Ionospheric Models and Measurement Comparison
There are several empirical ionospheric models that can be used for correction of ionospheric delay, ionospheric data ingestion or TEC (Total Electron Content) calibration. However, there is a limited knowledge about the models' performance. In the scope of the TRANSMIT demonstrator, we developed a method to analyze the TEC modelling performance of four models: Klobuchar, IRI2012, Nequick2 and NTCM, considering different geographical regions, time, season and solar activity. This method compares the modelled TEC data with a six year database of global ionospheric maps provided by CODE (Center of Orbit Determination in Europe), while the data were grouped according to the month, Universal Time and solar activity.
The results, in the form of global maps, show the distribution of models that are most similar to the reference TEC data for given conditions. In addition, the maps show the average TEC estimated by selected models according to the distribution map. This results therefore in a series of a global ionospheric maps created though by combination of the mention models. The maps also show how the 'combined global ionospheric map' differs from the reference TEC data (average modelling error). On average, all four models perform within the range of 1 TECU from each other, indicating that the models have very similar performance regardless on models' complexity, especially during the low solar activity. For high solar activity, the differences increase.
Figure 1: Distributions of the TEC difference between modelled TEC by four ionospheric models and measured TEC by TOPEX/Poseidon satellite. All the results have positive bias which is expected as the TOPEX/Poseidon satellite had orbit height of 1300 km. The arrows mark position of the histogram peaks.
The methodology is described in detail within the 3rd TRANSMIT workshop proceedings [Chapter 13: Performance Analysis of Empirical Ionosphere Models by Comparison with CODE Vertical TEC Maps], however, it is important to note some of the key assumptions and limitations of the method.
As mention above, the average difference between the model TEC estimates is within the range of 1 TECU. Keeping in mind such low modelling differences, it would hardly be reasonable to identify a single model as the most accurate overall (with respect to the CODE TEC), as model performance varies with the solar activity, location and time. In addition, the analysis deliberately used vertical TEC maps as the reference data to avoid influence of any mapping function error. In practice, this error would exist, potentially changing the resulting performance of the models for non-vertical TEC estimates. This level of success was achieved using multi-model performance distribution from different time period than TOPEX measurements. Using the distribution for the same time period the percentage of optimal model identification was higher. This means that the results are to certain extend time independent, even though there is still space for improvement.
The average difference between the TOPEX TEC and the TEC chosen according to the models' distribution is 4.7 TECU. In case the method would successfully identify the optimum model in 100 % (not 40 %) the difference would be 2.8 TECU, which is considered to be the minimum possible error.
The NeQuick model was driven by daily solar flux value, which gave him the possibility to react on the daily variations in solar activity, however, NeQuick version used in GNSS Galileo is driven by different parameter, so called Ionization level. For better information about the NeQuick performance, the same analysis should by done with version driven by Ionization level.