Entropic Man, you state that the US has been lucky because it has not been hit by any hurricanes, yet Climate Scientists predicted that more hurricanes would hit the US. It is just as likely that Climate Scientists made an exceptionally unlucky guess.

They said more hurricanes, not no change, or about the same, or even could be more/could be less. None, nada, zilch.

Do you wonder why faith in climate science has peaked?

Martin A

Your calculation looks over complex.

Why not just calculate the standard deviation of your CET linear regression?

EM - "the standard deviation of your CET linear regression?"

I'm not quite sure what you mean or what you have in mind. The 'standard error of the estimate' perhaps?

Here we are interested in the accuracy of the estimated slope, not (for example) how well the computed line represents the temperature data.

There are formulas for computing confidence limits for the slope estimate from a regression. See here for example. Confidence Intervals for Linear Regression Slope They depend on assumptions that don't seem necessarily valid here.

"Your calculation looks over complex" Looks can be deceptive. It's actually quite simple. I suggest reading it step by step to understand how it works.

Golf Charlie

Straw man alert.

This is what NASA say about climate change and hurricanes.

The one way in which global warming could impact hurricanes is by making them more intense. More heat and water in the atmosphere and warmer sea surface temperatures could provide more fuel to increase the wind speeds of tropical storms. Warming that has already occurred since 1980 has increased sea surface temperatures 0.3 degrees Celsius, which should increase the maximum potential wind speed of hurricanes by 1 knot, according to hurricane intensity models. But increases that small could not have been observed yet. “At present, hurricane intensity is measured only to an accuracy of plus or minus five knots, so it is not possible to discern any change that might have occurred owing to warming that has already taken place,” says Emanuel.

In summary

a) No change in the number of hurricanes is expected.

b) a small and presently too small to detect, increase in intensity is expected.

Radical Rodent

Perhaps the increase in named storms is due to our increased ability to detect and accurately measure them. It might also be because more storms are forming.

Please clarify. Do you regard this quiet year as evidence that climate change is reducing the number of storms, or is this normal variation?

Martin A

Perhaps we are pulling in slightly different directions.

You can calculate linear regressions or polynomials for any sample data. The question is what the sample data tells us about the actual behaviour of the system sampled.

For CET the system sampled is the English region centred initially on Oxford. For me the fascination is with what CET can tell us about the climate in England between the 1600s and the present.

Other records and anecdotal evidence suggest that the region cooled during the first half of the period and then warmed. What I would like the statistics to help with is to distinguish

a) How well can CET resolve any trends or patterns in the climate of central England? This is why I am interested in the standard deviations. It bears on whether CET is able to answer my second question.

b) whether there was a cooling/warming pattern or a near linear change. Essentially that asks whether CET or the other data is correct.

EM, you are trying to rewrite the dismal history of climate science now. Obviously NASA has a big financial interest in doing this too.

Your double standards and hypocrisy are a credit to climate science. Radical Rodent is far too polite.

If Climate Science has never made dire forecasts, projections or predictions, as you are keen to pretend, then you have just confirmed that climate science should not receive any further taxpayer funding. It is utterly devoid of any purpose.

EM, forgot to add that your link says that storms are likely to be more intense etc etc.

"......outcomes of an increase in global temperatures include increased risk of flood and drought, and increased risk of storms, including tropical cyclones with higher windspeeds"

Are you hoping to outclass Mann, in Climate Science honesty and integrity? It is there for anyone to read. Perhaps you only quoted the bit that you were told to, without having read it for yourself.

EM jumps the shark again.

Martin A,

I'm not sure what you are trying to determine with the procedure you outlined. Perhaps a more interesting test would be to model the CET record as an AR1 type process. This would give some information as to how likely the trend observed could be the outcome of a stochastic process.

Radical Rodent

You seem to have difficulty accepting that natural variation and long term trends can coexist. You might benefit from reading Tamino on the subject here.

Martin A,

many thinks for the detailed answer and context. I can see now what you were trying to do. Like you I'm not a fan of using confidence intervals in this situation. I'm not very happy with a linear regression of the data itself. It's making an assumption that there is a linear relationship between temperature which there is no a priori reason to expect.

EM also keeps asking for a polynomial regression. This makes even less sense to me. I'm sure he means some kind of moving average, or similar filter.

I think this discussion on geological records and palaeoclimates has just about run out of steam.

Paul Dennis - I think this discussion on geological records and palaeo-climates has just about run out of steam.

You are probably correct about that. What has come out of it?

It seems to me that the paleeo-climate data shows that even in the relatively brief period since the end of the LGM - say since the Bolling/Allerod interstadial there have been quite a number of episodes of rapidly rising and rapidly falling temperatures, ones which exihibit rates of temperature change, as derived from reasonably well correlated proxies, at least as fast as those which seem to have convinced some contributors that the only possible explanation for the recent very short period ( less than a fifth of a millennium) during which temperatures have risen at similar rates as being only explicable by the CO2 released into the atmosphere by the combustion of fossil fuels.

