Evaluation of IRI-2016 and IRI-Plas 2017 Total Electron Content Predictions against IGS-GPS Observations in Low-Latitude Regions during Solar Minimum (2008) and Solar Maximum (2013)

We present a systematic quiet-time evaluation of two widely used empirical ionospheric models, IRI-2016 and IRI-Plas 2017, for predicting total electron content (TEC) over ten low-latitude regions spanning two dynamically distinct longitudinal sectors: South America and Southeast Asia. The analysis uses GPS-TEC observations from the IGS network under geomagnetically quiet conditions (Dst > −30 nT) during the deep solar minimum of 2008 and the moderate-to-high solar maximum of 2013, covering all four seasons. IRI-Plas 2017 consistently outperforms IRI-2016, reducing the overall root-mean-square error by approximately 11%, with statistically significant improvements in nearly all season-year combinations (p < 0.05). The gain is largest during the equinoxes and more modest during the solstices, reflecting the enhanced sensitivity of TEC to the topside profile when the Equatorial Ionization Anomaly is well developed. Both models reproduce the seasonal error hierarchy, with the lowest errors in December solstice and the highest in June solstice, and exhibit a similar increase in error with solar activity of about 50% from minimum to maximum. The improvement is strongly region dependent: both models perform well over South America (correlation > 0.85, RMSE ≲ 28 TECU) but face challenges over the Southeast Asian–Pacific sector (RMSE up to 45 TECU). The systematic improvement of IRI-Plas 2017 is attributed to its plasmaspheric extension and GPS-TEC ingestion, which together provide a more complete topside and plasmaspheric electron content. Because the same GPS-TEC observations are used for ingestion and validation, the IRI-Plas results are not fully independent and should be interpreted as representing a constrained model rather than a pure forecast. These results establish a quiet-time benchmark for low-latitude empirical models and highlight the need for longitude-dependent EIA representations, expanded observational networks, and further validation against independent datasets for operational TEC prediction.