Thank you for your response to our review comments. As with your original submission, this update is well-written and clear, providing an interesting analysis of projected trends in normalized precipitation/meteorological drought for Poland. This update addresses many of the original comments; however, there are a few remaining issues which I would like to see addressed before final publication. I therefore recommend a minor revision.
Addressing the original major comments:
1. The new title is much clearer and more accurate. I appreciate this change. It is ok to refer to the SPI as a meteorological drought index in the text, I simply wanted to clarify that you are not specifically measuring drought. This now comes through clearly in the paper.
2. I remain uncomfortable with the description regarding the effect of bias correction on SPI1 in the Results. You do a very good job in the Discussion section of explaining that the effect on SPI trends decreases from choice of GCM >> choice of RCM >> bias correction. But, the Results section still describes bias correction as playing a significant role, particularly in the last 2 paragraphs of page 48. For instance, line 17 states that the number of grid cells with statistically significant trends “… strongly depends on the month, climate model, and also on whether or not bias correction has been applied”.
Including a test of Pearson correlation (Page 52) is interesting, but this test uses Ho: r = 0, Ha: r ≠ 0. When this test is statistically significant (p < 0.05), it means that it is highly unlikely there is no relationship between raw SPI and bias-corrected SPI. You are not testing whether the time series are the same, but rather if they have any similarities. With 128 years of data (1971-2099, depending on accumulation period), the Pearson correlation need to produce p < 0.05 is only 0.173.
It is ok to include this p-value test, but you should not lose sight of the extremely high correlation >0.9 for most models (Table 3) and nearly identical spatial patterns among model pairs (Fig. 8 to 9). Also, there doesn’t appear to be any consistent change (increasing vs. decreasing) due to bias correction among the models in Fig. 10. This is why I am concerned with the statement on Page 38, Lines 22-24 that “the use of bias correction slightly decreases the area with statistically significant trends in summer months (June, July and August) and slightly increases in the other months.” This could be confirmed with a paired t-test, but the models in Fig. 10 appear to be randomly distributed around 0, except for February, when they all show positive differences.
The effect of bias correction is an interesting topic, which is appropriate to highlight in this paper. But, given the analysis and the fact that this is only a minor portion of this paper, I recommend softening the language in the Results section and instead including a sentence or two in the Conclusions that recommends further study of this effect in the future.
I have only 2 additional minor comments that should be addressed in a revision:
1. The Methods section introduces the principles behind the modified Mann-Kendall test for serially correlated data. Based on your text, I assume you use the standard Mann-Kendall approach for SPI1, 3, 6, and the modified Mann-Kendall for SPI12 and 24. But, in the Results and Discussion it appears that you use the modified Mann-Kendall for all variables. Please add a short sentence in the Methods clarifying exactly which test you used for which time series/variables (SPI1-24).
2. The data in Table 1a. and Figure 10 do not match. Based on their descriptions, I believe they should. It appears that Figure 10 is actually plotting (corr - raw) divided by approximately 11.7. Using the high and low values in February as an example, Bias – Raw for KNMI-ECHAM and SMH-BCM are 427 – 240=187 and 38-10=28, respectively. These instead appear in Fig 10 as approximately 16 and 3. If you instead calculate relative difference by dividing this by Raw area, as described in the Figure 10 caption, the numbers change to 77.9% and 280%. This changes the relative order, making the previous highest peak the smallest value and vice versa. Please check this figure/Table 1a.