This then begs the question as to what was the cause(s) of the earlier excursions when anthropogenic emissions of CO2 can not have been involved. It also begs the question of why the temperature has been in decline for the last several millennia when it is argued that Milankovic cycles can not be invoked as the sole influence.

And this is without even addressing the pre Holocene, when equally or more rapid temperature changes have occurred within the cold glacial eras ( Heinrich events, Bond cycles and D-O events) in interstadials and episodes such as the 10.3, 9.3, 8.2 ky events and other excursions with high rates of temperature change (both warming and cooling )such as the oldest, older and younger Dryas.

It seems to this observer that the Anthroogenic CO2 "control knob" can not be invoked for any previous periods with high rates of temperature change and thus to invoke it as the sole explanation of a recent , and not a very marked upward excursion ( by paleo-climate standards), is to say the least, adventurous.

Paul Dennis

The reason for mentioning polynomials is that a polynomial regression allows you to find a best fit for a curve if one is present, which a linear regression cannot do.

EM,

You are deeply misguided here. What makes you think that a regular curve be present and that it should be represented by a polynomial. Do you expect it to be of parabolic form, or a higher order. How do you decide? Are you going to fit the polynomial to temperature represented as degrees C, K or perhaps anomalies. How are you going to represent time? It makes a difference. Once fitted what do the coefficients mean? Also, once you've fitted do you extrapolate to the future?

One might try a polynomial fit if one had a model or hypothesis that suggested temperature was a linear, or higher order function of time.

You won't convince me otherwise and I suggest you give this some thought.

Paul Dennis

Marcott et al (2013) shows cooling to the latter 1800is. Other records confirm a Little Ice Age of some type. The temperature record shows accelerating warming from 1880 on.

One would expect a similar trough to show in CET, yet it looks flat, if noisy, and Martin A's data suggests that a linear warming trend is quite a good fit.

When different sources give different answers I am always interested to know why. I am curious to know whether England followed the same pattern seen worldwide or not, and if it did cool and warm, why CET did not pick it up. Since CET is the longest continuous measurement record we have, it's reliability or unreliabilitymis of considerable interest

EM,

what you are proposing won't answer your question. Neither will a comparison between Marcott et al and the CET temperature record tell you anything about the reliability or not of the CET record.

Spreadsheets have a lot to answer for! I see students regularly who can't plot a graph and visually assess the data. Look at the plots Martin made and the filtered trend. This trend looks to be a near linear temperature rise since the 17th century at a rate for winter which is about 0.48K per century. If there is any acceleration it is negligible up to the present day. That is your answer!

OK so it does not show any acceleration in the rate of rise of temperature. There are many interesting questions none of which the data alone can answer. Maybe the record started after the depths of the LIA and has been recording a recovery since the?. Maybe the temperature is related to a gradual rise in population since the 17th century? It doesn't show any acceleration that accompanies the 20th century rise in CO2 - maybe CO2 has little effect on the CET? Maybe the Marcott et al modern (19-20th century) reconstruction is not very reliable?

What you should not do is fit a polynomial which implies some sort of functional dependency between an independent variable (time) and a dependent one (temperature). What we have here is a time series recording how temperature varies as time passes and not any functional dependency of temperature on that time. Again spreadsheets have a lot to answer for. Use common sense.

Pajul Dennis

Use common sense?

I am wary of common sense. It tells me that the Earth is flat, that the stars revolve around the Earth and that I could run the country better than the politicians.

That is why scientists are encouraged to use objective analysis tools rather than doing it by guess and by god.!

EM. Why such inane comments following Paul's strictures about misapplying an analytical technique? A warning that even a mathematical dunce like myself could follow. If this discussion has devolved to the extent that you can prolong the agony in this manner, I add to the clamour: close the suc***r down!!!

EM,

Have you never stood on a cliff and seen the curvature of the Earth, a ship disappearing over the horizon and why shouldn't patient observation of the rotation of the stars suggest Earth is spinning on its axis? All common sense observations. Who said visual inspection was guess work and what has god, whoever he or she may be, got to do with it.

Fitting a polynomial is not an objective analysis. It sounds subjective to me i.e. you want anything other than a zero or first order fit. I'm being deadly serious when I say your analysis should start with a visual inspection of the data.

There are a whole host of tools for analysis of time series data and these would be very much more preferable than a polynomial fit. If you really want to prove to yourself that the data is best described by a linear function plot the log of temperature against the log of time. I'd be very surprised after applying my common sense test of visually inspecting the data if the gradient of such a plot is anything other than one. I don't like fitting the data to a linear function. In this case, however, it does describe the long term secular drift in average temperatures. What it doesn't tell me is anything outside the time range of the series. i.e. it cannot be used to infer temperatures prior to or in the future. By the same token it doesn't tell me anything about the mechanism behind the temperature change.