Kobe University Repository : Thesis

学位論文題目Tit le

Offshore Wind Resource Assessment Using a Mesoscale Mode(メソ気象モデルを用いた洋上風力資源量調査)

氏名Author Shimada, Susumu

専攻分野Degree 博士（工学）

学位授与の日付Date of Degree 2011-09-12

資源タイプResource Type Thesis or Dissertat ion / 学位論文

報告番号Report Number 乙3169

権利Rights

JaLCDOI

URL http://www.lib.kobe-u.ac.jp/handle_kernel/D2003169※当コンテンツは神戸大学の学術成果です。無断複製・不正使用等を禁じます。著作権法で認められている範囲内で、適切にご利用ください。

PDF issue: 2021-08-30

Doctoral thesis

Offshore Wind Resource Assessment Using a Mesoscale Model

メソ気象モデルを用いた洋上風力資源量調査

Kobe University Graduate School of Maritime Sciences

Susumu SHIMADA

2011

i

Acknowledgement I would like to express my deepest gratitude to my supervisor, Assoc. Prof. Dr. Teruo Ohsawa at

Graduate School of Maritime Sciences, Kobe University whose comments and suggestions were of

inestimable value of my Phd study. If I did not have his dedicated and continuous support, I could

not finish writing my thesis. I also wish to thank all members of Air-Sea Laboratory in Kobe

University for giving me the opportunity to address such an interesting research for the last three

years. The quality of the thesis has been considerably improved thanks to the very helpful comments

from Prof. Dr. Hiroshi Ishida, Prof. Dr. Katsutoshi Kozai and Prof. Dr. Eiichi Kobayashi.

Furthermore I would also like to thank Dr. Detlev Heinemann who provided me an opportunity to

stay at ForWind, Germany for my research and all members of ForWind. Special thanks go to Dr.

Gerald Steinfeld and Mr. Michael Schmidt who supported me not only in doing researches but also

my life in Oldenburg. Finally I would also like to express my gratitude to my family and my wife,

Tomoko for their moral support and warm encouragements.

This project was supported by Research Fellowship of the Japan Society for the Promotion of

Science for Young Scientist, a Grant-in-Aid for JSPS Fellows 21-6911 and Excellent Young

Researchers Overseas Visit Program of JSPS from the Ministry of Education, Science, Sport and

Culture, Japan.

