Quality Control of Mean Temperature Measurements of DoM/WMO stations in

------------------------------------------Sri Lanka

 

A Brief Description
Station Temperature Data and Estimates from Other Stations
Map Showing Location of Air Temperature Observation Points
Quality Evaluation of Historical Mean Temperature Data in Sri Lanka

 

 

 

 

 

 

 

 

 

 

 

 

 

A Brief description

Quality assessment of mean temperature data at a near equatorial location such as Sri Lanka is challenging given the limited seasonal variability. Annual temperature in stations other than those in the hills such as Nuwara Eliya, Diyatalawa, Badulla and Kandy ranges between 22 and 29 oC. The inter-annual variation at all station is less than 2 oC.

Here we attempt to build reliable and more complete set of station mean temperature data for Sri Lanka. In addition, where the data does not pass quality control, we will attempt to reconstruct the discarded or missing data using neighboring stations records

[ to top]

 

 

 

 

 

 

 

 

 

 

Location of air temperature observation points

Click on the image to Zoom ---[to top]

 

 

 

 

 

 

 

 

 

 

Quality Evaluation of Historical Mean Temperature Data in Sri Lanka

 

Introduction

Analyses of historical variations of instrumental temperature records are important for many purposes including the investigation of regional effects of global warming. Such analysis is particularly important in locations with long and dense historical data such as in Sri Lanka.�� Systematic meteorological data acquisition was started in Sri Lanka by a British Military official, Colonel Fyers of the Royal Engineers at Colombo in 1853.� By 1901, there were 15 meteorological stations and 90 rainfall-measuring stations.

Data from eleven out of the fifteen stations have been incorporated into the Global Historical Climate Network (GHCN) database .Quality control has been carried out based on a globally applicable guidelines and suspect portions of the data have been omitted.

Quality assessment of mean temperature data at a near equatorial location such as Sri Lanka is challenging given the limited seasonal variability. Annual temperature in stations other than those in the hills such as Nuwara Eliya, Diyatalawa, Badulla and Kandy ranges between 22 and 29 oC.�� The inter-annual variation at all station is less than 2 oC.

Here we attempt to build reliable and more complete set of station mean temperature data for Sri Lanka. In addition, where the data does not pass quality control, we will attempt to reconstruct the discarded or missing data using neighboring stations records.

Stations that were used in the analysis

Station Name������������� Latitude(oN)���� Longitude(oE) Elevation(m)��� Duration of Record

Jaffna������������������������� 9 39���������������� 80 1���������������� 4��������������������� 1871-1994������

Mannar����������������������� 8 59���������������� 79 55 ������������� 4��������������������� 1870-2000������

Trincomalee��������������� 8 35���������������� 81 15 ������������� 3��������������������� 1866-2000

Anuradhapura������������ 8 21���������������� 80 23 ������������� 93������������������� 1870-2000������

Puttalam��������������������� 8 2������������������ 79 50 ������������� 2��������������������� 1869-2000������

Batticaloa������������������� 7 43 ��������������� 81 42 ������������� 3��������������������� 1869-2000

Kurunegala����������������� 7 28���������������� 80 22�������������� 116����������������� 1885-2000������

Kandy������������������������� 7 20���������������� 80 38�������������� 477����������������� 1869-2000������

Badulla����������������������� 6 59 ��������������� 81 3���������������� 670����������������� 1869-2000������

Nuwara Eliya�������������� 6 58 ��������������� 80 46�������������� 1895��������������� 1869-2000

Colombo�������������������� 6 54� �������������� 79 53 ������������� 7��������������������� 1869-2000

Ratnapura������������������ 6 40� �������������� 80 25�� ����������� 34 ������������������ 1969-2000

Hambantota��������������� 6 7 ����������������� 81 8���������������� 16������������������� 1869-2000

Galle��������������������������� 6 2 ����������������� 80 13�������������� 13������������������� 1869-2000

Diyatalawa����������������� 6 49���������������� 80 58�������������� 1248��������������� 1901-2000

Methodology

The following analysis was carried out:

  1. Exploratory data analysis was carried out on the time series to check for errors in transcription and data entry. The annual temperature time series for the different stations was constructed and its histograms, mean, variance and extreme values were examined. The time series was also examined for mean temperature values that exceeded 3 times the inter-quartile range. Such data was removed unless a similar anomaly was replicated in neighbouring stations
  2. An average �All Sri Lanka Temperature� (ASLT) series was constructed. A cross-correlation table was constructed of the mean temperature records of the15 stations and the ASLT.� In addition, correlation tables were constructed for parts of the record (1869-1894, 1895-1919, 1920-1949 and 1950-2000). When the correlations were lower, even smaller intervals were tested to see whether there was any correlation.
  3. The temperature variation at each station was reconstructed based on the data at the five stations that had the highest correlation during the years from 1900-2000.

The reference time series were estimated as

where

�� -� Standardized anomaly for jth year

Resultant weight for the ith station� �is given by

Here

���� -� Correlation between the station and the ith surrounding station

����� * ����-� Number of data used to derive the correlation coefficient for ith station

Years that showed significant discrepancies between the observed and reconstructed was identified as suspect. Based on a comparison of the observed and reconstructed data and the lack of any relationship between the observed and reconstructed data, data for the suspect period was discarded.�

4.�� Site-change effects were estimated using the cumulative sum technique� by the difference between the target station and weighted mean of neighbouring stations and monthly temperature values were adjusted.

Results & Discussion

Exploratory Data Analysis

Outliers: The outliers in the time series and histograms were identified and the data was checked from the various sources, such as the Department of Meteorology archives, the GHCN archives, reports of the Colombo Observatory, the Department of Meteorology and the reports to the Colonial office. In most cases, the data from the different stations were consistent but there were exceptions, which were corrected for. Even after these corrections, there were still outliers which were beyond a threshold of thrice the Inter-Quartile-Range of the distribution and which were identified as suspect. These were Jaffna and Ratnapura (1896), Galle and Hambantota (1901), Nuwara Eliya (1997) and Anurhadhapura, Batticaloa, Badulla, Galle, Hambantota and Diyatalawa�� (1998).��

Cross-Correlation

In summary, the data� whose correlation with ASLT were below 0.3 are listed below.�

Jaffna ������������ 1869-1894 (-0.1), 1981-2000

Puttalam��������� 1894-1905 (-0.23)

Batticaloa������� 1869-1880 (0.10), 1895-1905 (-0.16)

Hambantota��� 1869-1880 (0.28), 1895-1905 (-0.15)

Rathnapura�� � 1869-1880 (-0.30)��

Comparison of Data with Reference Time Series

The comparison of annual temperature at each station and the reconstruction of it using data of other stations that shows the highest correlation with it helps identify shifts in locations of the station, calibration errors and declines in quality of the data.�

 

References

Thomas Peterson.� The Global Historical Climatology Network (GHCN).

http://www.ncdc.noaa.gov/cgi-bin/res40.pl?page=ghcn.html

D.A. Rhoades and M.J. Salinger. 1993.� Adjustment of temperature and rainfall records for site changes. International Journal of Climatology.� 13:� �899-913