Watch Out for Tricky Graphs!

As the above example demonstrates, A bit of manipulation of a graph can be very misleading. There are two basic rules to be aware of to prevent yourself from being duped; these are:

1) Get the scale right

2) Get the slope right

If we want to see whether the temperature is rising or not, the scale we’re interested in is the change in temperature since a relevant point such as industrialisation or when the solar cooling trend began. If you look at the graph below which shows both instrumental datasets since 1850 as well as a 5-year average, the answer is quite obvious.

People that don’t want to believe that however will refuse to look at the graph and acknowledge the obvious fact that since industrialisation, temperatures have risen. The trick used is to simply change the subject - eg “yes, but the world was warmer x million years ago”, or “look, the temperature goes up and down over time”. No scientist disagrees with this - yes, the temperature does go up and down but the question we’re trying to answer is “has the temperature been going up since industrialisation?”

Think of it like a mountain slope. If you look at the slope below you can see that like the graph it is rising towards the right.

If you step back from the mountain though, the slope suddenly loses its significance.

Does this mean that it is not rising toward the right? Of course not, you’re now just talking about another subject. If you point that out, the person trying to avoid the point will say “ok, let’s go in close” and focus on a small part of the slope such as the small peak toward the right of the mountain, or the low descending ridge just to the left of centre, pointing out that the slope is actually decreasing toward the right there. True, but this is still avoiding the point. The overall slope is rising.

To effectively trick people in this way, the person needs to focus on a precise area of the slope. Some for instance choose to pick the year 1998 as his starting point. Why? Because it is a small peak along the slope. In comparison to that peak, temperatures fall for a little while afterward. If you look at the temperature graph you will see small peaks in just about every decade, but the temperature continues to rise. The small peak in recent years is in fact much smaller than most of these, but not if you block the larger ones from view and blow up the current one. Is the temperature rising overall? The answer is an obvious yes.

The second trick is also used when focusing on irrelevantly small areas. The graph below shows the HadCRUT temperature dataset cherry picked to start on a high temperature and finish on a low one.

Most people when looking at this graph will say “the temperature anomoly in January 1998 was 0.749 degrees C and in December 2007 was only 0.03 degrees; it’s fallen by 0.746 degrees!” This statement is true, but it actually twists the facts because it only considers 2 month’s data out of 10 years. In effect, it is like drawing a line from the 1st point to the last and deleting all of the other points in between. The way to consider all of the points is a mathematical technique called regression, which you can perform quite easily in Excel. If you perform a linear regression on this data, you actually find that while the starting temperature was much higher than the finishing point, the overall trend even for this small stretch of time is still actually still rising. The bottom line is that there is an established mathematical process. We use one formula to find the area of a circle and one to find the slope of a line like this one; when you object to using the right formula you are not engaging in debate, you are rejecting foundational mathematical principles.

NB: This trick only works when using the HadCRUT dataset. The reason for this is that both temperature datasets have been adjusted from the raw measurements to remove bias. Bias comes because some weather stations have been affected as buildings are erected around them, concrete laid to heat in the sun air conditioning ducts placed near them etc. Warming in the northern hemisphere is also much more pronounced than in the south, so as there are many more weather stations in the north the record could be distorted. Both datasets get around this by removing corrupted records and “interpolating” what the temperatures are in places between the weather stations. They do this in different ways and one of the differences is that the HadCRUT data does not interpolate temperatures at the poles. Unfortunately the north pole is where most of the warming has been occuring in recent decades, so this dataset underestimates the warming in comparison to the GISS dataset. The linear regression of the GISS data for the period 1990 and following tells quite a different story: