okhotsk sea and polar oceans researchŸ»読全体.pdf · sea and polar oceans research association...

Okhotsk Sea and Polar Oceans Research

Volume 1 (2017)

Okhotsk Sea and Polar Oceans Research Association

Mombetsu City, Hokkaido, Japan

General Information for OSPOR 1. Aims and Scope

Okhotsk Sea and Polar Oceans Research (OSPOR) is published by the Okhotsk Sea and Polar Oceans Research Association (OSPORA). Since 1986 the Okhotsk Sea and Cold Ocean Research Association (OSCORA) held the International Symposium at Mombetsu, Hokkaido, in Japan every February and has made its proceedings over 30 years. OSCORA changed to OSCPRA, because subjects of the Symposium have enlarged to the polar oceans (the Arctic and Antarctic Oceans), the Arctic passages, global warming and environment change in Polar Regions.

OSCORA starts a new system of reviewed papers from the 2017 Symposium. The papers are refereed by multiple reviewers, published in the proceedings of the Symposium with a title of “Article”, and opened on the OSPOR web site.

2. Subjects covered by OSPOR

1) Environment of Okhotsk Sea 2) Meteorology and oceanography in polar regions 3) Cold region engineering 4) Arctic sea routes 5) Global warming and environment change 6) Remote sensing 7) Snow, ice and human life 8) Other topics about Okhotsk Sea and Polar Oceans

3. Editorial Policy of OSPOR We intend to publish article papers, which should contain original scientific results and challengeable subjects about Okhotsk Sea and Polar Oceans, not submitted for publication elsewhere.

4. Editorial Board

Period: August 2016 - August 2018

Editor-in-Chief： Hiromitsu Kitagawa (Ocean Policy Research Foundation)

Editors：Hiroyuki Enomoto (National Research Institute of Polar Researches, Japan) Yutaka Michida (University of Tokyo, Japan) Humio Mitsudera (Hokkaido University, Japan) Koji Shimada (Tokyo University of Marine Science and Technology, Japan) Shuhei Takahashi (Okhotsk Sea Ice Museum of Hokkaido, Japan) Hajime Yamaguchi (University of Tokyo, Japan)

5. OSPOR website

Temporal website: http://www.o-tower.co.jp/okhsympo/top-index.html E-mail: [email protected]

Cover photo: Sea ice at Mombetsu, Japan (photo by Shuhei Takahashi, 2003) Back cover photo: Crab Scissors Monument at Mombetsu, Japan

Okhotsk Sea and Polar Oceans Research, Volume 1 (2017) Published by the Okhotsk Sea and Polar Oceans Research Association (OSPORA)


Volume 1 (2017)

Contents

The impact of data assimilation and atmospheric forcing data on predicting short-term sea ice distribution along the Northern sea route

Liyanarachchi Waruna Arampath DE SILVA and Hajime YAMAGUCHI

Improving numerical sea ice predictions in the Arctic Ocean by data assimilation using satellite observations

Dulini Yasara MUDUNKOTUWA, Liyanarachchi Waruna Arampath DE SILVA, Hajime YAMAGUCHI

Preparing an UAV for Drift Ice Observation

Kanichiro MATSUMURA

Inter-annual changes in responses of winter sea-ice motions to winds in the Arctic Ocean between 2003 and 2012

Eri YOSHIZAWA and Koji SHIMADA

‥‥ 1

‥‥ 7

‥‥ 12

‥‥ 16

Okhotsk Sea and Polar Oceans Research 1 (2017) 1-6 © Okhotsk Sea and Polar Oceans Research Association

1

The impact of data assimilation and atmospheric forcing data on predicting short-term sea ice distribution along the Northern sea route

Liyanarachchi Waruna Arampath DE SILVA1 and Hajime YAMAGUCHI1

1The University of Tokyo, Tokyo, Japan

(Received September 16, 2016; Revised manuscript accepted October 19, 2016)

Abstract With the recent rapid decrease in summer sea ice in the Arctic Ocean extending the navigation

period in the Northern sea routes (NSR), the precise prediction of ice distribution is crucial for safe and efficient navigation in the Arctic Ocean. Precise ice distribution prediction in the short-term (5–days scale) is one of the key issues to realize safe and efficient navigation in the NSR. Ensemble predictions of short-term sea-ice conditions along the Northern sea route have been carried out using a high–resolution (about 2.5km) ice–ocean coupled model that explicitly treats ice floe collisions in marginal ice zones. In this study, the ensembles are constructed by using forecasted atmospheric forcing data sets from THORPEX Interactive Grand Global Ensemble (TIGGE) project in 2015. We also discussed the influence of data assimilation on high-resolution model ice and ocean initial conditions estimated by the whole Arctic medium-resolution (about 25 km) model. The correlation score of ice–edge error and sea ice concentration distribution quantifies forecast skill. Skill scores are computed from 5–days ensemble forecasts initialized in each month between May 2015 to October 2015. Comparison of different ensemble atmospheric forecasts, using different months initial data sets, revealed that our ice–POM numerical model skillfully predicts the ice distribution during the NSR operational period. The average forecast skill of ice–POM model in the melting season is 9.28±2.68 km and in the freezing season without assimilated initial conditions is 15.43±6.29 km and with assimilation 13.85±5.77 km with the 15% thresholds of ice concentration for the ice edge. With data assimilation, there is 10% improvement of average ice edge error within 5-days simulation.

Key words: Arctic sea ice, Data assimilation, Northern sea route

INTRODUCTION

Past decades of satellite observations have shown a rapid decrease of summer Arctic sea ice extent. Furthermore, the Arctic sea ice cover is now thinner, weaker, and drifts faster. Those conditions increase the interest on the commercial use of Arctic shipping. However, the sea ice distribution varies with hourly time scales due to the atmospheric and oceanic conditions. Therefore, sea ice predictions and observations are important to protect the ships and offshore and coastal structures in order to utilize the Northern sea route (NSR). Global climate models have been employed to assess the predictability of Arctic sea ice. However, most of the available numerical models have shown high uncertainties in the short-term (about 5days) and narrow-area predictions, especially marginal ice zones such as the NSR (Hebert et al., 2015). Successful sea ice predictions are relying on comprehensive initial conditions of sea ice variables and ocean variables and an ability of forecast systems to capture high-frequency atmospheric variability and associated feedbacks.

There are 2 kinds of ensemble forecasting systems available for sea ice predictions. First, perturbing the sea ice or ocean initial conditions generates the ensemble members. Second, changing the boundary conditions, such as atmospheric forcing, generates the ensemble members. In this study, the ensembles are constructed by using forecasted atmospheric forcing datasets from THORPEX Interactive Grand Global Ensemble (TIGGE; Bougeault et al., 2010) project in 2015 and the ice and ocean conditions estimated by the hindcast model simulation.

The purpose of this study is to predict the short-term (5days) sea ice conditions in the NSR using mesoscale eddy resolving ice-ocean coupled model within the ice edge error of ±10km, which can meet ship crew requirement. The correlation score of ice edge error and sea ice concentration distribution quantifies forecast skill. Skill scores are computed from 5-days ensemble forecasts initialized in each month between May 2015 to October 2015. Also, we investigate the impact of sea ice predictions by different atmospheric forecasted datasets.

Article


2

Some researchers suggest that model alone predictions are prone to several errors such as uncertainties in initial conditions, uncertainties in the forcing data and limitations of the temporal and spatial resolutions (Lindsay et al., 2006; Mudunkotuwa et al., 2016). Therefore, in this study, we also investigated the impact of ice and ocean initial conditions into the short-term sea ice predictions by introducing data assimilation into the whole Arctic model.

