Record T, or not?

Sure pretty simple stuff. Look at the Milankovitch cycles of ice ages, if you trust in the alley compilation of the ice core data, the ice doesn't melt instantly when the global temperatures rise. If we look at the last interglacial I believe there was much less ice at the trough (of ice) (8m estimated higher sea level), yet carbon is much higher and global temperatures only slightly lower today. Higher global temperatures less, lower more, but there is a lot of hysteresis in the system. No nobel prize here. Others have written the papers and books.

Your first link gives error 403, forbidden. Your second is IPCC which uses thermometers and satelites, they don't try to use ice as a proxie in that chapter.

I don't know what the physics of water has to do with this, we are dealing with complex interactions of warming and cooling air and water, including ocean circulation and fresh and salt water mixing.

 

We agree that "ice doesn't melt instantly" but this is a 'red herring.' My interest is in sea level as a direct measure of global temperature which has two parts:

• temperature rise of the sea water
• land ice melt adding to sea water mass

From Wiki, the distribution:

In its Fifth Assessment Report (2013), The IPCC found that recent observations of global average sea level rise at a rate of 3.2 [2.8 to 3.6] mm per year is consistent with the sum of contributions from observed:

• thermal ocean expansion due to rising temperatures (1.1 [0.8 to 1.4] mm per year,
• glacier melt (0.76 [0.39 to 1.13] mm per year),
• Greenland ice sheet melt (0.33 [0.25 to 0.41] mm per year),
• Antarctic ice sheet melt (0.27 [0.16 to 0.38] mm per year), and
• changes to land water storage (0.38 [0.26 to 0.49] mm per year).

​

The reason of the thermometer is two fold: thermal expansion and the rate of land ice melt. So recently one of the Antarctic stations recorded a record high temperature over 60F. The "heat wave" did not melt all the Antarctic snow and ice that one day but some snow and ice did melt.

About 70% of our planet is covered by sea water and the orders of magnitude greater thermal characteristics means it is less subject to the 'noise' of surface temperatures. That is why I find this graph so useful:

There is 'noise' in both the air temperature and sea levels. Averaging them over a one year period does a great job of mitigating seasonal variability and showing long term trends using the local min/max technique. The 'noise' is another interesting problem but well beyond the scope of the original question.

If our planet were 70% covered with mercury or alcohol, the level would be identical to:

​

I appreciate the irony of liquid based thermometers measuring air temperature at different points when I can wade in the earth's largest thermometer at the nearest salt water beach.

BTW, the characteristics of water makes this work which is why I referenced the engineering data. When I studied mechanical engineering, the properties of materials, especially water, were critical to our studies. There is a reason why "I see math people."

Sorry about this:

The Wiley link was suspicious at the time so I'll see if if I can find a better link. But I'd like to suggest sea level is not a proxy but actually a thermometer:

• temperature expansion - a direct metric
• non-linear ice melt - indicates when polar regions are above 0C
• 'land water storage' comes from rain, evaporation from the oceans, net sum zero

Sea level is not a proxy but actually a direct measure with non-linear effects from ice melting. Temperature melts the ice. BTW, floating sea ice has a very minor effect on sea level because it already displaces the sea water of equal mass.

Bob Wilson

 

We have from the IPCC V and current research that sea surface temperatures lagged land temperatures. This makes some sense as the melting ice acts as a buffer to rising temperatures, as does mixing with colder deeper water. Now the way to discover this lag, to test the hypothesis is to look from satellites for ice extent and thickness and look at land and sea thermometers. Of course if your theory is ice extent should be trusted as much as thermometers, these lags should give you pause.

Here we have a place where warm ocean currents melt sea ice more when land temperatures are colder. This process may lead to faster sea ice melt than current models predict.
Large sea ice changes North of Swalbard

And here we have a negative feedback mechanism, where less sea ice in the antarctic brings about conditions that makes the polar vortex more likely, bringing colder conditions over some regional land.
Study links polar vortex chills to melting sea ice

Normally, sea ice keeps heat energy from escaping the ocean and entering the atmosphere. When there's less ice, more energy gets into the atmosphere and weakens the jet stream, the high-altitude river of air that usually keeps Arctic air from wandering south​

Yep absolutely but we have that thermometer that tells us within a small margin for error the temperature, or the melt waters that have a huge margin for error. If you have one instrument that is accurate within 1 degree and one that is only accurate within 20 degrees, do you try to use both, or ignore the second instrument that you know is inaccurate?

 

The problem with air temperature readings are not the individual instrument accuracy but rather they are point sources exposed to signals with large noise variability. If nothing else, weather effects. We can mitigate the noise with more sophisticated data analysis. But we're still trying to pull a global signal from a lot of noise. I admire those that do that work.

In contrast, the mechanics of sea level and ice inventory measurements from satellites is getting better and provides a global view with less noise than we see in the air temperature measurements. We really only need the loss of ice inventory to calculate temperatures above 0C over a year in a region. We can also use the melt to measure heat absorbed. Unlike evaporation that exchanges heat with condensation, lost ice is 'borrowing' the cold that helps keep this planet from a significant thermal jump.

