the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
At-scale Model Output Statistics in mountain environments (AtsMOS v1.0)
Abstract. This paper introduces the AtsMOS workflow, designed to enhance mountain meteorology predictions through the downscaling of coarse numerical weather predictions using local observational data. AtsMOS provides a modular, open-source toolkit for local and large-scale forecasting of various meteorological variables through modified Model Output Statistics – and may be applied to data from a single station or an entire network. We demonstrate its effectiveness through an example application at the summit of Mt. Everest, where it improves the prediction of both meteorological variables (e.g. wind speed, temperature) and derivative variables (e.g. facial frostbite time) critical for mountaineering safety. As a bridge between numerical weather prediction models and ground observations, AtsMOS help produce insights for hazard mitigation, water resource management, and other weather-dependant issues in mountainous regions and beyond.
- Preprint
(4337 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on gmd-2024-36', Anonymous Referee #1, 30 Apr 2024
The paper introduces an MOS (Model Output Statistics) method using artificial intelligence (machine learning) suitable for at-scale rapid deployment, for correcting deviations of numerical weather prediction (NWP) in complex mountainous areas. It uses the Mount Everest climbing meteorological service in the Himalayas as a pilot study to validate the method's feasibility. The paper also discusses the advantages, potential issues, and risks of this method. The main review comments and suggestions are as follows:
- As a manuscript intended for publication in GMD (Geoscientific Model Development), there needs to be a very detailed description of the described technical methods and model. In this manuscript, the technical details of AtsMOS need further refinement and organization. It is preferable to provide a more detailed flowchart than Figure 1, or to add detailed sub-flowcharts for each module, accompanied by text descriptions, especially for the implementation process, parameter settings of XGBoost and RF, etc., to enhance the practical reference value of this open-access paper.
- The paper only conducts simulated comparative analysis and verification based on observations and forecasts of Mount Everest in the Himalayas. On the one hand, for the verification of weather transition stages (rapid temperature decrease or increase, rapid increase or decrease in wind speed), a detailed analysis is needed. On the other hand, if possible, more experiments and comparative analyses can be conducted with richer observations and AtsMOS forecasts in other mountainous regions around the world (such as the European Alps, the Rocky Mountains in the United States) to strengthen the reliability and universality validation of this method.
- The textual presentation of the paper needs to be more rigorous. For example, some abbreviations need to be provided in full, or a list of abbreviations can be provided at the end of the paper. Sentence expressions need to be more rigorous, and writing needs to be standardized (such as subscript and superscript issues, unit measurement issues, meteorological professional expression issues, etc.).
- For Figures 5-7, it is recommended to extract the data segment that simultaneously includes observation and forecast results and redraw clearer graphs (or add a curve showing the difference between the two ones). The current figures do not clearly show the specific differences.
- The titles of all figures need to be further refined to increase clarity.
- The description of Kling-Gupta efficiency needs to be clarified. This evaluation index is mainly used in hydrology. Whether it is suitable for this work should be clearly explained.
- An interesting question is whether further verification and comparison can be conducted for forecast results similar to Figure 8. If the Everest climbing team(s) or guides have records or carry instruments with similar data, such verification and comparison can be conducted. I believe under such extreme geographical conditions in Everest, this serves as a meaningful validation and assessment of the AtsMOS method.
Citation: https://doi.org/10.5194/gmd-2024-36-RC1 - AC2: 'Reply on RC1', Maximillian Van Wyk de Vries, 29 Jul 2024
-
RC2: 'Comment on gmd-2024-36', Anonymous Referee #2, 07 May 2024
This paper describes a model output statistics (MOS) approach aimed at improving weather forecasts in mountain environments. The authors focus on Mount Everest and show example usage with high elevation station data from the region. In addition to the model description, they explain their procedure for pre-processing the station data. I agree with the author’s arguments on the importance of improved forecasts on Everest to improve climber safety and the need for combined NWP and in situ data to achieve this. The high altitude, telemetry equipped station network on Everest provides unique and interesting opportunities for model development targeting this goal. The example is instructive (though limited to one location) and the discussion section provides valuable context on strengths and limitations of the model. I agree with reviewer 1 regarding the need for a more detailed description of the model and I consider this the main area that should be revised/extended. I have a few additional suggestions below.
General comments:
- Given the nature of the journal, I would like to see a more detailed description of the various model components. The model code is available but the documentation of the code could also be extended to help potential users get started. For the paper, I would suggest a subsection for each of the main processing modules, with particular focus on the steps in the “core processing” section of Fig. 1. This should include an explanation of the various learning techniques that were implemented.
- A subsection in the methods on the performance metrics used for model evaluation would also be beneficial.
- Figures: Captions should be extended to fully explain the contents of the figures. Alternative visualizations for the contents of Figures 5-7 might be explored to better show the differences between learning methods.
Specific comments:
Fig 4: What are the black and red lines? Add info in a legend, extend the figure caption to explain what is shown in the figure.
Fig 5-7: hard to see differences between the learning methods. Can this be combined to show the different model results in the same plot? What do the colors in the mesh plot (lower right panel) represent? Please explain in the caption, add a colorbar.
Fig 8: Can the GFS forecast of temp and wind speed and the station data for the same time period be added for comparison? As is, the figure shows that the model outputs something and derivative values (WCT, frostbite time) can be computed. I believe more information could be added quite easily to enhance the contents of this figure.
L184 This seems to be the first mention of Random Forest. In my opinion this should be introduced in the methods section.
L187 Kling Gupta - This has also not been mentioned previously. Consider adding a subsection in the methods addressing your performance metrics.
Typos:
L135 “Figure X”
L194 “the estimated are more closely clustered” - missing word?
