Real Food Campaign Lab
for Jan 2018 – Aug 2018
The Real Food Campaign (RFC) emerged in 2018 from a partnership between the Bionutrient Food Association (BFA), Next7, and Our Sci LLC. The mission of the RFC is to identify the best ways to drive increased nutrient density in our food supply, specifically through a better understanding of soil health, food quality, human health and their connections. For the full description of RFC mission, goals and objectives go here.
RFC’s goals in 2018 are two fold: 1) to better understand the variation in food nutrient density in the food supply, and identify possible sources of variation that relate to soil health and farm practice, and 2) to attempt to correlate spectral reflectance of produce with standard lab nutritional measurements like antioxidants, polyphenols, proteins, various minerals, etc.
The 2018 season will end in November, and final results are expected by end of December 2018. This report is an in-process update for donors, BFA members, and the general public.
Overview of activities
The following activities and milestones were accomplished as of August 2018:
- A survey was designed to collect farm and market data from the NE and MW US, and soil and food testing methods were chosen to quantify food quality and soil health
- The Data Partner program was created with 7 collaborators who signed on to submit 6 – 18 samples to the lab per week
- A lab location was identified, equipped and staffed in Ann Arbor MI to measure soil quality and food quality
- An lab sampling/testing process was created and optimized throughout the summer
- A scalable, open source information management system was developed specifically for use within the lab, and plans laid for integration with other systems (like FarmOS) to collect more detailed farm-level data (source at gitlab)
- A handheld reflectometer (bionutient meter) was designed, tested, built, and used daily in the lab (source at gitlab)
- So far, 350 samples were processed from stores, farms, and farmers markets in 6 states from 7 data partners
We learned a huge number of small lessons in creating the lab, in terms of process improvements and optimization, as well as some larger overall lessons. These larger lessons are described below.
Identifying good, consistent Data Partners is essential to our success
Collecting samples from a wide range of locations over time is a key part of what makes the RFC effective and inexpensive. Sample collection, even when very simplified, does take both time and some experience to do correctly. We found that individuals sending a few samples was time consuming for the lab, produced low quality samples, and was frustrating for the individuals.
Also, not everyone is cut out to be a Data Partner. It requires visiting farms and stores every week, consistently and paying attention to detail.
In the future, we should expect that for every 2 people we sign up for the Data Partners program, we get 1 to stick around and send quality samples throughout the season.
Shipping is not easy, cheap, or without error
Early in the season several samples were lost or delayed in shipping, even using the higher level (USPS Priority Express) service.
We also were told that we should get a USPS corporate account, which would save us money on shipping, and allow us to print our own labels (you cannot print USPS Priority Express labels using a normal account). We will pursue this for 2019.
Overall, our shipping issues have been addressed, but caused lots of headaches early on.
We need to fix (no more changes!) our testing methodology by early spring
Once a sample arrives in the lab, it goes through a long and detailed set of procedures which must be done consistently to be confident that data is comparable over time. We had to change procedures several times this year to address problems with our testing methods and to improve efficiency. Every time a procedure or test changes its version, it causes a lot of headache when analyzing the data, and calls into question comparability of samples over time, effectively reducing the power of the data.
We should minimize these in-season changes in future years and spend the winter creating a process we are confident in as soon as the survey begins in the spring.
Committed lab staff (full time, living wage) are essential to our success
Initially, we hired a 3/4 time lab manager to process samples, write out procedures, and help us improve the lab process. While this person was good and in fact enjoyed the job, she changed jobs after 3 months because she needed more full time and long term work.
In order to bring professionalism and consistency to the lab, we need a full time position which has at least 6 months of funding. Retraining lab personnel is time consuming, causes errors and/or inconsistency in the data, and puts us in a position of reworking old ground rather than making our process better in the future.
We are also looking into internships, credit hours, and collaborations with university labs at U of M to support this full time person with student hours. We hope that students also benefit from learning about the RFC and being inspired by its mission.
Below are actual and project expenses for the RFC Lab in 2018.
Detailed report of contract software, hardware, and consumables (excluding XRF, includes part lab labor but not all) here.
Projected cost per sample in 2018 ($212.85), is about half of a quoted commercial alternative for a similar suite of tests, and includes all costs including software, hardware, and equipment.
The projected cost per sample for 2019 (116.95 – 60.04 per sample), which removes the up-front costs associated with the XRF and hardware/software and assumes some efficiency gain, is a more realistic long-term range for per-test cost.
At this point, only initial results are available. These include total numbers within each category (organic, conventional, etc.), histograms to give a sense of the number of values, and some initial relationships between soil and food properties.
This dashboard is continuously updated as new data comes in, and includes a link to a combined CSV with all of the data (soil, food, and metadata).
While the numbers of samples in each category are relatively small (somewhere between 10 and 60) , it does appear that results from locations using ‘regenerative’ management (which includes things like no till, cover crops, biological amendments, etc.) have a negative correlation with antioxidants and, less so, polyphenols. There were only 8 samples from completely conventional farms, however, so these numbers are definitely preliminary. In terms of soil, there is a huge difference between certified organic and organic but not certified. This seems odd, but is the largest single effect in both soil respiration and soil carbon.
Overall, while interesting to look at, these results are preliminary and visual in nature – no rigorous statistics work has been done. In addition, due to many lost samples or poor quality data early in the process, we expect at least twice as many valid samples as we have currently by the end of the 2018 season.
Expand sample range
We hope to expand our range of samples we can run, and will be performing an experiment this fall on a wider range of products (tomatoes, peppers, rice, etc.). The goal is to evaluate how transferable our existing methods (extraction, nutritional information) are to things besides carrots and spinach. This will also help improve our Data Partners experience because they can get a wider range of samples from a single location (rather than hunting for the one farm which has carrots in June).
Identify more Data Partners + collaborators
We will have a strong presence at the Soil and Nutrition Conference, and will be actively seeking participation in the Data Partners program and beta testers for the Bionutrient Meter. We’ll also be connecting with larger institutions like farmer organizations or food processors who want to develop better systems to understand, track and measure nutrient density.
Improve lab process
We can significantly increase the efficiency of the existing lab with some focused time on lab process. Ideas we’ve identified so far include: when to start a process, how to move it along efficiently so anyone entering the lab can move samples along without prior knowledge, optimizing the sample intake process which is currently very tedious and time consuming, and using a pipetting robot to reduce errors and save time on the polyphenols, proteins, and antioxidants methods. We know we can increase our sample throughput significantly with dedicated time and effort in these areas.