About Climate Statistics

Introduction

The Climate statistics for Australian locations consists of information for more than 1000 sites. The tables prepared for each site provide averages and other statistics for a number of elements including:

Further information about each of these statistics can be found by clicking on the first column of each row in the statistics tables.

Sites have been included only if a minimum of 10 years of temperature data are available for the site. Thus, statistics for more than 15,000 rainfall-only stations are not currently available on this web site, but may be obtained by contacting the Bureau. Detailed information on the climate data available from all Bureau sites is available on-line, or you may contact the Bureau for specific information.

top

Availability of data

The basic station metadata provided for each site includes the year in which the station was opened and, if applicable, the year that the station was closed. Generally, but not always, observations will commence close to the start date and cease around the time the station is closed. However, the types of observations made over that period - wind, temperature and rainfall for example - may change according to operational needs. Thus the length of record for each element, and how complete that record is, may not be the same. Rainfall typically has the longest observation record at most sites.

The data field labeled ‘Years’ in each row of the climate statistics table contains two sub-fields: length of the record, and the first & last year of available data. The length of the record for an element is calculated by dividing the number of months used by 12, and does not mean calendar or complete years except for the rainfall decile values. It gives an indication of the amount of data used between the first date of occurrence and the last date of occurrence of the element. No statistic is provided if there are less than 10 years of data for the relevant observation.

The range of different meteorological elements observed is not the same at all sites. For example, sunshine duration, maximum wind gust and evaporation are only recorded at some stations. In addition, some weather stations will have ceased recording a particular element or elements during their period of observations. In these cases, there may be no data (for a particular month or element) shown on the statistics report.

top

Quality of data

Data quality control processes, and some of the methods used to derive climate parameters (such as the calculation of clear and cloudy days), have changed since records were first kept by the Bureau of Meteorology. This means there may have been quality issues with the climate statistics (calculated from historical records).
We will soon finish reprocessing many of the data which are used to calculate the climate statistics. This will improve the quality of the statistical information provided to you. In the meantime, the statistics will be based on the same (updated) datasets as used previously. This will mean:

• You may find small differences when comparing the superseded (pre 2007) version of the climate statistics with the most recent. These changes are likely to be associated with the calculation process or the removal of unsuitable data.
• Some of the previously identified climate extremes may disappear in this new product. This is associated with data quality issues.
• Sites will no longer be listed where the original data are not available to verify the statistics. This involves about 120 sites which generally closed by the mid-1900s.

There is a delay, which varies with the type of element and sometimes the site as well, between when an observation is made and when the data have completed the quality control process. This delay is typically greatest for rainfall data. Therefore, recent data may not be included in the statistics for a particular site.

top

Statistics used

Mean values

The mean value, also known as the average, is one of the most common statistics used to provide an estimate of what is most likely to happen. It is not necessarily equal to the most commonly occurring value, which is known as the mode, but for most elements it will be close. By itself, the mean does not provide any information about how the observations are scattered around the mean; whether they are tightly grouped or broadly scattered.

Median and Deciles

Figure 1. Definition of the median and deciles

Deciles are one of the statistics used to provide an indication of the spread of data in a data set (e.g.. a collection of rainfall observations at a site). To calculate deciles, we divide the ranked data set into ten parts. The median is simply that value which marks the level dividing the ranked data set in half. For example, 50 % of Januaries will have a total rainfall at or above the January median and 50% will have a total below it. The median is also known as the 5th decile, decile 5 and the 50th percentile - they are all the same thing. Decile 9 or the 90th percentile for January, means that 90 % of January totals will be at or below this figure. In other words there is a 90% chance of a January rainfall being at or below decile 9 (90th percentile), a 10% chance of it being above decile 9, and a 10% probability of it being below decile 1 (10th percentile). To get the annual decile value, you do not sum the deciles for the 12 individual months, but calculate it separately from the set of annual rainfall totals. However it is possible for the two values to be the same by chance.

The median is usually the preferred measure of 'average' rainfall from the meteorological point of view, particularly for the shorter timeframe. This is because of the high variability of rainfall - one extreme rainfall event will have less affect on the median than it will have on the arithmetic mean. For example, at Roebourne (site number 004035) in the north-west coastal region of Australia, the wet season and the number and path of cyclones can vary significantly from year to year. Figure 2 illustrates the January rainfall at Roebourne over the period 1961 to 1990. The median rainfall of 23 mm is more indicative of the majority of months than is the mean rainfall of 68.5 mm.

Figure 2. The median is particularly useful to represent 'typical' rainfall in areas where the rainfall is quite variable.

Extremes

While generally correct, the extreme values should be viewed with some caution. Extremes may have been recorded using non-standard equipment, particularly in the very early years of the Australian meteorological record; for example, pre-1910 temperature extremes. In addition, the extreme value may represent an erroneous measurement that escaped detection by the normal quality control processes.

Extremes can also be very location specific. A rain gauge at the centre of a concentrated, intense thunderstorm may accurately measure a record rainfall total while a location nearby observed no significant rainfall. Extreme values of minimum temperature tend to be more location specific than those of high temperature.

The extremes in the tables of statistics are drawn from the Bureau's computer archive of climate data. Some sites, which could have had more extreme values in the 1800s or the early 1900s, may not have been entered in the data base. In addition, very recent extremes may not have been processed at the time the tables were created.

top

Climate normals

Statistics calculated over standard periods (commonly a 30 year interval) are often called climate normals, and are generally used as reference values for comparative purposes. The period is long enough to include the majority of typical year to year variations in the climate, but no so long that it is significantly influenced by longer-term changes in climate. In Australia, the current reference climate normal is generated over the 30-year period 1 January 1961 to 31 December 1990.

Climate normals can be used to assess how typical of the current climate a particular event was. For example, the difference between the average temperature for a calendar year at a site and the climate normal average temperature - the anomaly - can be used to indicate whether that year was relatively ‘hot’ or ‘cool’. Normals are also useful in comparing changes in climate over a long period, as illustrated in Figure 3, which shows the increase in monthly mean minimum temperatures in Melbourne during the 20th century.

Figure 3. Comparison of climate normal mean minimum temperature for the two reference periods 1901-1930 and 1971-2000 at the Melbourne Regional Office site (number 086071)

top

Statistics and length of record

The statistics have been calculated over the available period of record, which may differ between elements. For all elements, climate statistics have been derived only when there is at least 10 years of data available. Because of the annual variability in rainfall, a period of less than 30 years of rainfall data may not produce reliable statistics and such information should be used with caution. As a comparison, some 5-10 years of temperature data will provide a reasonable estimate of the mean, although probably not of the extremes.

top

Means for specific hours (9 am and 3 pm) and daylight saving

Due to the effect of Daylight Saving, these values are only nominal for most Australian sites. The averages for 9 am are hence generally a combination of 8 am and 9 am (standard time) values, and those for 3 pm, of 2 pm and 3 pm values.

top