Tuesday, March 25, 2008

Global Warming...We're All Going to DIE!!!!

...and if you believe that I have some property you will love in Nevada that will soon be beach-front...get in while it's hot (well, at least I thought that was a funny pun).

Undoubtedly, you have already heard the hype - we humans are dooming our planet to imminent demise because of our air pollution. The exact consequence is not defined. Some say we will be paralyzed by drought. Others are certain we'll be plunged into the next ice age. Alternatively, many of our major cities could be flooded by rising ocean waters resulting from melting polar ice caps. Regardless of the assumed destruction it will be dire and it is clearly caused by human emissions...or so we're told.

We're also told that all scientists agree that Global Warming is harmful and is caused by mankind. However, reality is that over 20,000 scientists signed the Oregon Petition Project that made the following statement:

There is no convincing scientific evidence that human release of carbon dioxide, methane, or other greenhouse gasses is causing or will, in the foreseeable future, cause catastrophic heating of the Earth’s atmosphere and disruption of the Earth’s climate. Moreover, there is substantial scientific evidence that increases in atmospheric carbon dioxide produce many beneficial effects upon the natural plant and animal environments of the Earth.

Simply put, the alarmists currently garnering all the media and political attention have made the all-too-common error of confusing correlation with causation. Furthermore, the doomsday predictions are based upon climate models that attempt to predict complex weather patterns years into the future based upon estimates of future greenhouse gas emissions. Compound this with the vast number of assumptions and unaccounted for independent variables and one has to wonder how such models could ever be accurate years into the future, no less a few weeks from now. Similar models are used to bring you your weather forecasts every day...how reliable are those?

What most scientists actually agree with is a 1.2 degree Fahrenheit temperature increase since the 1880s. Not all of that can be attributed to mankind because other factors are important such as solar radiation, irrigation, and normal climate changes (that's right...imagine that the climate could actually change over time). Atmospheric CO2 has increased by 30% since 1880. In equivalency units, intended to account for the other greenhouse gases (e.g. water vapor, methane, etc.) there has been about a 60% increase in CO2. If we attribute ALL of the warming since 1880 to the increases in CO2, then another doubling of atmospheric CO2 (in equivalency units) would cause a 2 degree Fahrenheit increase in global temperatures. This is not the 3 to 11 degree increase predicted by the models.

I could spend all day discussing the relevance of the data, but the simple fact is that future predictions are based upon complex models. These models have not been validated and there is no agreement on the accuracy of the models within the scientific community. Furthermore, the actual consequences cannot be predicted by the models. Most importantly, the link between human activity and global temperatures has not been established; therefore, proposed reductions in CO2 emissions could cause more damage than good for all we know. More than likely, though, nature will be minimally impacted by human activity and all this discussion is moot.

Yet it is still intriguing to ask, "what about the consequences of Global Warming?" In other words, why would it really be a problem. Primarily, it is a gargantuan inconvenience for humans. We went off and built cities at sea level, they might become flooded, well then that was stupid. We might have to build levies, move, and adapt to the changing climate. But not all the consequences are negative. More CO2, for instance, could bolster plant growth. Warmer temperatures might save lives because more people die from cold exposure than from heat exposure. Shipping through the Arctic Ocean could become more feasible. Additional evaporation could bring more rain to drought ridden regions. Of course, we can't really predict the benefits any easier than we can predict the drawbacks.

So what should we do? I fully support economically responsible environmental action as I believe we should be good stewards of nature. However, we must be careful that what we do does not actually have negative net environmental consequences - such as recycling that requires more energy and resources than producing the materials from raw resources. Other environmentally motivated actions have imposed great hardship on developing nations and create unnecessary economic burden on developed nations and their citizens. Therefore, it is imperative that the known environmental benefit outweighs the known economic burden. In the case of Global Warming, scientists do not have clear evidence that the increase in global temperatures are linked to human activity much less they don't know the true consequences. So you tell me...what should we do?



  1. Joseph,

    I appreciate your comment. I have read through your analysis and some of the information on your referenced data sources.

    First, and most important, correlation does not mean causation. A correlation of 100% may confirm a relationship between two variables but it cannot prove causality. Correlations may be useful in developing further hypotheses which can be empirically tested to determine if there is a cause and effect relationship, but correlation alone is insufficient.

    There are some other issues with your analysis.

    1) The data is problematic for being the basis of a statistical analysis. The temperature data has already gone through statistical treatment. You are attempting to fit a polynomial curve to “data” that is the mean of other data; therefore, your curve fit (r-squared) will be better than a curve fit of the raw data because you have lost statistical information through the multiple data reductions. Furthermore, without reading through the technical papers (which I admittedly haven’t…yet) it cannot be determined to what basis the temperature data is considered anomalous; therefore, fitting a curve to mean values and reporting it as a “reasonable fit” is rather meaningless.

