Climate Change - The Scientific Debate
Discussion
I try to find (and acknowledge) agreement if I can - otherwise there's no point to debate of any kind
As for the question of mining data - that's effectively what any statistical analysis is, and argumengts about what te "correct" start dtaes, lengths of records and so on should be are a little pointless beyond excluding the extremes in either direction. I suspect we can both also agree that, say, 3 year "trends" and 50 millenia "trends" have no meaningful place in this context (altough they might in others)
I get the feeling your'e also misunderstanding my scepticism. sadly, the "denial" label seem to have done it's job over the years so that, if I point out a possible change point (which it IS to early to confirm in the data) in a spirit of being sceptical, it's autmatically equated to denying any AGW effect.
But its not the same. The data does show a possible change in trends around the turn of the century, and the longer "the pause" continues, the more likely that change in trend becomes compared to the alternative of it just being a blip in the warming road.
As for the question of mining data - that's effectively what any statistical analysis is, and argumengts about what te "correct" start dtaes, lengths of records and so on should be are a little pointless beyond excluding the extremes in either direction. I suspect we can both also agree that, say, 3 year "trends" and 50 millenia "trends" have no meaningful place in this context (altough they might in others)
I get the feeling your'e also misunderstanding my scepticism. sadly, the "denial" label seem to have done it's job over the years so that, if I point out a possible change point (which it IS to early to confirm in the data) in a spirit of being sceptical, it's autmatically equated to denying any AGW effect.
But its not the same. The data does show a possible change in trends around the turn of the century, and the longer "the pause" continues, the more likely that change in trend becomes compared to the alternative of it just being a blip in the warming road.
Variomatic said:
I try to find (and acknowledge) agreement if I can - otherwise there's no point to debate of any kind
As for the question of mining data - that's effectively what any statistical analysis is, and argumengts about what te "correct" start dtaes, lengths of records and so on should be are a little pointless beyond excluding the extremes in either direction. I suspect we can both also agree that, say, 3 year "trends" and 50 millenia "trends" have no meaningful place in this context (altough they might in others)
I get the feeling your'e also misunderstanding my scepticism. sadly, the "denial" label seem to have done it's job over the years so that, if I point out a possible change point (which it IS to early to confirm in the data) in a spirit of being sceptical, it's autmatically equated to denying any AGW effect.
But its not the same. The data does show a possible change in trends around the turn of the century, and the longer "the pause" continues, the more likely that change in trend becomes compared to the alternative of it just being a blip in the warming road.
Again you miss the point I'm afraid which was about what you conclude from your analyses and not about stop/start point choices. Pick any points you want, play with the data - test your conclusions to see if they make sense. Like a sceptic should.As for the question of mining data - that's effectively what any statistical analysis is, and argumengts about what te "correct" start dtaes, lengths of records and so on should be are a little pointless beyond excluding the extremes in either direction. I suspect we can both also agree that, say, 3 year "trends" and 50 millenia "trends" have no meaningful place in this context (altough they might in others)
I get the feeling your'e also misunderstanding my scepticism. sadly, the "denial" label seem to have done it's job over the years so that, if I point out a possible change point (which it IS to early to confirm in the data) in a spirit of being sceptical, it's autmatically equated to denying any AGW effect.
But its not the same. The data does show a possible change in trends around the turn of the century, and the longer "the pause" continues, the more likely that change in trend becomes compared to the alternative of it just being a blip in the warming road.
plunker said:
Silver Smudger said:
plunker said:
(We're concentrating on the surface/satellite data here of course. Ocean heat content shows an increase in recent years)
How far back does the ocean heat content record go, so we can see how much it has changed 'in recent years'?And you have fallen back on the 'I'm just an interested telephone engineer' excuse again.
Silver Smudger said:
plunker said:
Silver Smudger said:
plunker said:
(We're concentrating on the surface/satellite data here of course. Ocean heat content shows an increase in recent years)
How far back does the ocean heat content record go, so we can see how much it has changed 'in recent years'?And you have fallen back on the 'I'm just an interested telephone engineer' excuse again.
(anyone can play the bad-faith game)
plunker said:
You only want to dis-regard OHC data because it spoils the global warming has stopped narrative.
(anyone can play the bad-faith game)
C'mon plunks - why didn't the ocean swallow the "heat" before the pause? Why do you believe it is swallowing the "heat" now? Given the heat capacity of the oceans why is this not the end of the global warming threat?(anyone can play the bad-faith game)
Jinx said:
plunker said:
You only want to dis-regard OHC data because it spoils the global warming has stopped narrative.