MODEL DESCRIPTION

Ice–ocean coupled model used in this study is based on the model developed by De Silva et al. (2015). The ocean model is based on generalized coordinates, the Message Passing Interface version of the Princeton Ocean Model (POM; Mellor et al. (2002)). The level-2.5 turbulence closure scheme of Mellor and Yamada (1982) is used for the vertical eddy viscosity and diffusivity. The horizontal eddy viscosity and diffusivity are calculated using a formula proportional to the horizontal grid size and velocity gradients (Smagorinsky,1963); the proportionality coefficient chosen is 0.2. The ice thermodynamics model is based on the zero-layer thermodynamic model proposed by Semtner (1976). The ice rheological model is based on the elastic–viscous–plastic (EVP) rheology proposed by Hunke and Dukowicz (2002) and is modified to take ice floe collisions into account, following Sagawa and Yamaguchi (2006). Model domain is constructed using Earth Topography one-minute Gridded Elevation Dataset (ETOPO1) as shown in Fig. 1.

Fig. 1 Model bathymetries (m). (a) Whole-Arctic model.

To avoid the singularity at the North Pole, the

whole-Arctic model grid is rotated to place its North Pole over the equator. Red rectangle denotes the high-resolution domain in the Northern Sea Route. (b) High-resolution regional model domain of what in this study we call the Laptev Sea region, consisting of the Laptev Sea and part of the Kara and East Siberian seas,

with 50°E - 165°E longitudes and 68°N - 85.5°N latitudes. (De Silva et al. 2015)

The zonal and meridional grid spacing are approximately 25 × 25 km for the whole Arctic model and 2.5 × 2.5 km for the high-resolution regional model. To resolve the surface and bottom ocean dynamics, we use the logarithmic distribution of the vertical sigma layers near the top and bottom surfaces.

Mudunkotuwa et al. (2016) introduced the data assimilation into the ice-POM model. She claimed that assimilating sea ice variables improved the ocean and ice conditions. It is evident from the changes in sea ice extent, sea ice thickness, and ocean salinity. She also claimed that non-assimilated sea ice variables have also been indirectly improved by assimilation. In this study, we used the Newtonian relaxation (nudging) technique to assimilate the satellite observational (SSMI, AMSR-E, and AMSR2) sea ice concentration in 24hr intervals from year 2000 to 2016. During assimilation experiment, the model estimates concentrations are nudged to new estimate sea ice concentration with the following relationship (Eq.1)

Anew = Amodel+ K Aobs − Amodel( ) (1)

where, Anew newly estimated sea ice concentration, Amodel model derived sea ice concentration, Aobs satellite observational sea ice concentration and K is a weighting constant and this study it set to be 0.8. Some corrections are done to adjust the sea ice thickness, velocity, ocean temperature and salinity while assimilating sea ice concentration. When the assimilation creates ice, the ice thickness and velocity are set to be the average of the four neighboring cells while the maximum thickness of created ice is set to be 0.5m and the minimum is set to 0.1m to avoid immediate melting. When the assimilation removes ice, the values of other sea ice variables (sea-ice thickness and sea-ice velocity) are set to zero. Ocean temperature is also set to freezing temperature if the temperature is below freezing temperature.

High-resolution computations are initialized using interpolated whole-Arctic model results with and without data assimilation (e.g., sea-ice thickness, ocean temperature and salinity). Note that whole-Arctic sea-ice concentration is not used for high-resolution computations; rather, satellite observations Advanced Microwave Radiometer2 (AMSR2) are used. When the initial AMSR2 sea ice concentration is not zero in the open water areas of the whole-Arctic simulation, we interpolated the sea-ice thickness from neighboring grid cells. In this case, water surface temperature under those cells was set to the freezing temperature to avoid the rapid melting of sea ice. Moreover, when the initial AMSR2 observed concentration is zero and the whole-Arctic simulation concentration is not zero, we set the sea-ice thickness to zero in those cells. For those

De Silva and Yamaguchi

3

cells, the temperature under the ocean surface was assigned by interpolating the values only from the open water neighboring grid cells. Note that, in both situations, ocean salinity is unchanged and the same as the interpolated whole-Arctic model output value. In the marginal regions (open boundaries) of the high-resolution model, the whole-Arctic model results interpolated into the high-resolution model grids are applied for both the sea-ice and ocean open-boundary conditions with daily intervals.

The atmospheric dataset used in this ensemble forecast study comes from TIGGE (Bougeault et al., 2010) operational medium-range ensemble forecast project. The operational ensemble prediction systems used in this study include the China Metrological Administration (CMA), the Canadian Metrological center (CMC), the European Center for Medium-range Weather Forecasts (ECMWF), the Japan Metrological Agency (JMA), the France Metrological Office (FMO), the United Kingdom Meteorological Office (UKMO), and the US National Centers for Environmental Prediction (NCEP). The more details about the atmospheric forecasting and reanalysis are described in Bougeault et al. (2010). We used the 6-hourly atmospheric data outputs with the spatial resolution of 0.5-degree; air-temperature and dev-point temperature at 2m height, wind at 10m height, precipitation, sea-level pressure, and cloud cover. Using these data, surface fluxes, shortwave radiation, longwave radiation, sensible heat flux and latent heat flux are calculated according to the bulk formulation proposed by Parkinson and Washington (1979). DISCUSSION

To evaluate the sea ice predictions using different atmospheric datasets, we used the correlation score of ice edge error and sea ice concentration distribution. The ice edge error is defined as follows (Eq. 2). First, the difference in the ice areas between the model’s predictions and AMSR2 satellite observations are calculated, shown in Fig. 2. Note that area covered with ice concentration more than 15%, 30%, and 50% are taken into comparison. Because the definition of the opening of sea route highly depends on the ice level of an icebreaker. Next, dividing the length of the model predicted contour of the ice concentration of 15%, 30%, and 50%, we obtain the ice edge error with the dimension of length.

ice edge error= Area enclosed by both contours

Length of model predicted ice edge (2)

The results of the 5days in melting season (20 to 25 July 2015) forecasted ice edge errors and the hindcast ice edge error using ERA-Interim data are shown in Fig. 3. Ocean and ice initial conditions are initialized using the whole Arctic model alone run. There were no

significant differences of ice edge errors between different forecasted datasets and ERA-interim hindcast data. Within 5 days, average ice edge errors among seven-forecasted datasets are 9.28±2.68 km, 10.11±3.08 km, 10.04±3.51 km with respect to the threshold value of ice concentration 15%, 30%, and 50% respectively. There is a slight increment in the average ice edge error when the threshold values of ice concentration increase from 15% to 50%.

Fig. 2 Schematic diagram of model predicted (red) and AMSR2 (blue) observational ice edges and area enclosed by both contours (yellow)

Fig. 3 Ice edge error between different forecasted datasets

(and ERA-interim) and AMSR2 observations from 21- July-2015 to 25-July-2015 Top threshold value of ice concentration (a) 15% (b) 30% and (c) 50%

During the computation period, the JMA showed

best ice edge error prediction of 8.89±2.57 km. Overall, in melting season ice-POM model reproduced the ice edge error within the limit of 10km.

In addition to the quantitative comparisons of ice edge error, we compared the sea-ice concentration distribution qualitatively. Fig. 4 shows the difference between the model and AMSR2 sea ice concentrations after the 5th day of computation (25 July 2015). In the western part of the domain (Kara sea), the difference in sea ice concentration is higher compared to the other

Area enclosed by both contours

AMSR2 ice edge Model predicted ice edge

Ocean

Ice


4

regions. This discrepancy could be due to the underestimation of heat transfer process between ice and ocean. However, sea ice spatial distribution between different datasets has no significant differences in the melting season.

Fig. 4 Sea ice concentration distribution, Difference between model-predicted ice concentration and AMSR2 observation on 25-July-2015. Threshold value of ice concentration is 15%

Next, the results of the 5days in the period of Arctic

annual minimum sea ice extent (10 to 15 September 2015) forecasted ice edge error and the hindcast ice edge error using ERA-Interim data is shown in Fig. 5. Except for CMA dataset there were no significant differences of ice edge error between different forecasted datasets. The average ice edge errors with respect to the different ice concentration threshold values are 15% threshold value 10.15±3.45 km, 30% threshold value 10.56±2.8 km and 50% threshold value 12.87±2.15 km. There is a slight increment in the average ice edge error when the threshold values of ice concentration increase from 15% to 50%.

