75. One model to rule them all?

Technology trends for Food/Agtech leaders

Programming Notes: I interviewed Sarah K Mock about her writing process for the Foster podcast two weeks back. I will conduct a similar event with Janette Barnard (VC and creator of Prime Future newsletter) on October 8. The links to events are at the bottom.

This week’s edition includes the following topics,

  • Analysis: Tech stack for carbon and soil health, and the role of models (not the runway kind!)

  • Technology trends: Continued cyberattacks on food/agriculture infrastructure, performance guarantees (or warranties?) for crop plans, and supply chain improvements in Africa.

  • Research Review: An important paper which looks at the impact of cover crop on soy and corn yield, by using a combination of field data, and a comprehensive agroecosystem model (ecosys.)

  • Tidbits from the world of Ag/Foodtech: Robotics, AI, AgTech, Developing Countries, Sustainability. Will we get locally grown coffee in California?

  • Read, Listen, & Watch: My Foster podcasts, a podcast on rice and GHG emissions, and a video by Neil Halloran about climate change. And dogs and longevity!

Analysis

Tech Stack for soil health - one model to rule them all?

The (ir)rational exuberance for carbon markets has continued in 2021. Seana Day of Culterra Capital published a soil health tech stack, to explain different capabilities and how they relate with each other. It breaks down the tech stack into measurement, core calculations, and presentation.

Image credit and source: Culterra Capital’s Soil Health Tech Stack

Seana Day captures some key points about soil health, (highlights by me)

The holy grail in soil health will be integrating enough ground-truthed soil measurements to feed robust data models that are powered by machine learning and artificial intelligence (ML/AI) to interpolate the soil health of a particular farm field, pasture, paddock, or range land. In this way, remote sensors can be used in lieu of field-based measurement for verification.

The measurement layer is nascent today. The soil sample collection mechanism has not changed much (“follow a W pattern, collect some soil cores, send the samples to the lab, and do it every few years.”) Different labs use different methods to do component analysis. Soil test results are often not consistent due to different methods, and the wide variability of soil in a given field. The modeling of different soil parameters, even though you might have good measurements (which is not always true) is a challenge.

The added interest in carbon markets has spurred research and funding. For example, the University of Illinois was awarded $ 4.5 million from ARPA-E to develop commercial carbon credit tools. The project allows field quantification of carbon intensity for every field across the U.S. It combines field level observations, with satellite and hyperspectral data, soil sensing, and AI/ML modeling.

Earthoptics raised $ 10.3 million from Leaps by Bayer (and others) which reduces the number of manual samples. The sensor is mounted on a universal terrain vehicle, uses GPR (ground penetrating radar) and other sensors (EM conduction) with no soil contact or speed limitation. The data is used to train machine learning models, to help understand soil compaction, tillage prescriptions, and provides verification at a low cost.

We see this development in developing countries as well. India has done its own soil health card scheme for millions of farmers. In Africa, its leading lab for agriculture and environment testing, has launched an AI based soil testing and crop advisory service for small holder farmers. Currently only 0.5% or less smallholder farmers have used soil testing. This process eliminates the need for soil testing lab analysis, and correlates the soil fertility properties in the soil using AgViza's soil testing engine.

The above examples show efforts to reduce the cost of soil testing (measurement) or provide measurement capabilities at a lower cost, where none exist. But to monitor and improve soil health at scale, scalable modeling approaches will be necessary.

Seana Day writes,

Whether the market will ultimately arrive at a ‘measure-plus-model,’ ‘model-only,’ or ‘measure-only’ approach, it is too early to tell. But the potential power of ML/AI to improve model accuracy is clear. 

I don’t believe it is too early to tell. It will be a “measure-plus-model” approach. There are many parallels to draw from other fields, when we are talking about multi-dimensional issues like “soil health.” The credit worthiness check in the US is done by a FICO score model. Many lenders will use a combination of the FICO scores, with additional data from financial disclosures to make a lending and rate decision.

Within healthcare, BMI is a model of a healthy body, even though it is a calculation based on precise measurements. Nobody would be able to judge your body health by just looking at your BMI. They will combine a variety of measurements and models to assess your body health.

