the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Mar 2026
A database of objectively identified atmospheric rivers based on a multi-method fusion algorithm
Abstract. Atmospheric rivers (ARs) are long, narrow corridors in the atmosphere that transport immense amounts of water vapor poleward across the mid-latitudes. On one hand, ARs can act as one of the precipitation sources, end drought, accumulate snowpacks, and support ecosystems and the society. One the other hand, ARs also represent a type of hazard and are responsible for economic losses, such as by extreme precipitation and winds, rain on existing snowpacks, or causing debris flows and landslides. Given the importance of ARs, this paper proposes a multi-method fusion algorithm for more objectively identifying ARs on a global scale. The proposed algorithm, based on the vertically integrated water vapor transport (IVT), integrates advanced strategies from multiple existing algorithms and introduces a dual-axis test method to enhance the stability of AR identification. Using IVT data from ERA5, a global AR database is constructed at a horizontal resolution of 1° × 1° and a time interval of six hours from 1940 to 2024. Comparative evaluation against established AR databases reveals strong agreement in mid-latitude ocean basins where ARs are most active. The usefulness of the new AR database is also demonstrated by examining the role of ARs in two extreme events in the recent past: atypical AR activity during the East Asian Meiyu rainfalls in late June 2018, and rare AR activity during the Australian Black Summer in late January 2020. The results show that the new AR database helps to reduce the uncertainty in AR identification and to better understand extreme events and their variations in time and space.
- Preprint
(4232 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on essd-2025-836', Anonymous Referee #1, 12 May 2026
-
AC1: 'Reply on RC1', Jian Rao, 30 Jun 2026
The comment was uploaded in the form of a supplement: https://essd.copernicus.org/preprints/essd-2025-836/essd-2025-836-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Jian Rao, 30 Jun 2026
-
RC2: 'Comment on essd-2025-836', Anonymous Referee #2, 22 May 2026
Review of Chen et al: A database of objectively identified atmospheric rivers based on a multi-method fusion algorithm
This manuscript provides a mostly clear overview of a new Atmospheric River (AR) identification algorithm which uses two methods of identifying the AR major axis and therefore identifying the geometric extent of ARs. The figures are clear and help explain the dataset. Given there are a lot of existing AR identification algorithms, the authors could add more justification in the introduction on why their method is needed e.g. its value in distinguishing between ARs and other synoptic weather systems with high IVT.
Below are minor comments that would improve the clarity of the dataset description.
L11: change “can act as one of the precipitation sources” to “can cause precipitation”
L147: Why choose 15%?
L155: ‘the number of axis points exceeds 500’ – This upper threshold for the number of points is confusing. Presumably each point is located at a grid box? And if you are using a coarsened 1x1 degree resolution then 500 points would circle the whole world?
L187: ‘where l denotes the axis length’ units?
L212: “To overcome the respective disadvantage of the two methods, out algorithm integrates both…” It is unclear how this is done. Do both axis methods need to agree that there is an AR for the candidate to be considered an AR? If not, how are the methods weighted?
L257-8: Could cite Reid et al (2020) here which specifically discusses this problem in AR identification: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020JD032897
Table 1: The values in brackets would be easier to interpret as percentages rather than raw values.
L286-290: It would be useful to the reader to include a few sentences about how DB2 and DB3 differ from DB1 in terms of the detection method, IVT thresholds and geometric requirements.
Figure 5: There is a clear discontinuity at 180 longitude in g, h & i. You could pad the sides of the array when doing the identification/analysis to prevent the discontinuity i.e. copy 10-20 degrees longitude of data from one side of the array and attach it to the other side for the calculations and then remove the extra data for plotting.
Figures 6 and 7 are in the wrong places (Figure 6 should be the East Asia case and Figure 7 should be the Australia case).
L397: The fires lasted until late March, so I would be careful saying that this even extinguished them.
Citation: https://doi.org/10.5194/essd-2025-836-RC2 -
AC2: 'Reply on RC2', Jian Rao, 30 Jun 2026
The comment was uploaded in the form of a supplement: https://essd.copernicus.org/preprints/essd-2025-836/essd-2025-836-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Jian Rao, 30 Jun 2026
Status: closed
-
RC1: 'Comment on essd-2025-836', Anonymous Referee #1, 12 May 2026
The manuscript develops a multi-method fusion algorithm based on vertically integrated water vapor transport (IVT) for global atmospheric rivers detection. The proposed algorithm incorporates advanced strategies from existing algorithms, adopts a dual-axis method to improve stability, and introduces curvature-related criteria to filter out vortex-like structures. However, several concerns need to be addressed before the manuscript is suitable for publication.
Majors:
- Definition issue.
In the abstract, the manuscript states that, “Atmospheric rivers (ARs) are long, narrow corridors in the atmosphere that transport immense amounts of water vapor poleward across the mid-latitudes.”
Several keywords are embedded in the definition: (1) elongated geometry (long, narrow), (2) poleward transport, and (3) occurrence in the mid-latitudes.
The elongated geometry is enforced by the first two criteria (Lines 184–185), while the curvature criteria (Lines 185–187) are related to the poleward requirement.
However, it is unclear how the “mid-latitude” aspect is enforced in the algorithm. The manuscript discussed filtering certain tropical AR candidates (Lines 222–224: grids located in tropics must not exceed 95%), but I did not see an equivalent consideration for high latitudes. Please clarify whether the algorithm intends to apply the mid-latitude criterion. If so, please justify the 95% threshold for the tropics, which seems relatively permissive, and explain how the algorithm distinguishes or filters features at high latitudes.
