with those human communities having a carbon-neutral or even negative footprint, and an advanced, high-tech, eco-friendly lifestyle.
I've long thought that such could be the basis for a promising startup company.
i think your comments about the lack of integration of technology into this approach are spot on - bridging that gap could unlock a lot of value. i wonder what the barriers are? i see the cultural aspect of permaculture practitioners and their skepticism of technology as a clear obstacle, but there seem to be enough cutting edge yet nature-focused new farming pioneers that this should have been tackled by now, at least in my view. perhaps it is the systemic nature (technology is good at solving very specific problems but platforms/integrations exponentially increase complexity), or that individual technologies are still lacking? for example, how has technology touched on the issues of fungi and mycellium and their benefits that you raised, if at all?
If the present leadership are too averse to technology to see matters our way, I know some young permaculturists who have studied with them and who DO see matters our way. (In fact, one of them helped me with the permaculture section of my book.
So far as I can tell, very little has been done to bring together technology with fungi and mycellium. This despite Paul Stamets' informative talk a few years ago about how fungi can solve six of the world's most pressing problems!
To do this properly, it would be necessary to bring together a very wide range of disciplines. To my knowledge, no one has attempted this in a well-funded startup with excellent management and advisors. I know quite a few leaders in the relevant disciplines, but more would be needed.
I don't think that any of the necessary component technologies are lacking. I think it's a matter of integration. But you're spot on about the complexity of such integration, which IMHO is the major challenge in building a scalable startup here.
I'll have to take a look at the other thread. Thanks for mentioning it.