Real Food Campaign

Executive Summary / a launch framework

Tools for transparency

Tools for transparency in the food supply

Over the past several decades, concentrations of vitamins, minerals, and micronutrients have steadily decreased in fresh fruits and vegetables.1  During a comparable period, degenerative diseases such as Alzheimer’s, diabetes, and heart disease have risen to epidemic levels.2  Simultaneously, agricultural practices have polluted aquifers and ecosystems and led to degradation of millions of acres of land.

We understand that correlation does not prove causation and that grasping complex relationships requires collecting mountains of data over a long period of time. However, it is undeniable that human, crop, soil, and environmental health are deeply interrelated.

Nutrient-dense food comes from healthy plants, and healthy plants come from biologically vital, ecologically regenerative, and carbon-rich soils.

But what is the definition of healthy food? Consumers have had few cues at their disposal for determining the relative nutritional value of specific fruits and vegetables. The truth is we haven’t really known the quality of our food because nutrient density has not previously been measurable.

But imagine going to the farmers market, flashing a light at several different carrots, and comparing their nutritional value in real time. Readings might show that some carrots are nutrient dense, while others are not. Which ones would you buy? Would you start to choose food based on how good it was for you and your family?

〰

We need a comprehensive plan to understand and revitalize our food supply.

The Bionutrient Food Association is initiating the Real Food Campaign to make the nutritional density of food easily detectable and incentivize the entire food supply chain to focus on nutritional value as a key metric.

We believe this new paradigm has the potential not only to improve the quality of crops and human health, but also to nourish ecosystems and enable soil to sequester carbon to its fullest capacity.

[INSERT CIRCLE GRAPHIC OF CONCEPT]

〰

We are building partnerships.

The Bionutrient Food Association is coalescing a partnership of food, health, environmental, and climate movements for a three-pronged campaign:

  1. Create a handheld sensor to measure nutrient density.   Spectroscopy is a well-developed technology that can discern the makeup of materials using a noninvasive flash of light. Advances in spectroscopy now make it possible to produce inexpensive (~$200), hand-held consumer-oriented sensors that can measure nutrient density. Engineering a prototype will take 12 months and a consumer tool will be available in three years, following calibration through crop surveys. The hardware and software will be fully open source, and the long-term design goal is to integrate the hardware into cell phones so consumers won’t need to carry separate equipment.
  2. Reveal the spectrum of variation in the food supply and use it to calibrate the sensor.
    Documenting the variation of nutritional content in produce through crop surveys is critical to both calibrate the sensor and expose the unseen variation in crop quality. Once visible, this variation will drive public discussion of food quality and raise awareness of the project. The survey data will be held in a public domain database, spurring public and private entities to further develop tools and software to increase food quality.
  3. Identify management practices that produce the highest-quality crops by collecting and analyzing shared data on an open platform.   Historically, agronomists have been primarily engaged in single-factor analysis of a small number of nutritional compounds. This is not sufficient. Through exhaustive crop and soil sampling and analysis of multiple factors, we will identify strategies that are conducive to the growth of nutrient-dense crops in specific locales. In addition to fertility programs, cultivars, and other unique management practices, this open-source platform will track factors such as soil types, pest and disease pressure, microbiomes, and epigenetics.

By making data freely available to growers of all sizes, research institutions, organizations, and corporations, we hope to accelerate the progress of this campaign and inspire international collaboration that will support increases in food quality globally. This work obviously requires a major investment. With your help, we will raise $10 million over the next five years to achieve these ambitious goals.

Download a brief, 2-page overview.

1 Davis DR et al. “Changes in USDA Food Composition Data for 43 Garden Crops, 1950 to 1999.” Journal of the American College of Nutrition. 2004;23(6):669–682.

2 National Council on Aging: “Top 10 Chronic Conditions in Adults 65+ and What You Can do to Prevent Them.” https://www.ncoa.org/blog/10-common-chronic-diseases-prevention-tips/

Real Food Campaign
The Food system

Challenges we face

In markets for food, flavor and nutrition have taken a backseat to productivity and price, more often than not. As a result, we eat much less nutritious food today than our grandparents did 80 years ago.1  In addition, we more fully understand how significant erosion of food quality is inextricably linked with afflictions in soil, plants, animals, humans, the environment, and the climate.

A nonprofit dedicated to improving nutritional quality in the food supply, the BFA is initiating the Real Food Campaign to directly measure food quality and promote best farming practices. We believe this new paradigm has the potential to not only improve the health of humans, but to also nourish microbial ecosystems, enabling the soil to sequester carbon to its fullest capacity and helping to solve the existential crisis of our time.

These are a few of the challenges we face in our food systems today:

〰

Nutritional deficiencies

An estimated 3 billion people from both developed and developing nations have specific nutrient deficiencies, impacting us in ways we understand and in ways we do not.2  In the developing world, the results of poor nutrition are deadly: low birth weight, high rates of infection, and diarrhea, among others.3  In the U.S. and other developed countries, the effects of poor nutrition are just as real but much less obvious: Even subclinical4 nutrient deficiency results in higher rates of infection, increased risks of all types of cancer, and an increased risk of obesity.5  This is a slow but real drain on our health, our economy, and our happiness.

〰

Organics: only a partial solution

Addressing malnutrition in the developing world is notoriously complex, but solutions in developed economies are simpler, including better information and labeling at the point of sale, such as with organics. The organic movement has successfully reduced chemical residues on produce and resulted in more environmentally sustainable farming practices. However, studies are not conclusive about the health benefits of organic produce over conventional food, even though 89 percent of organic food consumers cite health reasons as a motivator for eating organic.6,7

〰

Variations in nutrient levels

At the same time, studies consistently show large and very significant variations in nutrients due to other factors (see up to 10x variation in Vitamin A from samples collected across the U.S.,8 variation in lycopene, antioxidants, phenolics, and ascorbic acid from 53 varieties of tomato9). In other words, a tomato is not a tomato is not a tomato – even if it’s organic.

〰

Nutrient density unknown

Nutrient density is the level of nutrients per unit calorie, reflected in flavor and aroma, not volume and visual aesthetics. Determining the variation in nutritional density of all crops is largely unknown because it is an expensive process and there have been no obvious economic drivers to incentivize this work.

