**6. Conclusions**

In this study, a two-step merging and downscaling method (OI-GWR) was proposed and used to downscale original IMERG precipitation data for the Tianshan Mountains from a resolution of ~10 to 1 km. First, the original IMERG precipitation data were merged with observed precipitation data using the OI method (https://github.com/tgq14/GWR-OI). Then, downscaling was performed using the GWR method, with the corrected CIMERG precipitation data as the initial values and the NDVI as an auxiliary variable. The performance of OI-GWR was assessed based on rain gauge observations, and the results were compared with those of other downscaling methods. The main conclusions include the following:

(1) The OI-based merging of the OIMERG data and the observed precipitation data greatly improved the precision of the original satellite precipitation product. After OI, the CC, RMSE, and MAE values for the OIMERG data were increased from 0.516, 28.83 mm, and 19.73 mm to 0.616, 26.56 mm, and 17.59 mm, respectively.

(2) An assessment of various downscaled datasets showed that the precision of DS\_CIMERG (based on OI-GWR) was much higher than that of DS\_OIMERG (based on GWR), indicating that the downscaling results for satellite precipitation data depend to a large extent on the quality of the initial precipitation product.

(3) Residual correction is a key step in global regression downscaling methods, such as ER and MLR. However, in the GWR method, residual correction tends to transfer the errors of the original satellite precipitation data to the final downscaling results. The statistical evaluation metrics of the GWR results obtained after residual correction were worse than those before residual correction, indicating that residual correction is unnecessary in GWR-based downscaling.

By improving the precision of the original IMERG satellite precipitation products and then downscaling the products thus obtained, the two-step OI-GWR method can serve as a more effective approach for the generation of regional precipitation datasets. Further efforts will be needed to extend the application of OI-GWR to the daily or hourly scale.

**Author Contributions:** Conceptualization, X.L. and G.T.; methodology, G.T.; software, M.W.; validation, X.W. and Y.L.; formal analysis, X.W.; investigation, Y.Z.; resources, Y.L.; data curation, X.W.; writing—original draft preparation, X.L.; writing—review and editing, G.T.; visualization, X.L.; supervision, G.T.; project administration, G.T.; funding acquisition, Y.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA20100306), the National Natural Science Foundation of China (U170310011), the Basic Research Operating Expenses of the Central Level Nonprofit Research Institutes (IDM2016002), and high-level personnel funding for the Xinjiang Uygur Autonomous Region (2017-41).

**Acknowledgments:** We are grateful to the scientists on the NASA science team for providing satellite precipitation and DEM data. We also thank the Xinjiang Meteorological Information Center for providing the gauge-observed precipitation data.

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