ii

謝 辞

本研究は(独)日本学術振興会特別研究員制度，特別研究員奨励費（課題番号：21-6911）

および優秀若手研究者海外派遣事業の支援を受けました．ここに感謝の意を表します．

本論文は，数多くの方々の御指導・御協力により，取りまとめることが出来ました．

指導教官である神戸大学大学院海事科学研究科，大澤輝夫准教授には，公私を問わず，

あらゆる面で御指導賜りました．大澤先生への感謝の気持ちは言い尽くすことができませ

ん．思えば，研究のエッセンスをご教授頂いたことは勿論のこと，卒業論文からこの博士

論文に至まで，私の稚拙な日本語・英語を常に丁寧に添削して頂き，論文執筆を通じて論

理的に道筋を付ける力をゼロから鍛えて頂きました．また，時折訪れる人生の分岐点にお

いて，常に適切なアドバイスを頂くと同時に，私の意思を尊重して正しい方向に進めるよ

う促して頂きました．現在でも，研究に対する積極的な姿勢から論文の文体に至まで，大

澤先生の全てをお手本にさせて頂いています．本当に有り難うございました．

神戸大学大学院海事科学研究科・海洋気象研究室，香西克俊教授，石田廣史教授および

林美鶴准教授に感謝の意を表します．香西先生には，在学中，常に暖かいお言葉をかけて

頂きました．研究室全体が暖かい雰囲気で，他の学生と同様に三年間非常に楽しく研究で

きる環境を与えて頂きました．石田先生には，博士課程の論文審査を通じて内容に関して

有益なご助言賜りました．林先生には，海洋研究開発機構での貴重な講義を受講する機会

を紹介して頂いたのと同時に，海事科学部からストレートで研究室に配属される学生と同

じにように分け隔てなく接して頂きました．香西先生，石田先生及び林先生とお話しする

機会を定期的に賜ったことで，在籍期間が博士課程の三年と比較的短いにもかかわらず，

自分自身が海事の卒業生の一人であることをはっきりと今後認識できると思います．また，

小林英一教授には，同様に，博士の論文審査を通じて有益なご助言賜りました．ここに感

謝の意を表します．

学位論文の一部は，神戸大学大学院海事科学研究科を単位修得退学後，岐阜大学環境エ

ネルギーシステム専攻で特任助教として研究に従事する傍ら取りまとめました．岐阜大学

大学院環境エネルギーシステム専攻・自然エネルギー研究室，安田孝志教授，小林智尚教

授および吉野純助教に感謝の意を表します．安田先生には，研究者として心構えをお忙し

い合間を縫ってお話して頂きました．安田先生からお話し頂いたことを常に頭の片隅に留

めておきたいと思います．小林先生には，学位論文の執筆や進路のことで精神的に追い込

まれている時に非常に心温まる優しいお言葉を頂きました．さらに，直接的に研究を継続

する環境を与えて頂き，博士論文の執筆に関しても最大限の配慮を賜りました．また，吉

野先生には，渡英前のお忙し時期であったにもかかわらず，着任後に研究がスムーズに始

められるよう，計算機環境一式を整えて頂きました．誠に有り難うございました．

iii

ドイツ，オルデンブルグ大学 ForWind 研究センター，デトレフ・ハイネマン博士，ゲラ

ルド・シュタインフェルド博士，ミヒャエル・シュミット研究員及び ForWind の皆様に感

謝の意を表します．ハイネマン博士には，オルデンブルグ大学での研究滞在を快く受け入

れて頂きました．また，シュタインフェルド博士及びシュミット研究員には，ドイツでの

生活を全面的にサポートして頂きました．風力資源量調査に関する先端研究を体感するこ

とで，自分自信が今後目指すべき研究レベルをはっきりと認識することが出来るようにな

りました．また，国境を越えて親密な交流を持てたことで，海外と自分との距離を大きく

縮めることが出来ました．

また，ここでひとりひとり全ての御名前を挙げることは出来ませんが，これまで所属し

た神戸大学海洋気象研究室，岐阜大学自然エネルギー研究室および社会人時代の上司・先

輩を目標として，同期・同僚とは競い合い，後輩からは刺激を貰い続け，それを糧に自己

研鑽に努めて参りました．数多くの方々の御協力を賜って何とか学位論文の提出に至った

と思っています．ここに改めて感謝の意を表します．

最後に，博士課程で勉強することを暖かく見守ってくれた家族，そして辛抱強く支え続

けてくれた妻，朋子に心から感謝の意を表します．

平成 23 年 7 月 嶋田進

iv

Abstract

For the purpose of evaluating the reliability and increasing the accuracy of offshore wind resource

assessment using a mesoscale model, the accuracy and characteristics of winds simulated by the

Weather Research and Forecasting (WRF) model, developed by the National Center for Atmospheric

Research (NCAR) and the National Center for Environmental Prediction (NCEP), are investigated

using in situ measurements taken in Japanese coastal waters. The contents of this thesis are

summarized as follows.

In Chapter 1, an overview of offshore power generation and the methodology of offshore wind

resource assessment are reviewed, and the background, purpose and contents of this thesis are

described.

In Chapter 2, in order to understand the characteristics of offshore winds in the waters of interest,

long-term wind measurements at the Shirahama research platform, located 2 km off the nearest

coastline in Tanabe Bay, Japan are analyzed. The analysis shows that the 14 years of measurements

from an anemometer at 23 m height have annual mean wind speed of 5.0 m/s and standard deviation

of 0.4 m/s. North-westerly winds associated with the monsoon prevail in winter, while easterly

winds associated with land-sea breeze circulation prevail in summer. The atmosphere is unstable for

approximately 80 % of the year, and neutral and stable conditions tend to occur in summer,

especially when southerly winds blow. With regard to turbulence intensity, winds from all directions

are found to be equivalent to turbulence intensity class B in the IEC (International Electrotechnical

Commission) standard. Comparing winds from the land and sea sectors, those from sea sectors tend

to have relatively smaller turbulence intensity than those from land sectors.

In Chapter 3, a WRF simulation is performed for the year 2005, and the accuracy of the simulated

wind speed is examined using in situ measurements from the Shirahama research platform. The

WRF surface wind speed is found to have an annual bias of +15.3 % of measured mean wind speed,

a RMSE (Root Mean Square Error) of 46.0 % and a correlation coefficient of 0.81. Moreover,

decomposition of the RMSE of the WRF wind speeds indicates that improvement of the large

positive bias is a key to increasing the accuracy of the WRF wind speed. In addition, comparison

with the Japan Meteorological Agency (JMA) wind profiler measurements at Mihama shows that the

bias in the WRF wind speed is negative at the top of the PBL (Planetary Boundary Layer), but it

becomes positive under a height of 1,000 m, considerably increasing toward the surface. This means

that the positive bias is not limited to the surface layer but also prevails in the lower PBL.

In Chapter 4, in order to remove the bias included in the WRF wind speed, the effect of the

accuracy of the Sea Surface Temperature (SST) dataset on the accuracy of wind speed simulated by

WRF is examined. In the mesoscale model, SST is used as a parameter for the calculation of

atmospheric stability, and the wind speed profile near the sea surface is parameterized using the

v

vertical momentum flux, sea surface roughness and atmospheric stability through the

Monin-Obukhov similarity theory. Consequently, the accuracy of wind speed near the sea surface is

considered to depend directly on the accuracy of SST. In the beginning of this research, NGSST-O

(New Generation Sea Surface Temperature for Open Ocean), produced by Tohoku University, had

been used as a lower boundary condition in our previous mesoscale model simulations, but it was

found later to have a positive bias of a few degrees at the Shirahama research platform. Therefore, in

order to reduce the large bias found in NGSST-O in coastal waters, a new higher spatial resolution

SST dataset is developed here based on the MODIS (MODerate resolution Imaging

Spectroradiometer) product provided by the Japan Aerospace Exploration Agency (JAXA).

Moreover, using both the NGSST-O and MODIS-based SST, annual WRF simulations on 2 km grids

are performed, and the accuracy of the simulated wind speed is quantitatively examined against in

situ measurements taken at five offshore research platforms. As a result, although some biases in the

WRF wind speed is not improved, the use of the MODIS-based SST instead of NGSST-O is found to

improve the bias and RMSE in the annual mean wind speed by 9.3 % (on five site average) and

4.5 %, respectively. The sensitivity of wind speed and wind energy density to SST is also discussed

by comparing the difference in annual mean SST and simulated wind speed and wind energy density

for the two SST datasets, respectively. The results obtained from this study suggest that the use of

accurate SST is a key factor not only for realistic offshore wind simulations near the sea surface, but

also accurate wind resource assessment with a mesoscale model.

