 |
|
The map shows the median surface temperature anomaly for the selected period (on the slider). Temperature anomaly is the deviation from the 1961-1990 mean. Click on an individual country on the map, or in the menu on the right to see its temperature anomaly over the past 150 years on the plot at the bottom. In accordance with the notion that warming is quickest in the northern hemisphere, particularly away from water, the highest temperature anomaly is in one of the most inland landlocked northern hemisphere countries, Mongolia.
One of the lowest is as far south as we can go, in Antarctica.
Some of the other countries to feature on the Top Five and Bottom Five lists owe their places to their small areas, which makes their temperatures fluctuate more, year by year - they just happened to have a warm or cold year in 2017.
We can see this by clicking on them to have their temperature series displayed in the plot at the bottom.
Tajikistan, for example, isn't experiencing a cooling trend at all, if we look at its temperature series - its 2017 was unusually cool.
This dashboard is a bit of an infodump, with plots of the most likely causes of the warming.
These are called forcings because they act to force the climate system away from its equilibrium.
They can all be quantified, and plotted over time.
On the left we show three anthropogenic forcings.
These are all the same, in a way - well-mixed greenhouse gases (GHGs).
They are "well-mixed" because they are uniformly spread throughout the atmosphere all over the world.
|
 |
|
An example of a GHG that isn't well-mixed is ozone, which doesn't last long enough to mix into the atmosphere globally - its concentration varies by location. This was, incidentally, why I did not include it in the analysis to follow, because the model I used is very simple, too simple to take account of geographic distribution. The three GHGs that I included - carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) - all show clear increases from the latter part of the twentieth century up to the present. Mouseover the info icons to reveal more about them. Also shown is human population growth, which also increases - more on this later. On the right are the leading natural forcings that influence the temperature. Carbon dioxide is listed here as well because, to be fair, it also exhibits natural variation over time. We show the prehistoric variation of CO2 concentration - the vertical axes of this plot and the more recent CO2 concentration plot are aligned, clarifying that natural CO2 concentration over the past 400 thousand years apparently never reached post-1960 levels (above 300 ppm). Again to be fair, however, the time sampling over this vast timespan isn't fine enough to draw this conclusion. But modern CO2 levels are unusually-high, to be sure. Solar irradiance is the energy received from the sun at Earth's surface. The variation is less than one percent, but this is thought to have a measurable effect. Apart from the 11-year cycle, there are longer-term trends like the flatlining in the seventeenth century, and the upward trend of the past few decades. The latter gives climate change sceptics hope for a natural explanation for warming, which we will examine. The concentration of stratospheric aerosols (the stratosphere being that layer of the atmosphere that sits above where you probably are, the troposphere) is measured as the transmission of 550 nm light. When a volcano erupts and spews vast quantities of aerosols into the air, light from the sun has a harder time penetrating the stratosphere and more of it is reflected back into space, resulting in a cooling effect. Which is a warming effect when there are no eruptions. A relative dearth of volcanic activity in the mid-twentieth century is thought by some to have resulted in effective warming. The El Niņo Southern Oscillation (ENSO) is a cyclical upwelling of relatively warm or cold water in the Pacific Ocean that affects rainfall in nearby regions like Australia.
It's also thought to have a measurable effect on global temperature.
A temperature anomaly spike in 1998 is thought to have been caused by ENSO.
This spike was disingenuously (just telling it like it is) used by climate sceptics to claim that the global temperature was falling, by using this as a reference year.
This dashboard is also something of an infodump, but builds more towards the analysis to be performed.
The top-left plot shows the temperature anomaly over time, as a single series.
The increase is clearer than in the country-by-country plot shown on an earlier dashboard.
In the top right we show the how the temperature anomaly and CO2 concentration have varied together in prehistoric times.
