Crowd-sourced location data from client devices is processed using enhanced filtering techniques in non-realtime by a centralized server process to improve the accuracy and reliability of georeferenced geodata. At a server computer, enhanced filtering techniques make use of future information to improve georeferencing of the geodata. In some implementations, the server computer can be configured to implement batch processing of raw GNSS data and other crowd-sourced location data (e.g., wireless scan data, inertial sensor data) using, for example, a smoothing filter or batch estimator. Augmentation data (e.g., satellite ephemeris data, clock offset and ionospheric delay) and sensor data independent of client devices can also be used by the server computer to augment the crowd-sourced location data to further improve georeferencing of geodata.