This post follows on from Part 1 where a discussion about urban heating was left unresolved in the comments.
- Flicking through many southern African temperature records I have observed two main trends. One is mainly flat, the other is clearly warming. I believed that rural records (towns with <10,000 inhabitants) displayed flat trends while urban records may be flat OR show warming. The theory I want to test in this post is that all warming in southern Africa is linked to urban heating.
- Warming records and flat records are geographically mixed (Figure 1). Southern Africa is either warming, which means that all the flat records are wrong, or it is not warming which means all the warming records are spurious.
- Lourenco Marques in Mozambique has a population of 755,000 and shows a clear warming trend. I placed Lourenco Marques at the centre of a NASA GISS search that returned 33 records (GHCN V2 unadjusted). ALL 16 rural records are plotted together with 7 urban records that show flat trends. Ten urban records that show clear warming are treated separately.
- The 23 rural + urban record group displays a flat to cooling trend. The 10 urban group displays clear warming of about 1˚C per century. It seems quite clear cut that the perception of warming in this part of southern Africa is down to urban heating that may be linked to deforestation, land use change, residual heat from energy consumption and badly sited thermometers. But not all urban sites are affected, for example Pretoria University does not display warming while Pretoria does.
- There are very serious issues with data structure. There are a great number of rural records that span 1960 to 1990. NONE extend beyond that. And there is only a single “flat” urban station that remained open beyond 1990 and that record is suspect (Durban, see part 1). Therefore, to define temperature since 1990 it would be necessary to use urban records known to be affected by urban heat. It seems odd that all of the urban stations displaying flat trends were closed.
- I would strongly advocate that urban records shown to be affected by urban heating SHOULD NOT be used in temperature reconstructions. And yet it seems that this is exactly what all of the main groups (GISS, NCDC, Hadcrut, and BEST) have done.
Figure 1 33 Stations used in this post centred on Lourenco Marques in Mozambique. In yellow (white text) are the 10 urban records that show warming. In red, the 16 rural + 7 urban stations that show no warming.
Figure 2 List from the NASA GISS web platform showing the 33 stations within 672 km of Lourenco Marques together with the populations. Data are GHCN V2 records from this NASA GISS source.
Figure 3 Temperature spaghetti plot for the 23 flat rural and urban records. Note how all bar one of these records stop around 1990. Only Durban, that has dubious quality, can be used from this group to define temperature trend 1990 – 2011. There is a similar problem with older records. It is only from 1917 that there is more than one record. Note the high density of records 1960-1990.
Figure 4 Temperature spaghetti plot for the 10 urban records that show clear warming trends. The populations are listed below:
- Lourenco Marques 755,000
- Volksrust 15,000
- Pietersburg 27,000
- Estcourt 15,000
- Jan Smuts 1,417,000
- Pretoria 573,000
- Johannesburg 1,417,000
- Vereeniging 250,000
- Bethlehem Air 30,000
- Rustenburg 35,000
Figure 5 Temperature anomalies (dT) are calculated for each of the 23 stations (Figure 3) using the average temperature for that station as a reference value. The average dT is then calculated for the stack. There is clearly little sign of warming. But beware. There is only a single record (Durban) pre-1917 and a single record (Durban) post 1991.
Figure 6 Same as Figure 5 but the time axis is chopped at 1917 and 1990 to remove the single record part of the stack and a 5 year centred running mean run through the data to reveal the structure. There is a temperature top in 1984 but the overall trend is slowly cooling, and these data, that are regarded as the most reliable, cannot be used to define the time – temperature trend since 1991.
Figure 7 Equivalent to Figure 5 for the 10 urban stations that show clear signs of urban heating.
Figure 7 Equivalent to Figure 6 for the 10 urban stations that show clear signs of urban heating. About 1˚C warming per century is evident. Note that in this part of Africa, it is only these stations, evidently affected by urban heating, that provide evidence for regional warming.
In southern Africa there is a serious issue with station closures that looks like it may have been selective. ALL rural stations that show no evidence of warming have been closed. All urban stations that showed no signs of warming bar one have been closed. Only urban stations that show signs of man made urban heating have been kept open and these stations should not be used in temperature reconstructions. This means there is no reliable record of temperature change in this part of southern Africa since 1990. What to do?
In order to get long records and recent records in southern Africa it is necessary to use urban records. This analysis shows that urban records that are warming are almost certainly affected by urban heat and need to be avoided. Hence, only long urban records that show no sign of warming have to be used. Imperfect I know. The keepers of the records have a lot to answer for.
Homogenisation of raw data may account for a small amount of spurious warming in southern Africa. Selectively using urban stations affected by urban heating accounts for a lot more.
Figure 8 The temperature anomaly stack for 26 widely spaced stations (urban and rural) presented in Part 1 are probably the most reliable representation of temperature – time in this part of southern Africa.
Figure 9 In the comments to Part 1 of this post, Roger Andrews produced this nice graphic that basically shows the same temperature trend as Figure 8 above and the regional distribution of long urban records that show no sign of regional warming.