The authors develop a method to reduce the sampling error in estimates of very extreme daily precipitation, based on a new dataset consisting of a vast number of weather forecasts.

The manuscript is of a very high standard in terms of clarity, and contains a thorough description of data, methods and results. However, there is one science aspect that needs further consideration. While statistical uncertainty in estimates of extremes is reduced in this new dataset, the evidence on their bias is limited to precipitation amounts for normal weather types (1-in-2 day, and 1-in-10 day) rather than the extreme weather conditions producing the most intense rainfall. A fuller evaluation is required to prove ensemble forecast data from a relatively short period of real weather situations can provide more accurate estimates of very extreme precipitation.

If a revised version included more appropriate validation, and considered the extra points below, then publication as a NHESS Highlight article may be suitable since it has potential to make a substantial contribution to a subject of growing importance to society.