The study investigates the relationship between soil moisture and precipitation, but a lot needs to be clarified. First, it is stated in the abstract that soil moisture is jointly affected by precipitation and evapotranspiration, but there is no description of evapotranspiration in the abstract. Secondly, the respective roles of Ridge regression models and Bayesian generalized non-linear multivariate multilevel models in attribution need to be explained. What causes the differences in dependencies between land cover types? How do these differences come about? There is no consensus on what common features these land covers have. Finally, the effects on seasonal scales and interannual scales look more like the usual conclusions, and it is not clear that this work finds something new based on these traditional results.

I suggest a major revision. Please see my comments below:

Major Comments

1. In the introduction, the linear or nonlinear relationship here is a model for estimating soil moisture by precipitation, whereas copula is a distribution function, they should not be compared together. Ridge regression is an important method in the abstract, but it is not mentioned in the introduction. What role does ridge regression play?
2. In the material and method, the joint probability of copula considers soil moisture and precipitation, ridge regression considers precipitation and evapotranspiration to predict soil moisture, Bayesian generalized non-linear multivariate multilevel models consider GPP, LST, and temperature to predict soil moisture and precipitation, what is the relationship between these three methods that seem to be simply spliced together. Why choose these models and how accurate are they in the simulation?
3. In Section 3.1, for example in northwest Africa, why is there a negative dependence between the soil moisture at the first layer and precipitation while a positive dependence between the soil moisture at the second layer and precipitation, and what causes the difference between the different layers? Is there any connection between the result expressed by λU/λL and Kendall’s tau, and why do many grids have no value in the result expressed by λU/λL?
4. In Section 3.1, the monthly scale and annual scale are used, but in Section 3.2, the monthly scale and seasonal scale are used, so it is recommended to unify the comparison scale.
5. There are too many descriptions in section 3.3, and scatters of different land cover types in figures 6 and 7 are not clear. The large number of listed results makes it difficult to distinguish the commonalities and differences between different land cover types, and why there are differences between different soil layers. Part of the discussion should be summarized in the results, and the discussion should add references.
6. The first paragraph in Section 4.1 repeats the results, which should add references to compare and explain why this is the case. The second paragraph of the discussion is more like an introduction to land cover types but does not explain why.
7. The meltwater discussed in 4.2 is even an important part of the abstract, but the meltwater is not used in the results. The discussion should be based on the main content of the results, and the discussion should also consider the geological conditions, such as karst landform, in addition to the influence of vegetation.
8. It is suggested that Section 4.3 be parted in different sections according to different mechanisms.

Minor Comments

1. What does dependence mean?
2. Line 287: What are the multivariate mixed effects, and why do these variables combine?
3. Line 519: The results about arid areas should be added after the reference to compare.
4. Line 532: The figures in the results should be marked here.

The authors examine the dependences between soil moisture and precipitation, and their drivers across the northern hemisphere. They find substantial negative dependences, which are mostly attributed to evapotranspiration and vegetation conditions. The topic is intriguing and the methodology interesting. However, I think this paper could benefit from refinement in areas like novelty declaration, method justification, presentation quality, and enhanced supporting evidence for conclusions. Therefore, I would recommend a major revision.

Below are major concerns that expect to authors to address in the revised manuscript.

1) The authors need to identify the research gaps and specify any novel findings or methodology not reported in earlier studies. Negative correlations between soil moisture and precipitation and their causes have previously been identified, a fact acknowledged by the authors (Line 66-73). The authors seem to claim their novelty in terms of climate change and climate extremes (Line 101-106). However, I find the two points only loosely related to this study.

2) There is a lack of quantitative summary of the findings throughout the paper, especially in the abstract and conclusion sections. For instance, the authors should indicate the proportion of positive/negative correlations across various soil layers/ecoregions and quantify the contribution of controlling factors.

3) The Results and Discussion section reads too imbalance. Currently, there is a lack of reasoning of the findings shown in the Results section, making the results a bit dull to read. The reasoning in Discussion is too spread and redundant, causing readers having to flip between the two sections. Also, I think a schematic diagram might help.

4) The ridge regression and Section 3.2 seem off topic, as the main scope is to study the dependences between precipitation and soil moisture as well as their drivers. As a key driver of the dependences, why ET is not added to the Bayesian model. The soil property, another key controlling factor according to the authors, is also not considered in the Bayesian model as well (Line 116).

5) The authors should justify their use of eco-region boundaries over the more well-known climate region, e.g., Köppen climate classification system.

6) The dependence between soil moisture and precipitation might not be concurrent, and could have a lag time. There are little consideration and discussion of this point.

7) I also have concerns about the time scale. I agree with reviewer #1 that the time scale (monthly, seasonal, annual) should be unified. Since the authors did not eliminate seasonal variations from monthly data, seasonal signals affect the monthly-scale results. The patterns and mechanisms during seasons appear clearer. I would suggest the authors to narrow the analyses by only focusing on one or two scales.

Minor comments:

1) Line 489: the “Arctic amplification” appears abruptly. How are the climate pattern associated with the dependences? Why not other climate patterns?

2) Line 517-519: not clear, need rephrase.

3) Line 747-748: need quantitative measures to support this point.

4) Line 764-769: these reasoning needs quantitative support.