Finally, the results of the 5days in freezing season (10 to 15 October 2015) forecasted ice edge error and the hindcast ice edge error using ERA-Interim data is shown in Fig. 6. There were significant differences of ice edge error between different forecasted datasets and ERA-interim hindcast data after the 2nd day (12- Oct. 2015) of computation. The average ice edge errors with respect to the different ice concentration threshold values are 15% threshold value 15.43±6.29 km, 30%

threshold value 16.56±7.08 km and 50% threshold value 19.76±8.98 km. There is a significant increment in the average ice edge error when the threshold values of ice concentration increase from 15% to 50%.


(and ERA-interim) and AMSR2 observations from 11-Sep-2015 to 15-Sep-2015 (a) threshold value of ice concentration 15% (b) 30% and (c) 50%


(and ERA-interim) and AMSR2 observations from 11-Oct-2015 to 15-Oct-2015 (a) threshold value of ice concentration 15% (b) 30% and (c) 50%

In most cases, the average ice edge error increases

with the threshold values of ice concentration increase

De Silva and Yamaguchi

5

from 15% to 50%. We speculate the reason for this issue as follows. In practice, the regions where ice concentration change from 15% to 50% are usually very narrow and subjected to the observational errors due to melt ponds and surface conditions of snow cover in melting and early freezing seasons. These issues might favorably affect the ice edged error calculations in less ice concentration areas.

It is also seen that time evolution of ice edge error is significantly large in the freezing season compared to the melting season and annual minimum ice extent. We believe this discrepancy could be due to the uncertainties in the model initial conditions.

The lower ice edge error values in the summer suggest ice-POM does a good job at capturing ice melt, while higher values during the freezing season suggest that ice-POM not produce ice as fast as actually occurs along the NSR.

Several possibilities exist to explain these discrepancies. The first reason could be because we used the bulk formulation proposed by Parkinson and Washington (1979) to produce the heat fluxes from the atmosphere to sea ice and those parameters may not have tuned into the latest Arctic conditions properly. The second reason could be coarseness of spatial and temporal resolution of forecasted datasets could not properly resolve the small-scale features of the atmosphere (Ono et al. 2016), which influenced the sea ice production and retrieve. These will be the topic of future study.

Fig. 7 Same as Fig. 6 but model initialized from whole

Arctic model with assimilation results

The third reason could be because we used ocean temperature and salinity data from the interpolated whole-Arctic model for our high-resolution

computations as an initial condition. Over prediction of ocean surface heat in the whole-Arctic model delayed the freezing of sea ice in the high-resolution models freezing season.

We have tested the third hypothesis in this study. We have run the whole Arctic model with data assimilation from year 2000 to 2016 and used the assimilated whole Arctic model data for high-resolution initial conditions.

The results of the 5days in freezing season (10 to 15 October 2015), initialized with data assimilated results, forecasted ice edge error and the hindcast ice edge error using ERA-Interim data is shown in Fig. 7. The average ice edge errors with respect to the different ice concentration threshold values are 15% threshold value 13.85±5.77 km, 30% threshold value 14.84±6.78 km and 50% threshold value 17.12±8.61 km. Compared with the model alone initialized computation there is 10% ice edge error improvement can be seen.

Fig. 8 shows the ocean temperature difference between data assimilated results and model alone results in the Laptev Sea on 10th October 2015. It’s very clear that along the ice edge (Fig. 8 black contours) less heat in the data assimilated model compared to the model alone.

Fig. 8 Ocean temperature difference (degree) between data

assimilated model run and model alone run in the Laptev Sea region on 10-October-2015 (freezing season initial date). Black contour shows the AMSR2 sea ice edge (concentration threshold 15%)


6

CONCLUSION Sea ice forecasted skill of different dataset (TIGGE)

is evaluated in the study. The average forecast skill of ice-POM model in the melting season is 9.28±2.68 km that is in good agreement with the requirement of an operational ice navigation system (10 km). Also, there is good forecast skill (10.15±3.45 km) in the ice-POM model when the Arctic sea ice extent hits its annual minimum. However, in the freezing season ice edge error become 15.43±6.29 km. But after introducing the data assimilation into the whole Arctic model freezing season ice edge error improved 10% from without assimilated results. However, to improve the model forecast accuracy, the further studies would be necessary for the freezing season. ACKNOWLEDGEMENTS

The authors wish to acknowledge support from the Green Network of Excellence Program Arctic Climate Change Research Project (GRENE) and Arctic Challenge for Sustainability Research Project (ArCS) by the Japanese Ministry of Education, Culture, Sports, Science and Technology and a Kakenhi grant (no. 26249133) from the Japan Society for the Promotion of Science. We also thank Ms. Dulini Yasara Mudunkotuwa for her useful comments. Their gratitude is extended to The National Snow and Ice Data Center for the gridded AMSR-2 data, Arctic Data Archive System (ADS) for providing sea ice thickness gridded data. REFERENCES Bougeault, P. and 21 others (2010): The THORPEX Interactive

Grand Global Ensemble. Bull. Am. Meteorol. Soc., 91, 1059–1072.

De Silva, L. W. A., H. Yamaguchi, and J. Ono (2015): Ice–ocean coupled computations for sea-ice prediction to support ice navigation in Arctic sea routes. Polar Res., 34, 18pp.

Hebert, D. A., R. A. Allard, E. J. Metzger, P. G. Posey, R. H. Preller, A. J. Wallcraft, M. W. Phelps, and O. M. Smedstad (2015): Short-term sea ice forecasting: An assessment of ice

concentration and ice drift forecasts using the U.S. Navy’s Arctic Cap Nowcast/Forecast System. J. Geophys. Res. Ocean., 120, 8327–8345.

Hunke, E. C., and J. K. Dukowicz (2002): The elastic-viscous-plastic sea ice dynamics model in general orthogonal curvilinear coordinates on a sphere-incorporation of metric terms. Mon. Weather Rev., 1848–1865.

Lindsay, R. W., J. Zhang, R. W. Lindsay, and J. Zhang (2006): Assimilation of Ice Concentration in an Ice–Ocean Model. J. Atmos. Ocean. Technol., 23, 742–749.

Mellor, G. L., and T. Yamada (1982): Development of a turbulence closure model for geophysical fluid problems. Rev. Geophys., 20, 851–875.

Mellor, G. L., S. Hakkinen, T. Ezer, and R. Patchen (2002): A generalization of a sigma coordinate ocean model and an intercomparison of model vertical grids. In Ocean Forecasting: Conceptual Basis and Applications., Springer, Berlin Heidelberg, 55–72.

Mudunkotuwa, D. Y., L. W. A. De Silva, and H. Yamaguchi (2016): Data assimilation system to improve sea ice predictions in the Arctic Ocean using an ice-ocean coupled model. 23rd IAHR International Symposium on Ice, Ann Arbor, Michigan USA, 1-8.

Ono, J., J. Inoue, A. Yamazaki, K. Dethloff, and H. Yamaguchi (2016): The impact of radiosonde data on forecasting sea‐ice distribution along the Northern Sea Route during an extremely developed cyclone. J. Adv. Model. Earth Syst., 8, 292–303.

Parkinson, C. L., and W. M. Washington (1979): A large-scale numerical model of sea ice. J. Geophys. Res., 84, 311–337

Sagawa, G., and H. Yamaguchi (2006): A Semi-Lagrangian Sea Ice Model For High Resolution Simulation. The Sixteenth International Offshore and Polar Engineering Conference, San Francisco, California, International Society of Offshore and Polar Engineers, 584–590.

Semtner, A. J. (1976): A Model for the Thermodynamic Growth of Sea Ice in Numerical Investigations of Climate. J. Phys. Oceanogr., 6, 379–389.

Smagorinsky, J. (1963): General circulation experiments with the primitive equations. Mon. Weather Rev., 91, 99–164.

Copyright ©2017 The Okhotsk Sea & Polar Oceans Research Association. All rights reserved.


