FoodEmissions

Context — Background

Multiple sources brought this dataset to life. I initially encountered this data through the dataset posted by user Vivek on Kaggle, Environment impact of food production, which identifies Our World in Data as its source.

I searched on Our World in Data (OWID) and found their series Environmental impacts of food production, which involves thoroughly researched data storytelling, compelling data visualizations, and the option to download the data behind each visualization as a separate .csv file. OWID also identifies the original source of published data: a study published in Science in 2018 in an article, Reducing food’s environmental impacts through producers and consumers by J. Poore and T. Nemecek.

I read the Science article in full and downloaded the supplementary material, including the original dataset aaq0216_datas2.xls which I found on beneath the 'Resources' section via the same page as linked above, Reducing food’s environmental impacts through producers and consumers.

When I downloaded the original dataset from Science I found it contained more precise values for each food product's global average GHG emissions than in the .csv files available for download from OWID (and subsequently from Kaggle user Vivek) include the same 'global mean average GHG emissions' I sought, but with values that are much more rounded.

As an example, let's take a look at the global average (mean) GHG emissions (kg CO2e per kg produced) from the Land Use Change (LUC) stage for Apples :
Value in the original dataset published in Science: -0.029 kg CO2e per kg produced
Value in the secondary dataset from Our World in Data: 0 kg CO2e per kg produced

What appears to be an unintentional rounding error can be found throughout OWID's downloadable .csv files, and subsequently the file I first found on Kaggle.

I chose to integrate the original source data (without rounding error) with the clarified structure of Our World in Data which was uploaded to Kaggle by user Vivek.

Content

For some of the most common food products around the world, Poore & Nemecek studied the greenhouse gas (GHG) emissions per stage of the food value chain from land (land use change or LUC) to retail (end-user purchase/acquisition point).

The stages identified from Land to Retail are commonly used to describe stages of global food systems. Note that variation occurs from one food product to the next and within a single food product, for reasons including the region and ecological profile of the system of production, size and type of production system, and more.

To maintain consistency with the format and content of data from OWID, and my selected focus of GHG emissions across stages of the global food system, I decided to narrow the focus of this dataset. It is limited to the global average (mean) GHG emissions per stage.

The original source of data (see above for context) provides multiple measurements such as global median GHG emissions, in addition to global mean GHG emissions. For the sake of simplicity and consistency with the dataset Environment impact of food production on Kaggle and its source, Environmental impacts of food production on OWID, I opted to use the global mean (average) for each food product and stage. (Note that this particular dataset does not include the global average GHG emissions for Consumer or Waste stages of global food systems.)

GHG emissions are measured in kg CO2e per kg of food produced.

• kg CO2e = kilograms of carbon dioxide equivalents
  Each food product is listed with global average GHG emissions from each stage of the food value chain from Land, Farm, Animal Feed, Processing, Transport, Packaging, up to Retail.