(anyone can play the bad-faith game)
C'mon plunks - why didn't the ocean swallow the "heat" before the pause? Why do you believe it is swallowing the "heat" now? Given the heat capacity of the oceans why is this not the end of the global warming threat?(anyone can play the bad-faith game)
http://www.nodc.noaa.gov/OC5/3M_HEAT_CONTENT/
plunker said:
I think you're referring to the idea that heat might be being squirreled away into the deep ocean, but I'm referring to OHC data:
http://www.nodc.noaa.gov/OC5/3M_HEAT_CONTENT/
Now convert from Joules to Kelvin and tell me how this is not the end of the threat?http://www.nodc.noaa.gov/OC5/3M_HEAT_CONTENT/
plunker said:
Again you miss the point I'm afraid which was about what you conclude from your analyses and not about stop/start point choices. Pick any points you want, play with the data - test your conclusions to see if they make sense. Like a sceptic should.
No, I'm afraid you're missing the point. I'm not concluding anything. Rather, I'm seeing (in the data) and accepting the possibility that something changed around that time. That can't be tested and turned into a "conclusion" without further years of data to see how it develops. Unlike the whole AGW bandwagon, which was kickstarted with far too little data (less than a decade at the time) to draw any valid conclusions from before promptly changing the rules so that nothing less than a magic "30 years for climate" would do to rebut those conclusions. Note that 30 years is an arbitrary number based on little except it being "the number" - the same as 2 deg C warming being the limit of armageddon.
Science really shouldn't start out with "think of a number, any number..."
plunker said:
Show me.
Why? You've already shown it, but you happen to place it around 2007 using oone dataset, whereas I would place it a bit earlier using another. But you've already agreed that it's there.Again, whether or not it's a permanent change is something that no amount of looking at the current data will answer - it either is or it isn't, and time alone will tell.
Variomatic said:
plunker said:
Again you miss the point I'm afraid which was about what you conclude from your analyses and not about stop/start point choices. Pick any points you want, play with the data - test your conclusions to see if they make sense. Like a sceptic should.
No, I'm afraid you're missing the point. I'm not concluding anything. Rather, I'm seeing (in the data) and accepting the possibility that something changed around that time. That can't be tested and turned into a "conclusion" without further years of data to see how it develops. Unlike the whole AGW bandwagon, which was kickstarted with far too little data (less than a decade at the time) to draw any valid conclusions from before promptly changing the rules so that nothing less than a magic "30 years for climate" would do to rebut those conclusions. Note that 30 years is an arbitrary number based on little except it being "the number" - the same as 2 deg C warming being the limit of armageddon.
Science really shouldn't start out with "think of a number, any number..."
30yrs is kind of arbitrary I guess but a suitably long period is required so that the ups and downs caused by ocean cycles (ENSO) cancel out and don't affect the trend. For example the period since 2007 is dominated by La Nina conditions so one would expect some reduction in the trend due to that (I'm not saying that's case closed though).
Variomatic said:
plunker said:
Show me.
Why? You've already shown it, but you happen to place it around 2007 using oone dataset, whereas I would place it a bit earlier using another. But you've already agreed that it's there.plunker said:
If you showed your basis for it being earlier somewhere I must have missed it.
Ok, let's start with your preferred dataset, hadcrut4. This shows a shorter pause / flatline / hernia than other sets so that works to "your" advantage.As we agree, the magic 30 years is arbitrary but intended to reduce or remove short-term influences. However, using it can't distinguish between a short-term 2blip" (that it's valid to suppress) or a turning point / systemic change that it's not valid to suppress.
So, let's look at the data with shorter (decadal) averages:
That's the same data with trends calculated, at 2 year starting points, for a decade at a time - about the same data length that was considered suitable evidence of the theory to start the ball rolling.
While that approach undoubtedly increases the risk of "capturing" transient changes, it also avoids the risk of discarding rapid systemic changes. Since a "turning point" is a rapid systemic change, it's reasonable to look for it using the data in a form that won't supporess such things.
Now, what we see is a trend that suddenly decreases when the start point is around 1998 (as we'd expect because of the El Nino that year) but continues at the decreased level once we pass the El Nino and move into the 2000's.
If we take each trend line as representative of its centre point year, that gives a sudden and (so far) continuing reduction in trend between 2003 and 2005. I suggested "around 2004" earlier in the thread, which seems like a fair compromise.
