37. South Africa - temperature trends STABLE to 1980

Over the next few posts I will look at the temperature data in Africa, starting with South Africa. Of all the countries in Africa south of the equator, South Africa has the most temperature data. However, the extent and quality of that data is still much less than that seen for even most individual states in Australia, let alone the entire country.

In total there are about 48 long and medium temperature records for South Africa with more than 480 months of data. Of these only five are long station records with more than 1200 months of data. The locations of these stations are shown below in Fig. 37.1 as well as the locations of similar stations in Namibia and Botswana. In South Africa the spread of the stations is fairly uniform. Sadly this is not true of Namibia and Botswana where the station densities are also much lower.

Fig. 37.1: The locations of long stations (large squares) and medium stations (small diamonds) in South Africa. Those stations with a high warming trend are marked in red.

If we average the anomaly data from the long and medium station records for South Africa we get the temperature trend shown below in Fig. 37.2. This has two main features. Before 1980 (when climate scientists first started claiming that global warming was happening) there is no warming at all. After 1980 there is a sharp rise of more than 0.7 °C. Details regarding how the anomalies in Fig. 37.2 were calculated have been outlined on previous blog posts, particularly Post 35 where the temperature trend for South America was calculated. This process involves calculating the mean reference temperatures (MRTs) for each of the twelve months of the year over a specific time interval (in this case 1961-1990) and subtracting them from the raw mean monthly temperatures to yield the anomalies.

Fig. 37.2: The temperature trend for South Africa since 1840. The best fit is applied to the interval 1857-1976 and has a gradient of +0.017 ± 0.056 °C per century. The monthly temperature changes are defined relative to the 1961-1990 monthly averages.

The number of stations in the average for each month in the temperature trend in Fig. 37.2 above is indicated in Fig. 37.3 below. As is clear from the graph, most of the station data was recorded between 1960 and 2000 with only one set of data continuing prior to 1880. Despite this, there is no evidence that the change in the number of stations over time has impacted the average temperature trend in Fig. 37.2 above in a manner that was seen for Indonesia (see Post 31). 

Fig. 37.3: The number of sets of station data included each month in the temperature trend for South Africa.

If we now compare these results with those published by Berkeley Earth we find similar patterns to those seen for data from other countries. Firstly, the average of the adjusted anomalies results in a temperature trend that is steadily increasing over time, not just after 1980, but also before (see Fig. 37.4 below).

Fig. 37.4: Temperature trends for all long and medium stations in South Africa since 1840 derived by aggregating and averaging the Berkeley Earth adjusted data. The best fit linear trend line (in red) is for the period 1884-2003 and has a gradient of +0.87 ± 0.02 °C/century.

The data in Fig. 37.4 agrees remarkably well with that published by Berkeley Earth and shown below in Fig. 37.5, even though I have only used stations with more than 480 months of data, and have not weighted any of the those stations differently when computing the average. This demonstrates that the fairly even geographical spread of the stations across South Africa eliminates the need to apply different weightings to each dataset. It therefore also justifies the use of the same simple averaging process for calculating the temperature trend in Fig. 37.2.

Fig. 37.5: The temperature trend for South Africa since 1850 according to Berkeley Earth.

The outstanding question is why does the data published by Berkeley Earth in Fig. 37.5 differ so markedly from that for the real anomaly data in Fig. 37.2. The answer is of course once again down to the use of breakpoint adjustments and homogenization by Berkeley Earth. The sum total of these contributions have been calculated and are shown in Fig. 37.6 below. They clearly show that large upward adjustments were made to most of the temperature data around 1890-1900 and 1930-1940.

Fig. 37.6: The contribution of Berkeley Earth (BE) adjustments to the anomaly data after smoothing with a 12-month moving average. The linear best fit to the data is for the period 1881-2000 (red line) and the gradient is +0.59 ± 0.03 °C per century. The orange curve represents the contribution made to the BE adjustment curve by breakpoint adjustments only.

