16 day forecast accuracy

maddog77's picture
maddog77 started the topic in Wednesday, 12 Feb 2014 at 7:59pm

I'm just wondering how accurate your 16 day forecasts really are.. I've noticed that even your 5 day forecasts change often so i'm a bit skeptical how you could predict the surf in 16 days time.. Have you guys tested your 16 day predictions against what the surf actually ended up being like? Out of 100% accuracy, how accurate were you guys??
thx :)

thermalben's picture
thermalben's picture
thermalben Thursday, 13 Feb 2014 at 8:24am

A great question MD, and one that will probably crop up continually over the coming years.

I don't have a great deal of hard data right now but that's the next phase of our forecasting product - creating the systems to automatically verify our output so that (1) surfers can have greater confidence in our forecasts, as well as understand the limitations of all forecast products, and (2) we (internally) have a better understand of where we need to tweak the model.

Long range forecasts are very useful for some locations, and less useful for others. Their accuracy - just like any long range weather forecast product - is dependent on a wide range of factors beyond our control. Under some complex weather patterns the atmospheric models (which drive our wave model) can be wrong just a day or two ahead. 

Our forecast model is also regularly under ongoing development, of which we're hoping to launch a number of secondary products over the coming months that will give you even greater confidence as to the forecast output. One of these tools will be a verification system where we can give accuracy ratings for periods (ie last week, last month, last season) at various intervals.

For example, I'd like to be able to display something like (using random percentages):

Surf location (A), for January 2014
Forecast 2 days out = 92% accurate
Forecast 4 days out = 84% accurate
Forecast 7 days out = 72% accurate
Forecast 10 days out = 67% accurate
Forecast 16 days out = 59% accurate

In my experience, long range forecasts are best used to assess upcoming broad trends, rather than specific heights/periods/winds for specific days. So rather than using the forecast for "I've got an RDO in 15 days time, what is the surf meant to be like on that day", you're better off loooking at the overall trend leading up to that period.