Measurements are expensive, time consuming, and not trivial to scale. The right models can be scaled to understand a big part of reality. In the soil health space, given the uncertainty in measurements, and the uncertainty in different models, it might be difficult to choose one model to explain some aspect of reality. Organizations might be better off to look at multiple models, and then run ensemble models to reduce the level of uncertainty in how closely models reflect reality.

Share

Technology trends

R-a-a-S (Ransomware as a Service) on thin margin businesses

The spate of ransomware attacks and hacks on the food and agriculture infrastructure has continued. This year has seen attacks on JBS, vulnerabilities exposed in John Deere and CNH systems, and this week saw an attack on New Cooperative by Blackmatter (supposedly a Russian group of hackers.) New Cooperative is one of the largest crop buyers, has 50 locations, and distributes fuel and crop chemicals. The agricultural cooperatives, and retailers are trusted advisors, and a major provider of inputs, services, and buying (in this case) for farmers in the Midwest.

As reported by Dark Reading (An online publication on cybersecurity),

BlackMatter is believed to be connected with ransomware-as-a-service (RaaS) group DarkSide, an affiliate of which targeted Colonial Pipeline in a major ransomware attack earlier this year.

The ransomware group has stolen financial, HR, and R&D data, including the “source code” for the technology platform called “SoilMap.” As more and more systems go online, the protection of the digital infrastructure is critical. More and more of the physical infrastructure is being managed by the digital infrastructure. Food and agriculture companies will have to step up their game and protect their infrastructure by using social engineering training, defense in depth design principles, and keeping their employees and infrastructure updated with the latest potential threats.

In early 2021, President Biden had presented a list of  industries to President Putin, which constituted critical infrastructure for the U.S. Food and agriculture was on the list of critical industries. An attack of any of these industries will be treated as a serious national security threat. Blackmatter has its own set of rules (!?!?).

The group's blog says it doesn't conduct attacks on organizations in industries like healthcare, critical infrastructure, oil and gas, defense, nonprofits, and government. Blackmatter does not consider New Cooperative as strategic as they claim it does not have large volumes and considers entities like New Cooperative as fair game for their attacks.

Older organizations with significant technical debt, struggle to keep up their digital infrastructure secure. Agriculture’s traditional focus has been on performance, and safety, but not security. Many of the food and agriculture organizations run a low margin business and so can struggle to allocate the right resources to cybersecurity. I had written about this prioritization challenge in edition 70 (A tractorload of vulnerabilities)

In my personal experience, resource and budget allocation for information security from cyber threats can be a challenge. These “Freedom to Operate (FTO)” budgets contend with budgets for new product development, and product improvements. Cybersecurity investments and improvements are largely invisible to customers. In case of competing priorities, it is easy to kick the cybersecurity can down the road.

I would not be surprised if these attacks become more frequent in the future.

Performance guarantees

For the last few years, we have been hearing about digital transformation in agriculture. The digital transformation is definitely happening. Different use cases, enabled by data and modeling have been prime drivers of these transformations.

Bayer has dabbled in an outcome based pricing model. Agtech companies have tried to bundle many services, with financing, by better understanding of the risk profile of a farming operation.

Growers Edge, a provider of agrifintech solutions has partnered with Farmers Cooperative Society (FCS) in Iowa to offer the first sustainability focused Crop plan. The Crop plan is warranty backed, and it helps farmers manage the uncertainty of adoption of sustainable agriculture practices for soil health, water quality, and long term economic performance.

The plan is not very different from what Bayer had tried a few years ago, with some interesting twists. The partnership with FCS leverages farmer relationships the co-op has with their members. The specified agronomic practices a farmer has to follow requires the use of products and services sold by FCS. The go-market strategy for Crop Plan is interesting as it is offered to agriculture retailers and input providers, with a four step process.

The program is designed to provide flexibility to the ag retailers. The retailer can tailor the program based on their customers needs. They can choose the benchmark (% of growers production) and the warranty details. The retailer sales team are armed with information to sell and implement the program, while Growers Edge does the data collection and analysis. Most importantly when the grower exceeds the benchmark yield, they keep all the upside, but are made a warranty payment, if they fall below the benchmark. The design of the program makes it a bit easier to sell the program to the grower.

Growers Edge is able to pass on most of the risks to the retailer and the grower, while they focus on the data, modeling, and risk management. The retailer can bundle and spread the risk across multiple product lines, and pool it across a number of growers. The retailer can push new products to market, build stronger relationships, and cross and upsell products which are not part of the Crop Plan.

