- JoNova - https://www.joannenova.com.au -

Mystery solved: Rain means satellite and surface temps are different. Climate models didn’t predict this…

A funny thing happens when you line up satellite and surface temperatures over Australia. A lot of the time they are very close, but some years the surface records from the Australian Bureau of Meteorology (BOM) are cooler by a full half a degree than the UAH satellite readings. Before anyone yells “adjustments”, this appears to be a real difference of instruments, but solving this mystery turns up a rather major flaw in climate models.

Bill Kininmonth wondered if those cooler-BOM years were also wetter years when more rain fell. So Tom Quirk got the rainfall data and discovered that rainfall in Australia has a large effect on the temperatures recorded by the sensors five feet off the ground. This is what Bill Johnston has shown at individual stations. Damp soil around the Stevenson screens takes more heat to evaporate and keeps maximums lower. In this new work Quirk has looked at the effect right across the country and the years when the satellite estimates diverge from the ground thermometers are indeed the wetter years. Furthermore, it can take up to six months to dry out the ground after a major wet period and for the cooling effect to end.

In Australia rainfall controls the temperature, which is the opposite of what the models predict, but things are different in the US. (In the US, temperature affects rainfall).*

In Australia maximum rainfall occurs in the summer but it is highly variable, whereas in the US, while the summer rain is heavier, it’s the winter precipitation where the big variations occur. This seasonal pattern makes a big difference. . Both the Australian pattern and the US pattern appear in other places around the world, but the models only have the one scenario. It appears the modelers figured out the situation in New Jersey and programmed it in for the rest of the world, but whole zones of the world are behaving quite differently.

Models predict that temperature affects rainfall — but in Australia the rainfall affects the temperature. No wonder these models are skillless at predicting  temperature and on rainfall — they are even worse.

As far as I know this is new and original research. Tom Quirk has run it past a few people, including John Christy of UAH who notes that this has been seen elsewhere. Let’s keep up with the peer review…

UPDATE: I’ve discovered Ken Stewart reported this correlation back in 2015. So for the record — his post was the first: “over three quarters of the difference between surface and atmospheric temperature anomalies is due to rainfall variation alone.” Some great graphs there….

— Jo

* Added for clarity. A more detailed post coming very soon.

In Australia, the bulk of the rain,
Falls in summer across its terrain,
With less heat above ground,
Where temp. readings are found,
Which the surface through drying would gain

                — Ruairi

Why Satellites and Surface Thermometers Don’t Agree: Explaining the Difference in Australia with Rainfall

Original Research and Guest Post by Tom Quirk

There is continuing questioning of the relationship of rainfall and temperature. Does temperature determine rainfall or is it the reverse…? The following analysis is a comparison of rainfall and near surface (BOM) and lower troposphere (UAH) temperatures for continental Australia.

This analysis shows that  rainfall modifies surface temperatures in Australia.

Figure 1 shows a temperature comparison. The BOM annual temperatures are averaged from 1979 to 2017 and then normalized to the UAH average, an adjustment of -0.33 0C so the two different time series can be compared.

The temperature increases are:

UAH   0.176 +/- 0.036 0C per 10 years

BOM   0.154 +/- 0.048 0C per 10 years

There is no significant difference in trends at 0.022 +/- 0.030 0C per 10 years.

Yearly measurements and analysis

While there is a good correlation of surface (BOM) and lower troposphere temperatures, there are two periods, 1999 to 2001 and 2010 to 2012 where the UAH satellite temperature anomalies are 0.40C above the near surface measurements of the BOM.

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 1: UAH and BOM Australian annual temperatures where the BOM anomalies have been normalized to the same mean value as that of the UAH measurements.

Bill Kininmonth, former head of Australia’s Climate Centre, suggested that this could be linked to periods of high rainfall as the dampened surface would lower the measured temperatures due to evaporation. This fits with other work by Bill Johnston showing a link between rainfall and temperature at individual sites.

A comparison of Australia wide rainfall sourced from the BOM (Figure 2) and the difference of UAH – BOM temperature anomalies (Figure 3) show that there is a correlation.

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 2:  Australian annual rainfall – source BOM

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 3: Annual UAH – BOM difference of temperature anomalies

This can be demonstrated in a scatter plot of the UAH – BOM temperature anomalies and the Australia-wide rainfall (Figure 4) where the slope on the scatter plot is 0.16 +/- 0.03 0C per 100mm rainfall. This shows rainfall has some effect on temperature. The increasing rainfall lowers the near-surface temperature below that of the lower troposphere.