Conclusions

1) There is no evidence of any warming of the climate in South Africa before 1980 according to the raw temperature data.

2) There appears to be a sudden sharp rise in temperature after 1980 that is not consistent with the gradual increase in CO₂ levels seen since 1850. 

3) Once again the temperature trend constructed by Berkeley Earth appears to be strongly affected by their own data adjustments via breakpoints and homogenization.

 

36. Lateral thought #2 - does human respiration cause carbon dioxide levels in the atmosphere to increase?

Does breathing contribute to a build-up of carbon dioxide in the atmosphere? This was the subject of an article on the Skeptical Science website that I came across recently that claimed to be debunking a climate myth. That supposed myth was that breathing contributes to a build-up of CO₂ in the atmosphere.

The article is not new: it was first published ten years ago. The central point of the article was to refute claims supposedly made by climate sceptics that breathing by humans adds carbon dioxide to the atmosphere, and so contributes to global warming. But after reading the article and many of the comments I realized that not only was the entire article wrong, so too were most of the comments. 

The motivation for the article appears to be a throw-away comment by Australian academic Professor Ian Plimer, Professor of Mining Geology (University of Adelaide) and Emeritus Professor of Earth Sciences (University of Melbourne), in an ABC radio interview regarding his latest book. The comment was a response to claims made in a green paper by Australian Climate Minister and Senator Penny Wong regarding the threat of climate change where she claimed carbon was a pollutant. In reply Professor Plimer said:

"If Senator Wong was really serious about her science she would stop breathing because you inhale air that's got 385 parts per million carbon dioxide in it and you exhale air with about ten times as much, and that extra carbon comes from what you eat."

I'm still not sure why that statement riled the people at Skeptical Science so much, other than it came from a climate sceptic attacking a supporter of global warming. To me it just seems like a statement of fact and a reference to the carbon cycle. It is therefore doubly puzzling that those same people at Skeptical Science then chose to use the carbon cycle to refute a claim that was not explicitly made, namely that breathing contributes to a build-up of CO₂ in the atmosphere. The argument outlined in the rebuttal by Skeptical Science basically came down to saying:

"Therefore, when we breathe out, all the carbon dioxide we exhale has already been accounted for. We are simply returning to the air the same carbon that was there to begin with."

The problem is this is not quite true. Actually, it is not true at all. In fact I will now explain why breathing by humans may have actually contributed to a build-up of CO₂ in the atmosphere over the last 100 years.

 Fig. 36.1 The carbon cycle.

The first problem with invoking the carbon cycle is that there is no such thing. There is no single carbon cycle. Instead there are multiple interlocking cycles as illustrated in Fig. 36.1 above. I've listed three possibilities below.

Atmosphere  ==>  plants  ==>  soil (bacteria)  ==>  atmosphere.

Atmosphere  ==>  plants  ==>  animals  ==>  atmosphere.

Atmosphere  ==>  ocean plants (algae)  ==>  oceans (bacteria)  ==>  atmosphere.

So the CO₂ doesn't just go round in a circle, as is claimed: it goes around multiple circles. 

The second problem is that the carbon cycle only describes the steady state. So you can’t use it to prove that human respiration isn’t increasing CO₂ levels in the atmosphere because the human population has grown exponentially over the last 100 years. It has almost quadrupled since 1920. That is not a system operating in the steady state or at long-term equilibrium.

In essence, the carbon cycle describes five competing carbon reservoirs or sinks (vegetation, animals, soil, the ocean and the atmosphere) all of which also act as carbon pumps. Moreover, these five reservoirs are all interconnected, and the pumping capacity of each depends on their size. Generally, the bigger they are, the more carbon they will pump. That interconnection means that changing the size of one will change the size of all the others in order to a) balance the pumping rates, and b) to ensure that the law of conservation of mass, as applied to the amount of carbon in the system, is never violated. These changes will happen as the system seeks to find a new equilibrium position or steady state. 

