TASTE Food: The Automated Scope 3 tool for Tracking Emissions from Food
The Culinary Institute of America, Stanford University, University of New Hampshire, AASHE, World Resources Institute
Project Overview
Currently, estimating purchased food emissions requires time intensive methods and cost prohibitive software which are often limited to estimating emissions based on spend. Previously, institutions that analyzed purchased food emissions underwent the slow and error-prone process of hand categorization. Due to the large amount of effort required by hand categorization, many institutions have been unable to prioritize quantifying and mitigating these emissions. Even for those institutions that have conducted hand categorizations, differences in how institutions categorize the same type of item, as well as high human error, make it difficult to compare results between institutions. To improve this process, The Automated Scope 3 tool for Tracking Emissions from Food (TASTE Food) was developed as part of a Master's research program at Stanford University. This tool uses a python-based software to categorize food purchases into 18 categories frequently used to analyze food emissions and is available for free online at https://food-emissions-categorization.wl.r.appspot.com/. The tool was validated using hand categorized data sets from 8 universities provided through a partnership with the SIMAP Nitrogen Working Group, and it was found that the tool-categorized data had emissions +/-5% the estimated hand categorized emissions. By using TASTE Food, institutions can quickly categorize food purchases, facilitating the process, and guaranteeing that the calculated emissions are determined consistently and can be compared from one institution to another. TASTE Food has also recently been updated to be able to tag plant-based foods for AASHE STARS Certification, increasing its utility to a larger range of institutions and applications. The Master's thesis written about TASTE Food and using this tool for a variety of analyses is available at https://purl.stanford.edu/dn946wd9147.
Background
A Master’s research project was focused on developing lessons learned from estimating scope 3 emissions associated with purchased goods. One main goal was to understand the impact of using various emissions factors on the estimated emissions from the same data set. Food was identified as the optimal subset of purchased goods to consider for analyzing the impact of using weight-based versus spend-based emissions factors. This is because it is commonplace to consider both the weight of food (i.e. a 6 oz steak) and the spend associated with food (i.e. $50 steak) as compared to other purchase categories (i.e. it’s hard to visualize $50 worth of chair or 50 kg worth of chair). To determine emissions from the purchased food from Stanford Dining Halls in 2019, I quickly realized that it was necessary to categorize the purchased foods into categories for which emissions factors exist. Determined to not do this by hand, I began to develop software which would eventually turn into TASTE Food which categorizes food purchases automatically into the categories for which emissions factors exist. TASTE Food is now available online for free at https://food-emissions-categorization.wl.r.appspot.com/. At first, TASTE Food could only categorize food purchases to the 18 categories used by the Sustainability Indicator Management & Analysis Platform (SIMAP), but as of October of 2024, TASTE Food has now evolved to be able to also tag Plant Based items in accordance with AASHE STARS certification. An additional new version of TASTE Food is currently in development which aligns with the 56 categories used by the World Resources Institute (WRI) Cool Food Pledge. Although the aim of the original research project was to determine the impact of different emissions factors on the same data set, the tool that was created for this goal was, in the end, the piece that attracted significant attention and interest from a variety of parties. I quickly realized that the need for fast, accurate, and replicable categorization of food purchases was not just unique to Stanford and this tool could be widely useful to many universities, and even beyond universities. There are many scenarios where TASTE Food may be helpful to an individual or institution. If an institution hasn’t started looking at food emissions yet because of lack of guidance or manpower, TASTE Food offers an opportunity to start looking at an institution's emissions from food purchasing with ease. If an institution has started looking at food purchasing emissions, but has been doing it by hand, TASTE Food offers an opportunity to continue the good work in measuring food emissions, but now without the large time commitment necessary for hand categorization which can be dedicated to other sustainability projects. An institution may also be interested in comparing food purchasing emissions to other institutions. With hand categorizations, different institutions may categorize things differently, making estimated emissions comparisons difficult between institutions. Since TASTE Food categorizes things the same across the board, the categorizations can be used to accurately compare among institutions. Finally, an institution may be interested in setting targets for their institution, and if this is the case, by using TASTE Food, institutions can set reasonable and achievable targets for the future by understanding their starting point.
