There is a sound theoretical basis that, all other things being equal, increased ghg’s should result in at least some warming. We have been burning fossil fuels, the CO2 has risen since the beginning of the industrial revolution, and the temperature have warmed. So, there is a sound theoretical basis, and some evidence.

Except there’s one problem with that (at least if one accepts there’s any validity to climate models at all). That is that when one sums the total anthropogenic forcings (mainly CO2 and sulfates), the cooling effects of the sulfates nearly balanced the increased CO2 until circa 1980. Almost all of the cooling prior to this point in time has to be attributed to changes in natural forcings.

See e.g., this figure.

That said, I basically agree with you. The climate is changing, and humans are partly responsible for it. However, I am really not interested in returning to the hard times of the 17th and 18th centuries. There was nothing “optimal” about that at all.

The data that I am generating has a built-in trend plus randomness to represent the yearly fluctuations that we see in nature.

And that is the point: if you do not build in a trend, the data generated looks nothing like the real temperature data that we have seen over the last 30 to 40 years in particular. However, if you do build in a trend, it looks very, very similar.

This seems to demonstrate that the world is indeed warming.

Andrew

If we assume that the models are approximately correct, how would you go about determining the likelihood that we would see a run of 12-13 data points below the projected trends?

I have run some modelling of this, and it appears that such a run is not an unusual occurrence; indeed, runs of below or above projected trends of 15 or higher (the highest in 10 random datasets being 22) occur in pretty much every single run of my model.

So, it seems to me that the above data is not sufficient for us to say that the models are wrong.

Now, I will admit that I am a novice at this and have possibly made errors. But I am willing to learn. I can provide further explanation if you wish – I can even email you the excel set-up I have for my model if you want. The source for my randomly generated data (no surprise: it is random.org).

The period I’d like the degrees C/decade trend for is the same as the two graphs in your top post. Are we talking about any other periods?

You & DeepClimate are simultaneously discussing two periods. The final point on my graph represents the 20 year trend from 1989 through 2009. The “about 2″ trend for the the first two decades of the century is either 2000-2020 or 2001-2021. These are two different time periods. We don’t have data for the second period discussed in the IPCC and I don’t show it on the graph above. I’ll show it later today.

Since you want numbers now:

Using the IPCC models, I get these 20 year trends:

Ending March 2009: 0.028 C/year and 0.024 C/year (Volcano only, all respectively.)

Ending Dec 2019: 0.22 C/year, 0.20 C/year (Volc only, all ) respectively.

but if you use the volcanic forcing lower values near the beginning of the computed series, and that pushes the trend up, isn’t the trend then an artifact of an anomalously low start point?

If you compare observed 20 year trends in a volcano dip, you will get an observed trend that is high compared to what you expect when there are no volcanos. It overstates warming due to ghgs because that trend shows combined warming from ghgs and recovering from Pinatubo.

Deep Climate appear to be suggesting that we should compare the observed 20 year trends ending in March 2009 to the projections ending in Dec. 2019.

That comparisons is apples to oranges. We know why the model trend ending in March 2009 are high: They are clearing off Pinatubo!

On looking at smoothed trends: I don’t like this method at all. I prefer to estimate uncertainty intervals for trends of any length and to compare observed simulated and observed trends matching the time periods. No matter what length of trend you pick, there is an uncertainty interval. You can get all sorts of weird things if you don’t match the time period.

The period I’d like the degrees C/decade trend for is the same as the two graphs in your top post. Are we talking about any other periods?

What are the trend rates/periods you gleaned from the mean of the 16 model runs for the two top graphs?

Dunno for sure, but if you use the volcanic forcing lower values near the beginning of the computed series, and that pushes the trend up, isn’t the trend then an artifact of an anomalously low start point? Bit like running a trend from 1998 and saying it’s an accurate reflection of a climate trend?

I note that the observed trend for the 25 yrs to 2005 was about 0.177C

(I realize you’re taking on someone else’s suggestion re 20 year time periods in god humour and playing out the results)

20o years must be the absolute minimum to derive a meaningful climate trend. I noticed someone on another thread stating that the WMO 30 year period is ‘arbitrary’, but from what I’ve read it’s selected to weed out noise – weather, ENSO, 11 year solar cycles etc. The PDO oscillation has only been discovered relatively recently, so perhaps the perod should be doubled (PDO is roughly 30 years).

Would you disagree/disagree with D/C’s contention that you’ve selected high-estimate model runs?

Nope. The 20 year lagging trends vary and are currently high. This is due to the volcanoe eruptions, not the models I picked. You’ll also see this in the graphs. The 20 year trends ending in 2030 will match.

