Gridded Decadal & Multi-decadal Rainfall Metadata

Dataset

Custodian

Description

Data Currency

Dataset Status

Access

Please note that the copyright for any data supplied by the Bureau of Meteorology is held in the Commonwealth of Australia and the user shall give acknowledgement of the source in reference to the data.

Apart from dealings under the Copyright Act 1968, the user shall not reproduce (electronically or otherwise), modify or supply (by sale or otherwise) these data without written permission from the supplier.

Please contact us (see details below) for more information.

Data Quality

The analyses (grids) are generated using statistical interpolation (Gandin 1963; Lorenc 1981), which is basically a form of least squares regression applied in two dimensions. Often referred to as optimal interpolation (Daley 1993), the technique yields estimate of grid-point values from a weighted sum of observation increments to a background value. In the analysis of monthly rainfall, the background value is related to the monthly climatology (a reasonable first-guess of the rainfall in a given month), and is set at 0. Correspondingly the increments are the anomaly-ratio of monthly rainfall, formed from the difference of the rainfall from the monthly climatology, relative to the monthly climatology. This way the background value has a zero mean.

The background error correlation and background error variance (BGE) form the most important elements of statistical interpolation, which are determined from interstation correlations for each month, across the climatology period 1981 to 2010. For a given realisation, the observation increment is spread out according to the spatial structure of the respective BGE for that month. This means that the level of accuracy is expected to be better in data dense areas and degrade to those defined by the BGE in areas of data sparseness.

To effectively account for low network density in parts of Australia, two passes of the statistical interpolation technique are applied (following the method of Glowacki et al. 2012), first at a low resolution (0.25 degrees) and a second time at the final resolution of 0.01 degrees. The first pass generates pseudo-observations in data sparse regions, which are limited in their inclusion, by sampling the grid at 2 degree intervals. Pseudo-observations have little or no evident impact on the final resolution analysis away from data sparse regions in Australia's interior. The resulting rainfall anomaly-ratio grids are converted back to rainfall totals by adding one then multiplying by the monthly rainfall climatology. See Evans et al. (2020) for full details of the analysis scheme.

Contact Information

Metadata date