Still, I think it makes sense to complete a sea-level based, temperature model that includes ice loss. Then we can compare its performance over the same intervals to land, Berkeley, and satellite thermometry. In engineering we often have more than one way to solve a problem and the use of sea level could be a very neat solution.

It was loss of glaciers that gave me some of the first, hard evidence of global warming. It is a small leap to realize the melt and thermal expansion of the seas would give a much easier global temperature metric for the 70% of the surface we don't live on/in.

Bob Wilson

 

Here are some responses from Curry that don't like the data analysis.

Understanding adjustments to temperature data | Climate Etc.
Berkeley Earth: raw versus adjusted temperature data | Climate Etc.

The responses are mainly for wattsupwiththat readers, that seem to get misinformed about accuracy. Definitely lots of people have looked into the biases, and we have fairly good error bars (we know the statistics for global warming errors). If you average these global temperatures over a decade errors are extremely small, but we are looking for a small signal.

The satellite era for ice didn't start until 1979. We also use satellites to gauge air temperature. Sure ice extent has gotten pretty good. I don't see less noise than algorithmicly adjusted thermometers. Ice thickness seems like its needed, again, not nearly as accurate. Sea level is the only one where we have a long record, but it has the same problems as thermometers, so how is it "less noise". At least now I understand the problem you are trying to solve. I don't think it helps though.

Well this ignores what is actually happening. In swalbard, warm Atlantic water driven north by currents is melting more ice and not allowing it to form when air temperature is colder. Similarly in antartic, we find an explanation that some thinning of the ice sheet found in satelites is probably caused by other warmer ocean currents, and geothermal heat sources melting the ice from beneath. Its not simple physics of a small freshwater lake, it is complicated interaction of currents, salt and fresh water, and other factors. We have an extremely accurate picture of the sea and land surface temperatures and yet scientists are still asking questions about the ice as predictions are off.

sure compare away. The the check on your model is the temperature record. If your model comes in with a different temperature than the temperature record, the odds will be the model will be wrong not the temperature record.

Glaciers forming and melting is a great way to understand how the earth's orbit cycle over millennium changes radiative heating. For understanding man's contribution to warming we need a lot more complicated temperature record.

 

She lost me with the first chart, "Global Temperatures, 5-year Smooth" because I deal with noisy, spiky network data today and going back to 1992. Even today, I manage a collection of network monitoring systems that have the following 'averaging' rules:

• 5 minute sample
• 60 minute roll-up
• 24 hour roll-up

The longer the 'averaging' interval, the more misleading the data. The network data is naturally noisy because of the large number of random demands and expanding the average hides these serious events. Professor Judith Curry lost me with "5-year" average but she has done this before.

It is something I'm concerned about too but I don't see that in the initial trend-line data:

The excel 'trend line' function is easy to apply but I've been fooled by it before. In this case, both are using a straight-line average of 1-year duration to coincide with the annual cycle. The risk is the left and right scales were chosen so both have the same, zero crossing lines and Y-axis division correlation between the 0.2 C and 20 mm sea level change. Changing the left and right Y-axis ratio can change the sea level slope which could be misleading.

If I wanted to mislead, the approach is trivial:

1. Observe the 2.5 division Berkeley range
2. Observe the 3.5 division sea-level range
3. Scale right axis by 2.5/3.5 to get the slopes parallel
4. Modify the right axis minimum value to bring zero scales to same line
5. Adjust the right axis division scales to match left and right

Our eyes see the 'parallel lines' BUT afterwards, the sea level slope is hidden in the ratio of the left and right Y-axis scales. Which raises the question, is the 0.2 to 20 mm ratio true or misleading? But I'm doing this for my curiosity, not advocacy. Knowing this trick means I can spot it in those who are trying to mislead . . . "I see math people."

Now I used a simple analysis because Doug was testing the 'hottest Jan-Mar' and 'hottest Mar' 2015 intervals. Sad to say, my inspection found the data too noisy and coarse to support any assertion about hottest.

So I proceeded to another analysis I am curious about, temperature vs sea level. I am using the simple trick of an annual average, trend line with both sets of data. But I prefer to use a Gaussian weighted average because it preserves the spikes. Climate spikes are important. For example "heat stress" and "extreme cold" events:

• 2003 - ~70,000 heat induced deaths in Europe
• 2012 - cold deaths in Russia

My analysis continues.

Late thought, Denial 101x will be addressing the paleo records next week. It is an area where I'm interested but not committed. I would rather work with current, satellite data. But then I've been working with satellite systems for most of my career.