7

Improving numerical sea ice predictions in the Arctic Ocean

by data assimilation using satellite observations

Dulini Yasara MUDUNKOTUWA1, 2, Liyanarachchi Waruna Arampath DE SILVA1, Hajime YAMAGUCHI1

1Graduate School of Frontier Sciences, University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa, Japan.

2Faculty of Engineering, University of Sri Jayawardenepura, 41, Lumbini Ave, Ratmalana, Sri Lanka.

(Received on September 16, 2016; Revised manuscript accepted on October 21, 2016)

Abstract

This study focuses on improving sea ice predictions in the Arctic Ocean by introducing data assimilation into an ice-ocean coupled Ice-POM model that is used to predict sea ice conditions in the Arctic sea routes. Ocean part of the model used in this study is based on the Princeton Ocean Model (POM). The ice model considers discrete characteristics of ice along the ice edge. The model domain consists of the Arctic Ocean, Greenland-Iceland-Norwegian (GIN) seas and the Northern Atlantic Ocean. The model grid is with 25km horizontal resolution. An improved nudging method that takes the observation errors into account is used in this study. Observation errors are varied in accordance with the season and the location. Sea ice concentration, sea ice thickness and sea ice velocity are assimilated simultaneously. Assimilation improved ocean and ice conditions significantly. This is evident from the changes in sea ice extent, sea ice thickness and ocean salinity.

Key words: Arctic sea ice, data assimilation, sea ice concentration, sea ice thickness, sea ice velocity,

nudging 1. INTRODUCTION With increased use of Arctic Sea Routes, accurately

predicting sea ice conditions in the Arctic Ocean is vital. Numerical simulations are considered to produce accurate predictions economically. However, the results of the simulations from the model alone are prone to several errors such as uncertainties in initial conditions, boundary condition and forcing data. Temporal and spatial resolutions are also limiting factors. Therefore, it is necessary to introduce satellite observations based correction method to model predictions. There are various data assimilation methods that are widely used in ocean modeling. Direct insertion, nudging, optimal interpolation, 3DVAR, 4DVAR and ensemble Kalman filter (EnKF) are commonly used methods. Methods such as 3DVAR, 4DVAR, EnKF methods comes with a high computational cost. Therefore, in this study an improved nudging method based on Lindsay (2006) is used. Sea ice concentration, sea ice thickness and sea ice velocity are assimilated simultaneously

2. MODEL DESCRIPTION The ice-ocean coupled model, ice-POM that is used in this study is the model used by Fujisaki et al. (2010) and De Silva et al. (2015) for ice-ocean coupled computations. The ocean model of the ice-POM is based on generalized coordinates, the Message Passing Interface (MPI) version of the Princeton Ocean Model

(POM) (Mellor et al. 2003). Zero-layer thermodynamic model by Parkinson et al. (1979) is used as the ice thermodynamic model.

Model domain contains the entire Arctic Ocean, Greenland-Iceland-Norwegian (GIN) seas and the Northern Atlantic Ocean as shown in fig. 1 The resolution of zonal and meridional grid are set to about 25km x 25km. The vertical grid is composed of z-sigma coordinates system with 33 levels. A z-coordinate system is used for top three layers with 1m depth in each layer. Remaining 30 layers are composed of sigma coordinate system. The bottom topography is created from Earth Topography 1 Arc-minute Gridded Elevation (ETOPO1) dataset. Radiation boundary condition is applied at the open lateral boundaries and no-slip boundary condition is used along the coastlines. Singularity at the North Pole is avoided by rotating the whole Arctic model grid to place its North Pole over the equator. The atmospheric forcing data are obtained from The European Centre for Medium-Range Weather Forecasts Re-Analysis Interim (ERA-interim) six hourly data. Precipitation is obtained by National Center for Environmental Prediction (NCEP) 6 hourly reanalysis data. First, the model is spun up for 10 years by providing the year 1979 atmospheric data cyclically. Then the model is integrated from year 1979 to 2013 with ERA-interim and NCEP realistic atmospheric forcing (Mudunkotuwa et al. 2015). The ice model and ocean model time steps

Article


8

are set to 4 minutes. The ocean and ice conditions resulting from this 33-year integration are used to initialize data assimilation experiments.

3. DATA USED

The ice concentration is obtained from the advanced microwave scanning radiometer (AMSR2) onboard the GCOM-W satellite. Daily gridded sea ice concentration data set is extracted from Arctic Data Archive System (ADS) from their website https://ads.nipr.ac.jp/. Daily sea ice thickness is calculated using (Krishfield et. al, 2014) algorithm based on AMSR-2 satellite data. The gridded daily sea ice thickness data set is also obtained from ADS. Sea ice velocity data set is extracted from KIMURA Sea ice velocity data set (Kimura et. al, 2013). Sea ice concentration data are available in a daily interval for the year 2013. Sea ice thickness and sea ice velocity data sets are only assimilated during first four months since the two data sets aren’t reliable in summer. To validate the study two independent data sets are used that are not used in assimilation experiments. Monthly averaged Cryosat sea ice thickness data and NSIDC Aquarius weekly sea surface salinity data sets are used for validation.

4. DATA ASSIMILATION METHOD

In this study the timespan of data assimilation experiment is set to year 2013. The Incremental analysis update (Bloom, S.C., et al. 1996) is used to assimilate sea ice concentration, sea ice velocity and sea ice thickness. In this experiment model is nudged

towards the observation with the following relationship. (1)

Where, C is the prognostic variable and is the time relaxation coefficient. is set to be 12 hours in this study. K is the nudging weight. Optimal least square value of the weighting is formulated as in the following equation according to Lindsay (2006).

(2)

Where and are the error variance of model and the observations respectively. Lindsay (2006) method uses to be the model error. One of the issues with this method is that it assumes observation to be the truth by assuming the model error to be

. Therefore as an improvement to the formulation the model error is assumed to be

. With this method, truth is assumed to be,

(3)

This method yields K as below,

(4)

(5)

As a result, two sets of values are used for K yielding a range as a forecast.

is calculated as,

(6) Unlike Mudunkotuwa et al. (2016) algorithm errors are also considered in this study. The values used for instrument errors and algorithm errors are presented in table 1. The values are obtained from Ludovic et al. (2014), Kimura et al. (2016), Ono et al. (2016), and JAXA (2014). An experiment is performed assimilating sea ice concentration, sea ice thickness and sea ice velocity simultaneously. Sea ice concentration is assimilated for the entire year 2013. Sea ice thickness and sea ice velocity are assimilated only during first four months. To avoid disparities between observation data sets and to improve the ocean conditions, some corrections are done to sea ice variables and ocean variables as described in detail in Mudunkotuwa (2016). 5. RESULTS AND DISCUSSION

According to Mudunkotuwa et al. (2016) Ice-POM model over predicts sea ice extent in the Barents Sea in

Fig. 1 Model domain. (The area enclosed by red square represents the polar area compared in the study.)

Mudunkotuwa et al.

9

the winter. This has been improved by the assimilation. Figure 2 presents the sea ice extent from the two experiments. An average is calculated from the upper and lower limits. It can be seen that both nudging weights produce similar sea ice extent values that are close to the observation. One of the criticisms of the Ice-POM model is that sea ice thickness near the North Pole and the Canadian basin is under predicted (Mudunkotuwa et al. 2016), however sea ice thickness shows improvement according to fig. 3.