2. Poleward transport requirement
Using the provided dataset, I plotted ARs for the monsoon case in a domain larger than that used in the manuscript. Attached are results at 06Z and 12Z 27 Jun 2018 (blue shading for ARs). Several features appear to deviate from the definition of an AR:
- At 06Z, a detected feature near 90N appears almost entirely zonal. Please clarify how this feature close to the North Pole adheres to the poleward transport requirement.
- At 12Z, an AR segment extending over Russia, Mongolia, and northern China shows IVT vectors pointing toward the tropics (Fig. 7c). Please justify its inclusion as an AR when the moisture transport is directed equatorward.
- At 12Z, the connection between the Northern European and Russian segments appears to rely on only one or two grid points. Are such tenuous connections physically meaningful?
Minors:
- Line 185: what do the authors mean by “the standard deviation of the two sums”?
- In Table 1, it may be clearer to present values in parentheses as percentage ratios rather than raw numbers, which would more directly illustrate the consistency between the proposed algorithm and the other two methods.
- Figs. 5g–i: There appear to be discontinuities near the dateline. Is this a plotting artifact, or is it related to how DB3 handles data crossing the dateline?
- Figs. 6 and 7 are swapped: the caption is not matched with figures.
- DB1/DB2/DB3 are used subsection 3.1, while DS1/DS2 are used in subsection 3.2. It would be better to use consistent abbreviations throughout the manuscript.
Citation: https://doi.org/10.5194/essd-2025-836-RC1 -
AC1: 'Reply on RC1', Jian Rao, 30 Jun 2026
The comment was uploaded in the form of a supplement: https://essd.copernicus.org/preprints/essd-2025-836/essd-2025-836-AC1-supplement.pdf
-
RC2: 'Comment on essd-2025-836', Anonymous Referee #2, 22 May 2026
Review of Chen et al: A database of objectively identified atmospheric rivers based on a multi-method fusion algorithm
This manuscript provides a mostly clear overview of a new Atmospheric River (AR) identification algorithm which uses two methods of identifying the AR major axis and therefore identifying the geometric extent of ARs. The figures are clear and help explain the dataset. Given there are a lot of existing AR identification algorithms, the authors could add more justification in the introduction on why their method is needed e.g. its value in distinguishing between ARs and other synoptic weather systems with high IVT.
Below are minor comments that would improve the clarity of the dataset description.
L11: change “can act as one of the precipitation sources” to “can cause precipitation”
L147: Why choose 15%?
L155: ‘the number of axis points exceeds 500’ – This upper threshold for the number of points is confusing. Presumably each point is located at a grid box? And if you are using a coarsened 1x1 degree resolution then 500 points would circle the whole world?
L187: ‘where l denotes the axis length’ units?
L212: “To overcome the respective disadvantage of the two methods, out algorithm integrates both…” It is unclear how this is done. Do both axis methods need to agree that there is an AR for the candidate to be considered an AR? If not, how are the methods weighted?
L257-8: Could cite Reid et al (2020) here which specifically discusses this problem in AR identification: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020JD032897
Table 1: The values in brackets would be easier to interpret as percentages rather than raw values.
L286-290: It would be useful to the reader to include a few sentences about how DB2 and DB3 differ from DB1 in terms of the detection method, IVT thresholds and geometric requirements.
Figure 5: There is a clear discontinuity at 180 longitude in g, h & i. You could pad the sides of the array when doing the identification/analysis to prevent the discontinuity i.e. copy 10-20 degrees longitude of data from one side of the array and attach it to the other side for the calculations and then remove the extra data for plotting.
Figures 6 and 7 are in the wrong places (Figure 6 should be the East Asia case and Figure 7 should be the Australia case).
L397: The fires lasted until late March, so I would be careful saying that this even extinguished them.
Citation: https://doi.org/10.5194/essd-2025-836-RC2 -
AC2: 'Reply on RC2', Jian Rao, 30 Jun 2026
The comment was uploaded in the form of a supplement: https://essd.copernicus.org/preprints/essd-2025-836/essd-2025-836-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Jian Rao, 30 Jun 2026
Data sets
A database of objectively identified atmospheric rivers based on a multi-method fusion algorithm Hongbin Chen and Jian Rao https://doi.org/10.5281/zenodo.18051602
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 284 | 119 | 28 | 431 | 16 | 23 |
- HTML: 284
- PDF: 119
- XML: 28
- Total: 431
- BibTeX: 16
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript develops a multi-method fusion algorithm based on vertically integrated water vapor transport (IVT) for global atmospheric rivers detection. The proposed algorithm incorporates advanced strategies from existing algorithms, adopts a dual-axis method to improve stability, and introduces curvature-related criteria to filter out vortex-like structures. However, several concerns need to be addressed before the manuscript is suitable for publication.
Majors:
In the abstract, the manuscript states that, “Atmospheric rivers (ARs) are long, narrow corridors in the atmosphere that transport immense amounts of water vapor poleward across the mid-latitudes.”
Several keywords are embedded in the definition: (1) elongated geometry (long, narrow), (2) poleward transport, and (3) occurrence in the mid-latitudes.
The elongated geometry is enforced by the first two criteria (Lines 184–185), while the curvature criteria (Lines 185–187) are related to the poleward requirement.
However, it is unclear how the “mid-latitude” aspect is enforced in the algorithm. The manuscript discussed filtering certain tropical AR candidates (Lines 222–224: grids located in tropics must not exceed 95%), but I did not see an equivalent consideration for high latitudes. Please clarify whether the algorithm intends to apply the mid-latitude criterion. If so, please justify the 95% threshold for the tropics, which seems relatively permissive, and explain how the algorithm distinguishes or filters features at high latitudes.
2. Poleward transport requirement
Using the provided dataset, I plotted ARs for the monsoon case in a domain larger than that used in the manuscript. Attached are results at 06Z and 12Z 27 Jun 2018 (blue shading for ARs). Several features appear to deviate from the definition of an AR:
Minors: