Maximising protein production by avoiding food-feed competition - parameterisation of feeds in a Swedish food systems model

Abstract

A major inefficiency in contemporary agriculture arises from food-feed competition, where limited agricultural resources are used to produce animal feed instead of human-edible crops. Due to factors like metabolic losses and waste, this is often less resource efficient. This thesis investigates the optimal number of cattle in the Swedish agricultural system when aiming to maximise protein production. To realise this, the biophysical food system model CIBUSmod was extended to parametrise feed rations, enabling modeling of food-feed competition across four strategic scenarios.

In collaboration with model stakeholders, maintainability and performance were selected as evaluation criteria for the changes made to the CIBUSmod software. The scenario development indicate that substantial increases in protein output could be achieved by optimising feed rations and reducing cattle numbers in favor of cultivating wheat and grain legumes. However, the findings also suggest that cattle do not necessarily need to be entirely removed from the Swedish agricultural production system.

The scenario modifications resulted in a 23-fold increase in problem size, yet only a 4-fold increase in execution time. Concurrently, iterative improvements were made to enhance the maintainability of the model framework’s optimisation module. Overall, the results challenge the narratives that increasing protein production in Sweden requires expanding cattle production. The extended version of CIBUSmod also provides a foundation for further explorations of food systems and food-feed competition.

Popular Abstract

Producing sufficient amounts of nutritious food without harmful environmental impacts, such as unsustainable emissions of greenhouse gases or destruction of biodiversity, constitutes a significant challenge. One key inefficiency contributing to this challenge is the use of finite agricultural resources, like land or fertiliser, to cultivate feed for animals rather than food for humans, a dynamic called food-feed competition. In many cases, growing human food directly is more resource efficient, as animal production incurs e.g. metabolic losses and feed-storage losses. Globally, over a third of all calories produced from growing crops are used for animal feed, a number which varies across regions (e.g., the corresponding value in the USA is over two-thirds).

In this thesis, I set out to explore how many cattle is optimal in the Swedish agricultural system, if we aim to produce the maximum amount of protein. Protein was chosen as the key indicator due to its strong association with animal products such as meat and dairy products. Across four scenarios, the production of cattle was compared to two crops: spring wheat and peas.

However, at the outset of this thesis, no Swedish food system model with the capability to optimise feed rations existed, i.e., allocating feed compositions in light of available agricultural resources (or, in other words, models with a parameterisation of feed rations). Hence, a major component of this thesis concerned extending the existing computational model CIBUSmod with a parametrisation of feed rations.

As such, the thesis has a dual aim: to strategically explore the optimal number of cattle in the Swedish agricultural system when optimising for maximum protein production, and to extend the model CIBUSmod to enable parameterisation of feeds. To ensure that the changes made to the software that constitutes CIBUSmod contribute positively, two evaluation criteria — maintainability and performance — were selected, and the changes made to the model iteratively evaluated against these.

The results suggest that significantly fewer cattle — but not necessarily zero — are optimal to produce the maximum amount of protein in the Swedish agricultural system. Furthermore, the protein-efficiency of cattle seems to be dependent on the presence of by-products in order to be more efficient than the two selected crops. The exact number of cattle, as well as the distribution of dairy- compared to beef cattle, changes between scenarios, highlighting the need for further and diverse scenario development. The modifications made to CIBUSmod increased the time required to set up the model by a factor of four, despite a problem size that increased by a factor of 23. Additionally, efforts were made to keep the ease at which the software system can be modified in the future as high as possible.

The thesis invites discussion on the assumptions made in modelling the scenarios. Recent literature has increasingly challenged protein being used as a key indicator when exploring food production in wealthy countries, as protein deficiency is rare, and not a result of insufficient supply. Modelling of production of a range of nutrients would introduce significant additional complexity, but also produce more nuanced results. Moreover, the results would benefit from inclusion of socio-economic perspectives, like impacts on livelihoods, as well as more extensive ecological indicators, such as impacts on biodiversity. Yet, I contend, the contributions of this thesis may provide a relevant addition to debates of what an increased food production can entail, and the modifications to CIBUSmod enable further exploration of food-feed competition across diverse scenarios.

• author: Wanecek, Wilhelm
• supervisor: Björn Landfeldt; Johan O. Karlsson; Elin Röös
• organization: Department of Electrical and Information Technology
• course: EITM01 20242
• year: 2025
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: Scenario modelling, Sweden, Design Science Research, sustainability, CIBUSmod, evaluation criteria, food production, agriculture
• report number: LU/LTH-EIT 2025-1048
• language: English
• id: 9187828
• date added to LUP: 2025-05-08 10:32:58
• date last changed: 2025-05-08 13:14:29