Is Climotology Overused?

[wkd]

One of my pet peeves when reading the weather forums is what I feel is the overuse of climotology in determining the outcome of an individual event.  Climotology if I'm not mistaken is just an average of the weather at any given location over a long period of years and as such would have absolutely no bearing on an individual event.  Certainly  it could be used to give a probablity forcast for the weather in the same location and period in future years, yet some posters seem to feel it is a major factor during a current event.  A classic example was a poster in the mid atlantic forum regarding the path of hurricane Irene.  He constantly insisted the models were incorrect and because of climotology it had to recurve out to sea and would not track up the coast.

 

Just to clarify,  I am not thinking of analogs here, which I think can be very useful.

[JBG]

One of my pet peeves when reading the weather forums is what I feel is the overuse of climotology in determining the outcome of an individual event.  Climotology if I'm not mistaken is just an average of the weather at any given location over a long period of years and as such would have absolutely no bearing on an individual event.  Certainly  it could be used to give a probablity forcast for the weather in the same location and period in future years, yet some posters seem to feel it is a major factor during a current event.  A classic example was a poster in the mid atlantic forum regarding the path of hurricane Irene.  He constantly insisted the models were incorrect and because of climotology it had to recurve out to sea and would not track up the coast.

 

Just to clarify,  I am not thinking of analogs here, which I think can be very useful.

You're quite right that climatology won't govern particular events.  The October 29, 2011 and November 7, 2012 events for example were outside of climo in the New York area.

 

However the working parts of climo, such as sun angle, upper atmosphere and ocean water temperatures can stack the odds mightily.  This is why it is most unusual to get a classic three-day over 90 heat wave in January.

[famartin]

One of my pet peeves when reading the weather forums is what I feel is the overuse of climotology in determining the outcome of an individual event.  Climotology if I'm not mistaken is just an average of the weather at any given location over a long period of years and as such would have absolutely no bearing on an individual event.  Certainly  it could be used to give a probablity forcast for the weather in the same location and period in future years, yet some posters seem to feel it is a major factor during a current event.  A classic example was a poster in the mid atlantic forum regarding the path of hurricane Irene.  He constantly insisted the models were incorrect and because of climotology it had to recurve out to sea and would not track up the coast.

 

Just to clarify,  I am not thinking of analogs here, which I think can be very useful.

 

There are different types of climatology.  Aside from the very important point made above about the "working parts of climo", climo can give you probability of an unusual event.  It can help you if guidance is not in agreement...  i.e., if you have several models showing a modest event and one showing a big event, what are the climatological odds of the big event? 

 

Yes, rarely seen events do occur from time to time... but far less regularly than model guidance shows.

[wkd]

You're quite right that climatology won't govern particular events.  The October 29, 2011 and November 7, 2012 events for example were outside of climo in the New York area.

 

However the working parts of climo, such as sun angle, upper atmosphere and ocean water temperatures can stack the odds mightily.  This is why it is most unusual to get a classic three-day over 90 heat wave in January.

But I would argue that these factors don't stack the odds in any current synoptic setup. They only stack the odds for any outcome for an event in the future.

[famartin]

But I would argue that these factors don't stack the odds in any current synoptic setup. They only stack the odds for any outcome for an event in the future.

 

But they do stack the odds when you've got a range of model solutions.

[wkd]

There are different types of climatology.  Aside from the very important point made above about the "working parts of climo", climo can give you probability of an unusual event.  It can help you if guidance is not in agreement...  i.e., if you have several models showing a modest event and one showing a big event, what are the climatological odds of the big event? 

 

Yes, rarely seen events do occur from time to time... but far less regularly than model guidance shows.

I think your argument may be somewhat valid for an event modeled 10 to 20 days out but the closer the event is the less climatology is a factor. Again climate is just a mathematical construct and is best used to predict the probability of outcomes of parameters or events in the future when they are not yet even modeled.

[famartin]

I think your argument may be somewhat valid for an event modeled 10 to 20 days out but the closer the event is the less climatology is a factor. Again climate is just a mathematical construct and is best used to predict the probability of outcomes of parameters or events in the future when they are not yet even modeled.

Even 3 days out:  You have one model showing a catastrophic snowstorm, one showing a modest snowstorm, one showing nothing.  Are you really going with the one showing the catastrophic snowstorm?  Certainly climatology would direct you away from that model.

[wkd]

Even 3 days out:  You have one model showing a catastrophic snowstorm, one showing a modest snowstorm, one showing nothing.  Are you really going with the one showing the catastrophic snowstorm?  Certainly climatology would direct you away from that model.