Goals
TASTE Food aims to lower the hurdle for universities and other institutions to begin analyzing their emissions from food purchasing, set goals for the future for emissions reductions, and report their emissions to appropriate reporting platforms. This goal is being achieved through a free online python-based categorization software which takes in food purchasing data (in a simple template) and categorizes those purchases into categories for which emissions factors are available. By easing the process of estimating emissions from food purchasing, TASTE Food hope to encourage conversation and action in the food purchasing space by bringing attention to this large source of emissions on campuses and at institutions.
Implementation
TASTE Food came to be as part of a Master's research thesis and was successful thanks to a large support network of research collaborators and partners. The original participants were: - Rebecca Grekin, the Master’s Student at the Energy Science and Engineering department at Stanford University - Sally Benson, a principle investigator from Stanford University - Stanford Residential and Dining Enterprises who provided the initial data for food purchasing for all dining halls in 2019 - The Stanford Office of Sustainability who were liaisons between Rebecca Grekin and the Stanford Residential and Dining Enterprises as well as project supporters and were available for research feedback As time has progressed, the Master's research officially concluded with the publication of the thesis, but work on TASTE Food has continued. After these initial partners, many others have been added to the crucial partners in the process including: - the SIMAP Nitrogen Working Group - the World Resources Institute Cool Food Pledge - the Menus of Change University Research Collaborative
As a Master's research project, TASTE Food research followed a common template for graduate studies. Weekly meetings for feedback and progress reporting happened with both the project principle investigator, Sally Benson, as well as with the Stanford Office of Sustainability. These sessions were crucial to ensuring continued progress and sanity checks to confirm research approaches were relevant and appropriate. Rebecca took it upon herself to seek out partners to solidify the applicability of TASTE Food and that is how the initial partnership with SIMAP came to be.
Timeline
The project is ongoing and improvements and more partnerships are in the pipeline. The project can be broken up into 3 phases. The first phase can be defined as the initial development of TASTE Food, which ended with the first version of the website going live in March 2021. The second phase of the project can be defined as continued improvement and ended with the newer, easier to use, website going live and 25 universities having used the tool which ended in May 2023 with the launch of the second improved version of the website. The third phase of the project is still ongoing and can be defined as continued improvement, developing additional partnerships, and increasing the functionality of TASTE Food.
January 2020: Project started with a Master’s research project at Stanford and partnership with the Stanford Office of Sustainability initiated January 2021: Partnership with SIMAP started/validation with the Nitrogen working group March 2021: First version of the website is launched - basic html website June 2021: Partnership with Menus of Change University Research Collaborative (MCURC) October 2021: Presented at AASHE 1st with SIMAP, The Carbon and Nitrogen Footprint of Food December 2021: Conclusion of active Nitrogen Working Group participation June 2022: Master’s Thesis Finished, Title:Scope 3 Emissions from Purchased Goods: Using Food Purchasing as a Case Study for Best Practices March 2023: Partnership with World Resources Institute (WRI) Cool Food Pledge May 2023: Launched second version of the website which is easier to use April 2024: Partnership with AASHE STARS Reporting October 2024: Presented at AASHE 2nd time and launched new AASHE STARS Reporting version of TASTE Food, Presentation Title: TASTE Food: The Automated Scope 3 Tool for Tracking Emissions from Food A Python-based Categorization Tool for Calculating Institutional Indirect Emissions from Purchased Food November 2024: New connections made for upcoming future partnerships December 2024: AGU Conference Presentation, Session: Climate Crossroads: Higher Education at the Intersection of Climate Action and Innovation Session, Presentation Title: TASTE Food: The Automated Scope 3 Tool for Tracking Emissions from Food A Python-based Categorization Tool for Calculating Institutional Indirect Emissions from Purchased Food Spring 2025: TASTE Food will be used as a module within the Stanford Living Lab Course to teach students about food emissions and to be used as an example of the types of projects that can happen on campus when campus is used as a Living Lab.
Financing
Because this project was associated with a Master’s research project, the work itself was funded by a Master’s fellowship which is external to Stanford. The fellowship was one that Rebecca applied for before applying to Graduate school and which covered all aspects associated with the Master’s course work and research. The research stipend amounted to approximately $50,000/year, but TASTE Food was not the only work the Master's student worked on during this time period thus this entire stipend cannot be attributed just to TASTE Food. The entire Master's took 2.5 years to complete, with TASTE Food development occuring during the middle 1.5 years.
After the Master’s was completed, the work to update and improve the tool is being done with free time and is thus unpaid, and the computing costs are being paid out of pocket. These are not significant at this time and thus are feasible to be paid for out of pocket (computing costs are less than $1/month).