L210 “The facial frostbit time briefly falls below 10 minutes this night also driven the the high wind speeds” - typos
Citation: https://doi.org/10.5194/gmd-2024-36-RC2 - AC1: 'Reply on RC2', Maximillian Van Wyk de Vries, 29 Jul 2024
Status: closed
-
RC1: 'Comment on gmd-2024-36', Anonymous Referee #1, 30 Apr 2024
The paper introduces an MOS (Model Output Statistics) method using artificial intelligence (machine learning) suitable for at-scale rapid deployment, for correcting deviations of numerical weather prediction (NWP) in complex mountainous areas. It uses the Mount Everest climbing meteorological service in the Himalayas as a pilot study to validate the method's feasibility. The paper also discusses the advantages, potential issues, and risks of this method. The main review comments and suggestions are as follows:
- As a manuscript intended for publication in GMD (Geoscientific Model Development), there needs to be a very detailed description of the described technical methods and model. In this manuscript, the technical details of AtsMOS need further refinement and organization. It is preferable to provide a more detailed flowchart than Figure 1, or to add detailed sub-flowcharts for each module, accompanied by text descriptions, especially for the implementation process, parameter settings of XGBoost and RF, etc., to enhance the practical reference value of this open-access paper.
- The paper only conducts simulated comparative analysis and verification based on observations and forecasts of Mount Everest in the Himalayas. On the one hand, for the verification of weather transition stages (rapid temperature decrease or increase, rapid increase or decrease in wind speed), a detailed analysis is needed. On the other hand, if possible, more experiments and comparative analyses can be conducted with richer observations and AtsMOS forecasts in other mountainous regions around the world (such as the European Alps, the Rocky Mountains in the United States) to strengthen the reliability and universality validation of this method.
- The textual presentation of the paper needs to be more rigorous. For example, some abbreviations need to be provided in full, or a list of abbreviations can be provided at the end of the paper. Sentence expressions need to be more rigorous, and writing needs to be standardized (such as subscript and superscript issues, unit measurement issues, meteorological professional expression issues, etc.).
- For Figures 5-7, it is recommended to extract the data segment that simultaneously includes observation and forecast results and redraw clearer graphs (or add a curve showing the difference between the two ones). The current figures do not clearly show the specific differences.
- The titles of all figures need to be further refined to increase clarity.
- The description of Kling-Gupta efficiency needs to be clarified. This evaluation index is mainly used in hydrology. Whether it is suitable for this work should be clearly explained.
- An interesting question is whether further verification and comparison can be conducted for forecast results similar to Figure 8. If the Everest climbing team(s) or guides have records or carry instruments with similar data, such verification and comparison can be conducted. I believe under such extreme geographical conditions in Everest, this serves as a meaningful validation and assessment of the AtsMOS method.
Citation: https://doi.org/10.5194/gmd-2024-36-RC1 - AC2: 'Reply on RC1', Maximillian Van Wyk de Vries, 29 Jul 2024
-
RC2: 'Comment on gmd-2024-36', Anonymous Referee #2, 07 May 2024
This paper describes a model output statistics (MOS) approach aimed at improving weather forecasts in mountain environments. The authors focus on Mount Everest and show example usage with high elevation station data from the region. In addition to the model description, they explain their procedure for pre-processing the station data. I agree with the author’s arguments on the importance of improved forecasts on Everest to improve climber safety and the need for combined NWP and in situ data to achieve this. The high altitude, telemetry equipped station network on Everest provides unique and interesting opportunities for model development targeting this goal. The example is instructive (though limited to one location) and the discussion section provides valuable context on strengths and limitations of the model. I agree with reviewer 1 regarding the need for a more detailed description of the model and I consider this the main area that should be revised/extended. I have a few additional suggestions below.
General comments:
- Given the nature of the journal, I would like to see a more detailed description of the various model components. The model code is available but the documentation of the code could also be extended to help potential users get started. For the paper, I would suggest a subsection for each of the main processing modules, with particular focus on the steps in the “core processing” section of Fig. 1. This should include an explanation of the various learning techniques that were implemented.
- A subsection in the methods on the performance metrics used for model evaluation would also be beneficial.
- Figures: Captions should be extended to fully explain the contents of the figures. Alternative visualizations for the contents of Figures 5-7 might be explored to better show the differences between learning methods.
Specific comments:
Fig 4: What are the black and red lines? Add info in a legend, extend the figure caption to explain what is shown in the figure.
Fig 5-7: hard to see differences between the learning methods. Can this be combined to show the different model results in the same plot? What do the colors in the mesh plot (lower right panel) represent? Please explain in the caption, add a colorbar.
Fig 8: Can the GFS forecast of temp and wind speed and the station data for the same time period be added for comparison? As is, the figure shows that the model outputs something and derivative values (WCT, frostbite time) can be computed. I believe more information could be added quite easily to enhance the contents of this figure.
L184 This seems to be the first mention of Random Forest. In my opinion this should be introduced in the methods section.
L187 Kling Gupta - This has also not been mentioned previously. Consider adding a subsection in the methods addressing your performance metrics.
Typos:
L135 “Figure X”
L194 “the estimated are more closely clustered” - missing word?
L210 “The facial frostbit time briefly falls below 10 minutes this night also driven the the high wind speeds” - typos
Citation: https://doi.org/10.5194/gmd-2024-36-RC2 - AC1: 'Reply on RC2', Maximillian Van Wyk de Vries, 29 Jul 2024
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
341 | 118 | 105 | 564 | 26 | 24 |
- HTML: 341
- PDF: 118
- XML: 105
- Total: 564
- BibTeX: 26
- EndNote: 24
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1