**5. Conclusions**

Our study aimed to advance temporal mangrove forest monitoring efforts to benefit from the potential that currently available satellite earth observation data and cloud-based high-performance computing can offer. The temporal domain of these information-dense datasets opens opportunities to apply data fusion principles to optimize classification outputs to be gap-free and temporally consistent with logical land use transitions as well as provide a means of cross-validation. The results of our case study on mangrove forests demonstrate how this information can be valuable in understanding the spatio-temporal dynamics, processes, and trends of land use changes and improve decision-making with detailed information. Thereby, our study builds on and contributes to a growing body of work advocating that accounting for temporal information in optimizing land cover classification is superior to temporally independent classifications in time series and other single-date methods.

Despite growing awareness, most mangrove forest cover classification studies are ye<sup>t</sup> to take full advantage of Earth observation's potential and the rich temporal information available from time series data. Implementation of temporal optimization, either post-classification or during the classification process, within future implementations or that can be automated on top of GEE's output as presented here, can hopefully contribute to advance mangrove monitoring studies towards fully unlocking the potential of data available as the field of earth observation keeps evolving. The integration of synthetic aperture radar remote sensing in addition to optical observations will be key in advancing the approach presented here and increase the temporal resolution by overcoming data gaps due to weather or lighting conditions.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-4292/12/22/3729/s1, Table S1: Links to the GEE code implementation used for this study, Table S2: Decision rules of the regression tree (CART) build for land use classification (2015) by Landsat-7 and Landsat-8 respectively.

**Author Contributions:** Conceptualization, L.T.H. and N.A.B.; methodology, L.T.H. and N.A.B.; coding, L.T.H. and N.A.B.; validation, N.A.B. and P.V.H.; investigation, N.A.B. and P.V.H.; resources, N.A.B. and P.V.H.; data curation, N.A.B.; writing—original draft preparation, L.T.H.; writing—review and editing, N.H.Q. and J.T.; visualization, L.T.H.; supervision, L.T.H. and P.V.H.; project administration, N.A.B. and P.V.H.; funding acquisition, P.V.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Vietnam Academy of Science and Technology, gran<sup>t</sup> number ÐLTE00.06/20-21 and the APC was funded by the Institute of Environmental Sciences (CML), Leiden University.

**Acknowledgments:** This research was partly supported by Vietnam Academy of Science and Technology through the project ÐLTE00.06/20-21.

**Conflicts of Interest:** The authors declare no conflict of interest.
