1. Chinese student work on storms done at HZG
von Storch Hans
Institute of Coastal Research
Helmholtz-Zentrum Geesthacht
Germany

2. Chen Fei 陈飞 Dissertation completed in 2013
Multi-decadal climatology of Polar Lows over the North Pacific by regional climate model
Publications:
Chen F., Geyer B., Zahn M., von Storch H., 2012: Toward a Multi-Decadal Climatology of North Pacific Polar Lows Employing Dynamical Downscaling. Terr. Atmos. Ocean. Sci. 23: 291–301.
Chen F.) and H. von Storch, 2013 : Trends and variability of North Pacific Polar Lows, Advances in Meteorology 2013, ID , 11 pages,
Chen, F., H. von Storch, Y. Du, and L. Wu, 2014: Polar Low genesis over North Pacific under different scenarios of Global Warming, Clim. Dyn /s

3. Xia Lan 夏兰 Dissertation completed in 2013:
Long-term variability of storm track characteristics
Publications
L. Xia, H. von Storch, and F. Feser, 2012: Quasi-stationarity of centennial Northern Hemisphere midlatitude winter storm tracks. Climate Dynamics, in revision
L. Xia, M. Zahn, K. I. Hodges, F. Feser and H. von Storch, 2012: A comparison of two identification and tracking methods for polar lows. Tellus A, 64,

4. Supervision: Hans von Storch and Beate Geyer
Since 2012
A multi-decadal high-resolution hindcast of wind and waves for Bohai and Yellow Sea
李德磊
Supervision: Hans von Storch and Beate Geyer

5. Motivations & Objectives
Despite numerous studiesvdevoted to climate change of areas like North Sea, Baltic Sea and Mediterranean Sea, only few attention has paid on the Chinese marginal sea Bohai and Yellow Sea;
A monsoon climate regime is dominant over Bohai and Yellow Sea, which is characterized by complex physiography and scarce observation network;
A climatology of wind and wave conditions are needed for many applications such as wind farming and risk assessment.
Objectives:
Describe (hindcast) atmospheric conditions over Bohai and Yellow Sea using the regional atmopsheric model CCLM with 7km resolution;
Describe the wave conditions of Bohai and Yellow Sea with the omogeneous wind forcing provided by the atmospheric hindcast;
Verify the reliability of wind and wave field sby comparing with observation, extract the statistical characteristics of wind and wave fields - with special attention paid on the extreme events

6. First step: A comparison of high-resolution modeled wind speed driven by different forcing datasets over Bohai and Yellow Sea
Methodology
COSMO-CLM V4.14 (CCLM)
Three simulations with 7-km resolution over Bohai and Yellow Sea (Fig. 1.) for the year 2006.
Different forcing datasets : CCLM 55km (hourly output, 0.5°, downscaled from NCEP1 1.8°); NCEP-CFSR (~ 55 km); ERAinterim (~ 80 km).
Comparison with observational data
9 land stations, including 8 airport stations; 8 offshore stations.
BIAS
𝑥 𝑚 - 𝑥 𝑜
Ration of short term Standard Deviation (RSD)
𝜎 𝑓𝑚 𝜎 𝑓𝑜
Normalized Mean Square Error (NMSE)
𝑀𝑆𝐸( 𝑥 𝑚 , 𝑥 𝑜 )/( 𝜎 𝑚 𝜎 𝑜 )
Correlation Coefficient (R)
𝐶𝑂𝑉( 𝑥 𝑚 , 𝑥 𝑜 ) /( 𝜎 𝑚 𝜎 𝑜 )
Brier Skill Score (BSS)
1−𝑀𝑆𝐸( 𝑥 𝑚 , 𝑥 𝑜 ) /𝑀𝑆𝐸( 𝑥 𝑟 , 𝑥 𝑜 )
Standard Deviation Error (STDE)
𝑀𝑆𝐸( 𝑥 𝑚 , 𝑥 𝑜 )− 𝐵𝐼𝐴𝑆 2
Fig. 1. Orography and station locations
Table 1. Statistical metrics and their definition

7. Results Evaluation of the wind speed time series
Positive biases at offshore stations;
Ratio oi short term variability RSD < 1 except for CCLM 7km at land stations;
Downscaled results add short term variability (RSDdyndown>RSDglobal);
Generally lower correlation R values for downscaled results;
Lower Brier skill score BSS values after downscaling in most cases (comparing with NCEP1)
Table 2. Statistical measures of modeled and observed wind speeds.
CCLM 7km
CCLM-CFSR
CCLM-ERAint
CCLM 55km
CFSR
ERAint
NCEP1
Land stations
BIAS
1.06
0.31
0.41
-0.13
-0.05
0.30
0.54
RSD
0.83
0.87
0.85
0.77
0.76
0.82
NMSE
1.19
0.84
0.79
0.81
1.51 R
0.58
0.61
0.67
0.64
0.68
0.50
BSS
0.01
0.44
0.49
0.42
0.45
Offshore stations
1.56
0.56
0.63
0.39
0.21
0.46
0.95
0.80
0.92
0.73
1.13
0.65
0.66
0.57
0.72
0.74
0.60
-0.45
0.29
-0.11
0.33
0.32

8. Fig. 2. Frequency distribution of wind speeds
Representation of wind speed distribution
Land stations
offshore stations
Modeled wind speeds underestimate the records of observed wind in the range of 0.0 m s m s-1,
and fit to the observation at high wind speeds.
Fig. 2. Frequency distribution of wind speeds

