the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Investigating the performance of Genetic Particle Filter in snow data assimilation across snow climates
Yuanhong You
Chunlin Huang
Jinliang Hou
Ying Zhang
Abstract. With the aim of reducing the uncertainty of simulations, data assimilation methodology is increasingly being applied in operational purposes. This study aims to investigate the performance of genetic particle filter which used as snow data assimilation scheme, designed to assimilate ground-based snow depth measurements across different snow climates. We employed the default parameterization scheme combination within Noah-MP model as model operator in the snow data assimilation system. And the feasibility of genetic particle filter used as snow data assimilation scheme was investigated at different sites, at the same time, the impact of measurement frequency, particle number on the filter updating of the snowpack state were also evaluated. The results demonstrated that the genetic particle filter can be used as snow data assimilation scheme and obtain satisfactory assimilation results across different snow climates. We found the particle number is not the crucial factor to impact the filter performance and one hundred particles can sufficient to represent the high dimensionality of the point-scale system. The frequency of measurements can significantly affect the performance of filter updating and a dense ground-based snow observational data always can dominate the accuracy of assimilation results. Finally, we concluded that the genetic particle filter is a suitable candidate approach to snow data assimilation and appropriate for different snow climates.
Yuanhong You et al.
Status: final response (author comments only)
-
RC1: 'Comment on hess-2022-350', Anonymous Referee #1, 12 Nov 2022
The topic addressed is within the scope of HESS. The manuscript is generally well organized and results are clearly presented. This manuscript investigated the potential of GPF used as a snow data assimilation scheme across different snow climates, the results presented in this manuscript will help develop new data assimilation scheme and improve the simulation accuracy of land surface model that leads to improve weather and climate prediction. In my opinion, this manuscript could be accepted for publication in HESS after the following comments are addressed.
Comments:
- Line 106: “Above studies” may need some recent references.
- Line 245: “The number of particles was set to 100” have been expressed in line 225, I suggest deleting one.
- Line 250: the variance scaling factor of the temperature was set to 2.0, why this value was chosen, 3.0 or other value can be used here?
- Line 259: What does the “SD” is refer to? Do you mean SD is the abbreviation of snow depth?
- The abstract should provide some numerical values from the performance metrics of the results.
- Line 236: Except for the air temperature and precipitation can be perturbed, whether other meteorological forcing variable can be perturbed, such as relative humidity and wind speed? As far as I know, the wind speed has great impact on the distribution of snow.
- The English writing has to be polished.
Citation: https://doi.org/10.5194/hess-2022-350-RC1 -
AC1: 'Reply on RC1', Yuanhong You, 19 Dec 2022
Dear Reviewer,
Many thanks for taking the time to review our manuscript and provide constructive suggestions. We have made our best efforts to respond to your questions and aspects you would like addressed in the paper in the attached response file. We will updated a revised manuscript, where we have made changes and edits.
With best wishes,
Yuanhong You, on behalf of authors
-
RC2: 'Comment on hess-2022-350', Anonymous Referee #2, 03 Dec 2022
Snow depth simulation is difficult in land surface models and the data assimilation for snow depth is of importance for cold regions hydrology and energy balance modeling. This manuscript tried to propose a new data assimilation scheme for the land surface model. I read it with high interest but did not find out the logic of this manuscript. Therefore I cannot give too positive evaluation at the current stage, I would suggest a thorough revision before it can be considered for publication in the journals.
first, the language is quite poor and the writing is difficult to understand especially in the introduction and results and discussion parts. I can't understand the importance of this work except from my own understanding of the cold regions modeling and data assimilation...
second, some of the references are not shown in the references part even they are put in the main text. This is awful and I feel that the preparation for this work was not serious and also not strictly following the journal's rules...
besides, the results and discussion are quite awful in writing, as I can't find out the useful information from this work concluded by the authors. This is a quite pity issue even the meaningful work was conducted...
given the above mentioned issues, I feel that the detailed comments are not necessary if the authors don't make a thorough revision on the whole story telling logic.
therefore, I suggest a rejection this turn and a chance for resubmission with a clear outline that focuses on the most interesting part of the work would be a good suggestion from my side. Sorry for being not too positive this time given the current version of the manuscript.
Citation: https://doi.org/10.5194/hess-2022-350-RC2 -
AC2: 'Reply on RC2', Yuanhong You, 19 Dec 2022
Dear Reviewer
Thanks for your sincere and constructive suggestions. You have pointed out the deficiency of this paper, and we sincerely agree the comments. We are very sorry to submit this awful version, and we realized that the experiment results were not clearly expressed. We will try our best to revise this manuscript and ensure a great improvement in next version.
With best wishes,
Yuanhong You, on behalf of authors
Citation: https://doi.org/10.5194/hess-2022-350-AC2
-
AC2: 'Reply on RC2', Yuanhong You, 19 Dec 2022
-
RC3: 'Comment on hess-2022-350', Eduardo Zorita, 16 Dec 2022
Summary: The study is focused on the simulation of snow depth with a simple energy balance model with assimilation with observations. The data assimilation scheme is based on a genetic particle filter algorithm. The main conclusion is that this algorithm performs well for snow depth simulations.
