Indicator: Weather: Data analysis and mapping techniques

In this section, we discuss three aspects of analysing and presenting rainfall and temperature data.

Rainfall and temperature analysis using GrafiX

J.A. Lindesay, K.M. Johnson and A.P. Heerdegen. Department of Geography and Human Ecology, School of Resource Management and Environmental Science, Australian National University Canberra, ACT 0200 johnson@anu.edu.au .

The analyses of rainfall and temperature data presented in this report were undertaken using an analytic and graphic display system known as GrafiX, a suite of teaching, learning and research software developed in the Department of Geography at the Australian National University using the S-Plus software system ( http://www.mathsoft.com/splus (External Link)).

The analysis is based on stl, a nonlinear nonparametric analysis, which is able to identify components of different duration and variation in time series (Tibshirani and Hastie, 1990). This robust exploratory technique deals with complex variation in time series data using loess to generalise variation. This analysis differs from traditional approaches which rely on identifying regular variations in data.

The aim of the analysis is to identify variation in:

  1. the seasonal cycle
  2. the longer term, once the seasonal pattern is removed from the data.

The GrafiX system of analysis and graphic display runs on UNIX platforms under a menu-driven system for ease and flexibility of use. Further information is available on the Internet site of the Department of Geography at the Australian National University: http://www.geography.anu.edu.au (External Link).

Reference: Tibshirani, R. and Hastie, T. (1990) Generalised Additive Modelling, Chapman and Hall.

Updating long-term records with new station data

Janette A. Lindesay, Department of Geography and Human Ecology, School of Resource Management and Environmental Science, Australian National University Canberra, ACT 0200. Email: janette.lindesay@anu.edu.au .

As the allocation of resources changes over time, the Bureau of Meteorology is moving to replace a number of observer-run stations with automatic weather stations (AWSs). Further information is available on the Bureau's web site at http://www.bom.gov.au (External Link) . Some of the impacts of this change on the collection of climate data are already evident in the Australian Capital Region, where a number of long-running stations (many with more than 100 years of high-quality continuous records) have been replaced in this way. The table below shows those stations used in this report that have been, or are being, replaced by AWSs.


Table 3. Comparing manual and automatic weather station (AWS) rainfall data
Station Station Number Date Opened Date Closed Correlation (r)
Batemans Bay 69001 1895 1996 0.67
AWS 69134 1991
Bega 69002 1879 0.89
AWS 69139 1992
Bombala 70005 1885 0.85
AWS 70328 1990
Braidwood 69010 1887 0.96
AWS 69132 1985
Young 73056 1871 1991 short overlap
AWS 73138 1990

The general procedure followed in changing from a station with manually recorded observations to an AWS is to run the two stations in parallel for a period of time, in order to allow the comparability of the two records to be checked. A relatively long period of overlap is desirable, to ensure that the full range of possible climatic conditions is included. The Bureau of Meteorology has run most of the stations in the Australian Capital Region for at least a further five years following the installation of the replacement AWS. Once the overlap period has passed, the original weather station is decommissioned and the AWS then provides the only source of climate data for that site.

Several statistical methods may be used to compare the manual and AWS records during the period of overlap in data collection. A simple technique is to perform a correlation analysis of the two time series; the higher the correlation (the larger the coefficient), the more confident one may be that records from the AWS accurately reflect the conditions recorded at the original, manual station. Such confidence is necessary if the new AWS records are to be used to extend the records for a place once the manual station has closed. The correlation coefficients for the five key stations in the Australian Capital Region that have been replaced by AWSs, and from which data are used in this report, are given in the table above.

It is clear from this analysis that, while most of the replacement AWS records may be accepted for extending the long time series necessary for monitoring climatic change and variability, there are AWSs whose records cannot be used in unmodified form with great confidence. In the Australian Capital Region, the record of monthly rainfall totals for Batemans Bay provides an illustration of this, with a relatively low correlation (r = 0.67) between the two data series for the overlap period. This means that only approximately 45% of the variation in rainfall at the old, manual station is mirrored in the rainfall data collected by the new AWS. By contrast, at Braidwood (correlation r = 0.96) approximately 92% of the variation in the old rainfall series is mirrored in the new series. Compatibility problems between recording stations are particularly evident in rainfall, which is highly variable in both space and time. Temperature data are less subject to such problems.

Fitting surfaces to climate data

M.F. Hutchinson, Centre for Resource and Environmental Studies Australian National University Canberra, ACT 0200. Email: hutch@cres.anu.edu.au .

The Regional climate maps for the Atmosphere section of the 2000 State of the Environment Report have been interpolated from data measured at meteorological stations across the Australian Capital Region, using the ANUSPLIN spatial interpolation package, which is maintained at the Centre for Resource and Environmental Studies of the Australian National University. Continent-wide monthly mean climate surfaces produced by this package have supported a wide variety of environmental analyses, including the National Forest Inventory.

The thin plate spline surfaces fitted by the package produce surfaces across the Region which have accuracy close to that of actual measurements. The key to this accuracy is the ability of the surfaces to incorporate various dependences of climate on elevation.

The elevation dependence of temperature is approximately constant across the Region. This has led to the sea-level plots, with accompanying lapse rates, of growing degree day sums for the growing season and flowering season maps. These maps readily permit accurate estimation of growing degree sums at any place where the location and elevation are known. These maps also show that about half of the variation of growing degree day sums across the Region is due to elevation.

The elevation dependence of precipitation is more complex, and varies spatially across the Region in a truly three dimensional fashion. Thin plate splines are able to calibrate this spatially varying dependence, in a way which is not supported by most two dimensional interpolation packages. Hutchinson (1995) has shown that the dependence of precipitation on elevation across the Australian Capital Region is one hundred times more sensitive than its dependence on geographic location.

Reference: Hutchinson, M.F. (1995) Interpolating mean rainfall using thin plate smoothing splines. International Journal of Geographical Information Systems, 9: 385–403.

Information on Bureau of Meteorology weather stations and climate data can be obtained from:

Contact
Organisation National Climate Centre
Contact Numbers
Phone (BH) (03) 9669 4082
Fax (03) 9669 4515
Email dstran@bom.gov.au
Web site http://www.bom.gov.au (External Link)

living sustainably

Click to expand sitemap