9. Is there added value to forcing datasets?
The bias for high wind speeds at land stations is strongly reduced by all downscaled results;
For high wind speeds of offshore stations, large bias reduction is generated for NCEP1 downscaled simulations CCLM 55km and CCLM 7km, and is reduced slightly by downscaling ERAint; but CCLM-CFSR generated larger bias;
Error variance is not reduced by all downscaled results.
Fig. 3. Comparison of wind speed bias (blue) and STDE (red) between forcing datasets and downscaled results: (a, b, c) for land stations and (d, e, f) for offshore stations

10. Which one rank the best among forcing datasets and among downscaled results?
CCLM 55km exhibits a somewhat less precision (standard deviation) than the other driving analyses at offshore stations;
CCLM-CSFR and CCLM-ERAint products are rather similar, and have much higher precision than CCLM 7km;
CCLM-ERAint is slightly better than CCLM-CFSR.
Fig. 4. Inter-comparisons of different forcing datasets and their downscaled wind speeds: (a, b) for land stations and (c, d) for offshore stations.

11. Conclusions and outlooks
The downscaling results are realistic with more details than their driving data sets.
The downscaling simulations driven by ERAint and CFSR are consistent with each other in reproduction of local wind speeds, with the one driven by ERAint slightly better. They generate local wind estimates superior to the one driven by CCLM 55km.
Due to dynamical downscaling short term variability of wind speed is added, and the biases are much reduced at land stations for high wind speeds while at offshore stations the reduction of biases are lesst so.
The error variance (bias2 + stddev 2) is not reduced by downscaling.
Outlooks
The long-term simulation driven by ERAinterim has been finished; the verification of wind will be done by comparing with satellite data;
Estimate the wave conditions of Bohai and Yellow Sea with wind forcing provided by the atmospheric hindcast;
Extract the statistical characteristics of wind and wave hindcast, especially the extreme events.
Publication almost done.
More wind data, in particular from Shandong province needed.

12. Since 2013
A multi-decadal high-resolution hindcast of storm surges for Bohai sea
冯建龙
Supervision: Jiang Wensheng and Ralf Weisse

13. Motivations & Objectives
Climate change will plays a critical role in increasing the intensity of a storm surges, however, little attention has paid on the Chinese marginal sea Bohai Sea;
Reliable and long enough time series of wind are available for establishing a long time storm surge simulating in the Bohai Sea.
Objectives:
Select atmospheric conditions over Bohai and Yellow Sea from different data sets;
Reconstruct the storm surge conditions of Bohai Sea with such homogeneous wind forcing data;
Verify the reliability of storm surge simulations by comparing with observation, extract the statistical characteristics of storm surge s, derive scenarios of possible future conditions.

14. A comparison of typhoon information of different data sets
Typhoon comparison
data sets:
Best Track Data (CMA)
NCEP1
ERA- int
Hindcasting wind data (OUC-Gao)
Six typhoons are selected from 1960
which impacted the Bohai Sea
199415
198509
198506
198211
197416
196705
The study domain and locations of Bohai Sea, Yellow Sea

15. Table1. The comparison of Vmax(m/s)
Results
Table1. The comparison of Vmax(m/s)
Typhoon
Best-tr
OUC-gao
NECP-1
ERA-int
199415 25
23.1
13.0
15.6
22.1
13.5
17.0 21
13.6 22
14.0
22.6
198509
21.8
14.5
14.2
20.3
15.3
14.6 30
27.3 15
16.5
24.2 14
16.9
198506
15.1
13.4
13.7
16.7
17.3 20
17.2
18.1
Typhoon
Best-tr
OUC-gao
NECP-1
ERA-int
198211 25
18.7
14.4
18.6
17.9
13.5
16.4
16.3
13.0 15 20
17.4
13.2
14.1
197416 30
24.1
13.6
26.7
15.6
23.5
15.4 14
196705 11
18.3
12.1
17.6
11.3
16.9
13.1

16. Comparison with the satellite data
Method
The Brier skill score S defined by
where denotes error variance between regional model data and verifying satellite observations, and is the corresponding error variance between a reference data and satellite observations. When S >0, then the winds downscaled by the regional model fit the observations better than the reference data;
Two regional model data : OUC-gao, HZG-7km (Li Delei)
Two reference data: ERA-int, HZG-55km
Satellite data: Blended wind data from NOAA

17. Results
Both regional model data sets are better than the reference data ERA-int. (white and red)
The HZG-7 km (delei) is better when considering all the wind than the OUC-gao, but when considering only the wind speed above 15m/s the OUC-gao is better than the HZG-7km.
The results for HZG-55km are similar
Skill scores of the OUC and HZG data sets: red and white – better than ERA-int

18. Storm surge case simulation
The OUC-gao and HZG-7 km data yield better results than then ERA-int data in all stations;
In the case OUC-gao performs better , while in the HZG-7 km performs better .
Two storm surge cases in Bohai Sea are selected to test the performance of the wind data from different data sets

19. Conclusions and outlooks
Both the OUC-gao and HZG-7km data sets are more reliably than the NCEP ERA-int and HZG-55km data sets;
The results from the two storm surge cases results indicate that the data are suitable for long time simulation of storms surge statistics in the Bohai Sea
Outlooks
Hindcast the long-time storm surge conditions of Bohai Sea with wind forcing provided by Gao Shanhong (OUC) and Li Deilei (HZG)
Extract the statistical characteristics of the storm surge hindcast.
Derive scenarios for the future with the help of downscaled climate change scenarios

20. Thanks for your attention!