the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Oct 2024
A Global Terrestrial Precipitable Water Vapor Dataset from 2012 to 2020 Based on Microwave Radiation Imager Measurements from Three Fengyun Satellites
Abstract. A global terrestrial precipitable water vapor (PWV) dataset has been developed using observations from the MicroWave Radiation Imager (MWRI) aboard the FY-3 satellite series (FY-3B, FY-3C and FY-3D) spanning 2012 to 2020. The dataset offers twice-daily PWV records at a spatial resolution of 0.25° × 0.25°, aligned with the ascending and descending orbits of the FY-3 satellites. The dataset was generated using an automated machine learning (ML) model that leverages MWRI-based features characterizing surface conditions and an enhanced Global Position System (GPS) PWV dataset as a reference. Trained on over one million sampling points from more than ten thousand stations worldwide, the model ensures a robust representation of global PWV variations. Independent evaluations against SuomiNet GPS and Integrated Global Radiosonde Archive Version 2 (IGRA2) PWV products yielded root mean square error (RMSE) of 4.47 mm and 3.89 mm, respectively, with RMSE values ranging from 2.90 to 5.49 mm across various surface conditions. The dataset effectively captures both spatial and temporal PWV variations, allowing for precise examination of localized and abrupt changes in water vapor induced by extreme weather events. Representing a significant advancement in global terrestrial PWV monitoring, the MWRI PWV dataset provides an all-weather, high-precision data record that bridges gaps in global coverage of passive microwave-based terrestrial PWV observations. It is a valuable resource for atmospheric research, climate modeling, water cycle studies, and beyond. The dataset is available at: https://doi.org/10.6084/m9.figshare.26712709 (Xia et al., 2024).
- Preprint
(3004 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on essd-2024-395', Anonymous Referee #1, 22 Oct 2024
This paper employs a Lightweight Auto machine learning framework to produce a global terrestrial precipitable water vapor (PWV) dataset based on the MicroWave Radiation Imager (MWRI) aboard the FY-3 satellite series (FY-3B, FY-3C and FY-3D) spanning 2012 to 2020. The training dataset for the machine learning model is the enhanced GPS PWV dataset. SuomiNet GPS PWV, IGRA2 radiosonde, and the enhanced GPS PWV are used as reference datasets for validation. The authors examined the product quality from three perspectives: statistical fitting, spatial distribution, and temporal variation, while also assessing performance over different land surface types. It is recommended to be accepted after major revisions:
- The authors have not explained why the enhanced GPS PWV dataset was chosen as the training data for the machine learning model. This raises questions about the rationale of the research method.
- Around line 245, the explanation for the bias between MWRI PWV and IGRA2 PWV is based on the argument that "the enhanced GPS PWV shares the same bias with IGRA2 PWV." This explanation lacks persuasive power and is not supported by relevant studies.
- It is recommended to include an analysis of the machine learning model's uncertainty or error, particularly focusing on how the model performs under different weather conditions.
- The dataset performs poorly under extreme weather conditions. It is recommended to consider increasing the variety of training data for machine learning in such regions. By categorizing rainfall events, the authors could select the dataset that performs best under specific rainfall conditions as the training data for the machine learning model.
- The MWRI has a limited number of channels and lacks high-frequency channels, which makes it less sensitive to precipitation compared to sensors with high-frequency channels. Could this limitation be mitigated by incorporating data from other FY-3 sensors?
- In Figure 7, the number of validation stations seems not enough, and the spatial distribution is uneven, with most stations concentrated in Europe. Is the validation in other regions reliable enough?
- It is recommended to include a quality comparison between the FY-3 MWRI Level 1C Tb dataset and other Tb datasets to highlight the innovation of the study.
- Many of the references cited are outdated. It is recommended to incorporate more recent studies in the literature review.
Citation: https://doi.org/10.5194/essd-2024-395-RC1 -
RC2: 'Comment on essd-2024-395', Anonymous Referee #2, 05 Nov 2024
Review of " A Global Terrestrial Precipitable Water Vapor Dataset from 2012 to 2020 Based on Microwave Radiation Imager Measurements from Three Fengyun Satellites" by Xia et al.
The authors developed a global terrestrial precipitable water volume (PWV) dataset from 2012 to 2020 by applying a machine learning model using Microwave Radiation Imager (MWRI) observations on board the Fengyun satellites series. The accuracy of dataset is evaluated by comparing with the products of SuomiNet GPS and Integrated Global Radiosonde Archive Version 2 (IGRA2) PWV. This work contributes to representing spatial and temporal PWV variations and providing valuable resource for atmospheric research. The manuscript may be considered for publication after being major revised in accordance with the following comments:
General:
- The introduction could benefit from a more comprehensive discussion of the significance of PWV dataset in the context. This could include a brief overview of existing challenges and gaps in PWV dataset construction, and how this dataset addresses them. Additionally, the literature review should be expanded to include more recent studies on PWV retrieval employing machine learning techniques. This may help establish the novelty and contribution of approach proposed in this study.
- In the method section, the authors choose Light Gradient Boosting Machine (LightGBM), Extreme Gradient Boosting (XGBoost) and Random Forest to train the model. The reasons for selecting these models should be supplemented.
- In the conclusion, it is essential to articulate not only the strengths of the dataset but also to elucidate its constraints and limitations.
- Please check the grammar in the manuscript to improve the text quality. For example, the subject of the sentence that “With the development of computer science, and in particular the proliferation of machine learning (ML), has led to the widespread adoption of ML by the remote sensing community” is missing.
Specific：
- Line 216-217, The full names of “WAT, WET, ENF, EBF, DNF, DBF, and MF” should be provide when they first appear in the manuscript.
- Figure 5: It is clear that the amount of data when MWRI PWV is compared to IGRA2 PWV is much smaller than when it is compared to SuomiNet GPS PWV and enGPS PWV. This should be supplemented in the manuscript as well as giving possible reasons for this discrepancy.
- Figure 8: The different colors of the solid dots in the figure should be clearly explained. Please include a color bar to indicate what each color represents for better clarity.
- The manuscript states that "the MWRI PWV exhibits a wet bias at low PWV values and a slight PWV underestimation at high PWV values." Could you provide possible explanations for these observed biases? Discussing potential reasons, such as instrument limitations, atmospheric conditions, or retrieval algorithm issues, would help clarify this point.
Data sets
A Global Terrestrial Precipitable Water Vapor Dataset from 2012 to 2020 Based on Microwave Radiation Imager Measurements from Three Fengyun Satellites Xiangao Xia et al. https://doi.org/10.6084/m9.figshare.26712709
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 198 | 51 | 10 | 259 | 6 | 8 |
- HTML: 198
- PDF: 51
- XML: 10
- Total: 259
- BibTeX: 6
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1