Not an email from a Nigerian prince

In most of the developing markets, supply chain and logistics infrastructure is a problem. Crops can go bad, crop quality can suffer due to inadequate storage or transportation facilities, resulting in post-harvest loss. Farmers lose money as they get paid less by processors, if their product does not meet the quality expectations of the processor. 

The oil palm market in Nigeria is $ 3 billion, with over 4 million farmers. Nigerian agritech startup Releaf is investing to build hardware and software solutions to make farmers and processing facilities more efficient and profitable. The founders of the company came through Y-combinator. Their story is a strong example of their ability to reality understand a customer's pain point.

They toured 20 states and studied the value chains for different crops to spot inefficiencies. Processing oil palm nuts requires serious processing power, and is not easily available to farmers. Farmers often employ outdated methods like rock crushing. Releaf provides a nut-cracking machine (“Kraken”) deployed close to farmers.

Releaf’s business model is simple. It buys nuts from farmers, and then processes them using their nut-cracking machine called Kraken. It sells the vegetable oil to processors and local manufacturers. Releaf produces oil which is higher standard than required by the processors, and so they will look for ways to monetize the quality difference.

The simple yet effective supply chain technology is already having an impact, as they are connected to 2000 farmers, who have supplied 10 million kilo of quality palm kernel nuts to processors.

As Jehiel Oliver, CEO of Hello Tractor said in edition 66, (Jehiel Oliver: Tractor-as-a-Service), in places like Africa, investing in infrastructure will be a win-win for customers, governments, and entrepreneurs.

Share Software is Feeding The World

Research review: Cover crops and ecosys model

In edition 26 (JCPenney and FBN), I had written about the Ecotone study done for General Mills, which highlighted the principles of regenerative agriculture, with cover crops being an important component.

Image from Ecotone study

According to latest research published in Field Crops Research, (Assessing the impacts of cover crops on maize and soybean yield in the U.S. Midwestern agroecosystems) evaluated the impact of cover crops by looking at field experiment data from 2013 to 2018 and the ecosys model. The ecosys model is a process-based agroecosystem model. The research summarized,

The impacts of cover crops on the productivity of the maize-soybean [Zea mays L. - Glycine max (L.) Merr.] rotation system in the U.S. Midwest still have large uncertainties based on results obtained from field experiments, specifically across different soil properties, climate conditions, and land management practices.”

The ecosys model simulates the carbon, nitrogen, and water cycles.

  • C cycle: plant C fixation, plant, and soil respiration

  • N cycle: N uptake, legume plants N fixation, and soil mineralization

  • Water cycle: Plant water update, and soil water flow

The ecosys model performs well with the simulation of crop yield. It does so by capturing the interannual and geospatial variation at a regional level for both corn and soy, with and without cover crops.

I found this detail interesting, as can be seen from Figure 9 from the study.

When no/low N fertilizer is applied in spring before maize, legume cover crops have a positive effect on maize yield. As N fertilization rates increase, maize yield no longer increases with the existence of cover crops.

Fig. 9. Relationship between yield response ratio (maize yield following legume cover crops/maize yield following winter fallow) and N fertilization rate (for maize) by ecosys simulation.

The summary from the study was as follows: 

  • Non-legume cover crops cause a 3.9 % yield penalty for maize, no penalty for soybean.

  • Legume cover crops have no significant impacts on maize and soybean yield.

  • Pathways for yield penalty are N and O2 deficiency, water stress and cooler soil.

  • Optimizing cover crop growth windows could maximize the benefit from cover crops. (for example, larger growing winter windows can achieve larger biomass with larger soil health benefits)

  • The ecosys model is an effective tool to quantify cover crop impacts on crop yield.

Reference: Ziqi Qin, Kaiyu Guan, Wang Zhou, Bin Peng, María B. Villamil, Zhenong Jin, Jinyun Tang, Robert Grant, Lowell Gentry, Andrew J. Margenot, Germán Bollero, Ziyi Li, Assessing the impacts of cover crops on maize and soybean yield in the U.S. Midwestern agroecosystems, Field Crops Research, Volume 273, 2021, 108264, ISSN 0378-4290, https://doi.org/10.1016/j.fcr.2021.108264

Share

Tidbits

Robotics and AI

Advanced Farm Technologies, Inc, raised a $ 25 million series B round, to support growth in strawberry harvesting, and adaptation of its technology to apple harvesting. The robot works with manual harvest crews, and can work autonomously for up to 24 hours. The big question is always whether the tech is transferable to other crop types, and farming operations.