We can grant that this alone does not demonstrate causation (this attribution has been made by researchers, not that we'll get into it), but it's intriguing.
|
 |
|
The remaining six lower panels plot the temperature anomaly against anomalies in the forcing quantities. This is the first bit of analysis we do. The temperature anomaly is relative to a 1961-1990 base period, over which it averages to about zero. We find the forcing anomalies in the same way: by subtracting from them their 1961-1990 averages. The thinking here is that if we believe the forcings to be responsible for the warming, deviations from the base temperature should result from deviations of the forcings from their base levels. Put another way, anomalies in the former should line up with anomalies in the latter - we might expect to this to be apparent in plots like these. For the anthropogenic forcings on the left, this is exactly what we see.
For the natural forcings on the right, it isn't so clear.
The trends are there, in the right direction, but the scatter is considerable.
Horizontal axes for stratospheric aerosols and SOI are reversed because it's the lower values of these that we expect to associate with higher temperatures.
This dashboard might be the centrepiece of the viz, and was half the reason I made it - if people aren't convinced by specialists in ivory towers who analyse the data for them, let them play with the data themselves in an interactive manner.
We do linear additive modelling (as in the title of this viz), by which I mean that the the response of the temperature anomaly to changes in the forcing quantities is assumed to be linear, and that its response to each is independent of the others and that the responses to individual forcings simply ``add up".
|
 |
|
The modelling here is very simple. For each of the forcing quantities, we calculate an average and a standard deviation over its values from the base period (1961-90) or part thereof, if some data are unavailable (this base period is that relative to which the temperature anomaly is measured). We then suppose that in any given year, each forcing influences the temperature anomaly by a product: its "impact parameter" multiplied by its deviation from the the base period average in standard deviations. An impact parameter of one means that a forcing influences the temperature anomaly by one degree Celsius per standard deviation. ANModel is a 6-parameter model with one parameter each for CO2, CH4, N2O, solar, stratospheric aerosols and SOI. NModel is a 3-parameter model with only the natural forcings. ANModel is tested over a shorter timespan because annual data on anthropogenic forcings are not complete for earlier years. There is no accounting for lag between cause (forcing) and effect (warming/cooling). In one line, NModel's temperature anomaly in the i-th year is: ANModel's temperature anomaly is: The table below shows what a standard deviation is for each of the forcings.
Attempt to make the models fit by experimenting with their parameters. Each forcing quantity has an impact parameter that can be fed to the model via the input boxes. To increase the impact of a forcing quantity, raise the value of its impact parameter. The best fits are generally obtained for values of the parameters between 0 and 0.5. Using the impact parameters, the models predict what the temperature anomaly should have been, and this is plotted at the bottom. The actual, observed temperature anomaly is also plotted for comparison. The deviations of the observed series from the model expectation is plotted as a column graph, and the overall deviation summarised in the statistics at the bottom of the dashboard (lower values indicate a better fit). ANModel is influenced by all 6 parameters, while NModel only has 3 (solar, stratospheric aerosols and SOI).
While the interactive modelling was a neat trick, when we want to know which parameter values drive a model to its best fit of the data, we ask a computer to try them all.
The root-mean-square deviation (RMSD) was used as the statistic to be minimised in these tests.
The RMSDs are between the model predictions and observed temperature anomaly (abbreviated to TA on this dashboard).
Using code written in GNU Octave, parameter spaces for both ANModel and NModel were sampled with a resolution of one hundredth of a standard deviation in each parameter.
An example of how this looks is shown on the heatmap here, with RMSD values from a slice of parameter space.
|
 |
|
ANModel is only ever tested for 1984-2012. I wasn't able to find methane data before 1984, nor was I able to find stratospheric aerosol data after 2012. So this period isn't very long, and probably can't be usefully divided into shorter periods. But data on all forcings in NModel are available from 1876 to 2012, and it's interesting to look at shorter periods within this timespan. In particular, before the 1961-1990 base period (1876-1960), and after (1991-2012). We are interested in whether the goodness of fit for NModel is affected by the introduction of anthropogenic GHGs, most dramatically after the base period. So the search for best-fitting parameter values was done over various time periods.