So in principle, any change to either the pumping rate or the size of a reservoir will have knock-on effects throughout the rest of the carbon cycle. That means that any increase in the human population will affect everything else. We can, however, estimate what some of these changes might be based on what we know about the change in human population over the last 100 years.

As the average 70 kg person generates about 1 kg of CO₂ per day, that means they transfer 100 kg of carbon to the atmosphere every year. This carbon comes from the food they eat. With nearly 8 billion people on the planet that equates to about 0.8 GtC per annum (GtC = gigatonne of carbon) being transferred into the atmosphere.

But that is not all. The average person probably eats their own bodyweight in meat every year. So the growth in the human population since 1920 must be reflected in a similar percentage growth in the number of farm livestock. If we assume there is about 2 kg of livestock per 1 kg of human (i.e. a 2 year supply of meat in production), then the overall CO₂ production from both will be about 2.4 GtC per annum. This is about a quarter of our fossil fuel CO₂ output so it is not insignificant. But is this directly increasing atmospheric CO₂ levels as some climate change deniers might claim (although I'm not entirely sure which)?

Some people have suggested that the increases in human and livestock CO₂ emissions are offset by increased crop production. Their argument is that, as all the carbon we breathe out comes from crops, any increase in the CO₂ produced by the human population will be offset by a commensurate increase in crop production required to feed the extra humans and their livestock. That in essence is the core of the original rebuttal from Skeptical Science outlined above. The problem is that this is not true either.

Increased crop production comes at the expense of other types of vegetation (e.g. forests). The total area under human cultivation may increase, but the total amount of land and vegetation won’t. All available fertile land is already fully occupied with vegetation, so any increase in farmland will be at the expense of wild countryside. Changing usage from one to the other does not increase CO₂ uptake because both types of land are already doing this. For example, deforestation in the Amazon region driven by the desire to grow crops and farm cattle does not increase the rate of CO₂ capture in the region. If anything, it decreases it. Forests, so we are told, are the best carbon dioxide scrubbers.

Also, increasing the number of animals does not increase the amount of vegetation or its growth rate. Instead it decreases the amount of carbon going into the soil. Animals eat plants before those plant can die and before they can decay in the soil. This means that animals replace the CO₂ producing capacity of the soil. That is where the substitution occurs. And if the pumping efficiencies of both animals and the soil were the same then nothing much would change as the animal population increases. But they aren’t the same. 

The carbon pumping efficiency of the soil is only 4%. As Fig. 36.1 indicates, the soil contains 1580 GtC globally but emits 60 GtC per annum. Humans store only 0.1 GtC but emit 0.8 GtC per annum. That is an efficiency of 800%. If we include livestock, the efficiency will be broadly the same (800%) but the size of the carbon reservoir and CO₂ emissions will both be about three times greater, for the reasons outlined above. This also means that the increase in CO₂ production from humans and livestock is the same as that produced by about 4% of the Earth’s soil. The consequence of this is that the volume of the soil must reduce by 4% over time as its pumping capacity is replaced by human and their animals, and the amount of carbon entering it from dead plants declines. 

So 63.2 GtC will be lost from the soil while only 0.3 GtC will be transferred to storage in humans and animals, and none to plants. There is only one other place that most of the 62.9 GtC can go: the atmosphere. This 62.9 GtC will increase the atmospheric CO₂ concentration by about 25-30 ppm. So the human population increase could have increased atmospheric CO₂ levels by up to 30 ppm over time, and about 20 ppm since 1920.

Fig. 36.2: A schematic illustration of the carbon cycle on land.

 

To understand this more fully consider the schematic diagram in Fig. 36.2 above. This represents the part of the carbon cycle involving exchange of carbon between the air and land in the case where initially there are no animals in existence. The terms T₁-T₄ are the flow rates of carbon between the three reservoirs, with the size of each reservoir indicated in parentheses. The four flow rates represent carbon capture in plants by photosynthesis (T₁), respiration from plants and animals (T₂), the transfer of dead plant and animal matter to the soil (T₃), and the decay of organic matter in the soil to release CO₂ back into the atmosphere (T₄).