As I've repeatedly said above, only further years of data will be able to confirm whether that's a transient change or a permanent one, but it undoubtedly shows the possibility of a change around that point.
Denying that possibility based on arbitrary trend lengths that specifically can't show a sudden change for another 20 years or so is, err, denial
Edited by Variomatic on Monday 24th November 22:28
Incidentally, that last line is intended as tongue in cheek towards you, Plunker, because we may have differing interpretations but at least you discuss based on interpreting the data (as I also try to do). As long as different (reasonable) interpretations of the available data are possible there's always room for that in any serious debate.
If, in another 10 years, the trend has resumed to a clear increase, then I'll be happy to accept that my interpretation was wrong - there's no shame at all in admitting a mistake if it's revealed by more data, provided your initial interpretation was based on a reasonable analysis of the data that was available at the time. I also trust that if, in 2024, the trend has remained flat, or even entered cooling, you'd do the same
If, in another 10 years, the trend has resumed to a clear increase, then I'll be happy to accept that my interpretation was wrong - there's no shame at all in admitting a mistake if it's revealed by more data, provided your initial interpretation was based on a reasonable analysis of the data that was available at the time. I also trust that if, in 2024, the trend has remained flat, or even entered cooling, you'd do the same
I'm afraid I closed the page but know what settings I used so I'll try to remember to reconstruct it this evening for you. Would do it now but I really have to stop finding excuses to avoid reassembling this chronograph
eta: here ya go:
http://www.woodfortrees.org/plot/hadcrut4gl/from:1...
Note that the last trend is a year short because we're not into 2015 data yet but it's unlikely to change matters much unless 2015 is the year that the oceans boil (or freeze)
eta: here ya go:
http://www.woodfortrees.org/plot/hadcrut4gl/from:1...
Note that the last trend is a year short because we're not into 2015 data yet but it's unlikely to change matters much unless 2015 is the year that the oceans boil (or freeze)
Edited by Variomatic on Monday 24th November 22:31
Here's the same approach applied to the RSS TLT data:
The interesting thing here is that the decadal trend reduces consistently for each decade (in 2 year steps) from 1992/2003 (centred on 1997) onwards, with the exception of 2000/2011, where there's a natural "rebound" from the effect of the 1998 El Nino, and 2004 / 2015 where the data isn't complete yet. Note that much the same holds if you plot in steps of 1 year, but the extra 9 trend lines make the graph pretty well unreadable
If 2015 is cooler than this year (which has been hovering on the edge of El Nino conditions, so that's fairly likely) then that'll be a reduction in decadal trend for all of the past 12 years except for a brief rebound after a historically large El Nino.
Note that I'm not saying the above is something that firm conclusions can be drawn from, but it is a valid way to look at the data if we're looking for a possible turn in the trend, and what we find is consistent with a change like that sometime (depending on dataset) between about 1997 and 2005.
The interesting thing here is that the decadal trend reduces consistently for each decade (in 2 year steps) from 1992/2003 (centred on 1997) onwards, with the exception of 2000/2011, where there's a natural "rebound" from the effect of the 1998 El Nino, and 2004 / 2015 where the data isn't complete yet. Note that much the same holds if you plot in steps of 1 year, but the extra 9 trend lines make the graph pretty well unreadable
If 2015 is cooler than this year (which has been hovering on the edge of El Nino conditions, so that's fairly likely) then that'll be a reduction in decadal trend for all of the past 12 years except for a brief rebound after a historically large El Nino.
Note that I'm not saying the above is something that firm conclusions can be drawn from, but it is a valid way to look at the data if we're looking for a possible turn in the trend, and what we find is consistent with a change like that sometime (depending on dataset) between about 1997 and 2005.
Edited by Variomatic on Monday 24th November 23:22
Variomatic said:
plunker said:
If you showed your basis for it being earlier somewhere I must have missed it.
Ok, let's start with your preferred dataset, hadcrut4. This shows a shorter pause / flatline / hernia than other sets so that works to "your" advantage.As we agree, the magic 30 years is arbitrary but intended to reduce or remove short-term influences. However, using it can't distinguish between a short-term 2blip" (that it's valid to suppress) or a turning point / systemic change that it's not valid to suppress.
So, let's look at the data with shorter (decadal) averages:
That's the same data with trends calculated, at 2 year starting points, for a decade at a time - about the same data length that was considered suitable evidence of the theory to start the ball rolling.