Relying solely on regulating farms to improve food quality, as organic certification does, will continue to fail. Consumers need measurements in the store to make the best purchasing decisions at the point-of-sale and to drive demand. And both conventional and organic growers need real-time measurements in the field to make production decisions that drive nutrient density.

〰

Dearth of research on nutrient density

Today most research is overly simplified, focusing on the effects of single nutrients on specific illnesses.10  Although the results are often significant, the public impact of each study is limited. Some studies indicate that fruits and vegetables with high levels of a broad range of nutrients produce better health outcomes. But many studies fail to account for potential differences in nutrient density, treating all fruits or vegetables as homogenous.11

Larger holistic studies with more broad and compelling outcomes are needed. Without compelling outcomes that can be communicated to the public, price and yield will continue to dominate breeding decisions and farmer practices.

〰

Lack of clarity about best practices

Finally, if and when markets can identify and demand nutrient-dense foods, there are no clear recommendations to farmers on ways to improve nutrient density in crops. It will take years to establish these recommendations and will require constant effort to keep them up to date.

As with the early organic movement, addressing these challenges requires sustained development, research, and communication.

1 David DR et al. J Am Coll Nutr. 2004;23(6):669–682. https://www.ncbi.nlm.nih.gov/pubmed/15637215, Mayer A-M. British Food Journal. 2006;99(6):207–211. http://www.emeraldinsight.com/doi/abs/10.1108/00070709710181540

2 Gram RD and Bouis HE. “Addressing micronutrient malnutrition through enhancing the nutritional quality of staple foods: Principles, perspectives and knowledge gaps.” http://www.css.cornell.edu/FoodSystems/AdvAgron4.html

3 FAO: “Human nutrition in the developing world.” http://www.fao.org/docrep/w0073e/w0073e03.htm#P522_61269

4 Not low enough to be diagnosed as a deficiency. Examples provided are effects for vitamins, trace minerals, and antioxidants, see reference 

5 Shenkin A. Postgrad Med J. 2006;82(971):559–567. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2585731/

6 Idda L et al. “The Motivational Profile of Organic Food Consumers: a Survey of Specialized Stores Customers in Italy.” 12 Congress of the European Association of Agricultural Economists—EAAE 2008. http://ageconsearch.umn.edu/bitstream/43946/2/152.pdf, FONA International: “Organic: A look at who is purchasing organics, why, where…and more!” https://www.fona.com/resource-center/blog/organic-look-who-purchasing-organics-why-whereand-more

7 Smith-Spangler C et al. Ann Intern Med. 2012;157(5):348–366. https://www.ncbi.nlm.nih.gov/pubmed/22944875, European Parliament: “Human health implications of organic food and organic agriculture.” http://www.europarl.europa.eu/RegData/etudes/STUD/2016/581922/EPRS_STU(2016)581922_EN.pdf

8 http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2621.1986.tb10851.x/full, Table 1

9 Hanson PM et al. J. Amer. Soc. Hort.Sci. 2004.129(5):704–711. http://journal.ashspublications.org/content/129/5/704.full.pdf+html, Table 1 

10 AHRQ Publication No. 09-E015: “Vitamin D and Calcium: A Systematic Review of Health Outcomes.” August 2009. https://www.ahrq.gov/downloads/pub/evidence/pdf/vitadcal/vitadcal.pdf

11 Slavin JL and Lloyd B. Adv Nutr. 2012;3:506–516. http://advances.nutrition.org/content/3/4/506.full

FORWARD LOOKING

Opportunity + innovation

Despite daunting challenges, technology and communities have evolved to a level where we can better communicate data and insights about food and farming practices. In addition, consumers have demonstrated their influence within the food marketplace.

〰

Proven consumer influence

Although the organic movement has not significantly improved nutrient density, it clearly shows the power of point-of-sale information and a strong market demand for healthy food. In only 15 years as a USDA-certified label, organic food is now a $43 billion industry in the U.S. (5 percent of all food sold in 2015), with 10 times the market growth of conventional food.

〰

Advances in technology

Advances in hardware have brought within reach low-cost sensors to measure nutrient density. Mini VIS-NIR spectrometers, handheld Raman, and decreased costs and complexity of all electronics are making lab-quality equipment accessible to anyone.1

This means the potential exists for consumers to non-invasively measure nutrient density in the store to find the best produce. Growers could also track the nutrient density of produce from their farms to the store – optimizing production, storage, and shipping practices to ensure the consumer receives the highest-quality product.

〰

Ready resources

In addition, inexpensive, open-source software for collecting, sharing, analyzing, and validating agricultural data now exists, including FarmOS and Our Sci (built on open-source code created for PhotosynQ.org). These platforms allow producer and consumer communities (farmer cooperatives, BFA, NOFA, and nutrition advocates, etc.) to organize large-scale experiments, develop and implement new methods and sensors, and track farmer practices and nutrient density. Using existing, open-source software saves millions of dollars and years of development time.

In turn, these advances may enable nutrition researchers to build studies on a mountain of public farm and consumer data to track health outcomes at a scale not currently possible.

〰

Collaborative food community

At the same time, the BFA has built a membership and network capable of catalyzing this vision and is ideally positioned to bring it to reality. Founded in 2010, and a 501c(3) non-profit educational organization, BFA’s goal is to improve food quality through agricultural methods that build soil vitality for better crop nutritional quality, vigor, flavor, and yields.

Growing it’s membership year over year, the BFA is in dialogue with an extended network of thousands of farmers, concerned consumers, researchers, academics, and experts in food and soil quality. Members include academicians at major universities (Cornell, Penn State, Michigan State, Chico State, and others), private-sector actors (Health Research Institute) and public-sector actors (USDA-ARS).

In short, the BFA has a passionate, cohesive community with the tools to collaborate efficiently – capable of producing scalable results and influence large, primed markets.

There has never been a better time to transform our food system and to:

  • Give consumers more choice for nutrient-dense foods with better health outcomes.
  • Support producers to get a premium for their products by improving food quality without the overhead associated with certification programs.
  • Enable researchers to create new tools, methods, and knowledge relating to human health and nutrition by using a new wealth of publicly available information.