Recommendation: the manuscript is in general terms poorly written. The English needs would need extensive copy-editing, and this deficiency often hinders the understanding of technical and scientific aspects of the study. In addition, the manuscript does not include important information that is critical to understand what has been done. One is that there is no description of the genetic particle filter algorithm itself, which is surprising. Unfortunately, I cannot recommend the publication of the manuscript. A revision would entail rewriting the manuscript almost completely.
Main points
1) The presentation is poor. The English needs very extensive revisions; acronyms are not defined (for instance SD, which I interpret to be snow depth!. SWE and GPF are not defined either) . The discussion is restricted to the own results and does not place the results in the framework of previous studies (what has been learned , what is the novelty?)
2) The destitution of important technical aspects is missing. The genetic algorithm is essentially not described - the paragraph starting in line 218 is so obscure that essentially nothing can be understood. The reader is not informed of many technical aspects. What are the 'particles' ? how are they genetically generated? what are the crossover and mutation operators ? why the genetic algorithm would improve on the deficiencies of other particle filters ?
3) I kept wondering of the utility and meaning of some of the mathematical assumptions. For instance, equation 9 seems to be unnecessary complicated. The distribution of the random noise w is just uniform in (-2,2), so there is no need for the additional complexity of equation 9. Also, why would the temperature errors be uniformly distributed ? why between -2 and 2 and which units represent those numbers ( I guess C ?). This is an example of a problem that goes through the whole manuscript.
Citation: https://doi.org/10.5194/hess-2022-350-RC3 -
AC3: 'Reply on RC3', Yuanhong You, 19 Dec 2022
Dear Eduardo Zorita
Many thanks for taking the time to review our manuscript and provide constructive comments. We have made our best efforts to respond to your questions and detail in the attached response file. You find our replies to your comments in the attached file.
With best wishes,
Yuanhong You, on behalf of authors
-
AC3: 'Reply on RC3', Yuanhong You, 19 Dec 2022
-
RC4: 'Comment on hess-2022-350', Anonymous Referee #4, 21 Dec 2022
This paper optimizes the snow depth data simulated by the model by using the snow depth data observed at the station. The main questions are as follows:
1. The title of the article is not accurate, and the purpose of assimilation cannot be obtained. The meaning of "snow" is too broad, so it needs to be specific;
2. Is the Ws in NOAH-MP optimized by snow depth observation? Ws is the snow water equivalent. Is there any difference between the two?
3. The observation operator and model operator mentioned in Flow Chart 1 are not seen in the text, and need to be clarified;
4. The optimized state variables are not seen in the paper. If only the simulation results of NOAH-MP are corrected, which can not be regarded as assimilation;
5. The calculation formula of fsnow, g needs to be given. Snow depth is also used in the calculation formula of fsnow, g. If snow depth is assimilated, has this been considered? If regional assimilation is carried out, how is fsnow, g calculated?
6. The introduction of assimilation process in this paper is not complete and detailed enough, and needs to be further improved.Citation: https://doi.org/10.5194/hess-2022-350-RC4 -
AC4: 'Reply on RC4', Yuanhong You, 07 Jan 2023
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-350/hess-2022-350-AC4-supplement.pdf
-
AC4: 'Reply on RC4', Yuanhong You, 07 Jan 2023
-
RC5: 'Comment on hess-2022-350', Anonymous Referee #5, 21 Dec 2022
Review on the “Investigating the performance of Genetic Particle Filter in snow data assimilation across snow climates” by You et al.
Snow is a critical hydrothermal variable in land surface that influences the surface energy and hydrological cycles. Snow data assimilation is also important to improve the modeling accuracy of snow and thus facilitates other related processes (e.g., albedo, snow melting runoff, and temperature). You et al. investigated the performance of genetic particle filter in snow data assimilation over eight stations and discussed the influences of different assimilating frequency and particle number. Although I am not very familiar with the data assimilation method especially the mathematical rule, I think the current research needs at least major revision to illustrate its novelty, introduce the method more clearly, analyzing the results in depth, and proofread the manuscript carefully. Detailed comments are below:
Major comments:
- The novelty of this research is not clear. If the genetic particle filter is a new method, then its advantages against other methods should be investigated directly. If the “across snow climates”is a novelty, then I do not think the eight stations can represent snow climates considering the highly heterogeneous snow distribution. Finally, “the higher assimilating frequency, the higher simulation performance” is not surprising. Thus, I suggest the authors to clarify the novelty more clearly, so as to help the reader get the importance of this work.