Fig. 2: Time series of sea ice extent for the whole

domain from nudging experiments using Eq. 4 and 5, their average, model and AMSR2 observation

There’s a considerable difference in sea ice thickness produced by Eq. 4 and 5 after sea ice thickness assimilation is seized in 4 months. The reason for this difference is that with K formulated as in Eq. 4, observation is considered to be underestimated (Observation error is added to the observation) making the model error larger than that of Eq. 5 formulation. Therefore, observation is weighted heavily with Eq. 4. Sea ice thickness has increased near the pole and the Canadian basin compared to the model (fig. 3). The average sea ice thickness calculated from these two methods is closer to the independent sea ice thickness data set cryostat’s values (fig. 3). Rise in sea ice thickness can be explained by the changes in sea ice velocity in the area. One of the reasons for ice-POM model to under predict sea ice thickness in the polar area is the over estimation in sea ice velocity in the area. However, with the introduction of satellite observations sea ice velocity in the Polar area has decreased (fig. 4), thereby it prevents advecting of sea ice away from the pole. This results in increased sea ice thickness in the

Polar area. Improved sea ice extent in the Barents Sea has led to improve ocean salinity in the area. Ice-POM model under predicts sea surface salinity in the Barents Sea, however assimilation removes sea ice in the Barents Sea in the winter (fig. 5). As a result freshwater in the Barents Sea is removed along with that. This increases sea surface salinity in the Barents Sea and along the marginal areas (fig. 6). 6. CONCLUSIONS The nudging experiments have improved sea ice extent predictability of the model. Sea ice thickness is also improved in Polar area due to the improved sea ice velocity in the area. Sea ice extent is specifically improved in the Barents Sea resulting in improved sea surface salinity in the area.

Fig. 3: Time series of mean sea ice thickness in polar area from nudging experiments using eq. 4 and 5, their average, model and AMSR2 observation

ACKNOWLEDGEMENTS The authors wish to acknowledge support from the Green Network of Excellence Program Arctic Climate Change Research Project (GRENE) and Arctic Challenge for Sustainability Research Project (ArCS) and a Kakenhi grant (no. 26249133) from the Japan Society for the Promotion of Science (JSPS).


10

Fig. 4: Monthly averaged sea ice velocity in the polar area in (m/s) from the nudging assimilation average, model and observation

Fig. 6: from left sea surface salinity in psu of model, assimilation avg, Observation NSIDC(area in black is where there is no data), salinity difference (assimilation avg-model) respectively in September

Fig. 5: Sea ice extent in Barents Sea from different

methods, Eq. 4 and 5 methods, model and AMSR2 observations in 2013

REFERENCES De Silva L.W.A. (2013): Ice-ocean coupled computations for the

sea ice prediction to support ice navigation in the Arctic Ocean. PhD thesis, University of Tokyo.

Fujisaki A., H. Yamaguchi and H. Mitsudera (2010): Numerical experiments of air–ice drag coefficient and its impact on ice–ocean coupled system in the Sea of Okhotsk. Ocean Dynamics 60, 377–394.

Kimura N., A. Nishimura., Y. Tanaka and H. Yamaguchi (2016): Influence of winter sea-ice motion on summer ice cover in the Arctic . Polar Research [Online] 32.

Krishfield, R. A. and 6 others (2014): Deterioration of perennial sea ice in the Beaufort Gyre from 2003 to 2012 and its impact on the oceanic freshwater cycle. Journal of Geophysical Research: Oceans , 119. 2, 1271-1305.

Lindsay R. W. and J. Zhang (2006): Assimilation of ice concentration in an ice-ocean model. Journal of Atmospheric and Oceanic Technology. Journal of Atmospheric and

Oceanic Technology, 23, 742-749. Brucker L., D. J. Cavalieri, T. Markus and I. Alvaro (2014):

NASA Team 2 Sea Ice Concentration Algorithm Retrieval Uncertainty. IEEE Transactions on Geoscience and Remote

Sea ice concentration Sea ice thickness Sea ice velocity

0.1~0.2 Not used in summer Not used in summer

0.01~0.08 15cm Zonal = 1.46 cm/s Meridional=1.35cm/s

0.05~0.15 42cm Zonal = 1.46 cm/s Meridional=1.35cm/s

0.1 Not used in summer Not used in summer

0.025 3.75cm Zonal and meridional =0.05cm/s

0.125 12.5cm Zonal and meridional =1.875cm/s

Table 1: Satellite observation error data

Mudunkotuwa et al.

11

Sensing, 11, 7336-7352. Mellor, G. L. (2015): Users guide for a three-dimensional,

primitive equation, numerical ocean model. Princeton University, 2003, 42pp.

Mudunkotuwa D. Y., L. W. A. De Silva and H. Yamaguchi (2015): Parametric study of assimilating sea ice concentration in a coupled ice-ocean model using nudging. Proc. 30th Intnatl. Symp. on Okhotsk Sea & Sea Ice, Mombetsu, Japan, 30, 66–69.

Mudunkotuwa D.Y., L. W. A. De Silva and H. Yamaguchi (2016): Data assimilation system to improve sea ice predictions in the Arctic Ocean using an ice-ocean coupled model. IAHR-ICE.

Mudunkotuwa D.Y. (2016): Data assimilation to improve sea ice predictability in the Arctic Ocean, PhD thesis, University of Tokyo, 149pp.

Parkinson, C. L. and W. M. Washington (1979): A large-scale numerical model of sea ice. Journal of Geophysical Research, 84, 311-337.

Ono, J. and 4 others (2016): The impact of radiosonde data on forecasting sea‐ice distribution along the Northern Sea Route during an extremely developed cyclone. Journal of Advances in Modeling Earth Systems, 292-303.



12

Preparing an UAV for drift ice observation

Kanichiro MATSUMURA1

1 Faculty of Bio-industry, Tokyo University of Agriculture, Abashiri, Japan

(Received October 7, 2016; Revised manuscript accepted November 3, 2016)

Abstract A cost effective, unmanned aerial vehicle (herein after UAV) is getting to be widely used in various field.

Especially in the civil engineering and agricultural fields, using a UAV is becoming a standard. The author has conducted surveys using an UAV since 2014. Aiming at observing drift ice observation along sea of Okhotsk, this research introduces fundamental method related to UAV technologies. Auto pilot software makes it possible for UAV to fly along a pre-calculated course as if automated cleaning machine flying. Series of pictures can be used as stereo-photogrammetry method. Very precise high-resolution digital elevation models (DEMs) are obtained easily and at low cost. Overlapping, geotagged images were obtained over grass land along sea of Okhotsk. Again, this research presents basic knowledge and techniques necessary for introducing UAV, its operation, aerial photography techniques and the possibilities of using an UAV to observe drift ice. Key words: Unmanned Aerial Vehicle (UAV), stereo-photogrammetry method, digital elevation models

(DEMs) 1. INTRODUCTION

UAV can get aerial data in remote, inaccessible regions. Series of two dimensional pictures can be transformed to three dimensional pictures so called Structure from Motion (SfM). SfM technology and multi copters technology build a 3-D model from pictures and applied for disaster areas from low altitude. (Inoue et al., 2014) Observation for a large outlet draining the Greenland ice sheet was conducted. (Ryan et al., 2015). 2. MATERIALS

DJI company has been produced a high performance multi copter since this company founded and its share is said to be up to 60 % or more. The author used DJI manufactured multi copter, Phantom2 and Phantom 3. Attaching an infrared camera on Phantom2 makes it possible to get Normalized Difference Vegetation Index (NDVI) images.

Fig. 1 DJI Phantom quadcopter

(Version2: Left and Version3: Right)

3. MISSION PLANNER, HUB, LITCHI

The open source software the Mission Planner (http’//plane.arudupilot.com/) makes it possible to conduct flight waypoint manipulation.

Fig. 2 Programming waypoint on Mission Planner

On the Mission Planner, choosing the area and put

flags to cover selected area. Flags are associated with coordinates and altitude above ground (not see level) and then a flight plan is obtained.

Table 1. Flight plan generated by Mission Planner

Article


13

Flight plan generated by Mission Planner should be transformed to csv style with the information latitude, longitude, altitude (m) using Microsoft Excel software.

Table 2. Csv format flight plan

Transformed csv style flight plan is uploaded to Mission Hub – Litchi (https://flylitchi.com/hub). Litchi's Mission Hub allows user to edit and share waypoint missions online and later execute them using the Litchi Android or iOS App.

Fig. 3 Mission Hub-Litchi

Flight plan on Mission Hub is transformed to auto

pilot software (Litchi) on iOS tablet machine makes it possible for user to let the UAV fly thorough the pre-programmed course.

