The paper introduces the new CARROTS method to adjust radar QPE with climate factors derived from ten years of radar data. The resulting product is compared to an operational as well as reference QPE of the Netherlands in terms of precipitation input to selected catchments as well as discharge measurements. The manuscript is well written and structured. The method is introduced, applied, validated and critically discussed. Relevant conclusions are drawn and the authors give proper credit to related work. Overall, I propose some but still minor revisions. Text-related comments are given directly in the attached pdf. More general comments, questions, and ideas to be discussed in the paper or interactive discussion are given below.

1. Do you think, MFB could be improved by dividing the NL into spatial segments (like the classical moving window) or even depending on the distance to the radar, or is the density of automatic stations to low? (The argument that MFB is limited to one factor for a whole country does not hold in general.)
2. Does the RA adjustment eliminate spatial dependencies of the resulting QPE? To be more precise: Is the quality of RA depending on the distance to the radar site? And if so - would CARROTS allow for the derivation of an improved adjustment procedure?
3. What’s the effect of temporally present spokes (positive and negative) in the historical data set?
4. I propose to add a figure showing the pixel-based differences between MFB and RA as well as RC and RA on a 5-min-basis, e.g. box-whisker (NL mean or catchments) and map with median and percentiles. This would help understanding the effects on discharges in different regions.
5. What is the performance in heavy rain situations? Despite mean numbers, please comment on the effects.
6. Could a dbz-dependent factor improve CARROTS?

The paper introduces a new method (Carrots) to improve the radar based QPE for the hydrological modelling purpose. Carrots proposes spatially and daily variable factors derived from the combination of radar, hourly and daily rain gauges on a 10-year period, that can be used for the adjustment of the radar based QPE. The proposed methodology is compared against Mean Field Bias correction of radar data and conclusions are drawn for areal precipitation estimation as well as for discharge simulation on various catchments in Netherlands. The methods are well applied and the results are well discussed. The paper is well written/structured and the readers would benefit from this study. There are however a couple of things that should be addressed but these are minor to moderate revisions. The comments and questions are given in the attached pdf, and also summarized below:

1.    Information about the location of the daily and hourly rain gauge should be included in Figure 1 in order to understand better the areal rainfall and discharge results. Additionally, in Table 1 rain gauges included per catchment (if applicable) can be added as an extra column. The role of the gauge density in some catchment (either hourly or daily) may explain the results of Figure 6 – for example catchment Deflfand where the MFB has slightly better results. 

2.    In Figure 5 only the areal annual precipitation of 4 catchments are given and there is not enough information to understand the results of Figure 6. Instead of annual volumes for each method, another Table or Figure may be added to summarize the annual bias of Carrot, MFB and Ru for each catchment. In this case a bias equation should be given in the paper so the reader can understand the results. 

3.    Another question open for discussion is the role of the catchment model type (either lumped or semi-distributed) and their calibration in the discharge errors (see comment on Figure 6). Are the semi-distributed catchments made of more than 2 sub-catchments? What data has been used for the calibration of these models?

4.    In section 2.3 it should be stated clearly that the method is not in “leave-one-out” or split-sampling validation, the same period of data is used for the RA, for the Carrot factors, for the MFB factors and for the model calibration (or was another period used for the model calibration). Explain also shortly at this section why the Carrot factor were not used in a “leave-one-out” validation (because of low sensitivity obtained from section 3.4).  

Review of Imhoff et al.: A climatological benchmark for operational radar rainfall bias reduction

The paper presents a novel method CARROT for climatological bias adjustment of radar data to be used in a real time framework as an alternative to hourly mean field bias adjustment. 



The paper is well written and structured. The objectives are clearly stated. 

I like the pragmatic approach of the developed method where it is acknowledged that limited data availability in real-time applications might be problematic and that a climatological approach might be as good or even better. In my own research, I have also seen how MFB adjustment at an hourly timescale can cause over or underestimation of QPE if data is limited and with high spatial heterogeneity. 

Please find a few comments below:

- In another study that the authors also refer (Schleiss 2020), we have seen significant differences between radar products based on single radars versus products based on composites of multiple radars. The latter being more reliable especially for the estimation of high-intensity rainfall. Can you maybe comment on how radar data from the two radars that you apply in the study are merged and how this relates to the larger biases as a function of distance from the radar. If there are significant range effects, I guess that this will be present both in Ra and Ru and therefore not necessarily lead to bias differences. Is there differences in the bias estimations depending on whether the location in question is covered by one or two radars.

- The idea of a climatological bias factor works if the drop size distribution and the Z-R relationship do not change with season. It is especially important here to distinguish between convective and stratiform precipitation. In NL the winter is probably dominated by statiform precip., but summer  precipitation is probably a combination of both…which would lead to more uncertainty in estimation of summer rainfall. Can you maybe provide an insight into the variability of Fclim depending on season and year. Like in figure 4 (a) where you show MFB for 2018, but Fclim for all  years with confidence bands or monthly boxplots describing the variability for each month and year. 

- In figure 7 (a) is the bias correction factor derived for all years or only 2015? Since the idea of Carrot is to use the climatological average, I would prefer to see data for the whole period 2009-2018

- To be consistent please indicate on figures 3, 5 and 7 that the “bias correction factor “ corresponds to Fclim