<< Click to Display Table of Contents >> Navigation:  Tutorials > Artefacts and stochastic realisations > Input data quality

There is a maxim about computer models: Garbage in, Garbage out.  However, because of the stochastic nature of PoreXpert, coupled with the considerations described in the previous section, for PoreXpert Garbage in can result in amplified Garbage out.  The software is extremely sensitive to changes in the input data, and errors in any input data are likely to be magnified in the output.  So the input data has to be of the highest quality - much higher, in fact, than normally recommended by manufacturers.  Suppliers of mercury porosimeters, for example, often trumpet the speed of their instruments - but fast intrusion can result in mercury heating effects which bulk up the intrusion and can cause physically impossible, and hence un-modellable, bulges in the intrusion characteristics.  Similarly for water retention data - these must be of the highest quality.  Our frustrating experience of that, when we first started out modelling soil, was to trust one of England's national centres of excellence.  The member of staff there guessed one of the experimental points, while he was with his wife while she had a baby.  However, he guessed that the soil would not release any more water during a tension increment, which was wrong, and we wasted several months trying to model the data before we heard the truth. We have since collaborated with the world-leading, and entirely trustworthy, Rothamsted Research. So, be sure to check your data, having read, if relevant, the further discussion about mercury porosimetry or water retention.  

 

If your input to PoreXpert is from mercury porosimetry or porometry, now move on to a consideration of void size range. For water retention and geometric input, skip to stochastic realisations.