Objectives The plasmasphere is an important region connecting the ionosphere and magnetosphere, and its plasmaspheric total electron content (PTEC) variations are essential for global navigation satellite system positioning, space weather monitoring, and studies of ionosphere-magnetosphere coupling. However, conventional observations suffer from uneven spatial coverage and limited temporal resolution, while empirical models have difficulty accurately describing the complex spatiotemporal variability of PTEC. This study aims to develop a global PTEC prediction model and investigate the multi-scale temporal-spatial variations and geomagnetic storm responses of PTEC.

Methods Using COSMIC-1 topside total electron content observations from 2006 to 2019, a global PTEC prediction model was developed based on the TabTransformer architecture. Geographic location, temporal information, solar activity, and geomagnetic activity parameters were incorporated as model inputs. The model performance was evaluated using an independent test dataset, and the predicted global PTEC distributions were further analyzed in terms of diurnal, seasonal, interannual, and storm-time variations.

Results The proposed model achieved high prediction accuracy, with an mean absolute error of 0.678 1 TECU, an root mean square error of 0.966 6 TECU, and a correlation coefficient of 0.940 8 on the independent test dataset, without significant systematic bias. The predicted PTEC exhibited clear multiscale spatiotemporal variations. Diurnally, global PTEC generally showed a single-peak pattern, with maximum values occurring mainly during 12:00:00-16:00:00 local time (LT) and minimum values during 03:00:00-06:00:00 LT. The diurnal variation was strongest at low latitudes, weaker at middle latitudes, and almost absent at high latitudes. Seasonally, PTEC showed the highest values around the December solstice and the lowest values around the June solstice, with evident hemispheric asymmetry. Interannually, PTEC variations were mainly controlled by solar activity, showing strong consistency with F10.7 during solar maximum years but weaker dependence during solar minimum years. During geomagnetic storms, PTEC responses exhibited significant dependencies on latitude, local time, and L-shell. During the storm main phase, global PTEC showed a rapid enhancement following the sharp decrease in Dst, and the background diurnal variations were substantially disrupted. During the recovery phase, positive PTEC disturbances gradually weakened, negative PTEC disturbances expanded at low and middle latitudes, while positive PTEC disturbances persisted at high latitudes.

Conclusions The proposed TabTransformer-based model enables long-term global PTEC prediction from COSMIC-1 observations. The results reveal the complex spatiotemporal characteristics of PTEC and its responses to solar and geomagnetic disturbances, improving understanding of plasmaspheric dynamics and ionosphere-magnetosphere coupling processes.