It is shown the development and preliminary results of operational ionosphere dynamics prediction system for the Brazilian Space Weather program. The system is based on the Sheffield University Plasmasphere–Ionosphere Model (SUPIM), a physics-based model computer code describing the distribution of ionization within the Earth mid to equatorial latitude ionosphere and plasmasphere, during geomagnetically quiet periods. The model outputs are given in a 2-dimensional plane aligned with Earth magnetic field lines, with fixed magnetic longitude coordinate. The code was adapted to provide the output in geographical coordinates. It was made referring to the Earth’s magnetic field as an eccentric dipole, using the approximation based on International Geomagnetic Reference Field (IGRF-11). During the system operation, several simulation runs are performed at different longitudes. The original code would not be able to run all simulations serially in reasonable time. So, a parallel version for the code was developed for enhancing the performance. After preliminary tests, it was frequently observed code instability, when negative ion temperatures or concentrations prevented the code from continuing its processing. After a detailed analysis, it was verified that most of these problems occurred due to concentration estimation of simulation points located at high altitudes, typically over 4000km of altitude. In order to force convergence, an artificial exponential decay for ion–neutral collisional frequency was used above mentioned altitudes. This approach shown no significant difference from original code output, but improved substantially the code stability. In order to make operational system even more stable, the initial altitude and initial ion concentration values used on exponential decay equation are changed when convergence is not achieved, within pre-defined values. When all code runs end, the longitude of every point is then compared with its original reference station longitude, and differences are compensated by changing the simulation point time slot, in a temporal adjustment optimization. Then, an approximate neighbor searching technique was developed to obtain the ion concentration values in a regularly spaced grid, using inverse distance weighting (IDW) interpolation. A 3D grid containing ion and electron concentrations is generated for every hour of simulated day. Its spatial resolution is 1° of latitude per 1° of longitude per 10km of altitude. The vertical total electron content (VTEC) is calculated from the grid, and plotted in a geographic map. An important feature that was implemented in the system is the capacity of combining observational data and simulation outputs to obtain more appropriate initial conditions to the ionosphere prediction. Newtonian relaxation method was used for this data assimilation process, where ionosonde data from four different locations in South America was used to improve the system accuracy. The whole process runs every day and predicts the VTEC values for South America region with almost 24h ahead.