The manuscript presents a method to parametrize the soil water retention curve. The authors build on their previous work, namely Rooij et al. (2021) (https://doi.org/10.5194/hess-25-983-2021), which is further modified in this work to let matric potential at oven-dryness fitted or fixed (log(-hd)= 6.8 cm H2O). The code and a detailed guidance for fitting the presented method is provided in Fortran and available from ZENODO (https://zenodo.org/record/6491979).
The presented fitting method aims to improve the description of the dry end of the soil moisture retention curve with a method that does not increase the number of parameters that have to be fitted, this way could be applied by many users. It would be important to add some information about the Twarakavi et al. (2010) classes under “3.1 Selected soils” paragraph, because it was used for selecting the 21 samples for the analysis. It would be very informative:
- to test the fitting performance on data that has theta-head pairs from the dry range of the soil moisture retention curve,
- add descriptive statistics of the 21 samples with information on organic carbon content, sand, silt and clay content, bulk density, pH, calcium carbonate content and information on soil type and summarize based on that which soils were covered in the analysis,
- to further analyse why the method is not accurate for clay loam and clay soils and possible further specific soils (e.g.: volcanic soils, salt affected soils, organic soils, etc.), for this many more samples from other soil hydraulic datasets could be used, e.g.: HYPRES and/or EU-HYDI and/or GRIZZLY and/or data from the tropics, etc. or at least mention that this has to be performed in the future.
The text includes many technical details, some parts could be explained more to ease understanding and use of the method by the readers. It would be important to add a separate section which:
1. highlights the advantages, disadvantages and limitation of the presented fitting method by comparing its “capabilities” to other existing models to fit the moisture retention curve;
2. lists what has to be fulfilled for the proper use of the fitting method – kind of very short practical guidance, e.g.: minimum number of theta-head pairs in the input data required for the fitting, how many theta-head pairs are required between specified matric head values – if that is important, computation of standard deviation of the error in the matric potential, etc.
Please find some further comments under SPECIFIC COMMENTS section.
After solving the above mentioned issues the paper can attract wide interest.
L15: why the fitting was performed for only 21 samples? What about the other samples of UNSODA, why were not the other samples used?
L16: Please specify what you mean under „The fits are good”.
L17: please shortly add why it is problematic/ highlighted: „For some soils, α is very large.”
L19-20: would be more informative if the last sentence of the abstract would be a statement which is based on the results, please highlight the main advantage of the method presented in the manuscript, why is it recommended to use instead of other models available from the literature?
L82-82: … realistic but large n (n > ∼ 1.4 and α > ∼0.01) require values …
L87-88: … For soil 2126 …
L116-118: please rephrase the sentence starting with „Exploring the parameter space” to ease understanding.
L137: please add the range in numbers if possible after stating the following: „α can vary over its entire range”.
L161: … Its value …
L208-210: please shortly add here how these SD were considered during the fitting.
L211: please explain why the following was done: „The sample height was set to zero for h ≤ –1000 cm and for h = 0 cm.”
L228: Please shortly add what the “complexes” mean.
L252-253: please check if the following is accurate: „The code returns a table of the fitted curve based on the best run”. The code might return the fitted parameters or computed theta-head pairs based on the fitted parameter. Is that correct?
L262-264: please describe the following sentences to ease understanding: “If these values are equal, the parameter is treated as a fixed value, and the dimensionality of the parameter space is reduced accordingly. The number of complexes is two (for 8 of fewer fitting parameters) or four (see Duan et al. (1994)).” E.g.: what do you mean by equal? What does number of complexes mean? Maybe you could provide examples.
L344-345: Please add possible reasons for the less accurate fit. Could the fit be improved if the multimodal version of the method is applied for the fitting?
L366-369: the availability of the code with detailed documentation and user manual is plausible, just a minor suggestion: use of the presented method might be easier if it would be implemented later on into Python or R as well.