**5. Conclusions**

This research proposed a method for the synergistic use of Sentinel-1 and Sentinel-2 features for oasis crop type mapping through a case study in a smallholder farming area in Northwest China. First of all, a SHP DSI algorithm was introduced for the de-speckling of SAR intensity and accurate estimation of interferometry coherence to improve the quality of SAR features. It was demonstrated that the use of the SHP DSI method improved the crop classification accuracy by 6.25% when only using SAR features. A variety of SAR features and optical features were derived from multi-temporal Sentinel-1 and Sentinel-2 images, including several InSAR products and red-edge spectral bands and indices. Secondly, based on the permutation importance of the random forest classifier, a recursive feature increment feature selection method was proposed to obtain the optimal combination of Sentinel-1 and Sentinel-2 features for cropland extraction and crop type classification. Finally, a crop distribution map was generated with an overall accuracy of 83.22% and kappa coefficient of 0.77. The contribution of SAR and optical features were explored thoroughly. Among all the Sentinel-1 features, the VH intensity held the biggest proportion, indicating the better sensitivity of VH polarization to vegetation changes. It was also noted that some of the InSAR products, such as the VH amplitude dispersion, the master versus slave intensity ratio, and the VV coherence in early April revealed good separability of certain crop types. As for Sentinel-2 features, we demonstrated the merits of using red-edge spectral bands and indices in oasis crop type mapping. The inclusion of red-edge features improved the crop classification OA by 1.84% compared with only using conventional optical features. This proves the superiority of Sentinel-2 data due to the increased spectral resolution. A comparison was conducted on the performance of oasis crop classification using four combinations of features. The results indicated that the integration of SAR and optical features achieved the best performance. We concluded that the integration of time series S1 and S2 imagery is advantageous, and thanks to the free, full, and open data policy, it can be further explored in the vast majority of regions for the monitoring of crop status.

**Author Contributions:** Methodology & investigation & writing—original draft preparation, L.S.; supervision, J.C.; Writing—review and editing, J.C., S.G. and X.D.; Project administration, S.G.; Validation, Y.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research has been supported in part by the National Key Research and Development Program of China [Grant No. 2017YFB0504203], the Strategic Priority Research Program of Chinese Academy of Sciences [Grant No. XDA19030301], and the National Natural Science Foundation of China [Grant No. 41801360, 41601212].

**Acknowledgments:** The authors thank J. Liu, Q. Liu, X. Zheng, Y. Shen, and Y. Xiong from SIAT, and Wang and her team from Fuzhou University for their work in the field investigation.

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