Peru Data Problems
This week Alan McCallum pointed out errant data problems in two Peruvian stations from the Jones 1994 data while browsing records in MS Access experimenting with Std. Dev. parameters. These were 846730 Humanga and 844740 Juanjui. Graphing these stations to show all the monthly means clearly reveals many examples of errant data (note all temperatures are in degrees C times 10, as in the Jones 1994 data files and most other data sets). Some questions arose out of the Humanga graph which lead on to looking at another seven stations from southern Peru and Bolivia, almost all of which have errant data.
Juanjui in the north of Peru, has an errant cold entry in February 1961, another in March 1976 and several months more than 10 degrees too cold in 1979. No follow up has been done in this area.
Humanga has a data entry mistake of -174 ( -17.4 degrees C )in October 1965 and other errant cool entries in the years 1964 to 1967. Several other questionable entries are obvious. There looks to be a warming jump in 1968 and to check on that several comparisons have been made. Huancayo, like Humanga an alpine station supports the idea that there is a warming jump in Humanga at 1968. Cerro de Pasco, also alpine and a little further away supports the 1968 warming jump in Humanga but has a jump of its own in the late 1960's, together with many data gaps and non climatic trends. Altogether an unsuitable station to pin any faith in.
The capital Lima is close by on the coast so curiosity leads the search for better data in that direction. As the graph for Lima / Campo del Marte shows, monthly numbers look to be in order but two annual figures are entered wrongly. 1929 is entered as 83 (8.3 degrees) and 1948 as 100. Callao Airport has data in two sections and several examples of errant data are obvious. For example records too cool in 1921, data gap and warm jump ~1976 and data gap with warm outliers 1985-1987. No corrections are proposed according to CRU 1991 station notes which is surprising considering the obvious non climatic warming trends.
As a check on the cool records in Callao AP in 1921 the longer term station of Arequipa in southern Peru is interesting. The entries in Callao AP for 1921 do look cool compared to Arequipa trends while the large cool jump in Arequipa around 1973-74 has a correction proposed by CRU 1991. Other errant entries are seen such as the cold March in 1989. However the warm jump at 1961 goes uncorrected by CRU 1991 and it is interesting to note many disorderly records post 1949 in Arequipa; look for the errant cool February numbers for example. A notable puzzling feature is the very small range encompassing the monthly means in the 1900-1925 section compared to the later data and other data from this region.. This could suggest that the early Arequipa records are from another place or collected in different apparatus, or subjected to some later data corruption. They could not be relied upon as representing Arequipa climate for those years.
To check on the Arequipa early years the nearby but lower lying Tacna was looked at and again aberrant data entries are obvious around 1984 however the range of the monthly means does confirm the strange early data in Arequipa. Also a warm jump in Tacna is suggested near 1956, particularly in the winter months while post 1973 a cool step is apparent in summer months.
As a final check, on Arequipa the high altitude station of El Alto near the Bolivian capital La Paz demonstrates a jump either end of the 1940's corrected by CRU1991 and the range of the monthly means confirms the abnormality in early Arequipa records. Post 1950 El Alto looks less well in order with several erratic cold summer months for example. Not a record to be used with confidence.
These nine records from Peru and one from Boliviar, contain a disturbing proportion of non climatic, mainly warming errors and aberrations.
You read it first here.
Posted 4, August, 2000
© Warwick Hughes, 2000
Back to Data Quality Page
Back to Front Page