Statistical Methods Used in Forecasting

[DelcoWx]

I was surfing through the forum recently and I noticed Tony made a reference to doing a chi-square test for a forecast. I took a stats course over the summer (required for business admin/accounting majors), so it got me thinking what kinds of stats tests mets would use when forecasting.

So just out of curiosity, what statistical methods do mets use and for what purposes? 

[Rainshadow]

16 hours ago, DelcoWx said:

I was surfing through the forum recently and I noticed Tony made a reference to doing a chi-square test for a forecast. I took a stats course over the summer (required for business admin/accounting majors), so it got me thinking what kinds of stats tests mets would use when forecasting.

So just out of curiosity, what statistical methods do mets use and for what purposes? 

MOS guidance is a linear regression stat program that uses model forecast elements as predicands.  That is probably the biggest one.  We had a met who double majored as a stats major and he developed a regression technique to predict tides based on past occurrences and current model guidance. He used to also have winter/summer temperature outlooks (within a confidence interval) based on Spring & Fall Temps and occurrence of first heat wave/cold snap (he would have this winter being above avg based on how Oct & Nov have gone so far in his prelim).  He also developed pcpn intervals based on model qpf forecasts (everyone would be surprised at how wide those ranges are for a first period .50" forecast.).

[DelcoWx]

6 hours ago, Rainshadow said:

MOS guidance is a linear regression stat program that uses model forecast elements as predicands.  That is probably the biggest one.  We had a met who double majored as a stats major and he developed a regression technique to predict tides based on past occurrences and current model guidance. He used to also have winter/summer temperature outlooks (within a confidence interval) based on Spring & Fall Temps and occurrence of first heat wave/cold snap (he would have this winter being above avg based on how Oct & Nov have gone so far in his prelim).  He also developed pcpn intervals based on model qpf forecasts (everyone would be surprised at how wide those ranges are for a first period .50" forecast.).

Interesting, Tony. Thanks! 

I'll probably never approach the level of mathematical understanding that mets need to do their job, but it's really neat to have a very basic understanding of useful applications for this stuff.