Why food and beverage LCAs are difficult to do
Compared to several other sectors, such as the packaging, household and personal care, transport and building sectors, food and beverage LCA are only just catching up. What is it about the food and beverage sector that makes it so difficult to do good LCAs?
Food and beverage LCA is booming
In the spring of 2014, several activities around food and beverage LCA were celebrated. To name a few: the start of the second wave of food and beverage PEFCR pilots, the release of the Agri-footprint database in SimaPro, and the taking place of the LCA Agri-food Asia 2014 conference. More recently, the LCA Food 2014 conference was held in San Francisco. There is no doubt that the food and beverage sector has adopted LCA as an important tool for measuring and reducing environmental impact. But why has this taken so much time?
Food and beverage LCAs are complex
Doing LCAs of products that involve little or no agricultural inputs – such as houses, cars, or cell phones – is relatively easy. The only real challenge is deciding how to divide the recycling burdens and benefits. For agricultural products, however, there are several methodological challenges. To name a few:
- Deforestation: who is to blame?
- Functionality: what is the main function of a food or beverage product?
- Multi-output processes: how do we allocate upstream environmental effects when co-products have different functionalities? And what about crop rotations?
- Emissions from crops and animals: how should we measure them?
- System boundaries: where do we draw the border between nature and a crop or animal production system?
- Regionalisation of water scarcity: where can you get data and tools that are detailed enough?
- Carbon balance: how do we account for carbon capture and release in different forms, and how do we account for the timing of these processes?
Margarine: a practical example
Let’s look at the example of margarine. The main ingredients of margarine are vegetable oils. Vegetable oils mainly come from rapeseed/canola, soybeans, oil palms, and sunflower seeds, collectively known as oilseeds. Soybeans and oil palms are known for being grown in ecologically fragile areas such as the Amazon and Borneo. But soybeans and oil palm production is not the only economic activity that impacts these areas. What about wood harvest as a driver of deforestation, what about other land use activities in the area, such as livestock rearing, and what about reforestation…? Once you answer these questions, you face another issue: you need to quantify important emissions, such as the direct and indirect formation of nitrous oxide from fertilisers and the emission of pesticide ingredients to the soil, water and air. These are virtually impossible to measure, because you can’t put a wall around a system to measure its influxes and outfluxes without disturbing its processes. And even if you could – how would you decide where the system stops and nature begins?
It can be difficult to model LCAs of the various oilseeds at the farm, but even after they leave the farm, the difficulty remains: we’re still dealing with a beautifully complex organism. Oilseeds are brutally crushed and then divided into the oils for our margarine and a protein-rich fraction, oilseed meal, which is generally used as animal feed. Compound feed manufacturers are willing to pay a good price for oilseed meal. This means that – from a market perspective – you cannot talk about oil as the main product of oilseed crushing. Applying the system expansion method to avoid allocation of upstream environmental impact is possible, but complex. For example, the production of the co-product of soybean oil, soybean meal, avoids the production of rapeseed meal, which means that the upstream impact of rapeseed meal is subtracted from the upstream impact of soybean oil. However, rapeseed meal is co-produced with rapeseed oil, which avoids the production of sunflower seed oil and sunflower seed meal avoids the production of palm kernel meal, et cetera. The complex chain of avoided co-products can only be solved mathematically with a lot of difficulty. Alternatively, you can allocate upstream impact by the fat content or the protein content, but then all impact is allocated to either oil or meal. Allocating by energy content or dry mass may be a solution, but these characteristics are not really representative for the different functions of the co-products.
Talking about functionality is interesting when it comes to food. What is the actual purpose of eating margarine? Is it flavour? Is it nutritional value, and if so, which nutrient are we talking about? Does the texture of margarine make a sandwich more pleasant to eat? Or do people use margarine as an essential ingredient for making pastry? Everyone may have different answers to these questions. Many types or brands of margarine are out there, each fulfilling a different need. There’s margarine for kids, for seniors, for making a sandwich or for baking a cake. This variety of function often makes it useless to compare the LCA results of two food or beverage products, even if they are very similar.
Moreover, if the studies are performed by two different LCA experts, you can imagine that the results may be very different if the two experts make different decisions in solving the methodological challenges I mentioned. Even choosing different databases for your LCA can have a large impact. For example, the carbon footprint of soybean oil from Brazil differs by a factor of ten depending on whether you’re looking at “Agri footprint economic allocation” or “ecoinvent 3.1 cut-off by classification” (see the figure below).
Continuing developments in food and beverage LCAs
Now that you’re aware of the challenges, let me give you an overview of several continuing developments in food and beverage LCA. The European Food SCP Round Table started in 2010 and has done a lot of work that is very useful for developing the product environmental footprint category rules (PEFCR) for the food and beverage product categories. One of the results of this round table is the ENVIFOOD protocol, which gives valuable guidelines on how to approach the challenges mentioned above and others. The Food and Agricultural Organization of the UN (FAO) started a similar initiative in 2012, working on the global level and limited to livestock products. This initiative is called LEAP, because of its goal to make the sustainability metrics of livestock supply chains take a leap forward.
There are also several initiatives developing life cycle inventory databases that specialise in the food and beverage sector. As mentioned before, the Agri-footprint database by Blonk Consultants was released earlier this year. Early next year, an important update for this database will be released, with improved datasets and more products. In France, ADEME developed the Agribalyse database, a good source of background data and sensitivity analyses for food and beverage LCA studies. The last database initiative I want to mention is the World Food LCA database, founded by the Swiss Confederation, Agroscope and Quantis. It is not yet available outside the partner consortium, but will be really valuable when it is, because it is being co-developed by several important players in the sector, such as Kraft and Nestle.
Now the food and beverage LCA initiatives and databases are maturing, it is time for some serious decisions on how to deal with the extreme complexity of food and beverage LCA. The technical secretariats of the food and beverage pilots in the Environmental Footprint project and the working group on cow-related issues, have been discussing the challenges behind closed doors, but several technical secretariats have already published their drafts on the scope of the product category rules. This can already be called a milestone, because now the results of the discussions in the Food SCP Round Table and LEAP initiatives are finally being shaped into concrete rules. It may not, after all, be unfeasible to get consensus on the issues related to food and beverage LCA.
Tommie worked for PRé as a Technical Consultant from 2012 until 2015. As a part of the Consultancy Team, he worked with databases and methods. Tommie collaborated in projects such as Prosuite and improving the ReCiPe method.