In Chapter 5, in order to investigate the causes of the large positive bias apart from the inaccuracy

of SST, it is examined whether the large positive bias in wind speed in the lower PBL mentioned in

Section 2 is peculiar to Mihama or is present at other wind profiler measurement stations. From the

comparison with wind profiler measurements at ten JMA stations, the WRF-simulated wind speed

profile is found to have outstanding positive biases in the lower part of PBL not only at Mihama, but

also all other stations. Moreover, the positive bias is unlikely to be caused by either error in the wind

profiler measurement or the objective analysis data input into WRF. In addition, the author compares

the wind speed profiles simulated by WRF with seven different PBL schemes for a month. The

results show that the positive bias cannot be reduced simply by using alternative PBL schemes. The

author finally concludes that the positive bias is most likely caused by a deficiency in the WRF

model itself, such as the deficiency in the formulation of the PBL or surface layer parameterization

rather than the accuracy of external data input into WRF and the model settings.

In Chapter 6, the general conclusions obtained from this study are summarized, and future work

related to this study is mentioned.

vi

要 旨

本論文は，メソ気象モデルを用いた洋上風力資源量調査手法の信頼性の評価及びその高

精度化を目的として，日本沿岸における海上風現場観測値を用いて米国大気研究センター

及び米国気象局共同開発のメソ気象モデル WRF（Weather Research and Forecasting）の計算

精度及びその特性についての検討したものである．各章の概要を以下に示す．

第一章では，洋上風力発電の現状及び洋上風力資源量調査手法に関する既往研究をレビ

ューし，本論文の背景，目的及び構成について示した．

第二章では，和歌山県西牟婁郡田辺湾沖に設置された京都大学防災研究所所有の白浜海

象観測所海洋鉄塔における海上風の現場観測値を用いて検討対象海域における洋上風況特

性について検討した．その結果，1994 年から 2007 年までの計 14 ヶ年の白浜海洋鉄塔にお

ける年平均風速は 5.0 m/s（23 m 高）であり，その年々変動の標準偏差は 0.5 m/s であった．

風向別出現頻度は冬季は季節風に伴う北西風，夏季は海陸風循環に伴う東寄りの風が卓越

し，年間の大気安定度の約 8 割は不安定であり，夏季に南寄りの海風が吹送する場合に中

立から安定になり易いことが分かった．乱れ強度については，IEC（International

Electrotechnical Commission）乱れ強度基準 B 以下に相当し，さらに海上を吹送する風は陸上

からのものに比べて若干乱れ強度が小さいことを示した．

第三章では，メソ気象モデル WRF による年間海上風シミュレーションを行い，第二章で

解析した白浜海洋鉄塔の現場観測値を用いて WRF による海上風計算精度を検証した．その

結果，WRF の海上風計算精度は，年間のバイアスは+15.3%（観測年平均風速比），RMS（Root

Mean Square）誤差は 46.0%，相関係数は 0.81 であった．また，RMS 誤差の成分分解の結果，

計算値に含まれる RMS 誤差は初期値，境界値及び四次元データ同化値として用いる客観解

析値の精度に大きく依存するため大幅な改善は見込めないものの，バイアスさえ取り除く

ことができれば取得可能な洋上風力資源量を高精度で推定可能であることを示した．加え

て，白浜海洋鉄塔近傍の気象庁ウィンドプロファイラ美浜地点の観測値と WRF 計算値の風

速鉛直プロファイルの比較により，白浜海洋鉄塔で見られる大きな正のバイアスは海面近

傍だけではなく大気境界層下部全体に含まれることを示した．

第四章では，WRF 計算値に含まれる正のバイアスを改善するため，メソ気象モデルの下

面境界条件の一つとして与える海面水温データの精度が海上風計算精度に及ぼす影響を検

討した．白浜海洋鉄塔における WRF の計算精度検証の過程において，これまで著書らが海

面水温データとして用いてきた東北大学提供の外洋域新世代海面水温データ（以下，

NGSST-O：New Generation Sea Surface Temperature for Open Ocean）は白浜海洋鉄塔の現場観

測水温に対して年平均で 1，2 度の高い値であることが分かってきた．メソ気象モデルの海

面近傍の風速鉛直プロファイルは，大気安定度の効果を考慮したモニン・オブコフの相似

則によって定式化され，海面水温データは海面近傍の大気安定度を決定するための一つの

vii

パラメータとして用いられる．それ故，実際に比べて高い海面水温をメソ気象モデルに与

えた場合には，海面近傍の運動量輸送を過大評価することに繋がり，結果としてそれがメ

ソ気象モデルによる海上風計算の誤差要因になるものと考えられる．そこでまず，NGSST-O

に含まれる沿岸での正のバイアスを改善するため，宇宙航空研究開発機構（JAXA：Japan

Aerospace Exploration Agency）より配信されている MODIS（MODerate resolution Imaging

Spectroradiometer）による海面水温プロダクトを用いて，新たに日本列島全域を含む高精度

且つ高空間解像度の海面水温データセットを開発した．

更に NGSST-O 及び MODIS ベースの 2 つの海面水温データを用いて WRF の年間シミュ

レーションをそれぞれ行い，白浜観測鉄塔および大阪湾内 4 地点の海上風現場観測値を用

いて海面水温データの精度が WRF の風速計算値に与えるインパクトを定量的に検討した．

その結果，WRF 計算値に含まれる全ての正のバイアスは改善されないものの，MODIS ベー

スの高精度な海面水温データを用いることにより計算値の年平均風速に含まれる 9.3%のバ

イアス（5 地点平均）および 4.3%の RMS 誤差が改善されることを示した．また，計算領域

の全格子点における NGSST-O 及び MODIS ベースの海面水温データ，及びそれ基づく WRF

の風速計算値の差をそれぞれ比較し，海面水温データに対する WRF による海上風計算値の

感度を解析したところ，検討対象海域における海面水温データの両者の平均的な差の約 2 °C

では，10 m 高において年間約 4%の風速差を引き起こすことを示した．また，それを風力エ

ネルギー密度に換算にした場合，約 2 °C の海面温度データの差は風車ハブ高度に相当する

60 m 高において年間約 5%の差を引き起こすことを示した．これらの結果は，メソ気象モデ

ルを用いた洋上風力資源量調査において，高精度な海面水温データを与えることが一つの

重要な要素であることを示唆している．

第五章では，WRF 計算値の海面水温以外に起因するバイアスについて，その改善の方向

性を検討するため，第二章で示した気象庁ウィンドプロファイラ美浜地点で見られた大気

境界層下部全体の正のバイアスについて，それが美浜地点特有の誤差なのかそれとも他の

沿岸観測点でも同様に含まれる誤差なのかをより詳細に検討した．WRF の年間計算結果と

西日本地域における合計 10 地点のウィンドプロファイラと WRF 計算値の鉛直プロファイ

ルの比較の結果，WRF のよる風速計算値は全ての地点で美浜地点と同様に地表に近づくに

つれて徐々に大きくなる正のバイアスを含むことが明らかとなった．さらに，この正のバ

イアスはウィンドプロファイラの観測誤差や初期値・境界値及び四次元データ同化値とし

て用いる気象庁メソ客観解析値によるものではないことが確認された．最後に，WRF に実

装される 7 つの大気境界層スキームを用いて鉛直プロファイルを再度比較したところ，こ

の正のバイアスは大気境界層スキームを変更するだけでは単純に解決できないことが示さ

れた．これらの結果は，WRF による海上風計算の高精度化を図るためには，今後，大気境

界層スキームのモデル係数のチューニングや定式化の見直し等，WRF モデル自体の改善が

今後，必要不可欠であることを示唆するものである．

第六章では，本論文で得られた研究成果を総括するとともに，今後の課題を示した．

viii

Contents

1 Introduction ......................................................................................................... 1-1

