Validating Ionospheric Models Against Technologically Relevant Metrics

New, open access tools have been developed to validate ionospheric models in terms of technologically relevant metrics. These are ionospheric errors on GPS 3D position, HF ham radio communications, and peak F-region density. To demonstrate these tools, we have used output from Sami is Another Model of the Ionosphere (SAMI3) driven by high-latitude electric potentials derived from Active Magnetosphere and Planetary Electrodynamics Response Experiment, covering the first available month of operation using Iridium-NEXT data (March 2019). Output of this model is now available for visualization and download via . The GPS test indicates SAMI3 reduces ionospheric errors on 3D position solutions from 1.9 m with no model to 1.6 m on average (maximum error: 14.2 m without correction, 13.9 m with correction). SAMI3 predicts 55.5% of reported amateur radio links between 2-30 MHz and 500-2,000 km. Autoscaled and then machine learning "cleaned" Digisonde NmF2 data indicate a 1.0 x 1011 el. m3 median positive bias in SAMI3 (equivalent to a 27% overestimation). The positive NmF2 bias is largest during the daytime, which may explain the relatively good performance in predicting HF links then. The underlying data sources and software used here are publicly available, so that interested groups may apply these tests to other models and time intervals. Multiple research groups are developing models of the ionosphere to address effects on technology that depends on radio signal propagation. Here we present tests of these models that capture the model performance as it relates to some relevant applications. These are GPS position, long-distance HF communication and ionospheric critical frequency (which is proportional to the square root of the peak ionospheric density). All the data sets, testing code and model output used are made available in the public domain for reuse and development. For the test of critical frequency, we use machine learning to "clean up" the input data, removing errors introduced in the data generation process. New, open access tools developed to validate ionospheric modelsMetrics are GPS position, HF communications, and critical frequency/NmF2Machine learning applied to "clean up" auto-scaled ionosonde data, removing 94% of badly scaled NmF2 values