    For instance, I took the same temperature data you used (annual average temperature anomaly) and fit a 3rd order polynomial curve with a resulting R2 of 0.67. However, when I use the monthly temperature anomaly data (not average annual data) and fit a 3rd order polynomial I get an R2 of 0.30. Imagine what would happen if I attempted to fit a curve to all the data original data points! I’m not suggesting that a high R2 value is essential to modeling, rather I’m pointing out that the data have a lot of irreducible randomness; therefore, making predictions based upon your model is a questionable practice.

    2) Using the emission data and attempting to correlate it to the temperature data is likewise troublesome. The temperature data is based upon actual observations (measurements) whereas the emission data is estimated. Essentially, you are trying to model a model. There cannot be direct measurement of the CO2 emission data because it would be impossible to determine whether the emissions were from natural or man-made sources. Do you really believe estimated emissions should be used as predictive determinants of global temperature change? Furthermore, do you believe when your model under-predicts the original model of CO2 there should necessarily be a corresponding rise in temperature?

    3) Consider the data by itself, without the statistical treatment. If your assumption that CO2 emissions are cumulative relative to their affect on global temperature anomalies is correct, then how can you address the negative values seen in the 1960s – 1970s? Looking only at residuals of the CO2 emissions curve fit lends the impression that CO2 emissions were lower then previous years during this time frame. However, the emissions were only lower than predicted and using your logic there should still be a net accumulative effect whereby temperatures still increased compared to the baseline during this period of time. Looking at the residuals errantly suggests there were fewer emissions, but in reality there were just less than what you predict was predicted. Likewise, looking only at the residuals of temperature would suggest the temperatures were less than predicted but provides no information at to whether the temperatures were less or greater than the baseline.

    4) Adjusting the CO2 emissions estimated data by 10 years gives better correlation, but on what basis can you make this adjustment? Is there scientific evidence that proves 10 years is the right amount of time? This sounds like an untestable hypothesis to me. You can’t simply change the time scale because you like the correlation better – you must understand the actual relationship between the data (NOTE: the cited references in your data sources are not conclusive, i.e. a lot of “could be, should, might, possibly”).

    5) For the sake of argument, and as a converse to point #4 above, I could claim that the CO2 emissions are lagging temperature changes. I might shift the emissions data back 10 years assuming the predictions are based upon rough measurements of CO2 which are inaccurate because the CO2 was emitted 10 years prior and only detected after sufficient accumulation made the change notable. In this case, there appears to be an inverse relationship between temperature anomalies and emissions; thus, I can infer that CO2 emissions increase as a result of the higher demand for heating when temperatures decrease. In this case, a strong inverse correlation might lead to an equally improper conclusion enabled by unfounded manipulation of the data.

    6) The standard deviation of the monthly data is 0.57 degrees, i.e. not average annual data and not the actual data (not reported). Most of your residuals fall within this standard deviation; therefore, how do you justify that the residuals are meaningful and not just normal variation within the expected range?

    7) Assuming for a moment that the residuals are significant, and using a 5th order polynomial to model the temperature data (averaged annually) you get a better curve fit. Plotting these residuals against the emissions residuals yields a positive slope, but with a possibility for a negative slope within a 95% confidence interval (slope = 2.4e-5 to -1.4e-5). Which one should I believe?

    8) Furthermore, your slope confidence interval applies only to the slope of the linear fit to the residuals and does not imply that the fit is good, i.e. the slope fits the least squares regression method within the specified confidence interval, but the slope is not necessarily a good predictor of the correlation. The coefficient of determination for the line fit is very poor indicating the line does not adequately address the randomness of the data.

    9) Just for “fun” I used the correlation function in Excel to calculate the correlations between the annual average temperature anomaly data and the cumulative CO2 emissions calculations. Based upon these results, a 30 year shift in the emissions numbers is the best correlation (by a small margin):

    Shift(years) Correlation coeff.
    0 0.797
    5 0.836
    10 0.841
    15 0.851
    20 0.850
    30 0.852
    40 0.830

    Can I now conclude that there is a causal relationship between CO2 emissions that occur 30 years prior to anomalous temperature events? Should I stick with 10 years? Why?

    Overall, the assessment you make is not substantially different than all the other correlation calculations already professed as evidence for a causal relationship between CO2 emissions and global temperature anomalies. Resorting to correlations is meaningless without understanding the relationship between the variables – how one determines the other. Only evaluating correlations between CO2 emissions and temperature anomalies is short-sighted without simultaneously evaluating the correlations between other variables, e.g. solar insolation, ocean currents, natural greenhouse gas emissions, normal temperature cycles, etc. Furthermore, if you are correct in that we temperature changes lag behind CO2 emissions by 10 years then we are too late and our major cities and ecosystems are already in danger because we cannot affect a change quickly enough.