Putting food testing in the hands of consumers and increasing awareness of food crop variations will encourage farmers to focus more on quality with the potential to radically transform agricultural practices and ecosystems:

  • When consumers can choose crops based on qualitative metrics, supermarkets and food companies have an incentive to purchase more nutrient-dense crops from the supply chain.
  • As demand increases for high-quality crops, the market will respond by rewarding farmers with preferable practices.
  • As farmers begin to change their practices to achieve quality standards, the need for and application of synthetic fertilizers, pesticides, and insecticides will decrease.
  • As crops become healthier, their net effect on the ecosystem will become positive, increasing microbial biodiversity and carbon sequestration.

1 Hamamatsu: “Mini-spectrometers.” https://www.hamamatsu.com/us/en/4016.html, Photosynq: MultispeQ v1.0. https://photosynq.org/buy-multispeq, BWTEK: “Handheld & Portable Raman Spectrometers.” http://bwtek.com/technology/raman/

AN OVERVIEW

The Real Food Campaign

Outlined below are the three key elements of the Real Food Campaign:

1. The Bionutrient Meter to measure nutrient density

“Nutrient density” is a general term encompassing many components, from minerals to vitamins to organic compounds. Current measurement methods are lab based, expensive, and time-consuming, and measure a single or a small number of nutrients.

We propose to create a fast, inexpensive, field-based device – what we are calling (for now) the Bionutrient Meter – to measure a broad range of nutrients of interest in crops. Because many nutrients such as minerals come directly from the soil, we will also identify fast and inexpensive soil quality measurements to help drive improved farmer practices.

The techniques to calibrate such a device using existing technology require a library of reference data. In other words, it is necessary to have both the device and the data simultaneously. This “chicken-and-egg” problem – needing the device to get the data, but needing the data to make a useful device – has stalled previous efforts due to the scale of in-house development required.

However, we will combine the hardware development, country-wide collection and sharing of data from BFA members and partners and the Health Research Institute, a nonprofit lab capable of collecting the reference data with expertise in nutrition and human health. This partnership will reduce the development costs and yield a more appropriate, accessible product developed by and for the community.

〰

2. Crop surveys to identify ranges of nutrient density in produce

The Real Food Campaign will include a survey of farms, markets, and stores across the United States, generating a fully public database of food quality using established lab methods.

In the short term, the surveys will help build a nutritional database to calibrate the handheld nutrition sensor for the Bionutrient Meter. It will also provide a completely unique dataset for public health and nutrition researchers to understand the sources and drivers of food quality. And it will track the quality of fruits and vegetables in a more granular way so that future changes in quality can be measured.

In the long term, the surveys will result in faster, low-cost testing methods, both in the lab and in the field, so that consumers and farmers can quickly and easily identify high-quality produce. The ultimate outcome we seek is to inform consumers and farmers about food quality in real time, resulting in improved supply and demand of nutritious food.

〰

3. Open-source platform to track food & farming data

We propose to create a platform on which to collect and analyze data and to provide feedback and collaborate about food quality and farming practices. This platform will provide a springboard for organizing research that can further enhance food quality and farming practices. Because the information will be freely available, the broader public will also have easier access to nutrition-related data.

To achieve these goals, we will build on two existing open-source projects called Our Sci and FarmOS. Each project is already independently successful: Our Sci is a software + hardware platform for collecting, analyzing, and sharing data so researchers and communities of practice can create successful, global research projects. FarmOS is a farm management system (asset tracking, field maps, etc.) to save farmers time and money, and to simplify compliance with certification programs such as USDA Organic Certification.

Our expectation is that through funding and establishing this platform, hundreds of millions of research dollars will ultimately be spent to create data held and exchanged in this public commons. Similar to the evolution of the Human Gene Project, its role will likely change over time.

Phase 1

Proof of concept

Measure nutrient density

Expensive lab testing is not a sustainable way to provide nationwide nutrition information. We must find cheaper, faster, and easier ways to measure nutrient density in produce. The technology is now emerging to achieve that goal, but we need help developing it.

Currently there is no way to directly and inexpensively measure a broad range of nutrients quickly in the field. First, you must perform a survey of nutrient density in crops and build a reference database of known nutritional data, relating it to simple measurements like reflectance to calibrate a handheld nutrition sensor.

Key design parameters

Measuring nutrient density is complex because “nutrient” is a vague and all-encompassing term. Depending on how you define it, nutrients might include hundreds of compounds that interact with each other and the body in complex ways. Furthermore, each compound is measured using a different lab method, and each individual method can cost hundreds of dollars.

In reality, however, we care less about specific compounds and more about the cumulative impact of a food on human health. In fact, a myopic focus on specific compounds is ultimately counter-productive, leading to a shifting carousel of “superfoods” that drive sales but not health outcomes.

Therefore, a nutrition sensor should measure the broad range of nutrients and compounds that impact human health. Measuring individual compounds may be representative of a class of compounds, but is not required to achieve our goals.

These are required design features for our nutrition sensor:

  • Inexpensive
  • Easy to use
  • Safe
  • Interpretable results
  • Measures the whole sample (not just the skin)
  • Correlates to classes of nutrients of interest
  • Open source

These are optional design features:

  • Detects specific compounds
  • Non-invasive
  • Inexpensive to calibrate
  • Incurs no recurring costs (consumables, reagents, etc.)
〰

Available technologies

Although technology is always shifting, these hand-held technologies fulfill some or all of our design constraints:

  • Raman spectroscopy
  • Laser-induced breakdown spectroscopy (LIBS)
  • Reflection in the ultraviolet (UV), visible (VIS), and near infrared range (NIR)
  • Microfluidics

Here is a breakdown of how each of these technologies fits this application: (For a detailed description of these technologies, see Appendix A).

Although Raman and LIBS are able to identify some specific compounds, they are too complex and expensive for our purposes. Microfluidics is also attractive but requires significant research and development to convert current lab nutritional measurements to a microfluidics platform. Although this is a worthwhile endeavor to explore in the future, it is impractical to apply today.

Currently, only reflection in the UV/VIS/NIR ranges is technically possible at a reasonable price. It provides a uniquely low-cost option for measuring nutrient density. Improvements in LED and sensor manufacturing are driving down the cost and widening the range of measurement. Reflection is already used to successfully identify food fraud, but making correlations to classes of nutrients (which is significantly more complex) remains untested.