- The method needs further introduction. Although the genetic particle filter data assimilation scheme is introduced in section 2.3, I am still confused that which variable you assimilate into the land surface model? If you only assimilate the snow depth, then how do you deal with other snow variables (e.g., snow water equivalent, snow density and snow age)? Also, how do you deal with the potential inconsistency between snow and ground temperature (for example, when the model shows no snow and the ground temperature is above zero, then how to assimilate the observed snow depth)?
- Actually, I am very concerned about the assimilation and evaluation. In figure 3, it seems you assimilate the observation every 5 days and then evaluate the model simulation at the same step? If this is the case, then will the direct insertion method show higher performance than the genetic particle filter?
- The spatial difference. It seems the spatial difference among different stations is not strong and few information can be get (except the robust of the result, may be). Some insightful analysis on the spatial difference may help improve the manuscript.
- The writing needs careful proofreading. For example:
L48: “succeeds in catching snow dynamics is” may be “succeed in catching snow dynamics is”
L51: “is aimed at investigating ... and obtain the ...” may be “is aimed at investigating ... and obtaining the ...”
L60: “However, this method possible result in ...” may be “However, this method possible results in ...”
L68: ”this method does not require a model a model linearization.” what do you mean?
Citation: https://doi.org/10.5194/hess-2022-350-RC5 -
AC5: 'Reply on RC5', Yuanhong You, 07 Jan 2023
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-350/hess-2022-350-AC5-supplement.pdf
-
EC1: 'Comment on hess-2022-350', Carla Ferreira, 28 Dec 2022
Dear authors,
This manuscript uses field data to improve snow modelling through the genetic particle filter algorithm. The topic is highly relevant and fits nicely within this SI. However, the manuscript requires significant improvements in order to be considered for possible publication in HESS. The Introduction section should be shorted, more concise and better highlight the research question and the need for this study, and use more updated references. The study sites must be better described, namely regarding the snow differences between the sites, so that we can understand the real application of your proposed method. The methodology requires relevant improvements to describe the genetic particle filter algorithm, and all the mathematical assumptions performed. The Results must be clearly presented and discussed. Discussion should clearly present the advantages of the proposed method comparing with others, and the limitations linked with the assumption performed. It is also important to compare the results with those from previous studies, and bring more references into this section. The Conclusions section must clearly identify the novelty and main messages of this study, and clearly identify why the proposed method is better than the available ones. Language editing is also required. More detailed comments have been provided by reviewers and must be considered in improving the manuscript. There was one late reviewer whose comments are provided bellow and should be also considered.
Reviewer 6:
In this manuscript, You et al. set up a particle filtering framework using the Genetic Algorithm to avoid particle filter-typical degeneracy and sample impoverishment issues. They apply this framework to snow depth measuring sites in different climatological regions, expecting to learn about particle filter performance at each of the sites. They analyze the assimilated snow depth with respect to the suitability of their particle filtering algorithm for application in different snow climates, the influence of the used particle number on performance metrics, and the influence of assimilation window length on the performance.
The manuscript is structurally well-organized. The topic is in general very interesting and the effort to push the field of data assimilation forward is very much appreciated. However, in my opinion there are some significant weaknesses in this manuscript:
- a lacking motivation of the research question(s') relevance
- a superficial description of the measurement sites and their properties, making a meaningful interpretation of the results with respect to the research question(s) difficult
- a lacking presentation of the Genetic Algorithm and stressing why this method is the most suitable for the analysis
- a superficial interpretation of the results, in particular with respect to the overarching hypothesis (different filter performance in different "snow climates")
- issues with the used literature in the References section and in general a rather scarce literature selection
- difficulties in the use of English, which makes some sections of the manuscript hard to understand
- an intransparent (or simply not listed?) choice of model parameter values and meteorological values to perturb; unclear or not explained error distribution choices
- a results and discussion section that partly loses contact with the research questions
If this manuscript is accepted for a major revision process, it should be largely rewritten and then undergo line-by-line comments in a second review iteration. The focus should first be on the following aspects:
- reworking the manuscript research questions (is it about the filter performance in different climates as the title suggests or about the three questions formulated at the end of the introduction, or both?)
- a more comprehensive literature review on the technical literature regarding the research questions
- a more detailed description of the used particle filter method and why this filter is chosen to be the most suitable to answer the research question
- a more critical questioning of the results, in particular with respect to the 100-particle threshold (e.g. why in Fig. 7 a minimum exists at 100 particles).
Citation: https://doi.org/10.5194/hess-2022-350-EC1 -
AC6: 'Reply on EC1', Yuanhong You, 07 Jan 2023
The comment was uploaded in the form of a supplement: https://hess.copernicus.org/preprints/hess-2022-350/hess-2022-350-AC6-supplement.pdf
-
AC6: 'Reply on EC1', Yuanhong You, 07 Jan 2023
Yuanhong You et al.
Yuanhong You et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 541 | 173 | 30 | 744 | 6 | 5 |
- HTML: 541
- PDF: 173
- XML: 30
- Total: 744
- BibTeX: 6
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1