GNSS-based ionospheric-sounding (GNSS-sounding) allow us to measure total electron content (TEC), which is equal to the number of electrons along the line of site between a satellite and a receiver. Nowadays, the GNSS-sounding is widely used to  monitor the ionosphere, and, in particular, to detect ionospheric disturbances generated by earthquakes (Co-Seismic Ionospheric Disturbances, CSID). The main disadvantage of the GNSS-sounding is the imprecise estimate of the exact altitude of initial signal arrival (Hion) of an ionospheric TEC disturbance. From physical principles, the Hion is presumed to be around the altitude of the maximum of ionization, i.e., between 250 km and 400 km, depending on solar, geomagnetic, seasonal, and diurnal conditions. Changing the Hion value changes the geographic coordinates of the observed TEC disturbances, which in turn influences the ability to infer properties about the source of the disturbance. For many applications, the exact value of the Hion is not necessary, whereas it is crucial for the Ionospheric Seismology that uses CSID to find the location of the seismic source, which can  also be the source that generates tsunamis.  In this work, we present the results of our detailed investigation of the altitude of GNSS-detection of CSID generated by 20 earthquakes with different magnitudes. Our analysis is performed using GNSS-observations and three simulation techniques. We propose that two observation-based methods enable a useful estimate the Hion, and  are  applicable in near-real-time. We demonstrate  that, depending on the GNSS satellite geometry and the fault position, the Hion is different for each line-of-sight, and can vary between 150 and 300 km.