If 200 g of given food item creates 1 kg of CO₂, is that good or bad? Grades are how we answer this question in an approachable way.

If a recipe creates 2.0 kg of GHG impact, but it would yield only 1.0 kg if its ingredients were replaced by efficient baseline items, then we can say its efficiency is 50%. In other words, this food delivers its constituent elements (e.g. protein, carbs) half as well as an efficient baseline food product's ability to provide those same elements.

Armed with the efficiency of a given recipe, we then reduce it to 4 grades:

• 

  50% or greater efficiency

• 

  25-50% efficiency

• 

  12.5-25% efficiency

• 

  Below 12.5% efficiency
• These thresholds are heuristic, and we may modify them over time.

The categories are Starches (carbohydrates), Proteins, Fruits & Vegetables, Root Vegetables, Oils, Refined Sugars, Caffeine, and Cocoa. To set the baselines, we use the most GHG efficient food source in each category. If a category has more than 5 distinct items in it, we use the second best one. This is done to mitigate extreme data points in larger categories. For categories with only a single impact datapoint (e.g. Cocoa), or for which an in-category comparison doesn't make sense, we use a mass-based impact baseline selected from the mass-based impacts of the entire dataset. As this comparison set has more than 5 distinct items, we use the second best one. This is done to provide a reasonably proxy for: Cocoa/Chocolate, Beer, Wine, and Liquor. We may bring our approach to alcohol more in line with the other groups in the future, although the merits of treating alcohol or cocoa as distinct food groups is debatable.

The comparables we show on the GHG Calculator are derived from the CO₂ Equivalent as follows:

• Gas

  Gasoline generates 2.3 kg of CO₂ per litre. To calculate the equivalent gasoline usage per kg of CO₂ output, we divide it by 2.3. For example, if a recipe generates 6.0 kg of CO₂, it's the equivalent of combusting 6.0/2.3 = 2.6 litres of gasoline.

• Driving

  The 2018 average fuel economy in the United States was 24.9 mpg, or .0945 l/km. Continuing our example from above, those 2.6 litres of gas represent driving 2.6/9.45 * 100 = 27.6 kilometres.

• Global Per Capita

  Global CO₂ emissions were 33.1 Gigatons in 2018.

  CO₂ represents 76% of total GHG emissions, so total global GHG emissions were an estimated 43.6 Gigatons.

  Food represents 26% of global GHG emissions, or 11.3 Gigatons.

  Given a global population of 7.6 billion, per capita daily food-related GHG emissions are 4.1 kg.

• Regional Per Capita

  North America generates about 3-3.5x the global per capita CO₂ average, and Europe is approximately 2x. While their food shares of GHG may deviate from the global average, we estimate the per capita food related emissions to be 12.3 kg in North America, and 8.2 kg in Europe.

There are four primary anthropogenic Greenhouse Gas (GHG) categories: Carbon Dioxide (CO₂), Methane (CH₄), Nitrous Oxide (N₂O), and Fluorine compounds (CF₆, C₂F₆, CHF₃, CF₃CH₂F, CH₃CHF₂, and SF₆), with CO₂ accounting for about 76% of the total impact. Because the impact of the others differs from CO₂, when quantifying GHGs it is standard to refer to their CO₂ equivalent, which is the mass of CO₂ required to have the same greenhouse effect. We use the term GHG impact interchangeably with CO₂ equivalent.

Given the quantities of each ingredient in a particular food, you can calculate the GHG impact of that food simply by multiplying each item's mass (in grams) by the mass of CO₂ equivalent (per gram) associated with the production of that item, and then summing those up. So for instance, if you're eating 200 g of pasta, and pasta's CO₂E is 1.57 g/g, then the CO₂E for that pasta is 200 * 1.57 = 314 g. In plain English, producing that pasta generated 314 grams of GHG emissions.

To calculate this for the variety of things that people eat and how they are produced, we require 2 distinct datasets. We call the first Products, and it is a database of food ingredients. The second is Origins, a database of the GHG emissions associated with different Products based on where and how they were produced.

Our Products dataset includes the names (and alternate names), densities, food energy, and typical unit size for about 1,000 different basic ingredients like corn, beef, wheat, milk and butter. From this information The Products API is able to resolve ingredient values in any unit down to a mass in grams. We are continually expanding both the breadth and specificity of the Product database.

It is worth noting that for the purposes of GHG calculations, many ingredients are insignificant due to the minimal quantities in which they are used. We are primarily focused on understanding the foods with the largest impacts, as well as categories with diverse impact (such as seafood).

Our food energy (i.e. joules per unit of mass) values are generally sourced from the USDA's FoodData Central, and food densities are generally sourced from the Food and Agriculture Organization of the United Nations.

Origins represent where a product came from, either a region or a specific producer. Within an Origin, the Products it produces (for which we have data) each have an entry describing the GHG impact of producing that Product in that Origin. Our primary data source for this is Reducing food’s environmental impacts through producers and consumers (Science, 2018), which is a synthesis of over 1500 individual research articles covering over 38,000 farms. Currently, we default to the Global Origin, although for beef, cheese, and milk we have a separate Origin for North America.

The Origin impact data does not include any additional emissions associated with the preparation or storage of food once it leaves retail.

We are continually drilling down to more granular Origins. Populating an Origin with a given product requires published research covering a material percentage of its output of that product. For instance, if we have only one study covering the GHG impact of 2% of milk production in France, that's insufficient to populate a France Origin for milk. Conversely, with 23 distinct datapoints for North American beef, we are able to separate it from the global average.

While it is not currently within our capacity, eventually we hope to enable individual food producers to opt into an audit process to provide the most specific assessments possible. Having said that, the reality is that foods that today are problematic will almost certainly remain so regardless of where or how they are produced. The big drivers are less about farming technique than the fundamentals of input requirements, biological processes, and yields.