Uncertainty in Environmental Assessment for the Built Environment

Environmental assessment of the built environment has generally used spatially and temporally insensitive evaluation methods that do not include uncertainty. Aggregated mass emission values in a pollutant inventory do not consider the distribution nor the characteristics of the context, resulting in a loss of information regarding their location, and emission rate. Hence, the impact valuation methods that use such inventories cannot include these aspects. Although modeling spatial and temporal emission distributions in the context of a life cycle assessment (LCA) can be data intensive, judgements regarding the intensity, duration, and spatial distribution of emission releases should be considered in impact analysis. Furthermore, evaluating uncertainty in a LCA has been acknowledged as an important aspect of the analysis. Data regarding construction, operation and maintenance processes, properties of substances, ambient concentration limits, and manufacturing processes of functionally equivalent systems or materials are all variable and uncertain. In order to enable a more comprehensive evaluation of environmental impact of the built environment life cycle, appropriate methods and models implemented in computational design and evaluation tool is required. A multi-aspect simulation-based approach using multiple models for environmental assessment is presented along with demonstrative analysis results. The computational tool uses distributed simulation modules for projecting operational effects, and includes models of geophysical systems and evaluative environments to estimate the temporal and the spatial distribution of an environmental load for use in the indicator calculation. The impact valuation method is an extension of a traditional material-focused LCA, and is based on an allocation for emissions and resource usage. The recursive Monte Carlo sampling is based on a pseudo-random number generator, and the probability distributions are primarily log-normal and normal. Uncertain parameters are described by a mean value, standard deviation, and distribution.