Results
The outcomes of this project were much larger than originally imagined. Initially the intended outcome was for Stanford University to have an easy way to consistently measure food purchasing and track the impact of new sustainability initiatives including more plant-based dishes in the dining halls on food purchasing emissions for the university and help estimate Scope 3 Emissions from purchased foods. Today, TASTE Food has allowed for this and much more not just at Stanford but at more than 50 universities across the United States with several city and corporate partnerships now in the pipeline for the future.
Below are more details on the outcomes of specific sub parts of the project.
The partnership with the SIMAP Nitrogen Working Group gave me access to previously hand categorized data which were used to validate the categorizations. Due to human error and differences in categorization between institutions, I was not trying to get the differences in categorizations between TASTE food and the hand categorizations to 0, and in the end found that the estimated emissions between previously hand categorized data and TASTE Food were less than plus or minus 5%. Through this process I validated that TASTE Food is effective at categorizing food purchases into up to 3 categories for multi ingredient items accurately and repeatedly in a process thousands of times faster than the previously required hand categorization.
Currently, more than 50 universities have used TASTE Food which has effectively lowered the hurdle for universities and other institutions to begin looking at emissions associated with food purchasing many of which did not have the resources for hand categorization so have only been able to start thinking about food purchasing related emissions thanks to TASTE Food. TASTE Food has also allowed universities to start setting goals for emissions reductions. The food purchasing , data processed by TASTE Food up to now represents more than 18,130+ metric tons of food purchased, 107,500+ unique items, and $124.7 million in spend.
The partnership with AASHE STARS has also significantly opened up the potential impact of TASTE Food. I worked closely with Chris Pelton to develop the functionality to be able to accurately identify plant-based food items from a purchased foods data set and calculate the percentage of spend that is attributable to plant-based foods as well as organic foods, both things which can give universities a couple of more points for their AASHE STARS reporting. This is something that without the cool could be very labor intensive to complete, and now can easily be done if the underlying data is available, taking no more than an hour to run on the website instead of up to weeks to identify them by hand.
The new partnership with the World Resources Institute (WRI) Cool Food pledge will now open the usefulness TASTE Food to more directly impact city governments and corporations who have participated in the Cool Food Pledge, significantly increasing the impact of TASTE Food.
Lessons Learned
I learned that it is incredibly important to find partners and to continually search for partners. In fact, TASTE Food would not look at all like it does today if I hadn’t found partners that were willing to share data, provide feedback, and let me bounce ideas off of. SIMAP allowed me to validate the tool beyond Stanford and realize that there was actually quite a bit of interest in a step in my research which I did not originally believe would be the main attraction. I learned to listen, be flexible, and pivot. The tool improved significantly after I sought out feedback from others who had undergone this process in the past and had data they were willing to share. I also learned that what I thought initially was a very local to Stanford issue was an obstacle at many other universities. I would not have realized this had I not reached out to other universities and found an organization that focused on food purchasing at universities. Now with the WRI partnership that is in progress, I’ve learned that there is even interest beyond the university world and there are some city governments and companies that are also interested.
I also learned to be persistent, some of the partnerships took almost two years to fully get off the ground. A “no” in the past does not necessarily mean a “no” in the future. In some cases, the organizations I reached out to were not yet at a position where a partnership would have been fruitful, but when I kept track of the emails I had sent and sent a follow up 6 months to 2 years after the initial email, I sometimes found that now the organization was more developed and a new partnership bloomed and was incredibly successful.
I also learned that making the tool accessible was very important for more widespread adoption. The very original code was written in MATLAB, a software which requires a yearly license and would have required I share code with people who had licenses to be able to use the tool. I quickly realized that writing the tool in Python would make it much more accessible as a free tool, and also allowed me to create the website where even people who do not have a coding background could use TASTE Food by simply uploading their data in a simple template. Without the free and easy to use website, the amount of adoption that would have happened I’m sure would have been much less than what I have seen to date.
Finally, I learned that it is very important to do a lot of outreach. I have presented at several conferences about this work, and more specifically at conferences where people who may want to use TASTE Food would be in attendance. I also took the time to write explanatory documents and reach out to organizations that I thought would benefit from the use and knowledge of my tool. I’ve also run short workshops and currently even have a video walk through demonstration on youtube available for anyone to watch and follow along with.