This note reports a software test carried out from July to August 1997 for the Danish Environmental Protection Agency. It examines the statistical functionality available in beta-stage LCA software, rather than assessing commercial LCA software on the market. The authors tested the Danish EPA LCV system, KCL-EKO, TEAM and TEAMPlus, and PEMS. EPS and SimaTool were identified but could not be included because the relevant versions were unavailable.
All four tools were tested on the same computer using a deliberately simple life cycle comparison of a rapeseed-based lubricant and a mineral-oil-based lubricant. The case included only the most important unit processes and environmental exchanges for the test. This common example provided a practical basis for examining how each program handled uncertainty, simulation, output and software stability.
The review considered whether users could enter uncertainty for process exchanges and other data, which uncertainty formats and distributions were supported, and how easily inputs could be prepared. It also examined calculation methods, the number and duration of simulations, the way uncertainty results were reported, whether covariance could be represented, and whether data and results could be exported or traced to their contributing processes. The authors therefore compared the practical workflow and reporting support, not just whether a program offered a feature labelled “statistical analysis”.
The software differed substantially in what it could analyse. LCV and TEAM supported Monte Carlo simulation but had limitations in reliability reporting and in calculating uncertainty for comparisons between product systems. KCL-EKO offered quick and exact simulation methods, with the report noting accuracy and calculation-time trade-offs. PEMS provided interval-based sensitivity comparisons, but did not perform statistical calculations. The beta version of LCV was unstable in the test and could not complete more than 18 simulations for the case.
The authors could not make a direct comparison of the four tools’ numerical results because their approaches differed. Although LCV and KCL-EKO results should in principle have been comparable, LCV’s instability prevented enough simulations to produce a stable result. The conclusions describe software versions and a small test case from 1997; they should not be read as an evaluation of current LCA software or as a present-day ranking.