In equilibrium the flow rates into and out of each reservoir must balance. So 

T₁ = T₂ + T₄

(36.1)

 

T₁ = T₂ + T₃

(36.2)

and

T₃ = T₄

(36.3)

Only two of these equations are independent. In addition, the total amount of carbon in the system must remain constant at 2940 GtC (=1580+610+750).

Now suppose the ecosystem outlined in Fig. 36.2 initially contains only plants and bacteria in the soil. Then we introduce some animals. The effect of animals is to eat some of the plants and emit CO₂. This means respiration (T₂) must increase by an amount x and the amount of plant matter entering the soil (T₃) must decrease by the same amount in order for Eq. 36.2 to balance. For the case of the addition of humans and livestock we have already estimated that x = 2.4 GtC per annum. 

The problem is that both Eq. 36.1 and Eq. 36.3 now no longer balance. Only Eq. 36.2 remains balanced. So the soil will lose 2.4 GtC per annum and the atmosphere will gain 2.4 GtC per annum. There is a mass transfer of carbon from the soil to the atmosphere. This will only stop when the emission of CO₂ from the soil (T₄) decreases, as it will do gradually due to the slow and gradual reduction in its volume. When that happens both Eq. 36.1 and Eq. 36.3 will once more balance and the mass transfer will stop. That will happen when T₄ has also decreased by x. As T₄ was initially about 60 GtC per annum, this requires a 4% reduction in T₄, and therefore a 4% reduction in the volume of the soil, i.e. 63 GtC (the rate of decay of the soil and its rate of emission of CO₂ must be proportional to the soil volume). That amounts to a total mass transfer of approximately 63 GtC to the atmosphere, the same as in our preliminary calculation above.

Is this an upper estimate? Yes, probably. It assumes that the growth in the human population and farming livestock is a net gain in terms of animal numbers and that they do not merely substitute for the loss of other species. But we know this is not true. Humans and their livestock do displace other creatures to some extent. This analysis also omits any additional loss of CO₂ to the oceans and changes to vegetation volumes through loss of soil (down 4%) and increasing growth rates due to increased CO₂ levels in the atmosphere (up by 8%). But what it does demonstrate is that when the human population changes, everything else changes. 

 

Conclusion

What we have shown here is that changes to the ecological balance between plants and animals changes the concentration of CO2 in the atmosphere. So respiration by humans and other animals can contribute to a build-up of carbon dioxide in the atmosphere.

35. South America - temperature trends 0.5°C WARMING

The land area of South America is more than twice that of Australia while its population is more than sixteen times greater. Yet it has fewer high quality temperature records than New South Wales.

Overall South America has about 1000 temperature records, but only 21 have more than 100 years, or 1200 months, of data. Of these long station records, the longest is that of Rio de Janeiro (Berkeley Earth ID: 152852). Its earliest temperature data dates from 1832 and clearly shows that temperatures in the city at that time were higher than today (see Fig. 35.1 below). The average temperature then declined throughout the 19th century before recovering over the last 100 years. This behaviour is not unique to South America. We have seen it in both Australia and New Zealand previously, and if other countries in the Southern Hemisphere had longer records, we might have seen it in many other places as well. 

 

Fig. 35.1: The temperature trend for Rio de Janeiro since 1832. The best fit is applied to the interval 1921-1990 and has a gradient of +2.06 ± 0.15 °C per century. The monthly temperature changes are defined relative to the 1951-2000 monthly averages.

However, the extent to which the Rio data reflects the overall temperature trend in South America is harder to determine. Of the 1000 or so sets of station data from South America, only about 318 have more than 40 years, or 480 months, of data, and as I have shown in previous posts, even this length of data is insufficient to ascertain the recent overall temperature trend, let alone its longer term context. 