While that approach undoubtedly increases the risk of "capturing" transient changes, it also avoids the risk of discarding rapid systemic changes. Since a "turning point" is a rapid systemic change, it's reasonable to look for it using the data in a form that won't supporess such things.
Now, what we see is a trend that suddenly decreases when the start point is around 1998 (as we'd expect because of the El Nino that year) but continues at the decreased level once we pass the El Nino and move into the 2000's.
If we take each trend line as representative of its centre point year, that gives a sudden and (so far) continuing reduction in trend between 2003 and 2005. I suggested "around 2004" earlier in the thread, which seems like a fair compromise.
As I've repeatedly said above, only further years of data will be able to confirm whether that's a transient change or a permanent one, but it undoubtedly shows the possibility of a change around that point.
Denying that possibility based on arbitrary trend lengths that specifically can't show a sudden change for another 20 years or so is, err, denial
ok that aside the 11-year period starting in 1998 ends with 2008 - that's already got a big foot in the period that I'm claiming marks a shift. There was a large La Nina temperature drop that year. And then each successive period has a further year in plunker territory - I'm not sure you aren't in fact confirming my view of when the shift started! I'm also uneasy with the centre year of the trend line thing - it's like you're doing a centred rolling average but without calculating the values. It's a bit novel.
I'd suggest maybe a better test would be to do linear trends for each year with all of them starting in 1978 and then compare the slope values..
I started to do that but it appears if you load WFT with too many series it, err, crashes the site - oops
I'll try again tomorrow.
Yeah, sorry about the mis-counted dates, let's just pretend that a decade has 11 years - doesn't make a whole heap of difference cos using a decade was just an arbitrary choice
The logic behind using a rolling start date is that if there's a change point (or a relatively rapid change over a few years more likely) then using increasingly long periods which include the earlier trend will tend to obscure the change by weighting the trend with the earlier slope.
As a thought experiment, consider calculating trends along a sine curve (no, I'm NOT suggesting temperatures are following a sine) using a fixed start point.
If you happen to start at the "bottom" of the curve then the trend will increase as your end point heads towards the top, then start to decrease again towards zero as it drops down the other side, but it will never become negative, even though half the time the real trend in the data is downwards.
Similarly, if you happen to start at the top of the curve, you'll get the impression that the trend is never positive, even though the data spends exactly half its time increasing!
By using a "rolling" start date and fixed interval you're approximately taking a tangent to the curve in any given year. That's not quite the same as taking a rolling average because an average is more affected by any noise either side of the tangent point unless the interval is so long that noise cancels - we're intentionally using fairly short intervals to look for rapid changes, so that assumption of noise cancelling can't be relied on.
Treating each trend as representing the mid-point of the period in question is arbitrary but intuitively reasonable. If we specify each trend as "leading" from the start year then any change will appear to be earlier than it really is, if we specify the trends as lagging from the end year then they'll appear to be later.
The mid-point probably isn't mathematically accurate because the correct point will depend on the shape of the data, but as an approximation it makes sense for forming an impression of what might be happening.
The logic behind using a rolling start date is that if there's a change point (or a relatively rapid change over a few years more likely) then using increasingly long periods which include the earlier trend will tend to obscure the change by weighting the trend with the earlier slope.
As a thought experiment, consider calculating trends along a sine curve (no, I'm NOT suggesting temperatures are following a sine) using a fixed start point.
If you happen to start at the "bottom" of the curve then the trend will increase as your end point heads towards the top, then start to decrease again towards zero as it drops down the other side, but it will never become negative, even though half the time the real trend in the data is downwards.
Similarly, if you happen to start at the top of the curve, you'll get the impression that the trend is never positive, even though the data spends exactly half its time increasing!
By using a "rolling" start date and fixed interval you're approximately taking a tangent to the curve in any given year. That's not quite the same as taking a rolling average because an average is more affected by any noise either side of the tangent point unless the interval is so long that noise cancels - we're intentionally using fairly short intervals to look for rapid changes, so that assumption of noise cancelling can't be relied on.
Treating each trend as representing the mid-point of the period in question is arbitrary but intuitively reasonable. If we specify each trend as "leading" from the start year then any change will appear to be earlier than it really is, if we specify the trends as lagging from the end year then they'll appear to be later.
The mid-point probably isn't mathematically accurate because the correct point will depend on the shape of the data, but as an approximation it makes sense for forming an impression of what might be happening.
Gassing Station | Science! | Top of Page | What's New | My Stuff