North Atlantic polar mesoscale cyclones in ERA5 and ERA-Interim reanalyses

Transcript

1. North Atlantic polar mesoscale cyclones in ERA5 and ERA-Interim reanalyses Denis Sergeev, Ian Renfrew, Thomas Spengler, Annick Terpstra, Shun-ichi Watanabe IGP workshop | University of East Anglia, UK | 18 June 2019 @meteodenny
2. Take-home message 1. First polar mesoscale cyclone climatology using ERA5 shows improvements in reproducing key statistics and geographical patterns. 2. Cyclone occurrence in ERA5 is closer to that in satellite studies due to higher spatial resolution, which better captures wind gradients. 3. Cyclone tracking shows limited sensitivity to the time resolution, with a 6-hourly time interval being sufficient.
3. The End.
4. In this talk... • What are PMCs and why we use reanalyses for PMC climatology
5. In this talk... • What are PMCs and why we use reanalyses for PMC climatology • Case study of ACCACIA PMC
6. In this talk... • What are PMCs and why we use reanalyses for PMC climatology • Case study of ACCACIA PMC • Automatic tracking algorithm
7. In this talk... • What are PMCs and why we use reanalyses for PMC climatology • Case study of ACCACIA PMC • Automatic tracking algorithm • Verification of tracking
8. In this talk... • What are PMCs and why we use reanalyses for PMC climatology • Case study of ACCACIA PMC • Automatic tracking algorithm • Verification of tracking • Geographical patterns of cyclone occurrence
9. In this talk... • What are PMCs and why we use reanalyses for PMC climatology • Case study of ACCACIA PMC • Automatic tracking algorithm • Verification of tracking • Geographical patterns of cyclone occurrence • PMC characteristics
10. In this talk... • What are PMCs and why we use reanalyses for PMC climatology • Case study of ACCACIA PMC • Automatic tracking algorithm • Verification of tracking • Geographical patterns of cyclone occurrence • PMC characteristics • How to reproduce this study
11. What are polar mesoscale cyclones?
12. Polar mesoscale cyclones (PMCs) - maritime - tropospheric - short-lived & sub-synoptic-scale - poleward of the main polar front
13. Polar mesoscale cyclones (PMCs) - maritime - tropospheric - short-lived & sub-synoptic-scale - poleward of the main polar front a PMC event during IGP 8 Feb 2018 Greenland Sea Image courtesy G.W.K. Moore
14. Polar mesoscale cyclones (PMCs) - maritime - tropospheric - short-lived & sub-synoptic-scale - poleward of the main polar front a PMC event during IGP 8 Feb 2018 Greenland Sea Image courtesy Dundee Satellite Receiving Station Image courtesy G.W.K. Moore
15. Polar mesoscale cyclones (PMCs) - maritime - tropospheric - short-lived & sub-synoptic-scale - poleward of the main polar front Image courtesy Dundee Satellite Receiving Station a PMC event during IGP 8 Feb 2018 Greenland Sea Image courtesy G.W.K. Moore Image courtesy Dundee Satellite Receiving Station
16. Why do we need a PMC climatology?
17. Climatologies of polar mesoscale cyclones • Interaction with global climate system, e.g. the ocean circulation and sea ice • Socio-economic impact
18. Climatologies of polar mesoscale cyclones • Interaction with global climate system, e.g. the ocean circulation and sea ice • Socio-economic impact Bracegirdle & Gray (2008) Verezemskaya+ (2017) Stoll+ (2018)
19. Climatologies of polar mesoscale cyclones • Interaction with global climate system, e.g. the ocean circulation and sea ice • Socio-economic impact - Satellite-based: irregular retrievals, limited extent, subjectivity - Earlier reanalysis studies: low resolution, limited extent Bracegirdle & Gray (2008) Verezemskaya+ (2017) Stoll+ (2018)
