Review of manuscript “21st century change in precipitation on the Greenland Ice Sheet using high resolution regional climate models” by Boberg et al.

General overview of the manuscript

The manuscript looks at changes in solid and liquid precipitation over the Greenland ice sheet (GrIS) in an ensemble of high resolution future projections compared with those obtained with 34 CMIP6 global climate model in three different SSP scenarios (SSP1-2.6/SSP2-4.5/SSP5-8.5) and observations/reanalysis datasets. High resolution ensemble members are obtained in three different ways: (i) with the regional climate model (RCM HIRHAM, forced with EC-Earth3; (ii) with a set of # CMIP6 models downscaled with MAR and (iii) with the RCM RAMCO forced by CESM2-L. Total precipitation over GrIS is projected to increase up to 24% in the 2071-2100 period for the most extreme SSP scenario. The model spread and changes in precipitation characteristics (e.g. total precipitation versus snowfall/rainfall fraction) are analysed over the entire GrIS and over six different catchments. Overall, all models tend to underestimate the total amount of precipitation over the most part of GrIS during the validation period, with the exception of coastal areas. In general, I find that the manuscript is well written, although a few points need further clarification to ensure that the paper can be accessible to a broader spectrum in the scientific community. The research topic is relevant to the broad scientific community and the manuscript represents a good fit for Copernicus journal “The Cryosphere”. Most comments mainly aim to improve the readability and accessibility of the paper, and to better describe Figures. I would recommend publication in TC after current concerns have been successfully addressed.

Thank you for your positive and constructive comments.

Major comments

    Line(s) 95, table 1. The table caption reads “Numbers in blue/green/red are for the SSP1-2.6/SSP2-4.5/SSP5-8.5 scenario”, however all tables are in black and white in the pdf. Also I am not sure why some values are set to N/A, does it mean that for that model the scenario is not available? Please clarify. Moreover, at first glance it seems like the bias between the ensemble and CARRA over the 1971-2000 period is -142 Gt, and that for the 1991-2022 period the bias is -307 Gt. Again, it takes a while to understand that the last two columns represent the bias between the 7 chosen observation stations and the models (CARRA). Why is a small set of 7 weather station used to calculate the bias when the most of the anaylsis afterwards shows the comparison with CARRA? I would suggest to show in the table the bias with CARRA instead, or show both and remove the RMSE. Also that the CARRA value in the second column to the left is for a different period: this is misleading and I would at least mark the value with an asterisk, although I would suggest to re-think the table.

REPLY: Thank you for this valid comment. The original table had colors but these were replaced by different lines for different scenarios in the submission process. We subsequently forgot to update the table caption. This will be done in the revised text. Yes, the N/A refers to data not available, either missing scenario simulations (RCM) or time period not covered by the data (CARRA). This will be clarified in the new caption. An asterisk is added for the CARRA values. We do an undercatch correction of the observations and we only have data available for 7 stations for this correction as this requires wind observations as well as temperature. CARRA is treated as a model simulation and not a reference for the true precipitation. The models (RCMs and CARRA) are validated against observations. This will be clarified in the revised text.

    Line(s) 129-130. Figure 2 is described in too little detail. First, it is difficult in the figure to distinguish different colours from the colour scale: I suggest adding isolines on the plot and highlight one level every two or three, e.g. 15%, 45%, 75%. This way it would be easier to notice differences among different plots. Second, 12 different maps are described in two short sentences, with only a very general statement on where “positive” or “negative” biases are found, entirely leaving to the reader the task of trying to catch and interpret the values on the maps. Please revise the figure description adding values to the description of panels (b)-(l). See below a suggestion on how to improve the paragraph:

 “We see similar patterns for all models with large areas with a negative bias of about -15-30% in the central parts of the GrIS and smaller areas with a positive bias up to +40-50% along the Greenland coastline. The model with the highest bias are (add models), which show a deviation from the climatology up to -50%. Four of the MAR models (Figure 2f to 2i) show a large positive bias from the climatology, with values of about +40-50% in region NW and NO, which correspond to an increase in precipitation of about (add number) mm”. Note that the numbers provided above may be incorrect but this is currently what I can gather from the figure.

REPLY: We tried using contour lines but it made the figure very busy. We will reduce the number of colours instead, making it easier to distinguish the different regions. The following text is added to more clearly describing the differences between CARRA and the RCMs:

“We see similar patterns for all models with large areas with a negative difference of about -10 to -30% in the central parts of the ice sheet and smaller areas with a positive difference up to 50% along the Greenland coastline. The model with the largest positive difference (in regions NW and NO with values in the range 10-50%) over the ice sheet are MAR driven by CNRM-CM6, CNRM-ESM2 and MPI-ESM1.2 while the two HIRHAM simulations and the RACMO simulation have the largest positive differences outside the ice sheet with values in the range 30-70% with peaks in excess of 70%.”

 Note that we do not consider CARRA as the truth. In figure 2 we only compare CARRA to RCMs and look at differences.

    Line(s) 143-149. Figure 3. Same comment as for Figure 2. Please add values to the text and explain their meaning in the context of your analysis.

REPLY: We will update the paragraph to the following:

“For station 1 all model simulations agree with observations except MAR, driven by IPSL, showing a dry bias while RACMO shows a clear wet bias of about 60% (130 mm). For station 2 all MAR simulations show a dry bias of about -25% (-250 mm) while the other simulations are closer to observations. For station 3, CARRA is close to observations while the two HIRHAM simulations have a wet bias in the range 35 to 70% (250 to 500 mm), RACMO show a wet bias of about 110% (800 mm) and five of the MAR simulations have a dry bias of about -35% (-250 mm). The two HIRHAM simulations are close to observations for station 4 while CARRA and RACMO show a dry bias close to -20% (-350 mm) while all eight MAR simulations have a clear dry bias close to -65% (-1300 mm). For station 5, CARRA and one HIRHAM simulation are close to observations while the other HIRHAM simulation together with two MAR simulations and RACMO have a dry bias of about -20% (-180 mm). The other MAR simulations have a clearer dry bias in the range -30 to -60% (-330 to -530 mm). A similar pattern is seen for station 6 with the exception for RACMO here showing a wet bias of about 25% (90 mm). Finally, for station 7 one HIRHAM simulation is close to observations while the other HIRHAM simulation together with RACMO have a wet bias of about 15% (80 mm) while CARRA has a dry bias of -15% (70 mm) and all eight MAR simulations have a dry bias in the range -15 to -45% (-60 to -210 mm).

The two rightmost columns of Table 1 give the RMSE and mean bias for CARRA and the eleven RCM simulations, showing low RMSE values and low mean bias for the two HIRHAM simulations and CARRA while the MAR simulations have a relatively high RMSE and negative bias.”

    Line(s) 163-167. Figure 4. Same comment as for Figure 2. Please add values to the text and explain their meaning in the context of your analysis. I would also recommend shifting the panels from a 6 by 3 grid to a 3 by 6 to better suit the page format.

REPLY: We will change to a 3 by 6 matrix with one row per SSP and update the text to the following (note that the panel letters below match the new figure layout with a 3 by 6 grid):

“Figure 4 shows the model spread of the relative change in precipitation for a mid-century and an end of century period relative to 1971–2000 using the RCM ensemble for all three SSP scenarios. For the mid-century median for all three scenarios (Figure 4b, 4h and 4n) we see a positive change between 10 and 50% for most of Greenland except region SE with no clear change. For the mid-century 25^th percentile (Figure 4a, 4g and 4m) we mostly see a positive change (10 to 30%) for regions NE and NO and a negative change (-10 to -30%) in region SE. For the mid-century 75^th percentile (Figure 4c, 4i and 4o) all of Greenland shows a positive change that is most pronounced in regions NE and NO with values in the range 30 to 70%.

Towards the end of the century, we see an overall amplification of the mid-century signal. The ensemble median shows a reduction in precipitation compared with the mid-century values for the SSP1-2.6 scenario (Figure 4e) while SSP2-4.5 (Figure 4k) and SSP5-8.5 (Figure 4q) show an increase (except region SE) with values reaching 100% for region NE for SSP5-8.5. For the end of century 25^th percentile (Figure 4d, 4j and 4p), we see similar patterns except for region SE with a -10 to -30% change along the outer edge. For the end of century 75^th percentile (Figure 4f, 4l and 4r) we see a positive change for most of Greenland with values in the range 30 to 70% in region NE for SSP2-4.5 and values in the range 70 to 110% for SSP5-8.5.”

    Line(s) 187-194. Figure 5. Same comment as for Figure 2. Please add values to the text and explain their meaning in the context of your analysis. I would also suggest to add in each plot the period/SSP scenario as done in Fig. 4.

REPLY: We will put info into the panels as done for Fig. 4. The paragraph will be updated to the following text:

“Figure 5a and 5d show the ensemble mean relative change, using the eight CMIP6 GCMs dynamically downscaled by the RCMs (cf. Table 2), for the SSP5-8.5 scenario for 2031–2060 and 2071–2100 relative 1971–2000, respectively. The mid-century relative change is between 10 and 30% for regions SW, NO and NE with areas reaching 50% for region NE. For the end of the century, all regions except SE show positive change with areas reaching 90% in NO and NE. Figure 5b and 5e give the corresponding maps when using the full 34 CMIP6 GCMs for the SSP5-8.5 scenario. Compared with the eight-model ensemble, the full model ensemble has larger changes for all six subregions but most pronounced for region SE where we now see a positive change in the range 10 to 30%. Figure 5c and 5f show the number of CMIP6 models that give positive changes in precipitation for the mid-century and end of century, pixel by pixel. The CMIP6 ensemble show good agreement for all drainage basins except for region SE where about 6 of the 34 members show a negative change for the end of century, in line with Huai et al. (2025).”

