Research
What’s my argument for the methodology I’m proposing?
While there are many sources of uncertainty in an LCA, the LCI is a model created to calculate an output. This enables it to be analysed as a function. If the output of this function is considered a random variable. Well known mathematical techniques, like the ‘method of moments’ are used to approximate the probability distribution function of this function. The output of the model is also used in decision support. We want to have numerical measures of importance of our input parameters that have some intuition for communication to decision makers. Gaussian approximations aim to incorporate the uncertainty of the input parameter and the sensitivity of that parameter to convey, as a single number the importance of the variable. This is a numerical representation of the ‘importance’ defined by hauschild in their uncertainty chapter.
When to use MC sampling?
Objectively, the functions evaluated in this LCI are not complex. Monte Carlo sampling may be over kill in approximating the distribution for y as a random variable. What it is useful for is implicitly dealing with covariance. We trade computing power for labour in determining covariance between input parameters.
- At the stage now where I can do MC on the fuel production process. Just need to figure out the distributions to use.
LCI and building terminology
The functional unit for the fuel production is currently kJ of energy. This probably needs to be changed, but is just a multiplicative factor (hopefully). The output of the LCI is just related to elemental flows though. I think its not too distracting to generalise the output of the LCI. To consider the LCI as a function.
I’m constantly questioning whether what I say is understandable or has been established.
Each section should build on last.