1.1 Background and purpose of this study .......................................................................... 1-1

1.2 Outline of this thesis ..................................................................................................... 1-3

References ..................................................................................................................... 1-5

2 Characteristics of offshore winds at Shirahama Oceanographic Observatory

............................................................................................................................... 2-1

2.1 Introduction ................................................................................................................... 2-1

2.2 Shirahama Oceanographic Observatory ........................................................................ 2-2

2.2.1 Site description ...................................................................................................... 2-2

2.2.2 Observation devices .............................................................................................. 2-3

2.3 Characteristics of winds at Shirahama .......................................................................... 2-4

2.3.1 Data used in the analysis ....................................................................................... 2-4

2.3.2 Annual means and frequency distributions ........................................................... 2-4

2.3.3 Wind direction distributions .................................................................................. 2-6

2.3.4 Stability conditions ............................................................................................... 2-8

2.3.5 Turbulence intensity .............................................................................................. 2-9

2.4 Conclusions ................................................................................................................. 2-10

References ................................................................................................................... 2-11

3 Accuracy and characteristics of offshore wind speeds simulated by WRF .... 3-1

3.1 Introduction ................................................................................................................... 3-1

3.2 Data and methods .......................................................................................................... 3-2

3.2.1 Configuration of WRF .......................................................................................... 3-2

3.2.2 In-situ data used for evaluation of simulated wind speed ..................................... 3-4

3.3 Result and discussion .................................................................................................... 3-4

3.3.1 Accuracy of simulated surface wind speed ........................................................... 3-4

3.3.2 Evaluation and improvement of wind speed errors ............................................... 3-9

3.3.3 Accuracy of wind speeds above surface layer .................................................... 3-10

3.4 Conclusions ................................................................................................................. 3-11

References ................................................................................................................... 3-13

ix

4 Effects of Accuracy of Sea Surface Temperature on Offshore Wind Resource

Assessment using a Mesoscale Model ................................................................ 4-1

4.1 Introduction ................................................................................................................... 4-1

4.2 Data used for validation ................................................................................................ 4-3

4.2.1 Validation of SST .................................................................................................. 4-3

4.2.2 Validation of wind speed ....................................................................................... 4-4

4.3 Accuracy of NGSST-O in coastal waters ...................................................................... 4-7

4.4 High-resolution SST dataset based on MODIS ............................................................. 4-9

4.4.1 Methodology of objective analysis ....................................................................... 4-9

4.4.2 Bias correction .................................................................................................... 4-10

