long range forecast reliability
Donweather. Thanks for that, Yeh anything beyond a week I do take with a pinch of salt, but big swells forecasted a long way out generally appear as some sort of swell (usually downgraded from my observations) that is why I was surprised the model has two significant swells now, when before it was 4-6 ft. for about ten days following Thursday's (19/6).
I generally take it all with a pinch of salt really, I just look for pulses or trends up and down then see what the gods deliver. How does the upper atmosphere affect the forecast? Sorry I am pretty ignorant about all this, I pretty much just look at graphs now (pretty lazy really) I could read a synoptic etc. in the past but the internet has made it all too easy now, so I barely look at any other sources. I just calibrate in my head from experience what the WAMS say to what ever coastline I am looking at, works pretty good really, but you weather gurus are all over it, quite impressive. I think you need to be on the swell more on the East Coast because swell is such a critical component. On south and west facing coasts (oz) it is more about getting out of the wind, oh the fucking wind!
Ben. you say I had a lot of questions in there, I think there were two, and you answered with about ten! You love splitting hairs! Were you in the debating team at school? (jokes Ben just jokes)
but yeh I see your point/s, kind of hard to quantify, but like I said I just look for general trends and don't really split hairs about sizes, especially because waves sizes have become a really complicated issue of late, I believe. But the swell I am talking about arriving day out, was about the talk around the beach, you know the incessant talk when swells are coming, everyone was saying big tomorrow morning blah blah blah, you guys had a small jump in the arvo, then big the next day. It came up fast through the morning and was big by lunch, and was bigger next day, but it was also the dying swell you guys seemed to be day behind, over a couple of days. So I was just wondering if all the websites pump the same data into your own calibrated models.
I know you like to bag magic seaweed but I find them pretty good, you just have to know what each sites 3-4 ft. equates to for each piece of coast.
Magic seaweed is pretty vague as they do the whole world. You guys seem to have it down for Australia's East coast where you have obviously concentrated more. But again the nooks and cranny's of the SA coast take some interpretation, that I would rather not go into, having jagged some beauties around the edges of your predictions, avoiding the droves.
But Indo and the East coast in my opinion seem easier to time as there is wide open ocean abruptly striking the coast, maybe not, just my observations.
OMG that model is going crazy now, 8 days above 6ft. , lots of 8-10ft. days, a 12-15ft day, Faaarrrrkkk, should have got on that plane!!!
You must have the most coherent internal dialogue in the world.
Anyone correct me if I'm wrong here... Low surface pressure means that air is rising. As the air rises, it takes the path of least resistance (wind shear). In this way, the surface systems can kind of be puppeted by the upper winds. The upper winds can also provide ventilation for warm, moist, infeeding and rising air, in turn enhancing cyclogenisis.
Not sure about high lat systems. I don't understand how increased temperature differences plays a role, but I know it does.
Spykan, thanks for raising the topic. First of all there is no single answer for you here and I'm probably the least qualified on the forums compared to the other gurus. I reckon the 16 day forecasts give a great indication and are especially handy for the long range swell planning like Fiji Indonesia Tahiti. With regard to that Bali swell for last Thursday and the days (weeks) either side, I watched that swell a long way out checking the forecasts pretty much everyday. One of the real problems was the models bounced around as you say. This is one red flag for me that this is a problem and one of the secrets for using the long range forecasts. If things are moving daily there is uncertainty in the models. It bounced around a lot during that 2 weeks leading up to it and it eventually didn't arrive. Nusa Dua is a swell magnet and its possible given swell direction that it was double overhead compared to 4 foot Ulus.
In previous years lot of the forecast models were often a day late for Indonesia. You had to make an allowance for the extra day. I have gone on 4 swells to Indo in past year and tracked SN closely with the forecast versus actual and they have all been bang on. It looked like last week the swell didn't come in as big as predicted but it seems that all the forecasters were in the same boat.
The models as you have noticed have lit up with an XXL swell one of the benefits of the 16 day forecasts....be interesting if it arrives next week. If it does it will be the biggest waves in Indo for a long time. Lot of time between now and next week.
sypkan wrote:I know you like to bag magic seaweed but I find them pretty good, you just have to know what each sites 3-4 ft. equates to for each piece of coast.
There are many flaws in the MSW system (from my observations) but before discussing this it's best to know which coast you're referencing (are you on the Mid Coast in SA?).