The values of parameters that achieved the best fits are shown in the table below.
The results are shown in the series on this dashboard.
By splitting the NModel test period, a clear difference is revealed.
NModel stays in touch with the observed temperature anomaly when fit for 1876-1960, but is clearly discrepant over 1991-2012.
It actually predicts the short-term temperature variations correctly, but starts to consistently underpredict the temperature anomaly over this period, and the gap seems to widen with time.
In the final comparison at the bottom of the dashboard, NModel and ANModel are compared over the same period, 1984-2012.
Although we must remember that since ANModel is essentially NModel with some extra parameters and will thus always outperform NModel, the improvement in agreement between model and observation when anthropogenic forcings are included is noteworthy.
What we've done so far is look at the warming, visualise anomalies that might have led to it and fit models that might explain it.
We have found that a decent fit can only be obtained if anthropogenic forcings are considered.
On this dashboard we examine just how important each forcing is likely to be in the cause of warming.
With a Principal Component Analysis (PCA), we reduce our 8-dimensional parameter space (over 1984-2012) to just two principal component dimensions, while retaining 77% of the dataset's variance. The percentages in the principal component axis titles are how much of the dataset's variance resides in each principal component. The original data are plotted in this principal component space year-by-year. The original parameters are plotted as radial vectors; a smaller angle between two parameter vectors indicates that they tend to vary more together. These vectors are plotted on the component axes.
|
 |
|
In the principal component space, temperatures and the anthropogenic forcings all increase to the right. This shows that they tend to rise and fall together. Vectors for the natural forcings are essentially perpendicular to the others, so they have little to do with either human activity or with the overall trend in temperature. These inferences are confirmed by the projections shown, of forcings onto temperature and human population. The most extreme data point, 1992, followed the eruption of Mount Pinatubo in 1991. That did signifiantly affect temperature, but the natural forcings can only help to explain year-to-year temperature variation, and some trends over periods of a few years - not the overall trend. This is highlighted most clearly by the plot in the bottom left. Here we show the contributions of each forcing to warming, according to ANModel. Earlier in this commentary we spelt out the terms of this model in a formula. The contributions of specific forcings to warming are the corresponding terms in that formula. The natural forcings all oscillate, in some years causing cooling relative to the base period. The sense in which stratospheric aerosols ever lead to warming is indirect: when there are no significant volcanic eruptions, there is effective warming, relative to the base period. If this viz has a conclusion, it might be the column chart comparing the anthropogenic and natural contributions to warming. Here we compare the contributions of anthropogenic and natural forcings to warming, which vary over time. These are averaged contributions over the period of time set by the sliders on the right. To see the average contributions of anthropogenic and natural forcings for a particular timespan, type the start and end years, or use the sliders. There is one year in which, according to this, the natural contribution to warming was greater: 1987. But no others. If we look at 2010 by adjusting the start and end sliders to that year, the combined warming from anthropogenic greenhouse gases is 0.5 degrees. This is consistent with (at the bottom end of) the IPCC's assessed likely range for warming over the 1951-2010 period attributable to well-mixed greenhouse gases (Figure 1.9 in the IPCC's Climate Change 2014 Synthesis Report). With our minimal assumptions and relatively-unsophisticated analysis, we've managed to obtain a similar result.
The criticism that climate models are so complex that that they have enough fudge factors to be made to say anything and that's the reason they say what they do, doesn't hold up.
Our modelling is so simple that it has no right to be correct, but it obtains the same result, to first order.
To see the viz by itself, without the additional features on this page, look here. The data I used, along with code I wrote, can be found here. Numerical analysis (apart from the basic computations) was carried out with GNU Octave. The title of this work references a line from a David Bowie song, "Heat" (The Next Day). Illustrations on this page: still from the 1973 film Amarcord; still from a video accompanying "Heat", apparently from The Art of the Brick; another still from Amarcord. The Amarcord scene with the peacock is included in this passage. 
|