〰

Proposed strategy: reflectance

A low-cost, handheld nutritional sensor is not commercially available precisely because there are no perfect options. However, we believe that reflection in the UV/VIS/NIR is the best option to broadly measure nutritional density today.

The initial crop survey(s) will provide the data to test the first key hypothesis: reflection measurements correlate to nutrient density. (See Appendix B for why we feel this hypothesis has a high likelihood of being confirmed.)

〰

Correlating reference data with reflectance

Our Core Survey intends to collect at least 180 samples over 3 years. Using widely accepted laboratory methods, we will test each sample for a range of vitamins, minerals, and secondary metabolites. In addition, we will measure samples using a full spectrometer to maximize available reference data and the chance of identifying correlations. Using this data, we will identify the key wavelengths that drive the correlation from the full spectral response (see graph below). We will then build LEDs at those wavelengths into the handheld unit.

Benefits of the LED/photodiode strategy

We favor the LED/photodiode strategy instead of building a full spectrometer in the field because:

    1. A full spectrometer is expensive. The cheapest and smallest is $120 from Hamamatsu. A dramatic drop in the cost of LEDs at less than $2, an increase in wavelength specificity, and the availability of low-cost, high-quality pin photodiodes at less than $5 makes the LED/photodiode option relatively inexpensive.
    2. The LED/photodiode strategy can utilize pulse amplitude modulation (PAM), but a full spectrometer cannot. PAM eliminates signal from non-pulsed sources, such as the sun or background lighting, providing a more consistent and comparable signal of interest in real-world conditions. PAM uses short pulses of light to actively probe the sample and has been used successfully for years in handheld fluorometers to measure photosynthesis in plants, an application requiring high accuracy in wide-ranging environmental conditions.
    3. The proposed LED/photodiode designs are flexible in the wavelengths measured. Once we identify the wavelengths from the survey, we can easily modify the design to include those wavelengths.

Below are simplified visual descriptions of a full spectrometer and a device using the LED/photodiode strategy.

Future strategies

The LED/photodiode strategy today uses the same calibration information as a full spectrometer, so switching strategies in the future remains possible. If prices drop, the technology improves, and there is a clear benefit to including the full spectrum in the field, the reference data is still available to calibrate a full spectrometer.

Finally, both strategies have high-quality, fully open-source designs that are already available using standard equipment. The MultispeQ has firmware and hardware designs for the LED/photodiode option, and there are several options for the full spectrometer, including the GroupGets breakout board, the CoralspeQ, and breakout boards provided by Hamamatsu. These designs and firmware reduce the cost of development and are consistent with the public and open nature of the project.

Prototype builds on existing designs. We will build 8 beta prototype low-cost, handheld full-spectrum UV/VIS/NIR reflectance devices and use them with partner labs (HRI) to calibrate a reference nutritional database.

We will use the core design from the MultispeQ V1.0, available here, with modifications to reduce costs. We will also use the MultispeQ firmware with modifications. This dramatically reduces the design time and cost required to have a working prototype version. (For the sake of comparison, the original development of the MultispeQ cost about 1.5 million dollars and took 3.5 years to complete.)

The proof of concept for effective reflectance measurements is already completed. Modified versions of the MultispeQ have already been tested in reflectance mode by swapping the pin photodiodes to be positioned on the same side as their respective LEDs. This is an example project that correlates reflectance values to total carbon in soils across hundreds of samples with success. It is structurally very similar to correlating reflectance values to nutrition measurements.

The MultispeQ was designed to measure absorbance, resulting in sensors and LEDs on both sides of the sample, two separate circuit boards, and a relatively complex mechanical design. It was also designed to achieve very high accuracy to measure photosynthetic parameters, which are outside the scope of a nutritional sensor.

We will redesign the nutrition sensor specifically for reflectance, reducing the cost and improving usability as compared to the MultispeQ. An initial estimate of savings is described below. Total unit costs would drop by more than $100, and up-front costs would drop by $23,000. For more details, see below: 

The core MultispeQ V1.0 circuit was an improvement on the MultispeQ Beta. Both versions have been field-tested for more than three years on several hundred devices, in field and lab conditions, in more than 15 countries. Heavily involved in the design of the MultispeQ and with unique knowledge of its function, both Greg Austic and Jon Zeeff will perform the design work on the Bionutrient Meter.

〰

Design Process

The core requirements are outlined in the Design Parameters section above, but here we provide greater detail about the design process.

1) Evaluate application
We will evaluate how the device will be used, who will use it, and what the design requirements are. The evaluation stage is a collaborative process with developers, designers, and users to identify the nexus of what’s possible, what’s optimal, and what’s practical.

Users will include BFA core staff, collaborators as part of the Core Survey, including Kris McCue and Faith Reeves, and Dr. John Fagan from the Health Research Institute. The concept (how it works), features list (what exactly it does) and spec list (how well it does it) that emerge from this stage will define the rest of the process. The concept will include draft drawings and user stories.
Outputs:
    → Concept
    → Feature list
    → Spec list

2) Test assumptions
In this phase, we will create components to test any assumptions we’re not confident about from the feature list. If an assumption fails, we redesign and retest until we get it right. If we can’t get it right, we drop or adjust the feature. This is a “fast fail” stage, with breadboard, simple circuit designs, and 3D printed parts. We test assumptions both on the circuit and on any mechanical designs.

3) Debug design
Once we test the core assumptions, we will design and assemble a full draft circuit board and case. We’ll identify the first round of bugs, and produce and retest another version. Once we’ve produced a working device, we’ll test the spec list against the actual performance of the device. If something on the spec list fails, we will reevaluate and retest the design until that spec is achieved. 
Outputs:
    → Draft device (minimum 1 device)

4) Finalize design
If we achieve all specs and features on the draft board, then we can finalize the case with some aesthetic details in mind. Any design changes will likely require at least a few iterations to identify any remaining mistakes. Features and specs will be rechecked on the final design to ensure nothing was lost in the final design. Once complete, we will assemble and test eight devices. 
Outputs:
    → Final devices

5) Beta test
At this stage, we put final devices in the field for use. This almost always requires follow-up and sometimes hardware patches. Of the eight devices manufactured, at least two will be sent to the Health Research Institute (HRI) for testing in the Core Survey, alongside a traditional spectrometer. HRI may also use the devices with any of their other customers who would be willing to make their results public, potentially building partners and expanding the dataset. The remaining six devices will be split between the development team, BFA staff, early backers of the crowd-funding campaign who contributed at the $10,000 level, and other nutrition labs in industry or academia who may want to participate in alpha testing.