20. The new ECMWF reanalysis to rescue!
21. ERA-Interim ERA5
22. ERA-Interim ERA5
23. ERA-Interim ERA5
24. 1979 - 31 Aug 2019 1950 - present ERA-Interim ERA5 ERA-Interim
25. 1979 - 31 Aug 2019 1950 - present ~79 km (T L 255) ~31 km (T L 639) ERA-Interim ERA5 ERA-Interim
26. 1979 - 31 Aug 2019 1950 - present ~79 km (T L 255) ~31 km (T L 639) 60 levels up to 0.1 hPa 137 levels up to 0.01 hPa ERA-Interim ERA5 ERA-Interim
27. 1979 - 31 Aug 2019 1950 - present ~79 km (T L 255) ~31 km (T L 639) 60 levels up to 0.1 hPa 137 levels up to 0.01 hPa 6 h (plus 3 h forecast fields) 1 h ERA-Interim ERA5 ERA-Interim
28. 1979 - 31 Aug 2019 1950 - present ~79 km (T L 255) ~31 km (T L 639) 60 levels up to 0.1 hPa 137 levels up to 0.01 hPa 6 h (plus 3 h forecast fields) 1 h Input obs. like in ERA-40 and from GTS + various newly reprocessed datasets and recent instruments that could not be ingested in ERA-Interim ERA-Interim ERA5 ERA-Interim
29. 1979 - 31 Aug 2019 1950 - present ~79 km (T L 255) ~31 km (T L 639) 60 levels up to 0.1 hPa 137 levels up to 0.01 hPa 6 h (plus 3 h forecast fields) 1 h Input obs. like in ERA-40 and from GTS + various newly reprocessed datasets and recent instruments that could not be ingested in ERA-Interim No uncertainty estimate 10-member Ensemble of Data Assimilations ERA-Interim ERA5 ERA-Interim
30. 1979 - 31 Aug 2019 1950 - present ~79 km (T L 255) ~31 km (T L 639) 60 levels up to 0.1 hPa 137 levels up to 0.01 hPa 6 h (plus 3 h forecast fields) 1 h Input obs. like in ERA-40 and from GTS + various newly reprocessed datasets and recent instruments that could not be ingested in ERA-Interim No uncertainty estimate 10-member Ensemble of Data Assimilations ~100 parameters ~240 parameters ERA-Interim ERA5 ERA-Interim
31. 1979 - 31 Aug 2019 1950 - present ~79 km (T L 255) ~31 km (T L 639) 60 levels up to 0.1 hPa 137 levels up to 0.01 hPa 6 h (plus 3 h forecast fields) 1 h Input obs. like in ERA-40 and from GTS + various newly reprocessed datasets and recent instruments that could not be ingested in ERA-Interim No uncertainty estimate 10-member Ensemble of Data Assimilations ~100 parameters ~240 parameters Consistent SST and sea ice ERA-Interim ERA5 ERA5 wins!
32. 1979 - 31 Aug 2019 1950 - present ~79 km (T L 255) ~31 km (T L 639) 60 levels up to 0.1 hPa 137 levels up to 0.01 hPa 6 h (plus 3 h forecast fields) 1 h Input obs. like in ERA-40 and from GTS + various newly reprocessed datasets and recent instruments that could not be ingested in ERA-Interim No uncertainty estimate 10-member Ensemble of Data Assimilations ~100 parameters ~240 parameters Consistent SST and sea ice ERA-Interim ERA5 ERA5 wins!
33. What does a PMC look like in ERA5 and ERA-Interim?
34. What does ACCACIA PMC look like in ERA5 and ERA-Interim?
35. ACCACIA polar low 26 Mar 2013 Image courtesy Dundee Satellite Receiving Station
36. ACCACIA polar low 26 Mar 2013 Vorticity Image courtesy Dundee Satellite Receiving Station
37. ACCACIA polar low 26 Mar 2013 Vorticity Wind speed Image courtesy Dundee Satellite Receiving Station
38. ACCACIA polar low 26 Mar 2013 Scatterometer Vorticity Wind speed Image courtesy Dundee Satellite Receiving Station
39. ACCACIA polar low 26 Mar 2013 Scatterometer ERA5 Vorticity Wind speed Image courtesy Dundee Satellite Receiving Station
40. ACCACIA polar low 26 Mar 2013 ERA-Interim Scatterometer ERA5 Vorticity Wind speed Image courtesy Dundee Satellite Receiving Station
41. ACCACIA polar low 26 Mar 2013 More detailed analysis: Sergeev+, QJRMS (2017) ERA-Interim Scatterometer ERA5 Vorticity Wind speed Image courtesy Dundee Satellite Receiving Station
42. ACCACIA polar low 26 Mar 2013 More detailed analysis: Sergeev+, QJRMS (2017) ERA5 better reproduces • Asymmetric structure of vorticity patterns and mesoscale troughs • “Double-winged” wind field with sharp horizontal gradients ERA-Interim Scatterometer ERA5 Vorticity Wind speed Image courtesy Dundee Satellite Receiving Station
43. Let’s use objective automatic tracking
44. Automatic PMC tracking • Uses relative vorticity maxima • Suitable for detecting and tracking small cyclones embedded in large-scale cyclones View from above Code: github.com/dennissergeev/pmctrack
45. Automatic PMC tracking • Uses relative vorticity maxima • Suitable for detecting and tracking small cyclones embedded in large-scale cyclones Vorticity vs distance View from above Code: github.com/dennissergeev/pmctrack
46. Automatic PMC tracking • Uses relative vorticity maxima • Suitable for detecting and tracking small cyclones embedded in large-scale cyclones Vorticity vs distance View from above Code: github.com/dennissergeev/pmctrack
47. Automatic PMC tracking • Uses relative vorticity maxima • Suitable for detecting and tracking small cyclones embedded in large-scale cyclones Vorticity vs distance View from above Code: github.com/dennissergeev/pmctrack
48. Automatic PMC tracking • Uses relative vorticity maxima • Suitable for detecting and tracking small cyclones embedded in large-scale cyclones Vorticity vs distance View from above Code: github.com/dennissergeev/pmctrack
