There's a nice quote in Koester, 2008 on p. 309: "A human-driven consensus provides adequate and useful planning information. Only fools would trust a purely computer-derived POA map for the land environment. However, when used together, the best predictive result may be found". SMT is able to calculate initial POA values for segments from a human from consensus, or from a statistical model for distances subjects are found from the IPP.
Need to provide a method to allow users to set initial POA values from a merged combination of the human consensus and the statistical distances.
Best long term approach might be to use a Bayesian network as was done in the proof of concept SARBayes Probability Mapper, incorporating statisitcal distances, track offsets, etc with human consensus input. Not clear what weighting should be used - probably an area for research, so start with a simple weight - calculate initial POA for a segment as (humanPOAweight)+ (distancePOA()(1-weight)) where weight is between 0 and 1.
UI might be a wizard. Probably sets requirement to store history of probability distributions.
First cut at merging human consensus and distance from IPP statistics into POA initial added in revision [r78]. Working merge, but fragile workflow - easy for users to encounter problems if steps are not carried out in the right order. Also points to the need to store Mattson and other consensus data sets, and the need to develop a probability distribution model history.
Related
Commit: [r78]
Cleaned up inconsistencies in behavior, provided explicit reporting to users of whether or not prerequisites for providing probabilities have been met or not, and made process less workflow step order dependent in [r79]
Related
Commit: [r79]