*4.3. Future Implementation*

The case study presented here builds on and contributes to a growing body of work advocating that approaches that account for temporal information in optimizing land cover classification are superior to temporally independent time series classifications and other single-date methods [39,64]. The application of the post-classification optimization to stabilize land cover trajectories, mitigate unrealistic land cover transitions and overcome the limitations and costs of obtaining ancillary and field data are still rarely applied in land use time series studies [39,64]. GEE implementation of these methods within its standard functionality offers an opportunity to improve temporal consistency and promote temporally dense analyses as a common practice. In addition to the post-classification method presented here, other opportunities to fully benefit from the temporal information available would be a GEE implementation of methods combining the probability functions of land use trajectories and the propagation of classification uncertainties temporally to decide on the final classification outcome [33,39–41,46,54,65]. Furthermore, data fusion of multiple sensors have shown large potential to further increase the temporal resolution of analysis [34,37,38]. In particular, the integration of optical and LiDAR or synthetic aperture radar imaging offers opportunities for high-frequency temporal analyses that are independent of the weather and cloud cover [66].