4.4.3 Accuracy of the MODIS-based SST ................................................................... 4-10

4.5 Effect of accuracy of SST on offshore wind simulation ............................................. 4-13

4.5.1 Configuration of the simulations ......................................................................... 4-13

4.5.2 Mean wind speed fields simulated by WRF with NGSST-O and the MODIS-based

SST ...................................................................................................................... 4-15

4.5.3 Validation of the simulated wind speeds. ............................................................ 4-18

4.5.4 Effects of accuracy of SST on the mean wind speed and wind energy density .. 4-20

4.6 Conclusion .................................................................................................................. 4-23

References ................................................................................................................... 4-25

5 Accuracy of the wind speed profile in the lower PBL as simulated by the WRF

model ..................................................................................................................... 5-1

5.1 Introduction ................................................................................................................... 5-1

5.2 Data and methods .......................................................................................................... 5-2

5.2.1 WRF model ........................................................................................................... 5-2

5.2.2 Data used in the validation .................................................................................... 5-5

5.3 Results and discussions ................................................................................................. 5-5

5.3.1 Wind speed bias at 10 wind profiler stations ........................................................ 5-5

5.3.2 Possible causes of the positive bias ....................................................................... 5-8

5.3.3 Effects of the PBL scheme on the vertical wind profiles .................................... 5-10

5.4 Conclusions ................................................................................................................. 5-11

References ................................................................................................................... 5-11

6 Conclusions and summary .................................................................................. 6-1

6.1 Conclusions obtained in each chapter ........................................................................... 6-1

x

6.2 Summary ....................................................................................................................... 6-2

List of abbreviations .............................................................................................................. 7-1

Curriculum Vitae ................................................................................................................. 7-3

1-1

Chapter 1

Introduction

1.1 Background and purpose of this thesis

The number of facilities for offshore wind power generation is expected to increase in the future

as attempts has been made to reduce greenhouse gas emissions associated with energy production in

Northern Europe. The first offshore wind farm in the world, the Vindeby wind farm off the coast of

Denmark in the Baltic Sea, started operation with an installation capacity of 5 MW in 1991. In 2011,

45 offshore wind farms with the total installation capacity of 2 946 MW are operational in nine

European countries (EWEA 2011). The European Wind Energy Association (EWEA) predicts that

the development of offshore wind power generation will proliferate over the next few decades. The

total installation capacity in Europe will reach 150 GW in 2030, and 563 TWh/year of electricity,

equivalent to the 17 % of the annual electricity consumption of the EU, will be produced by means

of offshore wind power generation (EWEA 2009).

On the other hand, offshore wind energy is considered as an undeveloped wind energy resource in

Japan (Nagai et al. 1997; Ohsawa et al. 2009). For instance, Ohsawa et al. (2009) showed that

offshore wind power generation within the Japanese Exclusive Economic Zone (EEZ) has a potential

to produce 4.86 × 104 TWh of electricity annually, which is equivalent to about 50 times the annual

electricity production of Japan. Although a facility for offshore wind power generation is yet to be

installed, a government-led feasibility study of offshore wind power generation commenced in 2008

(NEDO 2008), and Japan has just taken the first step toward offshore wind power generation in the

near future.

As development of offshore wind power generation facilities continues, the characteristics of

offshore winds have been well analyzed using meteorological masts in some European countries

(Barthelmie 1999; Barthelmie and Palutikof 1996; Barthelmie et al. 1996; Coelingh et al. 1998;

Lange et al. 2004). These studies showed that offshore winds are not only stronger than those

onshore due to the lower surface roughness over water, but also have different characteristics.

Barthelmie (1999) and Lange et al. (2004) showed that the offshore wind speed profile near the sea

surface is more strongly affected by atmospheric stability than onshore. Moreover Barthelmie and

Palutikof (1996) and Barthelmie et al. (1996) mentioned that the wind conditions at the coast, where

initial offshore wind farms are likely to be sited, are complicated due to abrupt changes in surface

roughness, temperature and humidity that lead to the development of the Internal Boundary Layer

(IBL).

1-2

Assessing the offshore wind energy potential using in situ meteorological masts is usually

complicated by financial and technical constraints, and conducting feasibility studies for offshore

wind farms is generally more difficult than for wind farms onshore (Sempreviva et al. 2008).

Onshore, wind analysis applications such as WAsP (Wind Atlas Analysis and Application Program)

(Mortensen et al. 2005) have been successful for assessing expected wind energy potential. However,

most applications developed for onshore site assessment do not fully take into account the effect of

atmospheric stability on the wind speed profile and the influence of the Internal Boundary Layer

(IBL) on the coastal wind climate. According to Lange et al. (2004), wind speed profiles

corresponding to the stability condition found near the sea surface cannot be simulated, since the

WAsP program assumes a constant surface roughness over water and a slightly stable atmosphere.

Moreover Lange et al. (2001) pointed out that the difference between the WAsP wind speed and

measurements tends to increase in the transition zone between land and sea because the WAsP

program do not adequately take into account the influence of the IBL.

The more inconvenient problem with the previous methods is that in situ measurements taken at

offshore observation stations are needed for the accurate offshore wind estimation. Consequently, to

enable reduction of technical and financial costs for site assessment, satellite and numerical

model-based offshore wind resource assessment methods, which are less dependent on

measurements from in situ met masts, have recently been developed (e. g. Badger et al. 2006; Berge

et al. 2009; Hasager et al. 2008; Hasager et al. 2005; Suselj and Sood 2010). Using objective

analysis as input, mesoscale models can estimate high spatial resolution wind conditions including

dynamical and thermo-dynamical effects without in situ measurements, and offshore winds with

higher temporal resolution in comparison to satellite-based wind estimations. Therefore, mesoscale

models have been used for assessing expected offshore wind energy in several waters to date (Bailey

and Freedman 2008; Dvorak et al. 2010; Jimenez et al. 2007; Lavagnini et al. 2006).