We will use the open-source PhotosynQ web back end and mobile app during the Core Survey for collecting data from the devices. Although this is not the desired long-term solution for data collection and sharing, it will provide beta testers an easy-to-use interface and will store the data for future use. It also allows critical metadata generated by the survey to be combined with the sensor data in an easy and validated way. Analysis of the final data is also included in the project plan for the Core Survey. 

〰
Identify range of nutrient density

We plan to ultimately track nutrient density in hundreds of crops, but will begin with just one or two. A likely scenario is testing carrots and spinach in two regions: Connecticut and Iowa.

We will process each sample using standard AOAC and US Pharmacopia methods, as well as a simplified method, using the device for UV/VIS/NIR reflectance, to establish a correlation between the two and provide information that will allow calibration of the Bionutrient Meter’s sensor. Correlating these tests will also reduce the cost of future work by allowing us to use a simplified method in place of standard methods.

Initial sampling regions

We plan to initially test crops in two regions with contrasting soil types (for example, histosols in Connecticut and mollisols in Iowa). The soil profile, microclimate, and soil type in Connecticut and Iowa are also relatively consistent, so farm-to-farm variation in soil quality should be low. More regions would introduce too much sample variation from differences in collaborators and soils.

Vegetables to test

To begin, we favor testing carrots and spinach for several reasons:

  • They are in season and harvestable at roughly the same time in all zones, lowering the cost and effort required to collect the samples.
  • They are widely consumed, major crops (in the top 20).
  • In previous studies, carrots have shown significant variation (100x) in beta-carotene, which is easily measurable using spectroscopy.
  • Both vegetables are physically consistent enough to use spectroscopic methods (unlike broccoli or peppers, for example, which have large air pockets and complex surfaces).
  • Having one solid vegetable and one leafy green provides a broader learning opportunity for method development.
  • Juice can be extracted easily from both, allowing for possible infield refractometry and spectrometry in the future.

Benefits of HRI/BFA involvement

BFA-trained members will collect the data and the Health Research Institute (HRI) will test the samples. This strategy leverages the testing expertise of HRI as well as the strengths, reputation, and integrity of BFA in the agricultural and food sector, positioning the BFA to perform the survey at minimal cost and outside support in the future. It facilitates the following:

  • Lower costs by:
    • Using BFA’s membership to collect data for free from across the country
    • Working with a non-profit lab (HRI) whose prices are 20 percent lower than commercial alternatives
    • Correlating a full (expensive) nutritional test with a cheaper, simplified test to drive down testing costs in the future
  • Extend the public discussion by:
    • Collecting data over three years so we can incrementally improve the survey and extend the public discussion about nutrition
  • Build long-term revenue streams by:
    • Providing fee-based testing services for farmers and consumers not part of the survey to increase the flow of data and offset the cost for the testing equipment
    • Identifying partner organizations (academic or industry) willing to pay to expand the survey to new locations or new produce for their own research needs
〰

Addressing variations in testing

Three suppliers will be sampled for each Management x Location x Region x Crop group. To reduce variation within each of these groups, each supplier’s sample will consist of five individual samples (for example, five individual carrots from the same supplier), which are combined before testing. Test-test variation is always present, but the methods being used have known test-test variation, which we will assume also applies to our samples.

〰

Sample size considerations

Given the lack of data on nutritional testing in stores and farmer’s markets, it is hard to estimate exactly the expected differences between groups and decide on a “reasonable” sample size. However, some historical data can set basic guidelines. We see the variation in beta-carotene between carrot varieties in Takahata et al from 0 to 25mg / 100g sample,1 while Pinheiro-Santana et al found a mean of around 7mg / 100g sample.2 Both saw standard deviations within the population at around 1mg / 100g.

Assuming we have a similar variance within our populations, we can roughly calculate what the expected minimum statistically significant difference between populations would be at any given sample size. (See Appendix C for calculations.)

Assuming we compared beta-carotene among organic carrots (24 samples) versus regenerative ag (12 samples) in year one and two, a difference greater than 0.8 mg/100g should be statistically significant (two tailed t-test using alpha = .05 achieving a p-value ~ .02). By year three, the same comparison would include 18 and 36 samples per population. In this case, differences as small as 0.6 mg/100g should be statistically significant. Given Takahata showed a varietal variation is 25 mg/100g sample, 0.6 mg/100g is reasonable for detecting differences in the population means.

Given these considerations as well as cost constraints, we have determined sampling sizes for our crop survey. Even at reduced rates provided by our nonprofit laboratory partner, the cost of testing limits the number of samples we can test. Therefore, we will perform 30 nutritional tests from 30 samples for each crop in each year. For each crop in each year, 18 soil samples from participating local farms will also be tested for soil quality. See “Testing Methods” section below for details about soil and nutrition methods.

〰

Variables

The crop survey may include the variables below. In addition, we will collect other relevant metadata that may impact the dependent variables (who collected the data, the name of the store/market, name of the producer, time of day collected, etc.)

Independent:
Parameters to be varied within the experiment.

Crop(carrots, spinach)
Region(Iowa, Connecticut)
Location(supermarket, farmers market)
Management(conventional, organic, regenerative ag)
Management detailsfor samples from local growers only (management questions like no-till, fertilization details, etc.)

Controlled:
Parameters to be held constant across all samples.

Sampling method(cold packed, next day shipped, tested within 48 hours)
Sampling season(sampled during peak harvest)

Dependent:
Measured outputs from each sample.  For details on each test type, see Sampling Methods section below.

“Nutritional test”Vitamin A, K, Amino Acids, Total Phenolics, Minerals (36), Beta Carotene, UV/VIS/NIR and refractometry
“Soil test”Soil carbon, soil biology, cations, Minerals (36)

〰

Sampling methods

It is widely known that ripe produce loses nutritional value quickly after harvest. It is therefore important to have consistent sampling methods from the marketplace to the lab to keep data quality high. Data collectors will have kits for collecting samples, and they will collect all samples at the same time to reduce sampling and labeling error. Data collectors will immediately label samples, seal them in a bag, place them in an insulated container with dry ice, and send them to the lab for analysis. They will be stored at -80C in the lab until testing. The time from collection to analysis will be held constant at 48 hours.

