Rainfall Quality Control–A purpose-built semi-automated, interactive graphical application

The National Climate Centre of the Australian Bureau of Meteorology has developed a new data visualization technique for quality control of rainfall observations. The method identifies ‘suspect’ observations in a pre-process based on buddy and range-checking, then presents this information graphically using an IDL application, ‘RainQC’. Satellite data is used in both the pre-process and RainQC. The aim is to help assure the quality of available data on rainfall, which is arguably Australia’s single most important environmental parameter, particularly for the huge areas where conditions for agriculture are marginal and highly variable.

The large spatial and temporal deviations and discontinuities in rainfall observations make fully automated quality control virtually impossible. Further, there are typically 1000 real-time and 4500 other observations every day, and in non real-time use, a month of data is usually processed together, so the challenge is to present information to an operator in a way which allows effective decisions on the accuracy of data to be made quickly and easily.

The rainfall observations are mapped onto a high-resolution topographic map of Australia with the ‘suspect’ observations highlighted. Rainfall data for selected areas of the country are closely examined by the operator (‘zooming’). Within the zoom window, rainfall time series around a suspect observation may be viewed, from which rainfall entered against the wrong day for example may be quickly identified. Changes to data quality flags are made using mouse clicks, while more sophisticated changes are made through an edit table. Rainfall totals accumulated over several days may be objectively distributed by day with accuracy automatically estimated.

This development has reduced the time taken to process one month of data from about a month to under one week with significant improvements in the accuracy and objectivity of the quality control process.