Fig. 4 Auto Pilot software LITCHI

4. STEREO-PHOTOGRAMMETRY Auto pilot software makes it possible for UAV to fly

along a pre-calculated course. Series of pictures can be used as stereo-photogrammetry method. Very precise high-resolution digital elevation models (DEMs) are obtained easily and at low cost. The author conducted auto pilot system at the grassland along sea of Okhotsk.

Fig. 5 Picture of Observatory using GoPro Camera

attached to phantom 2

The author let phantom3 manually fly to the start point and then let it fly automatically using Litchi tablet software attached to controller. It is impossible for manned aerial vehicle to fly near the surface. In this research, pictures are taken every two seconds over grass land above 50 meter from ground level.

Manned aircraft must fly 300 meters high above ground level to avoid down burst. However, A UAV can fly 1 meter above ground level. It is impossible for manned aerial vehicle to fly such a low altitude. Using a UAV has a great advantage to get very precise surface data and generate 3 dimensional images.

Fig. 6 Series of pictures 50 meter above ground level

Matsumura

14

ESRI recently released Drone2Map for ArcGIS. This software makes UAV into an enterprise GIS productivity tool. Using this software, the author has tried to create orthomosaics, 3D meshes using drone-captured still imagery.

Fig. 7 Orthomosaics from drone-captured imagery

Fig. 8 High-resolution digital elevation models

This makes it possible to obtain more clear images and figures out the shape of structure such as building shown in Figure 9. Using three dimensional images, shape of buildings can be generated, that means that height of each plant can be calculated.

Fig. 9 Comparison between visible and DEMs

Fig. 10 Three dimensional images


15

5. CONCLUSION In this paper, the fundamental methodologies to let

UAV fly automatically and stereo-photogrammetry method are introduced. There must be potential and promising future applying proposed methodology to observe drifting ice along the sea of Okhotsk. The author let phantom2 to fly over drifting ice along the sea of Okhotsk (Drifting ice, 2015) as preliminary study during winter 2015.

To let UAV to fly under low temperature, battery must be pre-warmed up in advance. Once operation started, heat comes up from battery itself. To conduct operation above drifting ice, not quad copter but also aircraft shaped material can fly over much longer distance and collect more information.

The author conducted a drone work shop at Graduate school of Hokkaido University Oct the 3rd, 2016. Graduate student from Philippine, Nigeria, Tanzania, Indonesia, Russia, Malaysia, Bangladesh, Nepal participated. End of workshop, presentation entitled “How we can apply UAV to research field?” Three teams gave presentations there (See presentation movie files, 2016). A graduate student coming from Tanzania is interested in using air-craft shaped drone to observe Serengeti and Ngorongoro national park. The size of each park is more than 14,000 square km. The author proposed Tanzanian the method to let aircraft shaped drone to fly one way and change the battery and return to starting point. The author is interested in using aircraft shaped drone “Firefly6” and “Parrot”. Each aircraft shaped drone can gain 80km per hour and battery lasts more than 45 min. Applying series of proposed methodology and aircraft shaped drone for drifting ice observation leads to producing an innovation.

Fig. 11 Firefly6 (above) and Parrot Disco (below)

ACKNOWLEDGEMENTS We thank to Dr. Issei Watanabe, Research Institute for

Humanity and Nature, providing series of information related to UAV automatic flight and Firefly6, Parrot Disco. We also thank to Dr. Satoshi Inoue, National Agriculture and Food Research Organization providing agriculture information and Ms. Yuko Nakamura, ESRI Japan company to use Drone2 map software. This research is supported by Grant-in-Aid for Scientific Research from Ministry of Education, Culture, Sports, Science and Technology 2016 to 2018, entitled 「ドローンを使ったお

いしい牧草生成のための施肥システムの構築」 REFERENCES Inoue, H, S. Uchiyama and H. Suzuki (2014): Multicopter Aerial

Photography for Natural Disaster Research, NIED Research Report.

Ryan, J., C. and 7 others (2015): "UAV photogrammetry and structure from motion to assess calving dynamics at Store Glacier, a large outlet draining the Greenland ice sheet".

Litchi, App software (2016): available at https://itunes.apple.com /jp/app/litchi-for-dji-phantom-inspire/id1059218666?mt=8.

Drone2 map, Esri Company (2016), available at http://www.esri. com/products/drone2map.

Matsumura, K., K. Okuno (2015): Drifting Ice, available at https://www.youtube.com/watch?v=paB119vybNY.

A workshop at Hokkaido University, October (2016): https://www.youtube.com/watch?v=OyGo_Pkh5w8

https://www.youtube.com/watch?v=yMN4Uufikng https://www.youtube.com/watch?v=bKfyQYdvy_w.

Firefly6 Pro (2016): https://www.birdseyeview.aero/products /firefly6.

Parrot Disco (2016): https://www.parrot.com/fr/drones/parrot- disco-fpv#-parrot-disco-fpv.

Summary in Japanese

和文要約

無人航空機の流氷観測への可能性

松村寛一郎 1

1 東京農業大学オホーツク沿岸域でのドローンによる自動飛行の設定と

遂行の方法を紹介し，牧草地上での自動飛行による連

続写真の撮影を遂行した．連続撮影された一連の写真

群から，オルソ画像を生成し，精細な標高データ，建物

の判別が立体的に可能となることを示した．夏季に蓄積

されたノウハウを厳冬期のオホーツク海沿岸域での流氷

観測へ適用する際の注意，固定翼機材の利用可能性に

ついて言及した．著者自身による北海道大学大学院で

の留学生向けのドローンワークショップを通じてタンザニ

アの留学生による国立公園での固定翼機材による観測

の情報交換を続けている．オホーツク地域における厳冬

期の流氷観測に対するイノベーションを生み出す可能性

が見えつつある．



16

Inter-annual changes in responses of winter sea-ice motions to winds in the Arctic Ocean between 2003 and 2012

Eri YOSHIZAWA1 and Koji SHIMADA2

1 Division of Polar Ocean Sciences, Korea Polar Research Institute, Incheon, Republic of Korea 2 Department of Ocean Sciences, Tokyo University of Marine Science and Technology, Tokyo, Japan

(Received September 30, 2016; Revised manuscript accepted December 27, 2016)

Abstract

The response of winter (November-April) sea-ice motions to winds between 2003 and 2012 showed large year to year variations in the southern Canada Basin and the region along the Transpolar Drift Stream. We introduced the anomalies of the areal mean of sea-ice velocity curls averaged for the southern Canada Basin as a proxy of the intensity of the clockwise sea-ice motion in the basin. The composite analysis based on the sea-ice velocity curls showed that the clockwise motion was enhanced when the sea-ice motions in the Alaskan coastal region increased with less sea-ice concentration and the along-shore wind in November-December.

Key words: Arctic Ocean, sea-ice motion, mobility of sea ice

1. INTRODUCTION In the ice-covered Arctic Ocean, sea-ice motion modulates momentum transfers from atmosphere to ocean. The mean sea-ice velocity in the Arctic Ocean was accelerated anomalously in 2000s, compared to those in 1980s and 1990s, but the wind velocity only partially explained this acceleration (Rampal et al., 2009; Spreen et al.., 2011; Kwok et al., 2013). Such changes in the response of sea-ice motions to winds can have a large impact on the underlying upper ocean circulation and associated oceanic heat transports leading to the reduction of winter sea-ice formation (Shimada et al., 2006). Therefore, to investigate what have caused the changes in the kinematic coupling between atmosphere and sea-ice is important for understanding the ongoing sea-ice change. This study examines the inter-annual changes in the response of sea-ice motions to winds, using passive microwave measurement data from 2003 to 2012. The previous study suggested that the reduced internal stress caused by less sea-ice concentration along the Alaskan coast enhanced the sea-ice motion in the Canada Basin without the significant change in the wind forcing (Shimada et al., 2006). The wind direction against the Alaskan coastline can also affect lateral friction between the sea-ice cover and the coastal boundary, but it was not discussed in the previous study. Thus, we examine the relationship among the response of sea-ice motions to winds, sea-ice concentration and wind field, with a particular focus on the sea-ice motion in the Canada Basin.