Testing methods

Nutritional analysis is performed on all samples, and includes both standard methods (AOAC, etc.) and simpler methods (UV/VIS/NIR, refractometry).

We chose standard methods based on the following parameters:

  • Efficiency (cost per unit)
  • Impact on public health
  • Public awareness of their meaning
  • Variety of nutrient type
  • Likelihood of correlating to simpler test methods

We chose simplified methods based on the following parameters:

  • Cost per test
  • Equipment requirements (expensive machines)
  • Simplicity (anyone could do it)
  • Testing time

Soil sampling

Samples originating from local farms (⅗ of all samples, total of 108 samples over three years) will also have soil samples collected. Soil sampling methods were chosen to help identify overall soil health (carbon and microbial activity) and mineral content, which may relate to mineral content in the produce eaten. Soil data could identify differences in soil quality based on different types of farm management, and provides a link between soil and produce quality.

Desired Outcomes of the Crop Surveys

  1. Build capacity within the BFA staff and membership to perform large-scale nutrition surveys.
    As an advocacy organization, the BFA has a large, motivated, and engaged base of members. This project will test the capacity of members to collect high-quality samples and data. It will also build knowledge among BFA staff and partners to build and run a nationwide survey. Together, the BFA and its membership will have the capacity to expand the survey in the future.
  2. Identify statistically significant nutritional differences between organic, conventional, locally sourced, and non-local produce.
    The experimental design of this survey, although limited in scope, will provide opportunities to identify even small differences in food quality (see sub-section “Survey Design”). The differences identified in this survey will allow future surveys to be more targeted, decreasing cost and increasing impact.
  3. Communicate results effectively to the public.
    • The survey data will be posted on the BFA website, which receives ~10,000 page views per month, in a fully searchable form. Anyone will be able to map and analyze the data through the site or download it.
    • Yearly reports will be generated and distributed through the BFA membership, at the annual BFA conference (more than 800 members, and 500 conference attendees) through the classes and talks. These reports will be in PDF form, detailing the highlights from the survey as well as other relevant published data from the previous year.
    • The BFA will run a social media campaign after each report is released (for example, BFA Facebook page is 2,300 followers and growing).
    • The BFA will publish at least one peer-reviewed article about the results from the survey. Ideally this publication will be pre-published in year one, with the results emerging in year three.
  4. Lower the cost of the survey in the long term by identifying relationships between the expensive standard nutrition methods and less expensive methods.
    • The BFA will publish at least one peer-reviewed article relating standard nutritional tests with UV/VIS/NIR methods and refractometry, and identifying opportunities for future research.
  5. Leverage this survey to engage researchers, food activists, and additional funds to expand the survey in future years.
    • In years two and three, offer paid services to farmers and consumers for food quality testing. This expands the scope of the survey while generating income to offset survey costs.
    • Identify at least three researchers in food nutrition and human health to add modules to the basic survey to research their topics of interest.
    • Apply for at least two additional funding sources to expand and continue the survey beyond the third year.
〰
Tracking food & farming data

We propose to create a software platform to collect and analyze data about food quality and farming practices, and organize communities of practice around that data. This will be a place to provide feedback, collaborate, and organize research on nutrition and farming practices and a resource for the broader community. This platform will build on resources that are already available. By building on top of well-developed platforms, FarmOS and Our Sci with existing user bases, we avoid a “build-it-and-they-will-come” strategy, which is both expensive and unresponsive to feedback.

Establish Software Platform

Combined, FarmOS and Our Sci have already invested millions in software and hardware development and years of user testing, significantly reducing the cost to BFA in starting a similar project from scratch. In turn, BFA’s involvement and contributions will benefit their business models and contribute to their success.

Data collection is already possible through Our Sci and FarmOS, but communication between the platforms is not currently possible. In Phase I, we will need to develop application programming interfaces (APIs) (ways for programs to communicate information) to connect the two platforms. The APIs consist of three main components:

  • Data sharing. This allows consumers and producers to track nutrient density in markets and on farms, and enables researchers to utilize the data to improve farm practices and develop new sensors.
  • Method sharing. Successful farm practices should be delivered to the farmer natively through FarmOS, providing real-time suggestions and updates (think of a personal assistant that makes sure you are on track for your fitness goals).
  • Anonymization. Farm and consumer data can be anonymized to the county or state level so sharing doesn’t trigger privacy concerns. We are building a public database to enable research for improving tools, techniques, and methods, however, anonymization is also needed to ensure users feel comfortable.
〰

Build Research Communities

The software and hardware is the scaffolding on which work gets done, but the BFA and its partners will also contribute. This means expanding BFA membership, strengthening partnerships with research organizations, and training to ensure the highest quality research is being performed.

In Phase 1, we will build partnerships in these core areas of research:

  • Connecting farmer practices and nutrient density. The BFA has already been using low-cost tools like refractometers to estimate nutrient density for years. Now, using the platform described above, broader and more rigorous research is possible. Using its own members, the BFA itself will create a research agenda to build a data pipeline of farm practice and food quality. Once established, research partners like the Rodale Institute, UNH, PASA, and other partners can leverage the data pipeline to answer research questions.
  • Connecting nutrient density and human health. BFA research partners at Penn State, Ohio State, and the University of Montana, along with the Health Research Institute, can access the data and network of participants to produce high-quality human health research. Specifically, the breadth and depth of the network will allow for multi-factor studies between human health and quantitative values for a range of nutrients of interest. Previously, multi-factor studies have not typically been performed and reported on in the literature because of the cost associated with collecting the data and the number of samples required to get statistically significant results.
    goals.

1 Takahata Y et al. Japan. J. Breed. 1993;43:421–427. https://www.jstage.jst.go.jp/article/jsbbs1951/43/3/43_3_421/_pdf, see Table 1.