The IPCC specifically states a projection of 0.2C per decade for the next two decades – from 2006 (I presume).

“For the next two decades, a warming of about 0.2° per decade is projected for a range of SRES emission scenarios.” – AR4 SPM

Could you reference the table time-lines, please?

Not having the tools and knowledge to run statistics, I have to resort to reasoning from simpler metrics. If your selection of model runs produces a rate that is higher than the projection mentioned by the IPCC (cited above), and the period you’re examining predates the (2006 – 2025?) period, isn’t there something wrong here? Shouldn’t the rate be less than 0.2C?

Would you disagree/disagree with D/C’s contention that you’ve selected high-estimate model runs?

Again, forgive me if the questions seem naive (and repetitive). It’s just that you haven’t answered them directly as yet.

(I know about the volcanic eruptions)

I will check out your new graphs. Cheers.

The graph is the rate from the 16 models I downloaded from The Climate Explorer (except Essence which would mostly drive the rates higher. Essence wasn’t used in the IPCC.)

I did use the volcano runs for reasons stated: i.e. the volcanoes really did erupt on earth. I’ll show both tomorrow.

The IPCC does not projected 0.4C over two decades from 2006. They a) have specific numerical values in a table discussing the changes for two specific time periods, b) provide a graph of projected anomalies over time and c) discuss the projections are based on the multi-model mean from a specific set of models. For the early part of the century the trend is “about 2C/century”, but this is relative to a start point based on the average from 1980-1999. Trends computed using values in tables are all a bit larger than 2C/century.

The contention at Deep Climate was that you had selected an upper-estimate group of model runs, resulting in a decadal trend of 0.28C.

Is that actually the rate you gleaned from the 16 model runs to compare to surface records?

If this is so, I don’t see how that gels with IPCC projection of 0.4C over two decades from 2006. As the IPCC says trends have accelerated, presumably the trend would be a little bit less than 0.2C p/dec if the start point was earlier in time (with or without volcanic forcing).

My questions are no doubt naive. Just trying to understand in simple terms.

IPCC postulates a rise of 0.2C per decade (over the next two decades – from 2006 I presume).

20 year trends vary over time. The IPCC projections compare the average surface temperature of two periods. The periods are: 1980-2000 & 2011-2030.

The distance between the center points is 30 years, and the fact that it’s based on center points of two long periods makes it behave sort of like a 40-50 year average.

If I compute the multi-model mean using all models downloaded from The Climate Explorer, , I get the same result the IPCC got.

However, the 20 year trends vary over time. Factors like Pinatubo affect the 20 year trends in both the models and the observations. Because Pinatubo erupted in 1991, and the dip occurs afterwards, that “dip” falls on the 20 year timelines you are seeing in the graph above. It’s currently moving toward the ‘left’ (i.e. back in time) and tends to create a rise the 20 year trends for for both projections and observations.

Because of these known effects, I prefer to compare like-to-like. That means, I compare 20 year trends to 20 year trends with matching end points. I could compare the current trend 20 year trends to IPCC 40-50 year average, but I prefer apples-to-apples whenever possible.

Chris Pearson suggests that the model values in the graphs are too high (0.28C per decade). IPCC postulates a rise of 0.2C per decade (over the next two decades – from 2006 I presume).

I read in the IPCC that the rate has increased. Were the graphs generated (back in time) with this in mind?

Could you help a non-statistician on this, Lucia?

Really VG?

Willing to wager some money on arctic ice this summer? I’d give you 70-30 odds on arctic sea ice being at least at the third lowest level ever recorded. There is also sketchier data before the satellite era (http://tamino.wordpress.com/20.....n-and-now/) that helps put things in perspective.
—————————————————————-Endquote

When I look at the start of this Arctic melting season, I note that the current melting rate is lower that most of the 20+ years, It appears even lower than the 1979-2000 mean melting rate.
Do you have a suggestioin for plausible mechanisms for this? – I assume we do agree that the current melt is exclusively of first year ice from the polar periphery.

Cassanders
In Cod we trust

Thanks for your follow-up comments. I don’t think governments are necessarily any good at picking winners–but at building roads and internets they do okay. Building a smart grid is something that only a government could do, don’t you think? Building a successful launch capability for space based power satellites is again, something only a government could do.

A lot of current market choices are shaped by signals governments gave in the past that we may no longer want to give. Tax breaks for oil companies, for example. The government not only funds over half the research in this country, it sets broad directions for much more. I’m glad we’re not sending out the same signals about stem cell research as we were last year. I’d like to send out strong signals about energy efficiency research instead.