    Line(s) 196-203. Figure 6. Here I would say the colour shading is clear enough and values of the snowfall/precipitation fraction are reported in the text. However, I would explicitly include at the beginning of the paragraph the definition of the snowfall fraction (snowfall/total precipitation) and include an example to show what a certain value means.

REPLY: On suggestion from another reviewer, we will change to showing rainfall fraction instead of snowfall fraction. We will add context to the paragraph, e.g. that a value of 50% means that rainfall and snowfall contribute equal amounts to the total precipitation.

    Line(s) 222-227. Figure 7. It took me a while to understand why the total precipitation doesn’t seem to match the basin values. Please write explicitly in the text that this is the sum of the other values and the y-axis of reference is the right one. Also see previous comment for Figure 2. Please add values to the text and explain their meaning in the context of your analysis.

REPLY: We will add a clarification that the numbers for the full ice sheet refers to the right y-axis. Also, from a comment by another reviewer we have changed the bars into boxplots and also changed the color scheme. The paragraph will be updated to:

“Figure 7 shows the annual total precipitation for the six drainage basins as well as for the total ice sheet for CARRA and the RCM simulations. The vertical extent of each line represents the model spread for each basin while the boxes show the 25^th and the 75^th percentiles of the ensemble. For the two scenario periods the bar is split into three segments to show the difference between the three SSPs. The largest uncertainties are, as expected, found in region SE. There is no clear difference between the SSPs for the 2031–2060 period while the 2071–2100 period has a clear difference in the ensemble median. The CARRA values match the higher end of the model spread for each basin, giving an annual amount for the total ice sheet that is significantly higher than any of the other models.”

    Line(s) 244-297. In general, the discussion is missing a paragraph outlining the limitations of this study and of the used datasets.

REPLY: Good point! We will add the following paragraph about limitations of this study:

“A clear limitation of this study is that RCMs often exhibit large biases in simulating point-wise precipitation, while the observations also have large uncertainties, especially when undercatch is prominent. The large-scale circulation and sea surface temperatures from forcing GCMs often have errors that are subsequently inherited by the RCMs. Furthermore, small-scale convective processes are not (fully) resolved by the RCMs used in this study. Also, the lack of a gridded observational product for Greenland makes it difficult to perform bias adjustment of the modeled precipitation. These limitations are, to some degree, compensated when looking at relative changes and not absolute values and the use of an RCM ensemble instead of looking at a single RCM.“

    Line(s) 264-267. The importance of the discrepancy between the runoff and the increase in precipitation is prominently present in the abstract, however only two brief sentences on the topic are present in the discussion with no explanation on what could cause the increase in runoff. Runoff ultimately contributes to sea level rise and is therefore a key element of Greenland research. While I understand that the scope of this study is to analyse precipitation changes, it is important to elaborate a bit more the implications of these findings for GrIS runoff in the discussion, and explore potential reasons why this discrepancy is present in the first place. My uninformed assumption would be that surface temperature increase and surface melting may play a big role into explaining the runoff, but I suppose other explanations like model parametrisation or fixed ice sheet characteristics may play a role as well.   

REPLY: We have decided to remove all text about runoff/melt etc, based on a comment by another reviewer, and only present numbers on precipitation. We hope this is acceptable.                                                                                                                                      

Minor comments

    Line(s) 110-111. Could you please add one sentence to clarify what “undercatch corrected” means. This seems a quite specific/technical term that many readers may not be familiar with.

REPLY: We will add a clarification to the sentence:

“The station data are undercatch-corrected (i.e. positive correction of the amount of water caught by the rain gauge due to winds) using the mean correction factors and mean rainfall fractions from Huai et al. (2021).”

    Line(s) 249-254. Figure 9. I would suggest to move the description of this figure to the results section and focus on the interpretation/comparison with other literature in the discussion.

REPLY: We will move Fig. 9 to the Results section together with the text on lines 250-254.

    Line(s) 269-271. What about Greenland blocking? Please see

 Davini, P.,Weisheimer, A., Balmaseda, M., Johnson, S. J., Molteni, F., Roberts, C. D., Senan, R., and Stockdale, T. N.: The representation of winter Northern Hemisphere atmospheric blocking in ECMWF seasonal prediction systems, Q. J. Roy. Meteor. Soc., 147, 1344–1363, https://doi.org/10.1002/qj.3974, 2021.

REPLY: Good point! We will add the following sentence on the Greenland blocking at the end of the paragraph on lines 269-272 (and add the Davini et al paper to the References list):

“Furthermore, Greenland blocking events are difficult to estimate by climate models (Davini et al. 2021) and can have a large impact on precipitation, especially during winter.”

Citation: https://doi.org/10.5194/egusphere-2025-4360-RC2