Fig. 1 (a) A map of study region. Schematic seafloor

topographies are denoted by gray contours. (b) Winter (November-April) mean fields of sea-ice velocity vectors (vectors; cm/s) and SLP

(contours; hPa)

Article


17

2. DATA We use sea-ice velocity data from November 2002 to April 2012, which are calculated by the maximum correlation method using brightness temperature images obtained from passive microwave measurements (Kamoshida and Shimada 2010). In the calculation, brightness temperatures observed from the Advanced Microwave Scanning Radiometer for Earth observing system (AMSR-E) are used. In the period other than the AMSR-E operational period, brightness temperatures observed by the Special Sensor Microwave Imager/Sounder (SSMIS) are used. Here, we adopt sea-ice velocity data mapped onto the 43.75 × 43.75 km grid. We also use sea-ice concentration data provided by the National Snow and Ice Data Center (NSIDC), which is calculated by NASA team algorithm (Cavalieri et al., 1984). We also use sea-level pressure (SLP) and 10 m wind from the National Centers for Environmental Prediction-National Center for Atmosphere Research (NCEP-NCAR) reanalysis dataset (Kalnay et al., 1996).

Fig. 2 Winter (November-April) mean field of

iur×∇ (contours; 10-7 1/s). Thin solid (dashed)

contours represent positive (negative) values. Thick solid lines represent the zero lines of

iur×∇

3. RESULTS In this study, we focus on the period from November to April, when the mean sea-ice velocity in 2000s reached to its maximum and minimum values (Spreen et al., 2011). Hereafter, the period is referred to as "winter", e.g., the period from November 2002 to April 2003 is referred to as "the 2003 winter". First of all, we check briefly characteristics of winter mean sea-ice motions. The map of the study region is shown in Fig. 1a. Figure 1b shows the winter mean fields of sea-ice velocity vectors and SLP in 2003-2012.

The sea-ice velocity vectors exhibit the Transpolar Drift Stream (TDS) of sea-ice motions, which shows the sea-ice advection from the Chukchi Sea to the Fram Strait. Moreover, the clockwise motion is found over the Canada Basin. The sea-ice motions in the region from the southern Canada Basin to the Chukchi Sea are faster than those in the TDS region, but the SLP gradients in the former region are weaker than those in the latter region (contours in Fig. 1b). Figure 2 shows the winter mean field of sea-ice velocity curls ( iu

r×∇ ). The most of the Arctic Ocean is occupied by the negative iu

r×∇ , except the Laptev Sea, where the positive iu

r×∇ is evident. The negative iur×∇

value shows the local maximum in the southern half of the Canada Basin with flat sea-floor topographies. Next, we examine the inter-annual variations in the response of sea-ice motions to winds. Kimura and Wakatsuchi (2000) introduced the ratio of sea ice velocities, which are tangential to wind vectors, to wind velocities, in order to reveal the relationship between sea-ice motions and winds. In this study, 10 m wind velocities are used instead of geostrophic winds. Hereafter, this ratio is referred to as "wind factor (WF)". Sea-ice motions are also affected by ocean surface currents. In this study, for simplicity, it is assumed that effects of ocean surface currents are negligibly small compared with those of winds. Figure 3 shows spatial maps of the winter mean WF from 2003 to 2012. In order to capture changes in the relationship between sea-ice motions and winds in large scales, a 3 × 3 (131.25 × 131.25 km) median filter is applied for the WF fields. The previous studies reported that the acceleration of sea-ice motions in 2000s was evident in the southern half of the Canada Basin and the region along the TDS, even the wind speeds were at the almost same level as that in 1980-1990s (Spreen et al., 2011; Kwok et al., 2013). In these regions, the WF values show the large inter-annual variations from 2003 to 2012, whereas the values in the region north of the Canadian Archipelago remain unchanged (Fig. 3). After the late 1990s, the spatial distributions of sea-ice type varied from year to year in the former regions, while the robust multi-year ice occupied the latter region (Maslanik et al., 2011). Kwok et al. (2013) noted that the acceleration of sea-ice motions in the former regions seemed to be associated with the reduction of sea-ice thickness. Especially, in the southern half of the Canada Basin, the WF values are relatively large (~1.6%) in the winters of 2003, 2004, 2007 and 2008 (Figs. 3a, b, e and f). In these winters, the SLP gradients across the Alaskan coastline from the mouth of the Mackenzie River to Pt. Barrow are stronger than those in the other winters, suggesting that the along-shore winds are predominant (contours in Figs. 3a, b, e and f). As mentioned above, the previous study suggested that the

Yoshizawa and Shimada

18

decreasing in sea-ice concentration in the Alaskan coastal region enhanced the basin-scale clockwise sea-ice motion (Shimada et al., 2006).

Fig. 3 Winter (November-April) mean fields of WF

(%) with SLP (contours; hPa)

Fig. 4 Time series of the iu

r×∇ index from 2003 to 2012. A black (gray) solid line is winter means (monthly means) of the index

Therefore, to extract the relationship among the

intensity of the clockwise sea-ice motion and other parameters (sea-ice concentration and wind direction against the coastline), we employ the composite analysis based on iu

r×∇ . We note again that there is the local minimum value of iu

r×∇ in the southern Canada Basin (dashed contours in Fig. 2). It has been reported that the clockwise sea-ice motion was predominant in the Canada Basin throughout the year except the late summer (McLaren et al., 1987; Asplin et al., 2009). Similarly, during 2003 to 2012, the monthly mean values of iu

r×∇ averaged for the southern Canada Basin (73-77oN, 130-160oW) show the negative value except the summers of 2003 and 2008 (not shown). Thus, we use anomalies of areal mean values of

iur×∇ in this region as a proxy of the intensity of the

clockwise sea-ice motion in the Canada Basin (hereinafter referred to as "the iu

r×∇ index”). In this study, it is assumed that the intensity of the clockwise motion is strong (weak) when the iu

r×∇ index is negative (positive). Figure 4 shows the time series of the winter mean values of the iu

r×∇ index (thick black line) with the monthly mean values (gray line). In the inter-annual time scale, the intensity of the clockwise sea-ice motion in winter reaches its maximum value in 2008 and then recovered to the same level of the early 2000s (black solid line in Fig. 4). Using this index, we make two composites of WF and SLP when the index is below (above) zero. The composite with the negative index is constructed from the data in the five winters (2003, 2007, 2008, 2009 and 2011), and the composite with the positive index is constructed from the data in the other five winters (2004, 2005, 2006, 2010 and 2012). Since the inter-annual variability of WF is large in November-December compared with that in January-April (not shown), we here use the WF values averaged in November-December. Figure 5 shows the two composite maps of WF with SLP. In the negative index composite, the WF values are relatively large in the Chukchi Sea and the southern half of the Canada Basin (Fig. 5a). In the positive index


19

composite, the values are also large in the Chukchi Sea, but the values are small in the Canada Basin except the coastal area just east of Pt. Barrow (Fig. 5b). The difference in the WF values between the two composites also indicates that the clockwise motion is enhanced when the WF values are large in the southern Canada Basin (Fig. 6a). In the region along the Alaskan coast, sea-ice concentration in the negative index composite is lower, compared with that in the positive index composite (Fig. 6b). The results confirm the relationship between the intensity of the basin-scale sea-ice motion and the lateral boundary condition along the Alaskan coast, which was pointed by Shimada et al. (2006). If the wind that is along (perpendicular to) the Alaskan coastline from the mouth of the Mackenzie River to Pt. Barrow is dominant, the basin-scale clockwise sea-ice motion would accelerate (deaccelerate) due to decreases (increases) in lateral friction between the ice-cover and the Alaskan coastal boundary. Focusing on the Canada Basin along the 75oN line, the mean SLP field denoted by contours in Fig. 5a (Fig. 5b) shows that the wind in this region is nearly parallel to (intersects with) the coastline when the WF values are relatively large (small) in the southern portion of the basin. This implies that the direction of the wind in the southern Canada Basin is one of the key conditions to promote effective kinematic coupling between sea-ice motions and winds in the Canada Basin.