 

We must have a different opinion of what climotology is.  In the case you present I don't think climotolgy is a factor at all.  If I were more knowledgable I would try to determine why the three models are interpreting the data differently.

 

In the example JBG gave above concerning temps of 90 in January I would say for sure climotology argues against it and I would bet based on that, that it will not occur next January.  But if 3 days before that time the models showed it likely to occur I would not use climotology as an argument against it.

[weatherbear5]

Even 3 days out:  You have one model showing a catastrophic snowstorm, one showing a modest snowstorm, one showing nothing.  Are you really going with the one showing the catastrophic snowstorm?  Certainly climatology would direct you away from that model.

Depends upon which model it is, known biases of each particular model, how the models initialize etc, etc...  In that particular scenario, given all else is equal, I would forecast for the moderate but make fully well sure that everyone knew that it really could be much more or much less depending on track or temps or whatnot

[wkd]

That approach makes a lot more sense to me rather than saying climbo argues for one or the other solution. It is very possible, depending on time of year, climatology would say none of the three outcomes could occur.

[Ian]

climo knowledge makes almost any forecast probably 10x better. every now and then you'll bust on a hecs if you hug it too hard.

[wkd]

climo knowledge makes almost any forecast probably 10x better. every now and then you'll bust on a hecs if you hug it too hard.

 

I don't know.  I think that is a pretty bold statement. Perhaps forcasting accumulation of snow on the roads in DC you may be right but with regard to synoptic scale features I would doubt it.

[wkd]

Maybe this is just a matter of semantics, I'm not sure.  What finally got me to post this topic was the current arguement by so many posters 5 to 7 days out that accumulating snow was very unlikely around the March 25th timeframe because climotolgy argued against it.  To me that argument makes zero sense.

 

And Ian, I think a forcaster that has a good knowlege of analogous setups and their outcomes may be able to make 10X better forcasts, but to me this is not the same as a knowledge of climotology.

[donsutherland1]

IMO, climatology in particular and history in general offer good risk management tools. If one is looking at forecasts beyond the near-term, some weight should be given to climatology. The farther into the forecast horizon one goes, the greater the weight that should be accorded climatology.

 

In the near-term strong model consensus clearly trumps climatology. Hence, given the model consensus in place it is not unreasonable to suggest that an area from Philadelphia across NJ might see 3"-6" snow with locally higher amounts from the upcoming storm. In NYC, where the guidance had been divided, climatology suggested a 50% chance less of less than 1" snow, 50% chance of 1" or more, and a 29% chance of 2" or more and 18% chance of 4" or more. Hence, in the face of the uncertainty, one could have suggested that a period of snow with some accumulation was possible. If pressed for numbers, one could have argued that it was likely that the accumulation would amount to less than 4 inches (82%).

 

As it turns out, the models have now converged toward a less snowy outcome there. In the meantime, the convergence of models regarding parts of eastern PA and NJ make it a reasonable call to go above to much above climatology when forecasting possible accumulations there.

[famartin]

IMO, climatology in particular and history in general offer good risk management tools. If one is looking at forecasts beyond the near-term, some weight should be given to climatology. The farther into the forecast horizon one goes, the greater the weight that should be accorded climatology.

 

In the near-term strong model consensus clearly trumps climatology. Hence, given the model consensus in place it is not unreasonable to suggest that an area from Philadelphia across NJ might see 3"-6" snow with locally higher amounts from the upcoming storm. In NYC, where the guidance had been divided, climatology suggested a 50% chance less of less than 1" snow, 50% chance of 1" or more, and a 29% chance of 2" or more and 18% chance of 4" or more. Hence, in the face of the uncertainty, one could have suggested that a period of snow with some accumulation was possible. If pressed for numbers, one could have argued that it was likely that the accumulation would amount to less than 4 inches (82%).

 

As it turns out, the models have now converged toward a less snowy outcome there. In the meantime, the convergence of models regarding parts of eastern PA and NJ make it a reasonable call to go above to much above climatology when forecasting possible accumulations there.

I'm tired, so instead of writing myself I'll just say "what Don said"

[wkd]

IMO, climatology in particular and history in general offer good risk management tools. If one is looking at forecasts beyond the near-term, some weight should be given to climatology. The farther into the forecast horizon one goes, the greater the weight that should be accorded climatology.

 

In the near-term strong model consensus clearly trumps climatology. Hence, given the model consensus in place it is not unreasonable to suggest that an area from Philadelphia across NJ might see 3"-6" snow with locally higher amounts from the upcoming storm. In NYC, where the guidance had been divided, climatology suggested a 50% chance less of less than 1" snow, 50% chance of 1" or more, and a 29% chance of 2" or more and 18% chance of 4" or more. Hence, in the face of the uncertainty, one could have suggested that a period of snow with some accumulation was possible. If pressed for numbers, one could have argued that it was likely that the accumulation would amount to less than 4 inches (82%).

