Ok, I've read the whole article. Interesting. I don't think he is entirely down on the SPC or Mesoscale Analysis tools, but I suppose wants everyone to keep perspective on what they represent. I agree they aren't necessarily 'forecasting' tools as any value in that regard needs to be proven experimentally. However I do believe they are reasonably good diagnostic tools. True some have limitations and it is up to us to know what they are. I think their true value though is in the diagnostic realm after a forecast is made and visually assessing the progress of that forecast. As stated, and with regard to sparseness of data sampling input and errors with modeling, etc there will always be assumptions made that may not always be 100% correct. The quality of the diagnostic likely will vary based on the particular scenario and the particular diagnostic. Often I tend to monitor trends with them. For instance temp, dewpoint, moisture advection increasing and moving to a particular spot or direction. Growth and concentration of STP, SCP, Sig Svr, EHI, convergence and lack of Cinh at a particular location. I agree that experimentally verifying their accuracy is a good idea. Perhaps this could be done automatically using a large database of historical data including tornado reports and just running the data through it retroactively. There are similar analysis systems for stock trading that go backward in time and compare a statistic, method, or variable for it's utility in predicting future stock price based on results.
I don't quite agree with him on combined variables / indices. The weather scientist in him is forcing him to think that if they don't directly relate to proven physical principles or processes then they are of no value. To some degree this can be true if you are trying to have them accurate in 100% cases. Take for example CAPE. Now perhaps it isn't proven that it is real or reliable. Sure tornadoes can happen with low or maybe even no CAPE, but does that mean that it is worthless? Far from it IMO. We know that in most tornadic cases CAPE is a usuable value as a future indicator of potential problems. I hear the same thing about EHI all the time, but I love that statistic and a most of the time it works fairly well as a rough guide to where you could have severe. Notice I said COULD and not WILL. Anyway I think you can combine different statistics and create something useful. A simple tornado parameter to help diagnose a large percentage of tornadoes would be to just take indice values that historically have been maximized in areas of tornadic outbreaks and combine them in a rational way that makes sense. These might take into account cape, lifted indices, shear at various levels, moisture, cinh, etc. It's tough to have one of these catch all tornado scenarios, but if it starts reading very high numbers you might want to check out why? Hmm, I'd kind of like to try my own hand one day at trying to build my own parameter. Hopefully the idea is with any of these parameters and combined indices that met scientists will continue to research and study and refine what these parameters do, and how good of a job they do at diagnosing or forecasting severe weather. Perhaps you need different tornado indices to reflect all the different ways tornadoes can form and the different type (ex cold core versus classic warm front/dryline). Same with supercells and supercell type, etc.
Personally I think weather modeling is proceding well, but sounds like we need finer (more) data input and ways need to be developed for computer modeling programs to recognise the same things that human forecasters look for when making a forecast (such as boundary analysis). If the input grid of data is fine enough, boundary analysis is of high enough quality, then all that remains is to learn and apply all the physical process interactions that the atmosphere can perform and have it combine that into a type of forecast or at least model output products. Probably statistics based on atmospheric simulation and observation / rule based would be the best way to model the output. However as the future progresses perhaps quantum physics / benefits of quantum computing, and principles related to uncertainty would help to anticipate Earth's complicated environment.
Hmmm.....ok, sounds like I've been rambling for awhile. Sorry, but this is pretty cool stuff.