2 Pinheiro-Santana H M et al. Food Science and Technology (Campinas).1998;18(1): 39–44. http://www.scielo.br/scielo.php?script=sci_arttext&pid=S0101-20611998000100009, see Table 4. Convert from ug/g to mg/100g

PHASE II

Build + Scale

Phase II depends heavily on results, feedback, and lessons in our proof-of-concept phase – so detailed planning now does not make sense. We know that our goal at the end of 5 years is to create an ecosystem of farmers, researchers, and consumers that is self-sustaining. The Bionutrient Meter and platform will pay for its own on-going development through sales. Consumers will buy the meter because they see value in choosing healthier food and finding like-minded farms. Researchers will use the platform because it is a rich source of data for publications. Farmers will use the platform to track their own crop quality, and find better production practices to improve their bottom line.

So Phase II should put the project on stable ground to accomplish these goals. As such, Phase II work will be focused on the following:

Expand data collection

Phase I builds a library of nutrition information on a small number of crops. Phase II should expand the survey to more crops and locations to improve the prediction algorithm and improve the Bionutrient Meter.

Commercially develop the Bionutrient Meter

Creating a commercial product (rather than a prototype or short-run product) requires up-front money for manufacturing large quantities to reduce costs and ensure the design is manufacturable at scale. Furthermore, the long-term maintenance and improvement of the meter will require a relationship with an established company.

Improve platform

The mobile app for collection of nutrient density information is just a small part of the broader platform. To engage a broader group of farmers, researchers, and consumers, we need to fund additional features, including:

  • analyze the data more quickly and comparably
  • allow researchers or farmers to create and share farming practices which can be easily implemented by others
  • provide social features to help connect users, identify the most nutritious nearby food, connect measurements directly back to farms from the store, etc.

Communicate results

A great deal of research will be produced throughout the project, and the hope is that research sheds new light on our relationship to our food and farming systems. The BFA will need a team of skilled communicators – skilled in traditional media, social media, and scientific publications – to make sure this information gets out as broadly and accurately as possible.

STEWARDING RESILIENCE

Project Sustainability


Attempting to truly improve health outcomes on a global scale is inherently challenging. We believe our proposed plan must effectively address barriers to achieving the following goals:

  1. Keep incentives aligned.  Many stakeholders with different goals are involved in making this proposal successful in the long term. Each is involved for different reasons:

     

    Our Sci,
    FarmOS    		⟶· Strong software development partner with BFA
    · Increase size of user base
    BFA research
    partners    		⟶· Access to network of global collaborators
    · Access to large public databases
    · Strategic partnerships for pursuing grants and funding
    Farmers⟶· Ability to measure nutrient density on their farms
    · Increased value of produce by increasing nutrient density
    Consumers⟶· Ability to measure nutrient density of produce
    · Ability to be part of a movement

    These incentives are interconnected, and must grow organically over time. In addition, any features or proposals within this project will be evaluated based on how they impact these incentives.

  2. Ensure the platform stays responsive to nutrient density and human health.  A decentralized system cannot be fully controlled. Even with point-of-sale sensor data with nutrient information, one could still imagine gimmicky instruments, diet fads, or poor-quality research all steering consumers and producers away from the real goal of long-term health benefits. Like all communities, cultures change over time in ways that cannot be determined from the outset.

    The BFA should act as a buffer against these kinds of negative changes. As a central, respected, and active player in the platform, the BFA will help establish and reinforce a culture of scientific rigor, honesty, and focus on human health. Therefore, long-term BFA funding is not to maintain the platform itself, which will be financially self-sustaining, but to ensure the mission is maintained as the platform grows and changes.

  3. Balance data sharing and privacy.  Large-scale data sharing is central to the success of this proposal. Yet real privacy concerns exist – farmers who want to maintain private production information and researchers who must collect human health data according to ethical research standards. How will we balance these needs?

    Technically, FarmOS and Our Sci already have privacy levels to provide users options for their data. In practice, we need design rules to ensure the best outcomes as new features are added. Sharing is the default, anonymized is optional, and private is available when needed. When provided with all options up front, users tend to choose the private option. However, when the default is to share, users are rarely concerned enough to seek private data options. Users who are concerned should be able to find those options.

    Foundation to maintain and expand open-source assets

    Open-source software and hardware create immense value in the global economy. In 2012, 29 percent of all deployed code is open source, and 50 percent of companies used open-source code.1 The benefits include preventing vendor lock-in, improved security, higher-quality code, and a code-base on which to build business and create value.2

    To ensure the longevity and availability of our open-source assets (software and hardware), we need a foundation to maintain, promote, and expand those assets. While the foundation maintains the code, companies and non-profits (like Our Sci and Farmier) will create sustainable commercial implementations and provide services directly to users. The Apache Foundation, the Open Containers Initiative, and the Linux Foundation are all very successful examples that follow this model.

    1 Vienna Advantage: “7 Interesting Facts about Open Source Software.” http://viennaadvantage.com/blog/technologies/7-facts-about-open-source-software/
    2 Red Hat is a good example: https://en.wikipedia.org/wiki/Red_Hat

