The study aims to illustrate a method to identify important input parameters that explain most of the output variance of environmental assessment models. The method is tested for the computation of life-cycle nitrogen (N) use efficiency indicators among mixed dairy production systems in Rwanda. We performed a global sensitivity analysis, and ranked the importance of parameters based on the squared standardized regression coefficients (SRC). First the probability distributions of 126 input parameters were defined, based on primary and secondary data, which were collected from feed processors, dairy farms, dairy processing plants and slaughterhouses, and literature. Second, squared SRCs were calculated to explain the output variance of the life-cycle nitrogen use efficiency, life-cycle net nitrogen balance, and nitrogen hotspot index indicators. Results show that input parameters considered can be classified into three categories. The first category (I) includes 115 input parameters with low squared SRCs (<0.01), which are less important and can be established with default or regional averages. The second category (II) includes 5 important input parameters, with squared SRCs between 0.01 and 0.1; that can be established with country specific data. The third category (III) includes 6 input parameters with squared SRCs >0.1; that contribute most to the output variance of at least one of the life-cycle nitrogen use efficiency indicators. These most important parameters need to be established with accuracy thus require high data quality. The input parameters of category II and III include emission factors and coefficients that are specific for a region as well as activity data that are specific to the livestock production system. By carrying out such analysis during the scoping analysis, any LCA study in food sector can cut on the cost of data collection phase by focusing on input parameters that can be fixed through good practices in data collection. Further work on global life-cycle nutrient use performance will benefit from these results to generate analysis at lesser data collection cost.

Presented at LCA Food 2016, Dublin, Ireland.