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 4: Scatter plot of rainfall against the UAH-BOM temperature anomaly. The straight line is a least squares fit with a slope of 0.16 +/- 0.03 0C per 100mm rainfall

So there is a relationship of rainfall with temperature.

Monthly measurements and analysis

However the monthly rainfall in Australia shows large variations from month to month with the peak rainfall in summer being four times greater than the winter rainfall. An example of this is shown in Figure 5 for the period 2006 to 2014.

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 5: Example of the large variations in Australian monthly rainfall – source BOM


The seasonality is best removed by expressing the variations as monthly rainfall anomalies. The mean monthly rainfall is shown in Figure 6 for 1979 to 2017 along with the standard deviations for each month. The large rainfall variations are where there are the largest standard deviations in January, February and March.

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 6: Mean monthly rainfall for 1979 to 2017. The error bars are the standard deviations for monthly rainfall.


So after the removal of the mean monthly rainfall, the rainfall anomalies are shown in Figure 7 along with the UAH and BOM temperature anomalies in Figure 8.

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 7: Example of the rainfall anomalies for Australian monthly rainfall


Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 8:Monthly UAH and BOM temperature anomalies. Note the 1 0C differences of temperature anomalies in 2011.

In Figure 8 the monthly temperature anomalies for 2010 to 2012 show the near-surface temperature is 1.0 0C below the lower troposphere values in the summer months of 2011.

The relationship of the UAH and BOM monthly temperature anomalies to the Australia-wide monthly rainfall anomaly for 1979 to 2017 (but omitting 1996[i]) is shown in Figure 9 as scatter plots for the months of February, March and April. This again demonstrates the influence of rainfall where the relationship to temperature anomalies shows monthly variations.


Relationship to rainfall anomaly for February, March and April

0C per 10 mm rainfall anomaly

BOM  surface temperature anomaly

-0.12 +/- 0.02

UAH   lower troposphere anomaly

  0.04 +/- 0.02


Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 9: Scatter plots for BOM (Left) and UAH (Right) temperature anomalies against rainfall anomalies along with straight line fits.


However, rainfall in one month may well leave moisture on the surface for a longer period. This can be seen by using a sliding correlation test of rainfall against temperatures.

A sliding correlation calculates the correlation coefficient in time series by first calculating the correlation coefficient by matching month with month, that is for example, January rainfall with January temperature and matching all succeeding months. Then calculate the correlation coefficient matching January rainfall with February temperature and likewise a one month shift for all the following months. This process is continued with succeeding shifts.

The results of this approach are shown in Figures 10 and 11 for rainfall from 1981 to 2015, a 420 month period.

In Figure 10 there is no correlation of rainfall with temperature until the series shows a sharp negative correlation coefficient when there is no monthly shift. There is also a delayed effect after the rainfall month of six months before the correlation is lost…

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 10: Correlations coefficients found by sliding monthly rainfall time series against BOM near surface temperature anomalies…


On the other hand, Figure 11 shows there is no correlation of rainfall with UAH lower troposphere temperatures.

Australian temperatures, UAH, Satellite, Bureau of Meteorology.

Fig 11: Correlations coefficients found by sliding monthly rainfall time series UAH lower troposphere temperature anomalies…


This final test shows that the major change in surface temperatures in Australia is a result of rainfall and consequent evaporative cooling of the surface.


There are real and significant differences between near surface and lower-troposphere temperatures. The Australia-wide temperature and rainfall data are a clear demonstration of the interaction between temperature and rainfall. The sensitivity of the UAH – BOM anomaly to the rainfall anomaly has a variation which is naturally dependent on rainfall variations and the period of time that the moisture remains on the surface. The temperature changes can be explained by evaporation having a cooling effect on near-surface temperature which is not seen in the lower troposphere. Thus for above average rainfall the BOM temperature anomaly may be less than the UAH temperature anomaly while with below average rainfall the reverse occurs with the BOM temperature anomaly above the UAH temperature anomaly. This can be seen in Figure 1.

This is of course another difficulty for climate models, particularly regional model predictions.



[i]  In 1996 there is no correlation of UAH and BOM temperature anomalies with a correlation coefficient of -8%. This is at a time when automatic weather stations were introduced.

9.6 out of 10 based on 58 ratings