Initial offshore wind farms will be sited in shallow waters near the coast due to technical and

financial constraints (Barthelmie and Palutikof 1996). However, shallow waters, where the water

depth is less than 100 m, are limited in Japan to within only several kilometers of the coastline.

Offshore wind climates in the coastal waters in Japan are probably more complicated than those in

Northern Europe due to the dynamical and thermo-dynamical effects by complex topography on

land and stronger solar radiation in comparison to Northern European sites. Thus, mesoscale models

are expected to be a promising method for regional scale site assessment in Japanese coastal waters.

However since there are few in situ measurements taken offshore of Japan so far, the offshore winds

simulated by mesoscale models have hardly been compared with measurements except for a few

case studies (Ohsawa et al. 2007; Shimada et al. 2005; Yamaguchi and Isihara 2007). Thus, whether

the use of mesoscale models is a reliable method for offshore wind simulation or not, and the

mesoscale models can mitigate problems with the previous methods developed for wind resource

1-3

assessment on land has not previously been determined.

The author believes that the method of offshore wind resource assessment which is less dependent

on in situ measurements is needed not only for the reduction of measurement costs but also assessing

suitable sites for offshore wind farms in order to promote the development of offshore wind power

generation in Japan as in Northern Europe countries. Therefore, the main purpose of this thesis is to

quantitatively investigate the accuracy and characteristics of offshore winds simulated by a

mesoscale model using in situ measurements, which is seldom examined so far, and increase the

accuracy of simulated winds as much as possible. For this purpose, the author first analyzes the

characteristics of offshore winds using long-term in situ measurements taken at an offshore research

platform in the Japanese coastal waters. Secondly, an annual offshore wind simulation are performed

with the advanced research Weather Research and Forecasting (WRF-ARW) model (Skamarock et

al. 2008), which is a fully compressible, non-hydrostatic mesoscale model developed by the National

Center for Atmospheric Research (NCAR) and the National Center for Environmental Prediction

(NCEP), and the simulated winds are compared with the analyzed in situ measurements. Moreover, a

research for increasing the accuracy of the simulated winds is conducted, and the causes of the

remaining error are finally investigated.

1.2 Outline of this thesis

The outline of this study is shown in Fig. 1-1.

In order to understand the characteristics of offshore winds in the waters of interest in this study,

measurements taken at the Shirahama offshore research platform located in Tanabe Bay, Japan, are

analyzed in Chapter 2. The results presented in Chapter 2 are based on Shimada et al. (2009).

Then, the accuracy of wind speed simulated by WRF and causes of error are investigated from

comparison of the simulated wind speed with measurements from the Shirahama research platform

in Chapter 3. The results presented in Chapter 3 are based on a research by Shimada and Ohsawa

(2011).

The results obtained from Chapter 3 shows that the WRF wind speed includes a large positive bias,

and the positive bias causes a critical error for the wind energy estimation using WRF. Therefore, a

study to reduce the positive bias found in wind speed simulated by WRF at the Shirahama research

platform is carried out by means of improvement of the accuracy of Sea Surface Temperature (SST)

in Chapter 4. The results presented in Chapter 4 are based on a research by Shimada et al. (2011a).

The use of accurate SST data in the WRF simulation is partially successful to reduce the positive

bias in the WRF wind speed. But the positive bias cannot be entirely removed. Thus the

WRF-simulated wind speed profile is compared with Japan Meteorological Agency (JMA) wind

profiler measurements at ten stations in order to examine the causes of the positive bias in the WRF

wind speed apart from the use of inaccurate SST in Chapter 5. The results presented in Chapter 5 are

1-4

based on Shimada et al. (2011b).

Finally, conclusions obtained each chapter and summary are described in Chapter 6.

- Basic parameters representing characteristics of offshore winds at the target water

- Analyzed in situ measurements for validation

Fig. 1-1. Flow chart of this research

- Accuracy of the WRF wind speed offshore - Problem of the large positive bias in the WRF wind speed for wind

resource assessment

- New SST dataset developed for the reduction of the positive bias in the WRF wind speed

- Impact of the use of accurate SST on offshore wind simulation by WRF

- Characteristics of wind speed profiles by WRF - Possible cause of the positive bias in the WRF

Chapter 2: Characteristics of offshore winds at Shirahama Oceanographic Observatory (Shimada et al. 2009) Contents: Analyzing in situ offshore wind measurements at the Shirahama offshore platform

Chapter 3: Accuracy and characteristics of offshore wind speeds simulated by WRF (Shimada and Ohsawa 2011) Contents: Validation of the mesoscale model, WRF using in situ measurements

Chapter 4: Effects of Accuracy of Sea Surface Temperature on Offshore Wind Resource Assessment using a Mesoscale Model (Shimada et al. 2011a) Contents: A study for reducing the positive bias in the WRF wind speed by the use of accurate SST data

Chapter 5: Accuracy of the wind speed profile in the lower PBL as simulated by the WRF model (Shimada et al. 2011b) Contents: Examination of the cause of the positive except for the SST

Chapter 6: Conclusions and summary

1-5

References

Badger, J., R. J. Barthelmie, S. Frandsen, and M. B. Christiansen, 2006: Mesoscale modelling for an

offshore wind farm. Proc. of EWEC2006, Athens, Greece.