Fig. 5 Spatial maps of WF (%) and SLP (contours;

hPa) in November-December averaging when (a) the iu

r×∇ index is below zero and (b) above zero

4. CONCLUSIONS This paper indicated that the response of sea-ice motions to 10 m winds in winter varied largely from year to year in the southern half of the Canada Basin and the TDS region from 2003 to 2012. In this study, we particularly focused on the clockwise sea-ice motion in the Canada Basin, and attempted to investigate what controlled the intensity of the clockwise motion. We introduced the anomalies of the

areal mean values of iur×∇ averaged for the southern

Canada Basin, where the iur×∇ value showed its

local minimum, as a proxy of the intensity of the clockwise sea-ice motion. The composite analysis based on iu

r×∇ indicated that the intensity of the clockwise motion was large when the WF values were large in the southern half of the basin near the Alaskan coast compared with those in the other regions. In the case that the intensity of the clockwise motion was strong (weak), sea-ice concentration near the Alaskan coast was low (high) and the wind in the southern Canada Basin was nearly parallel to (intersected with) the Alaskan coastline in the early winter (November-December).

Fig. 6 (a) Differences between the two composites (the

negative index composite minus the positive index composite) in (a) WF (%) with SLP (contours;

hPa) and (b) sea-ice concentration (%) in November-December

REFERENCES Asplin, M. G., J. V. Lukovich and D. G. Barber (2009):

Atmospheric forcing of the Beaufort Sea ice gyre: Surface pressureclimatology and sea ice motion, J. Geophys. Res., 114, C00A06, doi:10.1029/2008JC005127.

Cavalieri, D. J., P. Gloersen and W. J. Campbell (1984): Determination of sea ice parameters with the NIMBUS 7 SMMR, J. Geophys. Res., 89(D4), 5355–5369.

Kalnay, E. and 21 other authors (1996): The NCEP/NCAR 40-year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437–471.

Kamoshida T. and K. Shimada (2010): Long term sea ice motion dataset in the Arctic from SMMR, SSM/I and AMSR-E. In: Proceedings of the second international symposium on the Arctic research, Tokyo, Japan, 2010, p 125.

Kimura, N. and M. Wakatsuchi (2000): Relationship between sea-ice motion and geostrophic wind in the Northern Hemisphere, Geophys. Res. Lett., 27, 3735–3738.

Kwok, R., G. Spreen and S. Pang (2013): Arctic sea ice circulation and drift speed: Decadal trends and ocean currents, J. Geophys. Res. Oceans, 118, 2408–2425, doi:10.1002/jgrc.20191.

McLaren, A. S, M. C. Serreze and R. G. Barry (1987): Seasonal variations of sea ice motion in the Canada basin and their

Yoshizawa and Shimada

20

implications, Geophys. Res. Lett., 114:1123–1126. Maslanik, J. and 3 other authors (2011): Distribution and trends

in Arctic sea ice age through spring 2011, Geophys. Res. Lett., 38, L13502, doi:10.1029/2011GL047735.

Rampal, P., J. Weiss and D. Marsan (2009): Positive trend in the mean speed and deformation rate of Arctic sea ice, 1979–2007, J. Geophys. Res., 114, C05013, doi:10.1029/ 2008JC005066.

Shimada, K. and 7 other authors (2006): Pacific Ocean inflow: Influence on catastrophic reduction of sea ice cover in the Arctic Ocean, Geophys. Res. Lett., 33, L08605, doi:10.1029/ 2005GL025624.

Spreen, G., R. Kwok and D. Menemenlis (2011): Trends in Arctic sea ice drift and role of wind forcing: 1992–2009, Geophys. Res. Lett., 38, L19501, doi:10.1029/2011 GL048970.


Submission Information for OSPOR

Reviewing processes of OSPOR 1) When manuscripts have been received by the Editor-in-Chief, an acknowledgement of

receipt will be sent to the author(s) by e-mail. The Editor-in-Chief chooses an editor to handle the manuscript review.

2) The submitted manuscript will be subjected to screening review for its scope, novelty, completeness, English level, and conformation to the OSPOR policy. A manuscript not passing the screening review will immediately be returned to the authors.

3) The editor in charge will select expert reviewers to evaluate the manuscript. 4) As to results of review, if the editor decides that the paper needs revision by the author(s),

the manuscript will be returned to the author(s) for revision. 5) Manuscripts returned to author(s) for revision should be resubmitted promptly. If the

revision cannot be finished within a month, the manuscript will be regarded as having been withdrawn.

6) The Editor-in-Chief will finally decide whether to accept the manuscript for publication.

Paper Submission Submission Guideline

All manuscripts should be submitted in digital format (PDF or WORD) with the OSPOR submission sheets (PDF or WORD, offered from OSPOR) by email to the OSPOR Editorial Board

OSPOR Editorial Board

Polar Oceans Research Association (OSPORA) Address: Kaiyo Koryukan, 1 Kaiyo Koen, Mombetsu, Hokkaido 094-0031 Japan E-mail: [email protected] Phone : +81-158-26-2810 (Japan 0158-26-2810) Fax: +81-158-26-2812 (Japan 0158-26-2812)

Publication Charge

Authors of their institutions are requested to pay the publication charge according to the following rate when paper is accepted. 5,000 Yen / (paper)

Copyright

Copyright for an article submitted to OSPOR is transferred to OSPORA when the article is published in OSPOR in any form.

Preparation of manuscripts

The manuscript should be formatted in the form of OSPOR template offered from the OSPORA office, which satisfies the following requirements. The maximum page in printing style is 6 pages. 1) Text

a) The manuscript should be in the international size A4 in camera-ready style according to the form of OSPOR template.

b) The first page should include: the title, the author(s) name(s) and their affiliations. If possible, a Japanese translation of the title and the name(s) of the author(s) should be provided in the end of manuscript. If they are not, the translation will be undertaken by the OSPOR editorial board.

c) An abstract not exceeding 250 words must be provided. d) Up to five keywords that describe the content for indexing purposes must be

provided.

2) References a) A list of cited references should be arranged alphabetically. Journal

abbreviations are better to use, but when the abbreviation is not known, the full title of the journal should be used in the list. In the case of many authors, the author name can be written in short as below.

Kawamura, K., F. Parennin and 16 others (2007): Northern hemisphere forcing of climatic cycles in Antarctica over the past 360,000 years. Nature, 448, 912-916.

b) References in the text will include the name(s) of the author(s), followed by the

year of publication in parentheses, e.g. (Clarke, 2003), (Li and Sturm, 2002), (Harrison et al., 2001).

3) Units Numerical units should conform to the International System (SI). Units should be in the form as kg m-3 not as kg/m3.

4) Tables A title and a short explanation should be located on the top of table. They should be referred to in the text.

5) Figures a) All Figures (illustrations and photographs) should be numbered consecutively. b) All Figures should be of good quality and referred to in the text. c) Figure captions should be located on the bottom of the Figures.

6) More information For more information about manuscript instruction, please ask to OSPORA office or see OSPORA home page in http://www.o-tower.co.jp/okhsympo/top-index.html.

Okhotsk Sea and Polar Oceans Research, Vol. 1 (2017, February) Published by the Okhotsk Sea and Polar Oceans Research Association (OSPORA)

Executive Committee of OSPORA: Chairman: Shuhei Takahashi (Okhotsk Sea Ice Museum of Hokkaido, Director) Secretariat: Eriko Uematsu

Address: Kaiyo Koryukan, 1 Kaiyo Koen, Mombetsu, Hokkaido 094-0031 Japan E-mail: [email protected] Phone : +81-158-26-2810 (Japan 0158-26-2810) Fax: +81-158-26-2812 (Japan 0158-26-2812) http://www.o-tower.co.jp/okhsympo/top-index.html


Published by the Okhotsk Sea and Polar Oceans Research Association (OSPORA)

O

kho

tsk Sea an

d P

olar O

ceans R

esearch

Vo

l. 1 2017

okhotsk sea and polar oceans researchŸ»読全体.pdf · sea and polar oceans research association...

Documents