As for "you just have to know what each sites 3-4 ft. equates to" - MSW forecast 'face feet' - so using that as a guideline, they are frequently way off. But if you're looking for overall trends, then sometimes their forecasts may be useful as the input data is essentially the same for Swellnet, MSW, Surfline, Buoyweather etc etc. And it is this input data - GFS 10m winds - which guide the overall swell forecast cycles.
sypkan wrote:Magic seaweed is pretty vague as they do the whole world. You guys seem to have it down for Australia's East coast where you have obviously concentrated more.
That is incorrect. Both websites have a global forecast database. Swellnet has an almost linear spread of locations, spaced roughly 50km apart across each coastline around the world (give or take, depending on popuation density and geographical attributes). MSW on the other hand have tried to itemise each worldwide "surf spot" with its own forecast. And this is where their system falls down - from both an accuracy level, and also a website navigability level.
sypkan wrote:But again the nooks and cranny's of the SA coast take some interpretation, that I would rather not go into, having jagged some beauties around the edges of your predictions, avoiding the droves.
As mentioned above, it's not really possible for global swell models to consistently resolve individual surf breaks to this kind of accuracy. And it is one of several reasons why we decided on our current spread of forecast locations. Using SA as an example, we have produced have one datapoint for each major surfing region in the state - this provides a forecast for 'exposed locations', and it's then down to the surfer to fine tune the forecast to their individual break requirements.
If you take Yorkes for an example, there's simply no way for a swell model to resolve the daily size variablity between Willys, Chinamans, Ethels, West Cape, Pondi, Tressies, Lizards, Salmon Hole, Daly Heads, Gleesons etc (plus all of the lesser known spots) within an individual size forecast. This is what MSW try to do and this approach is very hit and miss. And as a result, you have to check multiple pages to cross check the surf forecast. So we've tried to simply the process and have created one forecast page to check, which you can then use as a guide based on your own personal experience.
mick-free wrote:One of the real problems was the models bounced around as you say. This is one red flag for me that this is a problem and one of the secrets for using the long range forecasts. If things are moving daily there is uncertainty in the models.
Spot on Mick. The key is to look for trends within the trends. If the models have been unwavering in their forecast output over a particular period, then confidence is certainly increased.
However sometimes the models can wig out at even one or two days.. the forecasts are series of complex mathematical equations, and are only as good as the input source data. If there are problems or errors with the input data, then these errors will be magnified exponentially throughout the forecast period. This all depends on the kind of swell being forecast, and the destination beach.
Long range forecasts are inherently more accurate and reliable at west facing coasts, particularly Indonesia, South Pacific and also up through Northern South America/Central America/Mexico etc.. locations where the long range groundswell is likely to be the primary swell source.
Long range forecasts are also generally very accurate at west facing locations in the mid-latitudes, however these locations (ie West Oz, South Oz, Vic etc) also feel the influence of short range systems and this can sometimes contaminate the data and reduce the forecast quality.
Hi Ben, this is interesting reading. So based on the info above, I should take with a grain of salt that the morn pen forecast for Tuesday 24th is saying 15-18ft on the long range forecast?!
There's a strong system moving through next Tuesday GF and there'll be plenty of swell with it but yes, that is overcooking the size due to there being a lot of windswell contamination in the Wave Watch forecast. Ie the model isn't differentiating the separate groundswell and windswell components for this swell and is instead combining them, giving a wrong representation of the size.
At this stage I'd say you'd be looking at a stormy swell in the 10ft range, but as mentioned in my forecast yesterday, the models are still slightly divergent on the make-up of this system.
I wouldn't be surprised if the models dropped back the size next update. And this is what you look for, consistent trends or large variability to start to become confident in a forecast for a selected period.
Thanks for the reply Craig, I'll keep an eye on it.. Hopefully something happens from it
Good info everyone thanks,
Ben I was just making it up for the 3-4ft. thing as an example, yeh I pretty much agree with your comments about MSW, they are a little ambitious really. I just meant if I'm at yorkes I just check chis and know that their 3-4ft. equates such and such elsewhere, if I'm in Indo I just check Ulus and know that their 3-4 ft. there equates smaller waves elsewhere etc. You are right it is totally pointless scrolling through their surf spots I just have a spot for each coast that I look at and stick to, and work the rest out from there.
