This paper is related to study of the ion density recorded by the low altitude satellite DEMETER. It will present ionospheric perturbations observed during large seismic events.As the ionosphere is highly variable, the paper will show a statistical analysis performed on the plasma parameters during night time.An algorithm has been implemented to detect crests and troughs in the data before world-wide earthquakes.The earthquakes have been classified depending on their magnitude,depth,and location(land,below the sea,close to a coast). Due to the orbit,DEMETER returns above the same area every day(once during day time,once during night time) but not at the same distance of a given epicenter.Then,for each earthquake,data have been checked until 15 days before the shock when the distance between the trace of the orbit and the epicenter is less than 1 500 km.The results of the statistical analysis are presented as functions of various parameters.A comparison is done with two other databases where,on one hand,the location of the epicenters has been randomly modified, and on the other hand,the longitude of the epicenters has been shifted.Results show that the number and the intensity of the ionospheric perturbations are larger prior to earthquakes than prior to random events,and that the perturbations increase with the magnitude. This paper is related to study of the ion density recorded by the low altitude satellite. DEMETER. It will present ionospheric perturbations observed during large seismic events. As the ionosphere is highly variable, the paper will show a statistical analysis performed on the plasma parameters during night time.An algorithm has been implemented to detect crests and troughs in the data before world-wide earthquakes.The earthquakes have been classified depending on their magnitude, depth, and location (land, below the sea, close to a coast). Due the orbit, DEMETER returns above the same area every day (once during day time, once during night time) but not at the same distance of a given epicenter.Then, for each earthquake, data have been checked until 15 days before the shock when the distance between the trace of the orbit and the epicenter is less than 1 500 km. The results of the statistical analysis are presented as functions of various parameters. A comparison is done with two other databases where, on one hand, the location of the epicenters has been randomly modified, and on the other hand, the longitude of the epicenters has been shifted. Results show that the number and the intensity of the ionospheric perturbations are larger prior to earthquakes than prior to random events, and that the perturbations increase with the magnitude.