ODDS & ENDS

Appendices

Appendix A

  1. Raman spectroscopy – Measures specific compounds through Raman Scattering, which is largely unique by compound. Although powerful, this option has many drawbacks. High cost and complexity are the two most difficult to design around.
    • Expensive – $17,000 for commercial handheld version from BW-Tek, many core components of the technology (the laser and the detector) have requirements that limit their ability to drop in price.
    • Bulky – Even the handheld ones are still quite large and would limit the number of people who would actually use them in the field.
    • Dangerous – Raman requires a high powered lasers which could damage the eye, and require warning labels on the device.
    • Complex – Raman produces a spectra which must be de-convoluted to use.  Current successful applications are in the pharmaceutical industry, where very specific, consistent, and pure compounds are quantified.  Pharmaceuticals are a comparatively pure material, and therefore easy to analyze, as compared to a carrot which contains thousands of compounds.
    • Minimal penetration – Raman can be done at 3 wavelengths (532nm/785nm/1064nm).  Lower wavelengths (532nm) have minimal penetration, measuring only the surface of the object.  Longer wavelengths (1064nm) will penetrate the samples but will not identify many compounds of interest.  While you can build systems with more than one wavelength, the cost and signal complexity go up.
  2. Laser-induced breakdown spectroscopy (LIBS) – Detection of specific elements of the periodic table by burning the sample using a laser and measuring the resulting spectra.  This technology is dangerous, expensive ($50,000 currently), complex to interpret, and the interpretation requires a team of trained specialists for each type of sample.  Also, this does effectively identify organic compounds.
  3. Reflection – Measure the reflection of light in the UV, visible, near infra-red, and infra-red ranges (~250 – 2000nm).  Traditional spectrometers are expensive ($500), but costs are coming down and are expected to come down further.  For example, Hamamatsu’s mini-spec is $120 and measures 350 – 850nm.  It’s possible to design reflection spectrometers which are not affected by ambient light, bringing the total manufacturing cost below $100 and improving usability. While inexpensive and easy to use, reflection has not yet been proven to relate to broad ranges of nutritional compounds, though there are reasons to believe that it likely would with enough data to correlate nutritional data with the spectral output. (See Appendix B.)
  4. Microfluidics – Also referred to as “lab on a chip” systems, the strategy is to perform chemical reactions (wet chemistry, like that used in laboratories) using tiny amounts of reagents embedded on substrate.  Classic examples are paper strip soil tests or pregnancy tests, though now diseases and nutrient deficiencies are being identified with paper-based urine tests.  There are more flexible, small scale open source technologies like DropBot and OpenDrop as well. Microfluidics can potentially directly measure specific compounds of interest, and run (in miniature) well referenced and respected lab methods.  However, unlike the aforementioned technologies microfluidics has a consumables cost, though it can be as low as $1 per test in some cases. Additional tools, like a colorimeter, are also needed to quantify the results. Finally, the development time and cost to build the microfluidics platform and develop the methods for measuring any given compound is significant.

Appendix B

The use of UV, visible, near infrared, and infrared reflection is not new in the food industry. Reflection-based measurements have been used to determine the quality of fruits and berries. Reflection-based measurements have been proven to correlate well with the total soluble solids, titratable acidity, and flesh firmness in apples, apricots and strawberries.1,2,3  In addition to these quality parameters, reflection measurements have also been used to predict desirable traits in apples including roughness, crunchiness, mealiness, and both sweet and sour taste.4

These traits may be desirable traits for marketing fresh foods but they do not necessarily relate to either the nutrient density or potential health benefits derived from food.

Recently, however, reflection measurements have been correlated to various nutritional parameters important to human health. For example, reflection measurements of lettuce were correlated to the pigments anthocyanins and carotenoids, which are antioxidants, and chlorophyll, which has anti-carcinogenic properties.5

In another example, reflection measurements were used to predict lycopene content, another phenolic compound with health benefits, found in tomatoes.6  Furthermore, reflection has been used to identify negative quality issues in fruits and vegetables such as elevated nitrate concentrations in summer squash7 and blackspot in potato.8

In a somewhat related area, measuring reflection spectra has become a useful tool in predicting many soil parameters including organic C, pH, and soil texture. In this field, there are two approaches that have proven critical to improving the accuracy of the reflection-based predictions: 1) building up large databases of reference data9, and 2) incorporating key metadata into the models.10  When predicting soil properties, key metadata may include topography, vegetation, geological and climate data.10  In the area of food quality, it does not appear that any work has yet been done to identify metadata that may affect nutrient density.

Appendix C

Comparisons between populations in the State of Nutrition survey, and the estimated minimum statistically significant difference between the population means.  Populations are assumed to be normally distributed with a standard deviation of 1 mg/100g.  Calculations available on request in Open Calc (.ods) or Microsoft Excel (.xls) formats.

Nationwide comparison of supermarket vs farmer’s market carrots

In year 1 →

Supermarketn = 12
Farmers marketn = 18
Estimated min. statistically significant mean population difference ~ 0.9 mg / 100g

In year 3 →

Supermarketn = 36Farmers marketn = 54Estimated min. statistically significant mean population difference ~ 0.5 mg / 100g

Nationwide comparison of conventional, organic, and regenerative ag carrots

In year 1 + 2 →

Conventionaln = 24
Organicn = 24
Regenerative agn = 12
Estimated min. statistically significant mean population difference ~ 0.8 mg / 100g

In year 1 + 2 + 3 →

Conventionaln = 36
Organicn = 36
Regenerative agn = 18
Estimated min. statistically significant mean population difference ~ 0.6 mg / 100g

1 Martinez Vega MV et al. J. Sci. Food Agric. 2013;93:3710–3719. http://onlinelibrary.wiley.com/doi/10.1002/jsfa.6207/abstract

2 Amodio ML et al. Postharvest Biology and Technology. 2017;125:112–121. http://dx.doi.org/10.1016/j.postharvbio.2016.11.013

3 Berardinelli A et al. Journal of Food Science. 2010;75(7): E462–E468. http://onlinelibrary.wiley.com/doi/10.1111/j.1750-3841.2010.01741.x/abstract

4 Mehinagic E. et al. Food Quality and Preference. 2003;14;473–484. https://www.researchgate.net/publication/230725502_Relationship_between_sensory_analysis_penetrometry_and_visible-NIR_ spectroscopy_of_apples_belonging_to_different_cultivars
5 Steidle Neto AJ et al. J. Sci. Food Agric. 2017;97:2015–2022. http://onlinelibrary.wiley.com/doi/10.1002/jsfa.8002/abstract

6 Clement A et al. Quality Assurance and Safety of Crops & Foods. 2015;7(5):747–756. http://www.wageningenacademic.com/doi/abs/10.3920/QAS2014.0521

7 Sanchez M-T et al. Postharvest Biology and Technology. 2017;125:122–128. http://dx.doi.org/10.1016/j.postharvbio.2016.11.011

8 Lopez-Maestresalas A et al. Food Control. 2016;70:22–241. http://dx.doi.org/10.1016/j.foodcont.2016.06.001

9 Viscarra Rossel RA et al. Earth-Science Reviews. http://dx.doi.org/10.1016/j.earscirev.2016.01.012

10 FAO: “Prediction of soil properties with NIR data and site descriptors using preprocessing and neural networks.” http://www.fao.org/fileadmin/user_upload/GSP/docs/Spectroscopy_dec13/Aitkenhead.pdf

Scroll to Top

Real Food Campaign in 2 minutes