Robotics farming firm Iron Ox raises $ 53 million.

Iron Ox is one of a large number of companies looking to revolutionize farming for the 21st century, amid growing concerns over population growth, climate change, labor shortages and a slew of other mounting concerns. Solutions range from large-scale indoor farms to field robotics designed for a more plug and play approach that operates with a more traditional setup.

AgTech

Micro-climate predictions: Enabling hyper-local decisions for agriculture and renewables

How ag technology is embraced differently among different generations of farmers? Get off my lawn (farm) edition!!!

CPG & Food production

Agriculture accounts for nearly two-thirds of Nestle’s total greenhouse gas emissions, with dairy and livestock making up about half of that, according to the company.  Nestle will invest $1.3 billion in regenerative agriculture. Nestle will use its network of research and development personnel and agronomists to develop more environmentally friendly crops and production practices.

20 meat and dairy firms emit more greenhouse gas than Germany, Britain, or France, and also receive big public funds.

Why does Britain need to go on a crash diet to save farmers and supermarkets?

Developing countries

Senegal is experimenting with circular gardens - known as tolou keur to to help stave off desertification by stymying the Sahara as it wandered south. The inner rows include medicinal plants, whereas the outer rows are lined with baobabs and African mahogany. The idea is similar to a milpa from the Americas.

An aerial view of a newly built tolou keur garden in Boki Diawe, within the Great Green Wall area, in Matam region, Senegal. REUTERS/ZOHRA BENSEMRA

Only 4 out of 49 African countries reached the Comprehensive African Agriculture Development Programme’s (CAADP) recommended target of 10% public expenditure on agriculture. This speaks to the lack of investment in Africa, both from the public as well as the private sector.

The Solomon Islands are a high-forest-cover country, where there has been extensive deforestation in recent years. A 2018 report by Global Witness concluded commercial logging was contributing to forest clearance at nearly 20 times the sustainable rate. Training Solomon Islands farmers will be a form of responsible sourcing.

Sustainability

Sustainable coffee grown in Finland – the land that drinks the most coffee per capita produces its first tasty cup with cellular agriculture.

Will I be paying $ 12 for my almond milk latte, with almonds and coffee both locally grown in California?

Arabica and Robusta coffee make up 99% of the global coffee production. With current and future effects of climate change, researchers are looking for coffee species with greater tolerance for higher temperature. They have found a forgotten species called Coffea stenophylla, which is more high temperature resistant and has a superior flavor, similar to Arabica.

Are we eating ourselves to extinction? The human diet was more diverse a 1000 years back. In 1950, peat diggers discovered the intact body of a man who died 2,500 years ago. Inside the man’s stomach was a porridge made with barley, flax and the seeds of 40 different plants!!

Read, listen, and watch

🎧 Foster Podcast with Sarah Mock, hosted by me and DJ. “Writing about controversial (and niche) topics” (55 minutes). If you want to register for the live event with Janette Barnard, you can register here.

🎧 Growing up in India, with a family originally from the coastal region, rice was always a staple at our house. I didn’t know then rice production with flooded paddy fields is a big contributor to greenhouse gas emissions. A BASF podcast (“The science behind the salad” ) discusses the  paddy field system, direct seeded rice, and a way to cook perfect, plump, fluffy rice. (Is it Basmati though??) (29 minutes)

📺 “A skeptical look at Climate Change” by Neil Halloran. I enjoyed Neil’s analysis and his approach to presenting complex ideas in an approachable fashion. (24 minutes)

📚 I rarely post non-AgTech items. I am going to make an exception this time. I am a huge fan of dogs. Loyal is a San Francisco based startup, led by the amazing Celine Halioua. Loyal aims to give dog owners more time with their pets. Now this is what I call meaningful change! (4 minutes)

Here’s my dog Biscuit on his 2nd birthday. I want to celebrate many more birthdays with him.

So, what do you think?

💗 If you like “Software is feeding the world”, please hit the “Like” button at the bottom or share with a friend.

🙏 If you don’t mind answering 3 questions anonymously (2 are optional), I would love to get your feedback.