 

As it turns out, the models have now converged toward a less snowy outcome there. In the meantime, the convergence of models regarding parts of eastern PA and NJ make it a reasonable call to go above to much above climatology when forecasting possible accumulations there.

 

Thanks Don.  I was hoping for your input and I pretty much agree with you.  It really comes down to using climo in a probabalistic way.  My objection is that too many poeple use it as a shorter term forcasting factor no matter what the models or common sense would dictate.  Climotology would argue that NW  New Jersey will get higher yearly snowfall than south Jersey.  Sure would have been mistaken in the winters of 2009-2010 and  2010-2011.  Even so I would still bet that next winter NW New Jersey wins out. Every situation is different and every event changes an areas climate.

[famartin]

Thanks Don.  I was hoping for your input and I pretty much agree with you.  It really comes down to using climo in a probabalistic way.  My objection is that too many poeple use it as a shorter term forcasting factor no matter what the models or common sense would dictate.  Climotology would argue that NW  New Jersey will get higher yearly snowfall than south Jersey.  Sure would have been mistaken in the winters of 2009-2010 and  2010-2011.  Even so I would still bet that next winter NW New Jersey wins out. Every situation is different and every event changes an areas climate.

Here's one:  Climatology might argue that NW NJ will get the best snow tomorrow... but its entirely possible that the highest accums end up somewhere in S NJ instead.  That seems odd and against climo... but then you also have to consider the nature of spring snowstorms and how the outer edges often end up with light non-accumulating snow while it thumps in the core.  That's part of climo too, just not the classic, mundane sense that you see it.

[Ian]

I don't know.  I think that is a pretty bold statement. Perhaps forcasting accumulation of snow on the roads in DC you may be right but with regard to synoptic scale features I would doubt it.

 

i guess it depends on how you define climo. i see it as understanding how systems work based on history etc as well just a bunch of numbers for any given timeframe. there are always going to be outliers and you should be able to recognize them.. tho arguably knowing climo would still give you a better forecast than just running with exactly what modeling shows you. (we can revisit this later if there is a large widespread 1' snowstorm in the coastal plain tonight and tomorrow i guess).  you'd be a pretty poor forecaster if you ran to exact climo odds of an anomalous event while one is staring you down obviously.

[wkd]

Here's one:  Climatology might argue that NW NJ will get the best snow tomorrow... but its entirely possible that the highest accums end up somewhere in S NJ instead.  That seems odd and against climo... but then you also have to consider the nature of spring snowstorms and how the outer edges often end up with light non-accumulating snow while it thumps in the core.  That's part of climo too, just not the classic, mundane sense that you see it.

 

I would say in this case the reason south Jersey will probably get more snow is the synoptic setup and storm track.  This outcome, if the result is as forcast, will only climotologically add more snow to south Jersey and subtract from north Jersey.  And yes, climate is mundane, it is only an average over years.

[famartin]

I would say in this case the reason south Jersey will probably get more snow is the synoptic setup and storm track.  This outcome, if the result is as forcast, will only climotologically add more snow to south Jersey and subtract from north Jersey.  And yes, climate is mundane, it is only an average over years.

Your view of climate is very limited.

[wkd]

Your view of climate is very limited.

 

You are probably right. No offense.

[wkd]

i guess it depends on how you define climo. i see it as understanding how systems work based on history etc as well just a bunch of numbers for any given timeframe. there are always going to be outliers and you should be able to recognize them.. tho arguably knowing climo would still give you a better forecast than just running with exactly what modeling shows you. (we can revisit this later if there is a large widespread 1' snowstorm in the coastal plain tonight and tomorrow i guess).  you'd be a pretty poor forecaster if you ran to exact climo odds of an anomalous event while one is staring you down obviously.

 

To me this is more an understanding and use of analogy rather than climate.

[Hoosier]

Climo can be a great tool but like anything it can be overused and abused. If we're talking about the current storm, it is already doing something pretty uncommon for areas west of the east coast (how about 8.7" and counting at St. Louis and Indy looking at potentially a record breaking event for this late).

[WXheights]

Chuck Doswell cautions to watch for the "normalcy bias" that we all carry. We see a super Typhoon in the modeling over an uncharacteristic climatological area - we tend to diminish it as normal logic train.  After several model runs with the same thing locking in with other models we still carry a normalcy bias.  It's knowing when to hold'em when to fold'em when to walk away and when to run