Annual Temperature Plots - Histograms Tab

Annual Temperature Plots
Histograms Tab

Most presentations of global temperature data either deal with some type of average or display the info on a global map. Both of these presentations can be extremely useful and/or extremely misleading.

The info provided via the GHCN Temperature Plotter's Histogram tab is intended to help you understand what the data actually says.

There are so many options and configurations that using this can be a little confusing. (ok, extremely confusing.) The info on this page should help.

Main options

The Histogram tab displays the number of selected sites verses the selected parameter. These examples use the raw data for the entire planet and the default baseline.

Histogram options Full globe, unadjusted data, Baseline: 1961-1990

Trend - Bin size = 0.01 - This is the long term trend (°C/decade) for each site for the specified date range	Baseline Temperature - Bin size = 1 - The average temperature of the planet is supposed to be either 14°C or 15°C (depending on the reference)

Latitude - Bin size = 10 - Notice that most of the data is from the Northern Hemisphere	Longitude - Bin size = 10 - There are 3 peaks the Americas, Europe/Africa, China/Japan/Australia

NumOfYrs - Bin size = 2 - Number of sites with the indicated number of years - x-axis min/max set to 0-120

Only the Trend plots have the associated dates optionally added to the axis label
The Latitude and Logitude bin size defaults are different than the values used above
All the plots are based on what is currently selected

To save copies of these plots, simply right click the chart and select Save image as... or Copy image (the exact text depends on your browser). Chrome saves these as 180h x 360w *.png files.

Bin size

The application allows you to set the bin sizes. The defaults provide what I think are reasonable values.

When the bins are too large, the plot provides little or no information
When the bins are too small, all you see is noise

This example demonstrates the effect of this parameter - The default is in blue.

Effect of bin size
USA 1900-2014 trend (°C/decade) - unadjusted data

	Bin Size
Red	0.020
Blue	0.010 - default
Green	0.002

Multiple plots

There are a number of reasons to place similar plots on the same chart. Perhaps you want to compare the trend from 1900 to 1930 with the trend from 1980 to 2010. Or Europe to North America. etc.

There are 2 buttons that support comparing multiple datasets.

The Plot button erases the displayed data before plotting the current data. This is identical to clicking on the associated radio button.
The Add button adds new data to the existing plot using the specified color. (Simply changing the color performs the same function.)

These buttons were used to produce the multi-bin image in the previous section.

Raw vs Adjusted comparisons

It is possible to compare the raw and adjusted data using the Plot and Add buttons described in the previous section. However, I use this feature often enough that I added a checkbox to make it automatic.

The application provides 4 datasets

GHCN adjusted
GHCN unadjusted
GHCN adjusted+ocean
GHCN unadjusted+ocean

For this to work, both the raw (unadjusted) and associated adjusted data must be loaded. It does not matter which of the 2 related datasets are selected.

For a good comparison, make sure that the selection criteria are the same for both datasets - normally, you will want the same countries and/or same time periods. Since the number of sites is different, you may want to use the Raw vs Adj Filters tab to refine your selection.

Comparison of adjusted and unadjusted data
USA 1900-2014 trend (°C/decade)

Histogram of the USA adjusted and unadjusted data

Red	Adjusted
Blue	Unadjusted

From that graph, it should be obvious that a major part of the warming signal is produced by the data adjustment process - a process which is poorly documented and is very difficult to review. This particular graph suggests that a major part of the warming signal is produced by simply dropping those sites which are cooling. Of course, it is not that simple.

The ocean data is identical in the 2 associated datasets and provided as only an "anomaly" temperature - meaning that the actual ocean temperatures are NOT provided, only an offset from some unspecified "average" value. A general overview of that data, and an explanation of why I think it is mostly worthless, is provided here.

Other Controls

The user can select the colors used for each series.
Clicking on any radio button will produce the associated plot.
The date range for computing the Trend is set on the Map Trends tab.
The date range for computing the Baseline Average Temperature is set on the Basic Filters tab.
A checkbox is provided to hide the dates on Trend plots. This is needed when the chart contains data from 2 different time periods - such as when it compares data from the early twentieth century to that from a later period.
There are readonly fields that display the number of good and bad sites analyzed. With a limited date range, some sites will not have enough data to compute a trend. (At least 3 data points are required.) This indiaces how many failed that test.
The user can set the x-axis min and max values and the y-axis max value. This is used to remove bad data - for instance, at least one land site had a bogus cooling trend of more than -20°C/decade when the date range is 1900 to 1960! This is also useful for comparing plots where you don't want the scaling to change, and for producing overlays where the scaling can't change.

Comparison of different time periods

It is often claimed that "temperatures are increasing faster now than in the past". In my opinion, the histograms simplify this type of analysis.

These plots use all of the default filters to get only the 523 stations with the longest records.

Temperature Trend (°C/decade) 1900-1960 (blue) and 1960-2014 (red)
Global - unadjusted data, 523 stations	USA - unadjusted data, 328 stations

Global minus USA - unadjusted data, 195 stations	Global minus USA - adjusted data, 187 stations

It appears that globally, there has been a major increase in warming between the 2 periods, but not so much for just the USA. When the USA is removed from the global data (lower left), the shift in the slopes becomes more obvious. For a comparison, I have included the global adjusted data (also minus the USA) - 187 stations. In my opinion, the major difference between the USA and the rest of the globe indicates a serious data problem.

A better comparison might be a couple of 20 year periods - 1900-1920 vs 1990-2010. Using just the raw USA data with all the filters turned on shows no significant increase in warming. Because the spread of the data increased, the Trend bin size was set to 0.05 (the default is 0.01) to produce a smoother plot.

USA Unadjusted Temperature Trends (°C/decade) 1900-1920 (blue) 1990-2010 (red)

328 sites

Just for comparison, the next two plots use all 4561 adjusted global sites that meet the most basic baseline requirements. The first plot shows the trends from 1900 to 1960 - of the 4561 selected sites, 767 did not have enough data points to compute a trend, and several sites had trends larger than ±2°C/decade indicating obvious bad data. The trend histogram for the full 1900 to 2014 period is shown on the right for comparison.

Global Adjusted data Trend (°C/decade)
1900-1960	1900-2014

Of 4561 sites, 767 were not included due to insufficient data Trends larger than ą2°C/decade indicate bad data	Trend from 1900 to 2014 uses all 4561 selected sites

Remember, these plots are included just to indicate how to use the tools and to suggest questions than can be asked. It is up to you to decide which questions are appropriate and to help you decide if claims made in various sources make sense.

One of my primary complaints with respect to climate reporting is a complete lack of information on the data spreads (standard deviations). Hopefully, this tool will bring some sanity to the discussion.

Comment

You will often hear that "the United States is not the entire planet", or something to that effect. However, it has the best long term temperature record and, no matter how you cut it, an average of more than 300 stations spread over an entire continent has to be a better proxy for world climate than 12 trees in Yamal. (According to the article, out of 252 Yamal specimens, 12 showed temperature increases, 34 showed warmer temperatures in the Middle Ages.)

Author: Robert Clemenzi