Recent model skill

[Ken]

Here's a graph showing recent performance of models. They show each model's S1 skill score for their 24hr forecasts of MSLP. The lower the S1 skill score, the better (due to the way the score is calculated from the amounts of error). EC is still number one followed by JMA and UK almost neck and neck for 2nd place.

[Karl Lijnders]

Is this graph in reference to a particularly site (ie Melbourne AWS) or a general region bound by borders etc... (Central VIC, NE VIC etc)?

[Ken]

It covers the Australian region which stretches between about 55S and 15S, and 100E and 170E. There's also others that cover the tropics and other areas.

[Karl Lijnders]

And where would the average human skill set come in at in comparison to computer modelling?? Or is that too hard to quantify and chart?

[Ken]

That's hard to quantify because these days, almost all medium range forecasts and beyond, use models as their main form of guidance together with human intervention where needed. But the reality is that (official) forecast accuracy has had a quantum leap since the days before sophisticated models were used. There's a bit of a summary at: http://www.bom.gov.au/info/ftweather/page_26.shtml

There is one "direct" comparison that comes straight to mind though, and that's official min/max temp forecasts vs OCF. Stats have shown that OCF consistently outperformed official min/max temp forecasts at Day 1 overall (and from memory, I think it's also outperformed official forecasts for one of these parameters out to 7 days overall).