Evelyne Groen

1. Background

Food is essential to sustain and enhance human life. Nowadays, food production is generally acknowledged as one of the drivers of environmental pressure [Tilman and Clark, 2014]. Agricultural land, for example, covers approximately one third of the earth’s land that is not permanently frozen [Bajželj et al., 2014]. The majority of this land is used for production of animal-source food [Steinfeld et al., 2006]. The agricultural sector not only uses natural resources, such as land, energy, water and fossil phosphorus, but also contributes to climate change, acidification, water pollution and biodiversity loss. The global livestock sector, for example, contributes to approximately 14.5% of all anthropogenic greenhouse gas emissions, mainly via emissions of carbon dioxide (CO₂), methane (CH₄), and nitrous oxide (N₂O) [Gerber et al., 2013] . An increase in the demand for food products is still expected, because of the growth of the global human population, growing incomes and changes in diets. The challenge to produce this food in an environmentally friendly way, therefore, becomes even more urgent. To develop strategies to produce food with a low environmental impact, the environmental impact of food needs to be quantified.

Quantifying the environmental impact of agricultural activities, such as crop or livestock production, requires great care, since the input data naturally vary due to, for example, seasonal changes, geographical conditions and socio-economic factors. Moreover, uncertainties, such as measurement errors in observations, exist around modeling of emissions and technical input parameters. Once you acknowledge that variability and uncertainties exist, questions arise such as: how does variation of input parameters influence the results? Which input parameters explain most of the variation of the model output? Do input uncertainties have an effect on the comparison between the environmental impacts of two products? To answer the questions raised by the acknowledgement of variability and uncertainty, one is naturally led to the field of uncertainty and sensitivity analysis.

2. Quantification of environmental impact

Quantification of environmental impact originating from the production and use of materials, substances or products from a chain perspective, based on physical principles, can be divided into two main approaches [Wrisberg et al., 2002]:

1. Region-oriented approaches, where the system boundary of a specific area and time interval (e.g. The Netherlands, 01/01/2015 until 31/12/2015) is the starting point to determine which activities are included. Region-oriented approaches can concern a specific year and economic sector, or all processes controlled by one company, such as a farm. Four examples of region-oriented approaches are:
1. material flow analysis, concerning material flows through a region or company, e.g. iron or plastics;
2. substance flow analysis, concerning specific elements such as nitrogen or phosphorus;
3. nutrient balance analysis, concerning input and output of nutrients or substances through a region or a farm; and
4. environmental input-output analysis, focusing on pollution of industrial trade of, e.g. a country.
2. Function-oriented approaches, where the function of the product is the starting point and defines the system boundary, ideally from cradle-to-grave. Function-oriented approaches can be time independent, but often consider the production of the final product for a specific year, and only processes from power plants, companies or farms are considered that belong to the studied production chain. Examples of function-oriented approaches are:
1. life cycle assessment, concerning the environmentnal impact of a product over its life cycle; and
2. cumulative energy requirement analysis, concerning the energy requirements of a product.

A region-oriented approach widely used in agriculture is a nutrient balance analysis (NBA). A nutrient balance is computed as the difference between nutrients entering a system and leaving the system; the system itself is considered as a black box. The difference between nutrient inputs and outputs is generally called the nutrient surplus, and is assumed lost to the environment [Sutton et al., 2013]. Nutrient losses contribute to, for example, eutrophication, acidification, emission of greenhouse gas emissions [Sutton et al., 2013], and depletion of fossil phosphorus [Suh and Yee, 2011]. A nutrient balance can be computed at a range of levels, varying from crop or animal, to farm or, for example, the entire European Union [Aarts et al., 1992; Schröder et al., 2003; Lesschen et al., 2011]. Nutrients generally considered in agriculture are nitrogen and phosphorus, because these nutrients are essential for plant growth. A nutrient balance can be used to identify where and why losses occur, and to develop mitigation strategies to reduce these losses.

A function-oriented approach widely used in agriculture is life cycle assessment (LCA), which quantifies relevant resource use and emissions of a product over the entire production chain. LCA can be used to quantify all different types of environmental impacts (e.g. eutrophication, climate change, eco-toxicity), and is increasingly applied in agriculture [Guinée et al., 2010]. An environmental impact that is often considered related to livestock production is climate change, due to the large contribution of livestock to the global anthropogenic emissions of greenhouse gases [Gerber et al., 2013]. In such cases, one often speaks of the carbon footprint of an agricultural activity or product. An LCA can be used to identify where in the chain most emissions occur, and develop mitigation strategies to reduce these emissions.

3. Case study

There are two kind of models studied in this thesis: life cycle assessment and nutrient balance analysis. For each model, a (simplified) case study is given as an illustration. For both case studies, their matrix-based equivalents are used, to facilitate the implementation of the methods for uncertainty propagation and sensitivity analysis later on.

The Matlab and Python scripts provided on this website, to peform uncertainty propagation and sensitivity analysis, are applied to the (renowned) case study, first introduced by Heijungs and Suh [2002]. The case study respresents the production of electricity, but is a drastic simplification. Figure 1 shows a flow diagram of the case study.

The flows 1 to 6 in Figure 1 represent: (1): 10 kWh electricity; (2): 2 liter diesel; (3) 0 kWh electricity (there is no electricity used during the production of diesel); (4) 100 liter diesel; (5) 1 kg CO₂; (6) 10 kg CO₂.

The flow diagram can be converted to a matrix-based format containing two matrices: an A-matrix containing the production processes (flow 1 to 4) and a B-matrix containing the emissions (flow 5 and 6). For a complete introduction I would like to refer you to Heijungs and Suh [2002]. In case of a total demand of 1000 kWh electricity, the total emissions equal 120 kg CO₂.

---

Source: PhD thesis Evelyne Groen, An uncertain climate: the value of uncertainty and sensitivity analysis in environmental impact assessment of food, 2016

ISBN: 978-94-6257-755-8; DOI: 10.18174/375497

---