the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Jan 2026
Parameter estimation for land-surface models using Neural Physics
Abstract. The Neural Physics approach is used to determine the parameters of a simple land-surface model using PyTorch’s backpropagation engine to carry out the optimisation. In order to test the inverse model, a synthetic dataset is created by running the model in forward mode with known parameter values to create soil temperature time series that can be used as observations for the inverse model. We show that it is not possible to obtain a reliable parameter estimation using a time series of soil temperature observed at a single depth. Using measurements at two depths, reliable parameter estimates can be obtained although it is not possible to differentiate between latent and sensible heat fluxes. We apply the inverse model to urban flux tower data in Phoenix, United States, and show that the thermal conductivity, volumetric heat capacity, and the combined sensible-latent heat transfer coefficient can be reliably estimated using an observed value for the effective surface albedo. The resulting model accurately predicts the outgoing longwave radiation, conductive soil fluxes and the combined sensible-latent heat fluxes.
Status: open (until 11 Mar 2026)
- RC1: 'Comment on egusphere-2025-6015', Anonymous Referee #1, 06 Feb 2026 reply
-
CEC1: 'Comment on egusphere-2025-6015 - No compliance with the policy of the journal', Juan Antonio Añel, 06 Feb 2026
reply
Dear authors,
Unfortunately, after checking your manuscript, it has come to our attention that it does not comply with our "Code and Data Policy".
https://www.geoscientific-model-development.net/policies/code_and_data_policy.html
First, I am sorry that have to point that a previous comment by the Topical Editor regarding compliance with the policy of the EDI site was incorrect. We can not accept that you store data necessary to replicate your work in such site, which we can not consider a long-term trusted repository for scientific publication. Therefore, you must store the data linked from such site in one of the repositories we accept. Right now, it is unclear if they are included in the Zenodo repository, which you have provided us internally, but not published in your submitted manuscript, under the file on flux tower data. I am linking here the mentioned Zenodo repository, so that readers and the scientific community have access to it during the Discussions stage, which is mandatory and necessary:
https://doi.org/10.5281/zenodo.18004983
Beyond all this, you do not provide the exact input and output files of your computations, neither the NN4PDE which you use for your manuscript. Also, the provided Python notebooks rely on many third party libraries, for which you have not indicated their versions. This is important to ensure the replicability of your work, as the same algorithm could be coded in very different ways in different versions of them. Therefore, please, we are requesting you to identify the versions of Python and libraries that you have used for your work.
The GMD review process depends on reviewers and community commentators being able to access, during the discussion phase, the code and data on which a manuscript depends, and on having clear information on the software used. Please, therefore, address the issues pointed above and publish your code and data in one of the appropriate repositories and reply to this comment with the relevant information (link and a permanent identifier for it (e.g. DOI)) as soon as possible. We cannot have manuscripts under discussion that do not comply with our policy.
Also, please, note that the 'Code and Data Availability’ section of your manuscript must also be modified to cite the new repository locations, and corresponding references added to the bibliography.
I must note that if you do not fix this problem, we cannot continue with the peer-review process or accept your manuscript for publication in GMD.
Juan A. Añel
Geosci. Model Dev. Executive EditorCitation: https://doi.org/10.5194/egusphere-2025-6015-CEC1
Data sets
Eddy covariance data measured at the CAP LTER flux tower located in the west Phoenix, AZ neighborhood of Maryvale from 2011-12-16 through 2012-12-31 Winston Chow https://doi.org/10.6073/pasta/fed17d67583eda16c439216ca40b0669
Model code and software
SEB-model Ruiyue Huang https://github.com/RuiyueH/SEB-model
SEB model Ruiyue Huang https://doi.org/10.5281/zenodo.18004983
Viewed
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 168 | 0 | 0 | 168 | 0 | 0 |
- HTML: 168
- PDF: 0
- XML: 0
- Total: 168
- BibTeX: 0
- EndNote: 0
Viewed (geographical distribution)
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This study uses machine learning approaches to determine soil parameter values of a simplified land surface model, based on soil temperature data. A key finding of the study is that it is not possible to determine some soil parameters using observations of soil temperature at a single depth level, but that two depths can reliably estimate soil parameters including heat capacity, conductivity and air-surface heat flux transfer coefficients.
The manuscript is well structured and presented with conclusions that are well supported by the results. I commend the authors for introducing new methods to this field, generating novel results, and communicating/discussing results very clearly. I therefore have only a few comments. I am not a machine learning expert, so I leave the details of the machine learning approaches to other reviewers.
Page 3 states the land surface model “does not incorporate water in the sub-surface”, in which case I would have expected latent heat flux to be zero. However much of the paper is given to discussing the partitioning of latent/sensible heat fluxes, and the inverse model is asked to find the Bowen Ratio. Can authors explain from where latent heat fluxes in this scenario are originating or be explicit that latent heat fluxes are expected to be zero. If zero, can authors simplify the parameter inputs to exclude Bowen Ratio (i.e. all turbulent fluxes are partitioned into sensible heat in this scenario)?
On Page 8 the solar zenith angle is parameterised. A short explanation of the form of the parametrisation would be useful for readers.
The provided code (https://github.com/RuiyueH/SEB-model) does not include any instructions or README. In its current state I would not say this study is easily reproducible.
If authors wish to reach a larger audience, more consideration could be given to different reader backgrounds. I believe those with machine learning interests are well served, but authors could also consider users of traditional land-surface models and teams that observe land-atmosphere fluxes at flux tower sites. Both may find this machine learning technique interesting and potentially useful (although the current no-moisture LSM employed is a barrier, as authors have noted in the conclusion). Still, it could be beneficial for the reach of this study if authors further consider these users in the text and then provide an easier-to-follow code example. The current codebase includes no instructions for these users.
Authors could also take the opportunity to reconsider the abstract with a wider audience in mind (as above), for example indicating the potential of this approach for determining specific soil characteristics but mentioning current barriers (e.g. the current assumption of no soil water).