Performance of national scale smoothed seismicity estimates of earthquake activity rates

Four smoothed seismicity models were submitted for consideration to the 2018 National Seismic Hazard Assessment (NSHA18). Three of the models used fixed smoothing bandwidths, each with slightly different implementations, while the fourth used an adaptive smoothing bandwidth that varies spatially based on earthquake density. In this paper we assess the performance of these models at forecasting activity rates over decadal timescales by using part of the earthquake catalogue used for NSHA18 to develop the models and calculating the log-likelihood of the model activity rates against the remainder of the catalogue. We test the performance of the models at forecasting rates of earthquakes of different magnitudes over different lengths of training and forecast periods, and compare the use of non-declustered and declustered earthquake catalogues. We test for time-dependence by running the comparisons sequentially and using an earthquake catalogue randomised in time. The results are used to evaluate and inform the final implementation of the smoothed seismicity models for calculating 10% in 50 year probability ground motion exceedances for the NSHA18. This abstract was submitted and presented to the 2017 Australian Earthquake Engineering Society Conference (AEES) ( https://aees.org.au/2017-aees-conference/)