The approximate locations of the long and medium stations in South America are shown on the map below. It can be seen that they are fairly evenly spread throughout the continent, but with the Amazon region, not surprisingly, being less well represented. 

Fig. 35.2: The locations of long stations (large squares) and medium stations (small diamonds) in South America. Those stations with a high warming trend are marked in red.

In order to determine the overall temperature trend for South America we need to average the temperature anomalies from all the different stations in the region, as I explained in Post 5. These temperature anomalies are not the actual average monthly temperatures for each location, but the amount by which those average monthly temperatures change relative to a defined reference value for that month. That monthly reference temperature (or MRT), is usually an average of the values for that month over a given period, say 1961-1990. However, as I have pointed out in many previous posts, the way anomaly data used by climate groups like Berkeley Earth is calculated is not quite that straightforward. 

The anomaly data used by climate groups is also subject to adjustments via processes such as homogenization and breakpoint alignment. In homogenization, neighbouring sets of station data are compared and averaged to determine the monthly reference temperature values (MRT) for the target station. Once this homogenized MRT is subtracted from the monthly temperature to yield the anomaly for that month, a number of breakpoint adjustments are then made to different segments of the anomaly data in that dataset, supposedly to correct for bad data and other behaviours seen in the data that appear to be incongruous. The problem with both these interventions is that they are not neutral. Analysis from almost all the previous data I have examined so far in this blog has shown that both these interventions tend to add significant warming to the temperature trends after 1900. The data presented here for South America is no exception.

It is relatively easy to calculate the magnitude of the adjustments made by Berkeley Earth, because conveniently Berkeley Earth list both the adjusted and unadjusted anomaly data in each data file, together with the original raw temperature data. If we average the Berkeley Earth adjusted anomaly data from all the long and medium stations identified in Fig. 35.2 above, the result is the curves shown below in Fig. 35.3.

Fig. 35.3: Temperature trends for all long and medium stations in South America since 1832 derived by aggregating and averaging the Berkeley Earth adjusted data. The best fit linear trend line (in red) is for the period 1900-1999 and has a gradient of +0.85 ± 0.02 °C/century.

The temperature trends shown in Fig. 35.3 above clearly exhibit a strong warming of more than 1.0 °C after 1890. The other notable point is how well the data in Fig. 35.3 agrees with the official Berkeley Earth version shown in Fig. 35.4 below. This can be verified by comparing the pattern of peaks in the 12-month moving average in each case. This shows that using different weightings for each set of station data when calculating the average is not needed.

Fig. 35.4: The temperature trend for South America since 1850 according to Berkeley Earth.

So, what happens if we perform the same averaging of station data, but use the raw temperature data without homogenization and breakpoint adjustments? This means first constructing the anomaly data by calculating the MRT using only the actual station data. 

Ideally when calculating the MRT values it is best to choose as long a time interval as possible for the reference average so that those values have less uncertainty associated with them. This means that ideally you want to include as many years of data as possible in the calculation of each of the twelve MRT values. But this poses a number of problems.

(i) Different station records generally have different amounts of data.

(ii) Different station records often encompass different epochs.

(iii) Most station records have periods of missing data or gaps in their record. These may be a single month of data, or they could be several years in length, and there are often multiple gaps.

(iv) Different sets of station data have different temperature trends. Moreover, the longer the temperature records, the more the trend can vary over time within that station record, or the greater the difference in the mean temperatures will be at either end of the record.

The result is that you need to compromise by choosing a time interval for the MRT average that maximizes the total number of stations that can be included in the average, reduces the uncertainty in that average, but also addresses the four issues listed above. The solutions to those four problems are as follows:

(i) Use the same length of time interval for the MRT average for all stations. This is normally 20 or 30 years.

(ii) Choose the MRT time interval so that it overlaps with the epochs of as many station records as possible. In the case of South America this was achieved by using the time interval 1971-2000.