I'm OK with looking at currents but the 'geothermal heat sources' is something that appears to be off by orders of magnitude. We had an earlier discussion about this with 'Mojo' who copy-and-pasted 'volcanoes' without doing the math. It is the 'nature of scientists to ask questions' regardless . . . or rather burden their graduate students. <GRINS>

I'm a curious guy and have demonstrated stunts showing a reasonable understanding of how the Prius saves fuel:



I also know how to 'game' the system:

. . . and recognize when the Taylors, 'Green Human', Consumer Reports, or Judith Curry are 'gaming' the system. The worst are those who fool themselves including my own work.

Bob Wilson

 

That certainly gives me more insight onto your excellent data analysis on the prius.

Here though the question is what you are looking at, and what you are looking for in the data. To me a 10 year smoothing function would be better. What would I try to smooth out? The ocean oscillations effect on temperature that seem related to the 11 year solar cycle. Why 10 instead of 11, people seem to like round numbers, I prefer 11 but I don't want to fight it versus 10 which people like better. Climate scientists seem quite impatient though, and I guess that is why they HADCRUT had been showing the awful 1 year signal and pretending it didn't have this systemic noise from oscillations, leading to big problems with cooling from 1998 using their data. Then they started going along and smoothing. The log of ghg concentration changes fairly slowly, and that is the key figure to man made warming.

I think you though want a temperature signal, not a temperature signal with oscillations removed so climate warming and cooling can be seen more clearly. If you want temperture, and I would use adjusted temperature data to remove station noise, its all available. BEST the organization Curry works with make all that raw data available so you can play to your hearts content. You can even separate out sea surface temperature from land station temperatures. You probably need to hunt for deeper ocean temperatures.from other data sources though. Daily carbon dioxide concentrations are available from Mona Loa.

 

It sounds like you may be interested in the solar radiance. Let me suggest asking Mr. Google for the dates of solar minimum or maximum. Use those dates as the end points for averaging the temperature.

Actually I find the 'noise' more interesting. As we understand these fine details, we're approaching the boundary between climate and weather. For example, this afternoon, I added El Nino/La Nina index scaled to the earlier chart:

There is a general relationship until 2004-2005 when it inverted. This could be the signature of another forcing function but I have not researched this, yet.

A lot of my insights come from extracting signals from noisy data. It is a target rich environment.

GOOD LUCK!
Bob Wilson

 

The noise is not noise its weather, and really weather data. If you are interested in weather data you absolutely don't want a global average, but to look at most at the hemisphere level, while probably looking at monthly or at least quarterly data. You may want to separate the sst from the land temperatures, and only look regionally.

For climate change if you can not model weather enough to remove it, at least average by at least the number of years to smooth out the weather effects from the climate effects. In climate it is perfectly fine or even better to look at 10 year moving averages of yearly global data.

Assume you have a good climate model. If you had you could normalize the weather data on the climate change. If we use a 10 year moving average then subtract that climate signal from weather, you may get a clearer picture. Of course then we have to cut off the last 5 years of the series as we don't have enough data to construct a moving average there.

Again that signal has weather and climate included in it. I don't see much "noise" as measurement error. If you can properly model enso then you can remove its signal from climate data, unless climate change changes enso.

 

This is not the first time bright people have seen a different understanding and I am OK with that. For me, it means digging into the data. I don't ignore as much as I prefer to let facts and data drive my opinions and I remain open to alternate hypothesis based on data.

Now one thing I don't like about the Berkely data is the monthly average. I would prefer all global metrics have a one week interval with average and stNdard deviation after rogue station filtering. These could be used to improve the weather forecasting models.

But this is a hobby and way down in my priority list.

Later,
Bob Wilson

 

The data is not lost it is undefined. You can do a predicted series for years after today and use them, but note strongly in the graph that these are not calculated in the same way.

Now bob appears to be interested in short time frames, which we normally define as weather. The reason for smoothing is a naive attempt to remove the short term weather signal from climate.

If we are talking 30 years, losing the last 5 years is not that big of a problem. The averaging is not monte carlo, but simply a black box modeling technique. A white box technique would be to look at the data and determine whether it is climate or weather and adjust. Unfortunately even among climate scientists these adjustments are disagreed upon.

Now we can put it all out in a multicolored chart if people don't mislead themselves. 10 year averaged trend line that changes to a different color 60 months in the past to a projected trend line. Going back in the past in a lighter different color the variability in weather to the past, and as this gets older variability rises based on possible adjustment error. Starting where actual to projected variability in predictions can be shown in a 4th color, here the variability shown has to include both possible weather and climate. Records in the past of both cold and warm should be shown clearly.

A naive technique if we assume ghg is the main driver of determining climate in the last 60 months is to determine difference in ghg from mona loa 60 months ago from today. Decide that change will be double in the next 60 months. Use 3 (its round and close to average estimated sensitivity) for temperature sensitivity) for sensitivity and put a spot out 120 months. Draw a straight line from the trend, then use this data for future. This is a much better technique than say ipcc 3 that include a chart with short averages then actual data for 1998, that seemed a weather outlier even at the time, but this was pointed out in ipcc 5, after more data was collected.