49. Automatic PMC tracking • Uses relative vorticity maxima • Suitable for detecting and tracking small cyclones embedded in large-scale cyclones • Developed by Watanabe+ (2016, 2017, 2018) for mesoscale cyclones in the Sea of Japan Vorticity vs distance View from above Code: github.com/dennissergeev/pmctrack
50. Automatic PMC tracking • Uses relative vorticity maxima • Suitable for detecting and tracking small cyclones embedded in large-scale cyclones • Developed by Watanabe+ (2016, 2017, 2018) for mesoscale cyclones in the Sea of Japan Vorticity vs distance View from above Code: github.com/dennissergeev/pmctrack
51. Parameters in this study • Data: ◦ ERA-Interim (0.5°⨉0.5° grid) ◦ ERA5 (0.25°⨉0.25° grid) • Period: 18 extended winter seasons (1 Oct 2000 - 30 Apr 2018) • Extent: 65°N-85°N, 20°W-50°E • Time interval ◦ ERA-Interim: 3 h ◦ ERA5: 1 h • Other parameters - same as in Watanabe+ (2017)
52. Verification of PMCTRACK
53. Reference dataset: STARS • Intense PMCs (polar lows) in the Nordic Seas • Coverage: 2002-2011 • Compiled by forecasters on duty at the Norwegian Met Institute • Publicly available at http://polarlow.met.no/stars-dat • Has been used by for PMC tracking verification (Zappa+ 2014; Michel+ 2018; Stoll+ 2018) Exactly 100 STARS tracks are used
54. (Optional) How to match automatic tracks to the observed ones? Observed
55. (Optional) How to match automatic tracks to the observed ones? Observed Tracked
56. (Optional) How to match automatic tracks to the observed ones? Observed Tracked ?
57. (Optional) How to match automatic tracks to the observed ones? • Usually some arbitrary distance threshold • Requires interpolation in time etc. Observed Tracked ?
58. (Optional) How to match automatic tracks to the observed ones? • Usually some arbitrary distance threshold • Requires interpolation in time etc. • A better choice: a non-dimensional distance metric (Blender & Schubert, 2000) Observed Tracked ?
59. (Optional) How to match automatic tracks to the observed ones? • Usually some arbitrary distance threshold • Requires interpolation in time etc. • A better choice: a non-dimensional distance metric (Blender & Schubert, 2000) Observed Tracked ? (x 1 , y 1 , t 1 ) (x 2 , y 2 , t 2 )
60. (Optional) How to match automatic tracks to the observed ones? • Usually some arbitrary distance threshold • Requires interpolation in time etc. • A better choice: a non-dimensional distance metric (Blender & Schubert, 2000) Observed Tracked ? (x 1 , y 1 , t 1 ) (x 2 , y 2 , t 2 )
61. (Optional) How to match automatic tracks to the observed ones? • Usually some arbitrary distance threshold • Requires interpolation in time etc. • A better choice: a non-dimensional distance metric (Blender & Schubert, 2000) Observed Tracked ? (x 1 , y 1 , t 1 ) (x 2 , y 2 , t 2 )
62. Sensitivity to: Vorticity threshold
63. Sensitivity to: Vorticity threshold
64. Sensitivity to: Vorticity threshold Time interval
65. Sensitivity to: Vorticity threshold Time interval
66. Sensitivity to: Vorticity threshold Time interval
67. Sensitivity to: Vorticity threshold Time interval
68. Sensitivity to: Vorticity threshold Time interval
69. Sensitivity to: Vorticity threshold Time interval
70. • 88 are detected in ERA5, compared to 64 in ERA-Interim • Rate similar to those using Arctic System Reanalysis (Smirnova & Golubkin, 2017; Stoll+ 2018) • Detection rate sensitive to the vorticity threshold • Longer time intervals are sufficient Sensitivity to: Vorticity threshold Time interval
71. PMC climatology
72. Cyclone density spatial distribution
73. Cyclone density spatial distribution
74. Cyclone density spatial distribution ERA5 ERA-Interim Track density
75. Cyclone density spatial distribution ERA5 ERA-Interim Track density
76. Cyclone density spatial distribution ERA5 ERA-Interim Track density Cyclone formation
77. Cyclone density spatial distribution Cyclone decay ERA5 ERA-Interim Track density Cyclone formation
78. Cyclone density spatial distribution ERA5 ERA-Interim
79. Cyclone density spatial distribution ERA5 ERA-Interim
80. Cyclone density spatial distribution ERA5 ERA-Interim • ERA5 maximum: 20 PMCs per 104 km2 per winter • Western part of the Barents Sea • ERA-Interim maximum is close to Svalbard - likely due to orographic vorticity filaments • ERA-Interim: track density is too low Genesis density derived from satellite passive microwave data over 14 winters (Smirnova+ 2015)
81. PMC characteristics t 0 T = t N -t 0 t N
82. PMC characteristics ∑
83. PMC characteristics (∑d) / T t 0 t N
84. PMC characteristics MAX( 0 , …, N ) 0 N 1 N-1
85. PMC characteristics MEAN(A 0 , …, A N )
86. PMC characteristics MIN(SLP 0 , …, SLP N ) SLP 0 SLP N SLP 1
87. A typical wintertime PMC ERA5 ERA-Interim Frequency [PMC per week] 6 1 Lifetime [h] 21 21 Total distance [km] 900 700 Propagation speed [km h-1] 42 34 Approx. size [km] 150 250 Max vorticity [10-4 s-1] 4.5 2.5 Min SLP [hPa] 993 988 Take-home message: our results show North Atlantic PMCs as smaller but more intense and faster-moving vortices compared to most previous climatologies.
