A statewide grid surface (80m spatial resolution) delineating recent accumulated rainfall (millimetres since 9am) across Tasmania is produced in near real-time. The outputs are dynamic with maps updated at hourly intervals (there is a production time lag where outputs typically take 30 minutes to be produced from the true observation time). Refer to the following link for details of the latest map updates: https://sdi.tas-hires-weather.cloud.edu.au/shiny/ Map outputs are based on records produced from Bureau of Meteorology (BoM) Automatic Weather Station and Rain Gauge sites, in addition to Tasmanian government rain gauge and thrid party weather station sites. For operational real-time application, the mapping was fully automated in the R programming language and hosted on a cloud-based computing platform - via the high performance computing cluster provided by the Tasmanian Partnership of Advanced Computing (TPAC) of the University of Tasmania.

The outputs are an indicative display of current accumulated rainfall with interpolation carried out from observations derived from Bureau of Meteorology (BoM) Automatic Weather Station and Rain Gauge sites, in addition to Tasmanian government rain gauge and third party weather station sites. Further refinement is made by incorporating Bom Radar data and Himawari-8 and -9 satellite imagery. A composite modelling approach via machine learning, kriging and thin-plate spline inerpolation is used to harmonise the datasets with the observation data and produce hourly predictions using iterative model refinement techniques. Leave-one-out cross validation analysis of the interpolation procedure was performed to ascertain prediction accuracy and to understand potential quality of each map output. The root mean square error (RMSE) statistic was used to measure the prediction accuracy based on a subset of observation data. This analysis indicated that on average predictions are usually within 1-3mm of the true accumulated rainfall amount. It should be cautioned, however, that on occasions it was found that predictions can be erroneous by as much as 10mm during significant rain events. As such, the map products produced from this system should be treated as an indicative product and not relied as a true representation. Model cross validation analysis for accumulated rainfall predictions recently produced (as displayed in https://sdi.tas-hires-weather.cloud.edu.au/shiny/) can be accessed from the following link. This provides model evaluation metrics including the Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Concordance (CC) and Correlation (R2) coefficients. https://www.dropbox.com/s/0l304nqw71exy7s/ModelEvaluation_Rainfall.csv?dl=0 GIS datasets (i.e. geotiff raster files) for accumulated rainfall predictions recently produced can also be downloaded from here: https://www.dropbox.com/scl/fo/izet3gk0uh0hko1t2urk7/h?dl=0&rlkey=nbvlekulead5tly8tho1hy7o9 Alternatively, the following web map service (wms) can be used to import live updates into a GIS: https://sdi.tas-hires-weather.cloud.edu.au/geoserver/Precipitation/wms