(iii) Set a minimum threshold for the number of years of data required for the MRT for that month to be of sufficient quality. For the data analysed here that was set at 40% (i.e. 12 out of a possible 30 years).

(iv) Set the length of time interval for the MRT average to be shorter than the timescale over which most significant temperature trends are seen. This is normally 20 or 30 years.

Fig. 35.5: The temperature trend for South America since 1832. The best fit is applied to the interval 1900-1999 and has a gradient of +0.54 ± 0.05 °C per century. The monthly temperature changes are defined relative to the 1971-2000 monthly averages.

After the MRT has been calculated for each of the twelve months in each station record, these MRT values are then subtracted from the raw temperature data for that station to yield the monthly anomalies. The anomalies from all the stations are then averaged to provide the regional temperature trend. This trend for South America is shown in Fig. 35.5 above.

The data in Fig. 35.5 clearly shows an upward temperature trend since 1920 of about 0.5 °C. This is similar to trends we have seen in other regions such as Australia and New Zealand, but it is much less than the 1.0 °C or so that is expected based on the HadCRUT4 data. However the picture before 1900 is less clear. There is some evidence of higher temperatures in the late 19th century, but the data is not extensive enough to provide definite proof. There are only about a dozen stations in the whole continent with data that dates from before 1880. Nevertheless, there are definite similarities between the data shown here for South America and other Southern Hemisphere data from Australia and New Zealand. The fact that all three regions exhibit similar trends before 1900 and also after 1900 suggests that the higher temperatures seen in Fig. 35.5 before 1900 are probably real.

What is evident, though, is that the temperature rise in the trend based on the raw data shown in Fig. 35.5 is not as severe as that published by Hadley-CRU, or that published by Berkeley Earth and shown in Fig. 35.4 above. The differences between the two sets of data (from Fig. 35.5 and Fig. 35.3) are highlighted in Fig. 35.6 below.

Fig. 35.6: The contribution of Berkeley Earth (BE) adjustments to the anomaly data after smoothing with a 12-month moving average. The linear best fit to the data is for the period 1901-2010 (red line) and the gradient is +0.33 ± 0.02 °C per century. The orange curve represents the contribution made to the BE adjustment curve by breakpoint adjustments only.

What Fig. 35.6 demonstrates is that the adjustments made to the data by Berkeley Earth are (again) not neutral. In this case they add at least 0.4 °C to the warming since about 1900. But, this is not new. Similar impacts have been seen previously on the temperature trends from many other regions in the Southern Hemisphere (see my other country-based and regional blog posts 8, 18-26, 30-34).

Conclusions

1) There has been about 0.5 °C of warming in South America since 1920. This is much less than is claimed by climate groups and the IPCC.

2) There is some evidence of higher temperatures before 1900, similar to those seen in Australia and New Zealand, but it is heavily dependent on only one or two temperature records, principally Rio de Janeiro as shown in Fig. 35.1.

3) Adjustments made to the data by Berkeley Earth have added significant warming to the temperature trend since 1900.

Addendum

The temperature trend shown in Fig. 35.5 above is the result of averaging almost 300 temperature records. While all these records have over 480 months of data, very few extend back beyond 1900. In fact between 1860 and 1900 the number of temperature records involved in the average for the trend varies from 4 up to 18. This implies an uncertainty in the trend in Fig. 35.5 of between 0.25 °C and 0.5 °C, whereas after 1960 this falls to about 0.05 °C. Thus the uncertainty in the trend in 1860 is about ten times greater than in 1980. This is probably reflected in the differing amounts of natural variation in the 5-year moving average temperature trend for those different epochs. 

It should also be noted that the temperature data shown in Fig. 35.5 before 1860 is dominated by the temperature data from Rio de Janeiro shown in Fig. 35.1. Its reliability is therefore questionable.

Fig. 35.7: The number of sets of station data included each month in the temperature trend for South America shown in Fig. 35.5.