ABSTRACT Production of food contributes to climate change and other forms of environmental impact. Input data used in environmental impact assessment models, such as life cycle assessment (LCA) and nutrient balance (NB) analysis, may vary due to seasonal changes, geographical conditions or socio-economic factors (i.e. natural variability). Moreover, input data may be uncertain, due to measurement errors and observational errors that exist around modelling of emissions and technical parameters (i.e. epistemic uncertainty). Although agricultural activities required for food production are prone to natural variability and epistemic uncertainty, very few case studies in LCA and NB analysis made a thorough examination of the effects of variability and uncertainty. This thesis aimed to enhance understanding the effects of variability and uncertainty on the results, by means of uncertainty and sensitivity analysis. Uncertainty analysis refers to the estimation of the uncertainty attribute of a model output using the uncertainty attributes of the model in- puts. There are three types of sensitivity analyses: (I) a local sensitivity analysis addresses what happens to the output when input parameters are changed, i.e. the intrinsic model behaviour of a parameter; (II) a screening analysis addresses what happens to the output based on the un- certainty range of the different input parameters; and (III) a global sensitivity analysis addresses how much the uncertainty around each input parameter contributes to the output variance. Both the screening analysis and the global sensitivity analysis combine the intrinsic model behaviour with the information of uncertainty around input parameters. Applying uncertainty analysis and sensitivity analysis can help to reduce the efforts for data collection, support the development of mitigation strategies and improve overall reliability, leading to more informed decision making in environmental impact assessment models. Including uncertainty in environmental impact assessment models showed that: (1) the type of uncertainty analysis or sensitivity analysis applied depends on the question to be addressed and the available information; (2) in some cases it is no longer possible to benchmark environmental performance if epistemic uncertainty is included; (3) including correlations between input parameters during uncertainty propagation will either increase or decrease output variance, which can be predicted beforehand; (4) under specific characteristics of the input parameters, ignoring correlation has a minimal effect on the model outcome. Systematically combining a local and global sensitivity analysis in environmental impact assessment models: (1) resulted in more parameters than found previously in similar studies (for the case studies discussed in this thesis); (2) allowed finding mitigation options, either based on innovations (derived from the local sensitivity analysis) or on management strategies (derived from the global sensitivity analysis); (3) showed for which parameters reliability should be improved by increasing data quality; (4) showed that reducing the (epistemic) uncertainty of the most important parameters can affect the comparison of the environmental performance.