Comments on: Lake Lanier Water Level Projection

By: Dax

Dax — Mon, 24 Dec 2007 07:24:58 +0000

2007 is at or near a record low for rainfall. Lanier started the year at about 1063 and will end at 1050, a drop of 13 feet. So it seems that even if 2008 matched that record low rainfall, we still end the year above deadpool level.

Unless the drought extends into the summer of 2009, this is much ado about nothing.

By: rkolter

rkolter — Wed, 05 Dec 2007 19:09:57 +0000

Actually, since I only track deadpool and end of volume right now, that feature would be neat, but not necessary for me.

I think that I could do what you suggest as far as the prediction test goes - I can easily get the average depth loss or volume loss on any given day, and then just extrapolate where the level should be based on that, thirty days out. Then compare to the actual values.

Maybe when I’m bored this weekend I’ll look into doing that.

By: lf61

lf61 — Tue, 04 Dec 2007 20:29:55 +0000

It just occurred to me that some of these models may not provide you with a graph where you can see what they expect in 30 days. In that case there really isn’t much you can do. To make forecasts easier on mine I just added level to the days along the bottom. I could also add a form entry that gives you the ability to mark any depth point or day point just like deadpool if that would make things easier for you. Just let me know.

By: lf61

lf61 — Tue, 04 Dec 2007 17:07:14 +0000

rkolter said:
>What you want is for me to freeze the predictions at a point (like I did when I held the ACE to their “79 day” prediction) and then chart from there. I could do that for the volume predictions, but the depth predictions will necessarily change as the surface area contracts.

The basis for a good prediction is to take into account the changes that occur over time. If it doesn’t then you can’t really trust the results. A month by month or longer accuracy percentage would indicate to the readers which can be better depended (both volume and level based predictions). If the level predictions aren’t accurate, then it is a flaw in the prediction, not the fact that the lake shrunk.

By: rkolter

rkolter — Tue, 04 Dec 2007 16:24:05 +0000

Like an idiot, I responded to lf61’s message without indicating what parts I was responding to. D’oh. It’s paragraph by paragraph - the first of mine to the first of his, the second of mine to the second of his, and the rest to the third of his.

By: rkolter

rkolter — Tue, 04 Dec 2007 16:22:18 +0000

Now they can’t be - they just jumped 20 more days.

Actually yours for the deadpool fits right in there. The 0-volume prediction is way ahead of my 0-volume prediction; only time will tell about that one.

Well, the values PERFECTLY tract what really happens of course - at least mine will, since they’re based on averages that update daily, they will accurately chart the course. If they start releasing 100x the water, the predictions will adjust accordingly.

What you want is for me to freeze the predictions at a point (like I did when I held the ACE to their “79 day” prediction) and then chart from there. I could do that for the volume predictions, but the depth predictions will necessarily change as the surface area contracts.

I could extrapolate what the end of month values would be from the start of month values, and then show the difference - that’d be kind of fun.

By: lf61

lf61 — Mon, 03 Dec 2007 21:05:40 +0000

They are probably looking at your web page and averaging all the values you list :-).
You have put together a nice page. It is interesting that prior to adding the seasonal variation mine followed the other volume prediction values very closely but when taking into account the average rainfalls for Dec through Feb adds a good bit more time.

It will be curious to see how the values track what really happens. As you get more observation time with the various values you could add an accuracy indicator (how the previous predictions are tracking reality).

By: rkolter

rkolter — Mon, 03 Dec 2007 20:44:56 +0000

No Jay.

Not being rude, I just -can’t- figure it out. I really thought it was volume based until it jumped 30 days when my own calculation did not. Prior to that I thought it might be depth based and they were just using a smaller set of numbers (so that a very low flow day would screw with their average). But that wasn’t it either.

I get a value from it each day because it’s a public countdown, but I couldn’t even begin to tell you what numbers they use, if they’re keeping them accurate, or what. Right now their number is within range of the other predictions, suggesting that whatever formula they use, it’s not wildly off base.

I do know they are considering the deadpool, and not the whole of the lake.

By: Jay Randal

Jay Randal — Mon, 03 Dec 2007 19:30:36 +0000

rkolter > can you explain AJC newspaper countdown clock for Lanier? A while back it said 98 days and falling, then recently 168 days and falling? What basis are they using and why all the changes? Seems like the paper is playing some kind of game or they want to keep us all confused.

By: rkolter

rkolter — Mon, 03 Dec 2007 19:22:50 +0000

and lf61 - Nice! Many thanks on the changes!