**5. Conclusions**

We quantified snow albedo feedback and calculated snow albedo radiative forcing during the period of 2003–2016, with remote sensing, atmospheric reanalysis and radiative kernel data. The results were then compared with those from partially observation-based calculations and model-based estimations.

The results suggest that, excluding Antarctica, both snow cover and snow albedo radiative forcing are the largest in landmasses north of 30◦ N. Snow albedo radiative forcing peaks in spring, due to strong insolation and large snow cover extent. Snow albedo feedback over the NEL is estimated to be 0.18 ± 0.08 W·m−2· ◦C−<sup>1</sup> and the global mean is 0.04 ± 0.02 W·m−2· ◦C<sup>−</sup>1. Results were compared with partially observation-based studies first. The regional mean snow albedo feedbacks were consistent, while the spatial pattern of snow albedo radiative forcing was different. The differences probably originated from snow albedo data and albedo data, as well as resolutions of data. Compared to the 25 models that participated in CMIP5, a general overestimation of models was found, mainly due to the overestimation of surface albedo variation between snow-covered and snow-free surface in the models. Surface albedo change is also with the largest spread among snow albedo feedback determinants. Therefore, model parameterization should specifically focus on the constraint of this factor. Meanwhile, remotely sensed snow cover data with high spatial and temporal resolutions, the constrained albedo contrast data between snow-covered and snow-free surface, as well as the resulted high resolution of snow albedo radiative forcing in this study, can offer valuable information for model parameterization.

**Acknowledgments:** The authors would like to thank Karen M. Shell and Brian Soden for sharing radiative kernels of CAM3 from NCAR, which is freely available online at http://people.oregonstate.edu/~shellk/kernel.html. We would also thank the anonymous reviewers for the constructive suggestions and comments. This study was supported by the Chinese National Natural Science Foundation under Grant numbers 41690143, 91547210 and 41271356 and the Chinese Academy of Sciences Project under Grant KJZD-EW-G03.

**Author Contributions:** Lin Xiao and Tao Che designed the study and wrote the paper, Linling Chen, Hongjie Xie and Liyun Dai all contributed to the discussions, edits, and revisions.

**Conflicts of Interest:** No potential conflict of interest was reported by the authors.
