Honest feedback

bjc's picture
bjc started the topic in Thursday, 24 Oct 2013 at 8:24am

Hi Guys

The forecast for Sydney NB last night was for 1 foot. Basically flat and short period.

Report from swell et this morning:

Looking very fun this morning. We've got a peaky swell combo from the south and north-east, creating very nice waves in the 2ft+ range right across the open beaches. Winds are light offshore so conditions are smooth and it seems that most beaches will be delivering the goods this morning. Get in early before something changes!

Quiet a bit of disparity and led to me opting for the sleep in. Bugger.

I much prefer a localised forecast which has some written commentary to help interpret the data you are providing and also instill more confidence in the forecast.

I think I will stick to coastalwatch's Hurley app now as it provides this. Shame as I thought you guys were a great forecast site alongside CW.

thermalben's picture
thermalben's picture
thermalben commented Thursday, 24 Oct 2013 at 10:55am

Thanks for the feedback. The text forecasts are only temporarily off the site, and will return in the next week or two (we're a very small company with limited resources during an important event such as a major website relaunch).

As for the accuracy of the model forecasts - we're extremely happy with their output (having monitored them closely over the last couple of years).

That being said, they're not perfect - seems to be somewhere around 85% accuracy for swell events between 2ft and 10ft - but we are constantly refining the algorithms to make sure they don't miss anything at all. We surf every day too, so we're acutely aware when we (or the models) get it wrong, and we're always working to improve the product.

Unfortunately, the surf report system is only as good as the input data, and there are several scenarios that the WW3 model frequently misses (we'll detail these in another forum post). A classic example of this is the small local NE windswells either side of 2ft, and short range S/SE thru' SE swells. Under these circumstances you'll find that every model-derived surf forecast will also miss the mark.

Anyway, thanks for the honest feedback - it's what we want to hear as we continue to improve the service.

thermalben's picture
thermalben's picture
thermalben commented Thursday, 24 Oct 2013 at 11:15am

Also, it's worth me explaining why there was a disparity in the forecasts between last night and tonight.

Today's surf seems to be a mix of small NE windswell (from yesterday) and building S'ly swell from a weak front that entered the Tasman Sea last night.

Yesterday, the computer models didn't expect much strength in the southerly flow behind the front, so as a result they didn't pick up much of an increase in projected surf size for today. When looking at the Northern Beaches forecast yesterday afternoon - for today - the 6am forecast had 1ft (NE swell), building to 1-2ft in the afternoon with a new S'ly swell.

However, overnight the models updated and resolved the fetch behind the front a little better. This 'upgraded' the surf forecast overnight, meaning that - if you looked at the website just after midnight - today's 6am forecast was 1-2ft at south facing beaches, building to 2ft during the day. Only a small upgrade, but an upgrade nevetheless.

The model has updated again since then and resolved the southerly flow a little better, and has subsequently upgraded the surf forecast from 1-2ft this morning to 2-3ft this afternoon at south facing beaches. This is still below what eventuated (1-2ft forecast vs 2ft+ reported this morning, and it'll likely be a touch bigger than the 2-3ft forecast for this afternoon) however you can see how the system works.

Obviously, the problem here is that the user is not aware of these upgrades behind the scenes. That's something we've been working on for a while, which we'll launch soon - the ability to be alerted if the short term forecast is upgraded or downgraded, within specific parameters (ie "send me a message if the surf is going to be much better/worse tomorrow morning, than the forecast I viewed last night").

The other alternative is to log on 4 times per day when the updates some through.

Or, once we resume writtern forecasts Craig or myself will detail these possibilities in our analysis notes.

Hope that makes sense!

barstardos1's picture
barstardos1's picture
barstardos1 commented Thursday, 24 Oct 2013 at 2:28pm

......or set the alarm early. Wake up and check actual swell in the water http://new.mhl.nsw.gov.au/data/realtime/wave/Stn-sydney
Then decide go surfing / go for bike ride / go back to bed.

bjc's picture
bjc's picture
bjc commented Thursday, 24 Oct 2013 at 7:53pm

Hi Ben. Thanks for the response. Good to hear you will be also having the text forecasts and an iPhone friendly app. This will be excellent!

I should get up ride my home and check the surf but that's where laziness kicks in. If only I was a morning person.

Best of luck with the new site.

thermalben's picture
thermalben's picture
thermalben commented Saturday, 26 Oct 2013 at 2:02pm

Just a follow up comment - our forecast output has been very good for Sydney over the last three days, as per my observations.

This highlights some of the problems with automated forecasts - they get most swell events right, but not all of them. And the underlying problem is not in our algorithms but in the underlying WaveWatch III data.

We are aware of which swell events are poorly represented by the model but it's very dfficult to isolate them and (as an example) 'boost' those swells that frequently come in under forecast size (like the swell discussed for last Thursday AM).

However this is where the detailed forecast notes - once we reintroduce them in the next week or two - will come in handy, as we'll use the notes to point out the strengths and weaknesses of the model data in the upcoming forecast period.