Is there a single “best” way to display temperature anomaly data? The answer is obvious – NO! The best display depends on what we are trying to show. Statistical charts compare one variable with one or more other variables.
Since our display option affects how we interpret the data, it is important to be clear on what we are comparing. In this post I want to show 3 ways to display temperature anomaly data and the implications that the display method has on our interpretation of the data. I’ll use a map, a trend chart and a dot plot.
Map: 2008 Anomalies Compared With Location
Figure 1 shows a 2008 GISS temperature anomaly map I generated using an interactive application at this NASA site. The color coding displays the 2008 GISS temperature anomaly compared to the baseline period 1951-1980. The yellow – orange – red colors denote positive temperature anomalies, the green – blue colors denote negative anomalies. The white color denotes zero anomaly while the grey color denotes missing data.
Figure 1 compares 2008 temperature anomalies with location. It shows a pattern of increasing temperature anomalies as you move north from the equator. This pattern is known as Northern Hemisphere polar amplification.
This map shows a wide geographic variability, even along the same latitude. We can see areas in the US with negative, zero and positive anomalies. The oceans also show areas with negative, zero and positive anomalies. When we look at a map like this we realize how difficult it is difficult to realistically summarize global conditions with a single number or chart. As good as it is, this map doesn’t tell us about other years. Was 2008 unusual? What happened in the 1990s, early 2000s? Temperature maps display data for fixed time periods.
Users can convert the color codes to actual temperature anomaly values with an open source tool like ImageJ.
Trend Chart: RSS Anomalies Compared With Time and Zone
Figure 2 shows a 3 panel R trend chart that I developed to compare the RSS temperature anomaly trends for 3 zones:
- 70S to 82.5 N: reflects global mean conditions
- Equator to 82.5N: reflects northern hemisphere conditions
- 60N to 82.5N: reflects upper NH latitude conditions
Figure 2 compares the anomalies for 3 zones with time. It includes the 30 year trend line and slope so that the change rate for the 3 series can be compared. The Y axis scale is the same for each series to facilitate direct comparison between charts without having to compensate for changing scales.
This chart tells us about anomalies over time so that we can see what happened last year, in the 1990s, early 2000s. However, we can not see the locational variation that we saw in the map. By using zones to aggregate location data by latitude ranges, we can get a sense for variation between zones. However, zone summaries mask variation within the zones.
This chart shows that the upper northern hemisphere has a much higher rate of temperature increase than the global or northern hemisphere zones. The magnitude of the anomalies and the month to month variation for the upper NH zone are also considerably greater than the NH or global mean series. This is consistent with the northern hemisphere polar amplification pattern.
Dot Plot: Anomaly Trend Rates Compared With Zone
Figure 3 shows a Excel dot plot I developed to compare the temperature trend rates (C/century) with latitude zones.
This dot plot clearly demonstrates that temperature trend rates vary significantly by latitude zones. While the 3 panel trend charts gave us an idea of the variation by zone, the Dot plot provides a direct comparison of change rates with zone.
Figure 3 shows the dilemma of using a single global trend rate to assess climate change. The global rate overstates the temperature trend in the southern hemisphere and understates it in the northern hemisphere. The continental US rate, for example, is 1.44 times the global rate and the upper NH is 2.16 times the global rate.
No Single Chart Can Tell Full Temperature Change Story
Temperature change is a complex process that has significant spatial and temporal variation that can not be fully displayed in a single chart. Our analysis tools and displays need to reflect both the spatial and temporal aspects of temperature change.