88. Conclusions • First PMC climatology based on ERA5 & using new tracking algorithm Submitted to GRL!
89. Conclusions • First PMC climatology based on ERA5 & using new tracking algorithm • Overall, ERA5 provides a refined picture of PMC activity over the NE Atlantic compared to ERA-Interim Submitted to GRL!
90. Conclusions • First PMC climatology based on ERA5 & using new tracking algorithm • Overall, ERA5 provides a refined picture of PMC activity over the NE Atlantic compared to ERA-Interim • Time resolution - less crucial for PMC tracking once it is above a certain threshold, so can save computational resources Submitted to GRL!
91. Conclusions • First PMC climatology based on ERA5 & using new tracking algorithm • Overall, ERA5 provides a refined picture of PMC activity over the NE Atlantic compared to ERA-Interim • Time resolution - less crucial for PMC tracking once it is above a certain threshold, so can save computational resources • PMCs tend to form and develop over the Barents Sea, close to areas of high occurrence of CAOs (more info - in Annick’s talk!) Submitted to GRL!
92. Conclusions • First PMC climatology based on ERA5 & using new tracking algorithm • Overall, ERA5 provides a refined picture of PMC activity over the NE Atlantic compared to ERA-Interim • Time resolution - less crucial for PMC tracking once it is above a certain threshold, so can save computational resources • PMCs tend to form and develop over the Barents Sea, close to areas of high occurrence of CAOs (more info - in Annick’s talk!) Submitted to GRL! Thank you!
93. Conclusions • First PMC climatology based on ERA5 & using new tracking algorithm • Overall, ERA5 provides a refined picture of PMC activity over the NE Atlantic compared to ERA-Interim • Time resolution - less crucial for PMC tracking once it is above a certain threshold, so can save computational resources • PMCs tend to form and develop over the Barents Sea, close to areas of high occurrence of CAOs (more info - in Annick’s talk!) Submitted to GRL! P.S. Key aspects of PMC climatology are very sensitive to what tracking method is used and how cyclones are selected - but it’s almost impossible to know what code was used for the analysis Thank you!
94. Cyclone climatology suffer from reproducibility crisis too!
95. How this study can be reproduced 0. Data • Freely available on Copernicus • Downloading can be automated with Python libraries (cdsapi and ecmwfapi)
96. How this study can be reproduced 0. Data • Freely available on Copernicus • Downloading can be automated with Python libraries (cdsapi and ecmwfapi) 1. Cyclone tracking • PMCTRACK algorithm (Fortran library with netCDF interface) • http://github.com/dennissergeev/pmctrack
97. How this study can be reproduced 0. Data • Freely available on Copernicus • Downloading can be automated with Python libraries (cdsapi and ecmwfapi) 1. Cyclone tracking • PMCTRACK algorithm (Fortran library with netCDF interface) • http://github.com/dennissergeev/pmctrack 2. Postprocessing and analysis • Octant (Python package) • http://github.com/dennissergeev/octant
98. How this study can be reproduced 0. Data • Freely available on Copernicus • Downloading can be automated with Python libraries (cdsapi and ecmwfapi) 1. Cyclone tracking • PMCTRACK algorithm (Fortran library with netCDF interface) • http://github.com/dennissergeev/pmctrack 2. Postprocessing and analysis • Octant (Python package) • http://github.com/dennissergeev/octant 3. Plotting • Jupyter Notebooks and Python scripts • http://github.com/dennissergeev/mc_era5