Dear Authors,

After careful consideration of the split reviews and my own assessment, I am recommending revision of your manuscript. Reviewer 3 raises fair points about novelty and validation claims; at the same time, I recognise the value of your systematic exploration of CLM5’s sensitivity to soil parameter choices. The paper can make a useful contribution if the claims are tightened and the role of ERA5‑Land is represented accurately.

The central correction concerns ERA5‑Land. Throughout the manuscript it is treated as “observations” or an “observational benchmark,” and performance or validation language is built on that premise. This is not correct. ERA5‑Land is a model product—an offline land‑surface replay forced by ERA5—and it does not assimilate soil‑moisture observations. You have rightly added this clarification in Methods, but the same premise must be applied consistently across the paper. Please recast the terminology so that ERA5‑Land is described as model output used for pattern comparison, not as ground truth. In practical terms, any statement about “better performance,” “improved alignment,” or “validation” based solely on ERA5‑Land should be rewritten in terms of similarity of spatial and temporal patterns to ERA5‑Land.

Please keep a short, prominent statement in Methods that positions ERA5‑Land as a spatially complete, model‑based benchmark without land data assimilation and explains that it is used only for pattern consistency rather than validation of absolute levels. In the same place, note the forcing mismatch: your CLM5 experiments are GSWP3‑forced, whereas ERA5‑Land is forced by ERA5. Differences between CLM5 and ERA5‑Land therefore reflect both forcing and structural contrasts (CLM5 vs. HTESSEL) and should not be attributed to parameter effects alone.

The Abstract and Introduction should set this expectation from the outset. Please present the work plainly as an intra‑model sensitivity analysis within CLM5, not a validation exercise. The statement about study objectives that currently appears late in the Discussion should be moved into the Introduction. In the Results and figure captions, wherever ERA5‑Land is called “observed” or used to assert “better/worse performance,” please rephrase as “greater/lesser similarity to ERA5‑Land patterns” or “pattern differences.”

Please correct the figure and table cross‑references highlighted by the reviewer, and include a compact experiment summary table so that the distinction between Exp. 1 (single uniform parameter set) and Exp. 4a–d (separate uniform texture‑class sets, run individually) is unambiguous. It will also help to state plainly that neither your CLM5 configuration nor ERA5‑Land includes irrigation; readers should be cautioned not to over‑interpret agricultural hotspots on that basis. A brief Limitations paragraph that acknowledges model‑to‑model comparison, forcing mismatch, and structural differences will keep the conclusions proportionate.

I want to emphasise the positive core. The value here is not in “discovering” that soil parameters matter—this is well known—but in documenting, within CLM5, how parameter datasets propagate into regional patterns and variability, and in identifying where parameter uncertainty most strongly affects simulated soil moisture. Framed this way, the paper is useful for model users and developers, provided the claims are disciplined and the ERA5‑Land corrections are implemented.

Please also address the specific technical points raised by Reviewer 3 alongside these broader revisions. I will review your next revision with a focus on these issues.

Thank you for the careful work so far. I look forward to reading your next revision.

Best,
Ting

Dear authors,

Thank you for carefully addressing my and the other reviewer’s comments.
Having reviewed the revised manuscript, I am pleased to recommend it for publication.
Congratulations!

Best,
Ting