"You must have the most coherent internal dialogue in the world."
Not sure what you mean there Mitch, but you made me smile, if its about talking to myself yeh I do that a lot, doesn't everyone? don't believe that rubbish about first sign of craziness, it is good for you!
How confident are you guys with your 16 day forecast? I know it must change from swell to swell, but do you have an average percentage. I know the weather bureau wont go into too much detail more than four days in advance due to unreliability.
I noticed a couple of days ago it was basically going to be 4-6ft everyday following the big swell Thursday 19/6, now you have two days 8ft and two days 8-10ft. Big difference in my book. Craig s forecast notes are usually very good, not very impressed with the 16 day forecast.
I know you probably wont say due to 'commercial interests' or something, but do you just pump the swell charts available elsewhere into your calibrated modelling?
thermalben commented Monday, 16 Jun 2014 at 10:26am
sypkan, we're working on a verification system at the moment that'll be able to give you some running percentages. The difficulty is assimilating actual report input data however we're pretty confident that the output produced will be useful.
sypkan commented Monday, 16 Jun 2014 at 12:15pm
Cool it would be good to have, but you might be giving too much away with that Ben i.e. how unreliable long term forecasting is. I was just wondering about a general percentage.
Not bagging you guys, as I said the bureau wont even give two week wind data due to unreliability, one has to look elsewhere, but the bureau is conservative by its nature.
I find the 16 day wind data quite good, but the swell seems to jump all over the place, I guess that is what you get when it updates 4 times a day (I think I read that). That is why I was wondering how you come up with it. Clearly Craig uses many sources, hence the pretty pictures with his forecasts. But the long term forecasts would have to be automated, no hints there obviously?
Its ok, I know the swell charts elsewhere jump all over the place long term as well, so I didn't really expect much different from you guys for the 16 day thing, but I thought I would give it ago because you gave me a legrope with the subscription. While I am not overly impressed with the 16 day, it is still very useful, and I am impressed with the legrope!
donweather commented Monday, 16 Jun 2014 at 12:47pm
sypkan, any WAM is only as reliable as the wind data that is input into it. Nearly all WAMs freely available on the internet use GFS wind data as their inputs, hence the unreliableness of the WAMS is not a function of the WAMS, but more a function of the GFS wind data model. Its fair to say that any wind data models more than 7-9 days out can be taken with a grain of salt to some extent, particularly if there's some very active low pressure systems in the models, particularly around the tropics and temperate zones. Hence I can't speak from experience too well on the storms that create the Indian Ocean swells, but I'm assuming any variations on the tropics/temperate zone forecasts has some impact on the southern indian ocean storms. One also has to remember that it's not just MSLP (which we as surfers only really care about). Any small changes in the upper atmosphere can have a reasonable impact on the MSLP forecasts.
thermalben commented Monday, 16 Jun 2014 at 12:51pm
There are lots of questions contained within your post sypkan, and with limited time I'll just round up and say that we're happy to publish accuracy data (yes, it decreases over time, we're not hiding behind it). However there are many factors that are difficult to quantify in these verification models.
For example, if we forecast an 8-10ft swell at 14 days out, and it arrives a day late - but is spot on size wise, was the forecast correct or not?
How about if we forecast a swell to arrive at 3pm, and it arrives at 6pm - just as the sun goes down? The surf may have only technically been a few hours late, but if it's a brief event and peaks under the cover of darkness - when you can't surf - before fading rapidly before dawn - does it count as a hit or a miss? What kind of error margin do we allow?
And what kind of accuracy is required for wave heights? If the forecast is 2ft and the surf comes in at 3ft, it's fair to say that the forecast was under. But if the forecast is 8ft and the surf peaks at 10ft, is this a forecast hit or a miss? How about an 18ft forecast vs a 20ft verification?
For what it's worth, most other long range weather forecasts use similar data sources (so the overall trends are usually the same), and yes - all of our forecast graphs are automated.
However the strength of the Swellnet forecasts are (1) the proprietary algorithms we've created to calculate surf heights from the swell data, and (2) the way we display the forecast data on our graphs (mainly in relation to the East Coast and the dual-height system).
Happy to answer any other questions you have though.
thermalben commented Monday, 16 Jun 2014 at 12:58pm
Also - may be better starting a new forum thread, as it's not quite approrpiate to have them in this Indo forecast (it'll be harder to find them in the future).