Climate Change 2001: The Scientific Basis

Looks like the actual figure for anthropogenic emissions in a given year can only be off by about 6.5% off from the estimates.

That just about settles this. You were only able to show co2 rise could outweigh co2 emissions in a given year by assuming anthropogenic emissions could be off by about 30%.




That is true for one measurement, but not the average of many measurements which is where the co2 figure for a given year comes from. The error range for that figure will follow a normal disribution. co2 measurements are taken thousands of times during a year, so the error range due to measurement error is far lower for the average than for a single measurement.

I used your assumptions:
1. that the max co2 rise is 3.33 for 2006
2. that the error range of 1.33-3.33 for 2006 co2 rise has equal probability distribution.

Those are nice assumptions for me to use because they are the best case senario for you. If the error range is normally distributed (as I believe it is) in which case co2 rise above 3ppm in 2006 is unlikely, then the results will show the co2 rise is close to the actual average figure given of 2.33, and that makes it very difficult for co2 rise to be more than co2 emissions.

Right, so from what you are saying then at most the average of the 1997 measurements could be 1ppm off, and the average of the 2006 measurements could be 1ppm. That means subtracting the average 2006 level from the average 1997 level gives you the 10 year rise with a max of 2ppm off. That is what I am doing, and none of your arguments address that at all. I don't know what you mean by "we're not taking measurements every 10 years". Yes we are.

But most probably is not. You'd be lucky to find one year in 10 that did that.

That is not necessarily true I believe it follows a normal distribution where the further you get from the average the less liklihood there is.

But I did show you what happens by even assuming equal probability:

1.33-1.73 20%
1.73-2.13 20%
2.13-2.53 20%
2.53-2.93 20%
2.93-3.33 20%

That represents an equal probability across the error range for co2 in 2006. Any value has as equal chance of being the actual value as any other. But ranges of course have differeing probabilities. The probability of the actual value falling above 2.93 is just 20% for example.

No the source does not say that. It says that the actual value has 90% chance of falling somwhere between 6.0-8.2 has 90% chance, not that all those values inbetween have equal probability. You might find that the 50% confidence interval is just 6.8-7.4 for example (and it looks like from the above links that the error range is even less that this)

But the probability of the actual value being more than 2.93 is 20%, with 80% probability of it being less than 2.93.

1.33|-----20%-----|1.73|-----20%-----|2.13|-----20%-----|2.53|-----20%-----|2.93|-----20%-----|3.33

Ie

1.33|--------------------------------------------80%--------------------------------------|2.93|-----20%-----|3.33|

Only 20% probability of the actual value being higher than 2.93

We can be certain with just 10, more certain with 20. More than 50 and we have to start using ice core measurements for the pre 1950's period. But it would still work out as higher co2 emissions than co2 rise.

Why not use a period in which the error rate gives us good enough certainty of the result? 20 years does it for sure.

It's meant to be used exactly how they have used it. Observe:
co2graph

Notice they are perfectly happy to use the data to show long term trends. They even show the rise for the last 60 years. Yet you seem to be complaining that we shouldn't calculate the rise over the last 10...

And monthly, annually, and on decadal time scales. All of them, not just hourly. That's like claiming that because market levels are measured every few seconds, therefore they intend to see variation over only a few seconds and that they don't intend to perform analysis on weeks, months, or even 10 years of stock market trends.

You say "the next best thing" as if the monthly averages are somehow inferior to the hourly averages. You are aware that by measuring every hour they implicitly measure every month too? I don't see you argument at all.

If you only want to use measurements on the 1st hour of the 15th day of every month then that option is available. However it won't be as accurate as taking the average measurements over each month.

The intent of the dataset is not to take hourly readings, it's to compile hourly, daily, monthly, yearly and decadal trends. All of them, not just hourly.

Seeing as I am performing the same calculation you did, but for two year - 2006 and 1997, I don't see your point. I am not averaging an average of an average, I am taking the difference of two averages of averages.

With a very low probability. On the otherhand the probability that co2 emissions over the past 10 years are more than the co2 rise is very high to be beyond reasonable doubt.

The background noise affects the yearly figure, not a single measurement. Measurement error has very little impact on the yearly average, as by taking many measurements you reduce the overall measurement error. For example imagine co2 levels were static at 380ppm for an entire year, and you took 8760 hourly measurements over the course of that year.

A single measurement has an error range of 379.5ppm - 380.5ppm, and assuming equal probability across this error range any one measurement could fall anywhere between with equal probability. But the average of 8760 such measurements does not have the same equally distrbuted error range between 379.5ppm - 380.5ppm.

I did 1000 runs of averaging 8760 hourly measurements of random value between 379.5 and 380.5 and got these results:



The lowest average I got was 379.912, the highest was 380.083

That's an error range for an average of about +-0.085. rather than an error range of +-0.5 for a single measurement.

Also you can see it the probability of an average falls the further away it is from the actual value, ie a normal distribution.

It's overwhelmingly certain that this isn't the case.

Land use changes are technically not emissions, but because they result in loss of absorption they therefore result in extra co2 in the atmosphere, and that's why scientists do include them as anthropogenic emissions. Anthropogenic sources might be a better phrase, but that's just an argument over words.

When the 90% confidence interval is between 3.1 and 3.5, 0 is not going to be likely at all. Beyond reasonable doubt.

Your the one trying to use the table wrongly, not me. That table simply shows the partitioning of co2 emissions into natural reservoirs. It's probably based on ratios, ie they know the oceans absorb a certain proportion more co2 than forests, so they plug in the first set of numbers to get the second.

That is not true, it is based on false assumptions on your part that a figure of 0 into the atmosphere is just as likely as a figure of 2.9 going into the atmosphere. Just becuase both are outside the 90% confidence interval doesn't mean both are just as likely.