Bailey, B. H., and J. M. Freedman, 2008: A regional assessment of the US offshore wind energy

resource through the use of mesoscale modeling. Mar Technol Soc J, 42, 8-18.

Barthelmie, R. J., 1999: The effects of atmospheric stability on coastal wind climates. Meteorol Appl,

6, 39-47.

Barthelmie, R. J., and J. P. Palutikof, 1996: Coastal wind speed modelling for wind energy

applications. J Wind Eng Ind Aerod, 62, 213-236.

Barthelmie, R. J., M. S. Courtney, J. Hojstrup, and S. E. Larsen, 1996: Meteorological aspects of

offshore wind energy: Observations from the Vindeby wind farm. J Wind Eng Ind Aerod, 62,

191-211.

Berge, E., O. Byrkjedal, Y. Ydersbond, and D. Kindler, 2009: Modelling of offshore wind resources.

Comparison of a meso-scale model and measurements from FINO-1 and North Sea oil rigs. Proc.

of EWEC2009, Marseille, France.

Coelingh, J. P., A. J. M. van Wijk, and A. A. M. Holtslag, 1998: Analysis of wind speed observations

on the North Sea coast. J Wind Eng Ind Aerod, 73, 125-144.

Dvorak, M. J., C. L. Archer, and M. Z. Jacobson, 2010: California offshore wind energy potential.

Renew Energ, 35, 1244-1254.

EWEA, 2009: Oceans of Opportunity. http://www.ewea.org/offshore.

—— , 2011: The European offshore wind industry key trends and statistics 2010, Available at

http://www.ewea.org.

Hasager, C. B., and Coauthors, 2008: Remote Sensing Observation Used in Offshore Wind Energy.

Ieee J Sel Top Appl, 1, 67-79.

Hasager, C. B., and Coauthors, 2005: Offshore resource estimation from satellite SAR wind field

maps. Wind Energy, 8, 403-419.

Jimenez, B., F. Durante, B. Lange, T. Kreutzer, and J. Tambke, 2007: Offshore wind resource

assessment with WAsP and MM5: Comparative study for the German Bight. Wind Energy, 10,

121-134.

Lange, B., and J. Hojstrup, 2001: Evaluation of the wind-resource estimation program WAsP for

offshore applications. J Wind Eng Ind Aerod, 89, 271-291.

Lange, B., S. Larsen, J. Hojstrup, and R. Barthelmie, 2004: Importance of thermal effects and sea

surface roughness for offshore wind resource assessment. J Wind Eng Ind Aerod, 92, 959-988.

Lavagnini, A., A. M. Sempreviva, C. Transerici, C. Accadia, M. Casaioli, S. Mariani, and A. Speranza,

2006: Offshore wind climatology over the Mediterranean basin. Wind Energy, 9, 251-266.

1-6

Mortensen, N. G., D. N. Heathfield, L. Myllerup, L. Landberg, and O. Rathmann, 2005: Wind Atlas

Analysis and Application Program: WAsP 8 Help Facility. 335 Topics.

Nagai, H., I. Ushiyama, and Y. Ueno, 1997: Feasibility study of predicting offshore wind resources

and power generation in Japan. Proc. of Wind Energy Symposium, 168-171 (written in Japanese).

NEDO, 2008: Preliminary survey for the feasibility study of offshore wind generation.

https://app3.infoc.nedo.go.jp.

Ohsawa, T., M. Tanaka, S. Shimada, N. Tsubouchi, and K. Kozai, 2009: Assessment of Offshore Wind

Resources within Japan's EEZ using QuikSCAT Data. Journal of the Environmental Sciences, 18,

841-845.

Ohsawa, T., A. Hashimoto, S. Shimada, J. Yoshino, T. D. Paus, D. Heinemann, and B. Lange, 2007:

Evaluation of Offshore WindSimulations with MM5 in the Japanese and Danish Coastalwaters.

Proc. of EWEC 2007, Milan, Italy.

Sempreviva, A. M., R. J. Barthelmie, and S. C. Pryor, 2008: Review of Methodologies for Offshore

Wind Resource Assessment in European Seas. Surv Geophys, 29, 471-497.

Shimada, S., and T. Ohsawa, 2011: Accuracy and Characteristics of Offshore Wind Speeds Simulated

by WRF. Sola, 7, 21-24.

Shimada, S., T. Ohsawa, S. Serizawa, and I. Yoneda, 2009: Characteristics of offshore winds at

Shirahama Oceanographic Observatory. Proc. of ISOPE 2009, 424-428.

Shimada, S., T. Ohsawa, D. Heinemann, and G. Steinfeld, 2011a: Effects of Accuracy of Sea Surface

Temperature on Offshore Wind Resource Assessment using a Mesoscale Model. J Appl Meteorol

Clim, under review.

Shimada, S., T. Ohsawa, S. Chikaoka, and K. Kozai, 2011b: Accuracy of wind speed profile in the

lower PBL as simulated by the WRF model. Sola, 7, 109-112.

Shimada, S., T. Ohsawa, A. Hashimoto, K. Fukao, and T. Yasuda, 2005: Feasibility of offshore wind

power generation over Ise Bay. Wind Energy, 29, 92-97 (written in Japanese).

Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, and J. G. Powers,

2008: A description of theadvanced research WRF version 3. Tech. Note TN-475+STR, 1-96.

Suselj, K., and A. Sood, 2010: Improving the Mellor-Yamada-Janjic Parameterization for wind

conditions in the marine planetary boundary layer. Bound-Lay Meteorol, 136, 301-324.

Yamaguchi, A., and T. Isihara, 2007: An assessment of offshore wind energy potential using

mesoscale model and geographic information system. Journal of Wind Engineering, 32, 63-76

(written in Japanese).

3-1

Chapter 3

Accuracy and characteristics of offshore wind speeds simulated by WRF

Abstract

This chapter discusses the accuracy and characteristics of offshore wind speeds simulated by the

Weather Research and Forecasting (WRF) model. Firstly the accuracy of the simulated wind speed is

examined using in-situ measurements from the Shirahama offshore research platform for the whole

year of 2005. It is found that the WRF surface wind speed has an annual bias of 15.3%, which is

much larger than the National Centers for Environmental Prediction Final Analysis data (NCEP

FNL) used as input. Secondly, the possibility of accuracy improvement of the simulated wind speed

is discussed in terms of the decomposition of Root Mean Square Error (RMSE) and the estimation

error of wind energy density. As a result, it is shown that reducing the large positive bias is a key to

increasing the accuracy of the WRF wind speed. Without the large bias, the error in the estimation of

annual mean wind energy density drastically decreases from 47% to 4%. Finally, the accuracy of

wind speeds above the surface layer is examined using wind profiler measurements from Mihama. It

is found that the WRF wind speed has large positive biases not only near the sea surface but also in

the lower PBL. Thus, the large positive biases are speculated to be mainly due to inaccuracy of the

PBL scheme in WRF.

3.1 Introduction

Recently methodologies for the offshore wind resource assessment independent of in-situ

measurements have been enterprisingly researched in some Northern European countries as their

national projects, to reduce technical and financial costs of development of facilities for offshore

wind power generation. In these projects, the use of a mesoscale model is expected to be one of the

useful methodologies, and some related studies including not only validation of the model but also

development of suitable physical schemes for offshore wind simulation are already under way

(Jimenez et al. 2007; Suselj and Sood 2010).

In Japan as well, a government-led feasibility study on offshore wind power generation has started

since 2008, and the country has just taken the first step toward full-fledged installation of offshore

wind power generation in the future. For the installation it is crucially important to assess offshore

wind resources based on a reliable methodology. However, due to the lack of offshore in-situ

measurements that can be used for validation, there are still many unknowns about the reliability of

mesoscale models for offshore wind simulation. In particular, their performances for Japanese

3-2

coastal waters, surrounded by complex terrains and influenced by prevailing dynamical and

thermo-dynamical circulations, have hardly ever been discussed, except a few case studies (e.g.,

Ohsawa et al. 2007).

In this point of view, the offshore research platform of Shirahama Oceanographic Observatory,

located in Tanabe Bay in Wakayama Prefecture is one of the handful, precious stations that have

observed offshore winds for a long time. Thus this offshore research platform has been utilized as

the research base since 2007 (Shimada et al. 2009). In this chapter, the in-situ measurements from

the Shirahama research platform are used to investigate the reliability of a mesoscale model for

offshore wind simulation. The Weather Research and Forecasting (WRF) model is used as the

mesoscale model in this study.

Details of the simulation and validation method are described in Section 3.2. In Section 3.3,

one-year simulation is performed with WRF and the accuracy is examined using the in-situ

measurements from the Shirahama offshore research platform. The errors are analyzed through the

decomposition of Root-Mean-Square-Error (RMSE) and the estimation of wind energy density, and

then the possibility of improvement of the accuracy is discussed. Moreover, accuracy of the

simulated wind speeds above the surface layer is examined using the wind profiler measurements in

order to clarify the cause of errors in the simulated surface wind. Finally results of this study are

summarized in Section 3.4.

3.2 Data and methods

3.2.1 Configuration of WRF

The Weather Research and Forecasting (WRF) model version 3.0 (Skamarock et al. 2008) is

used in this study. WRF is a fully compressible, non-hydrostatic model with a large number of

physics options regarding cumulus parameterization, cloud microphysics, radiation, PBL

parameterization and land surface model. The WRF model also includes multiple nesting and

four-dimensional data assimilation (FDDA) options, which enable us to hindcast meteorological

conditions realistically.

The WRF simulation is performed for the whole year of 2005 with the 2-way nesting option for

three domains gradually focusing on the Shirahama area. The domains used in this simulation are

shown in Fig. 3-1 and model configuration is summarized in Table 3-1. Horizontal resolutions of the

three domains 1, 2 and 3 are 18 km, 6 km and 2 km respectively. Vertically 28 levels are configured

between the surface and the 50 hPa pressure level, and the vertical grid-size decreases towards the

surface. Heights of the lowest three levels are approximately 27 m, 94 m and 185 m, respectively.

The National Centers for Environmental Prediction Final Analysis (1° × 1°, 6-hourly) data (NCEP

FNL), which has 24 levels from surface to 10 hPa is used for the four-dimensional data assimilation

as well as initial and lateral boundary conditions. The detail of NCEP FNL is available at

3-3

http://dss.ucar.edu/datasets/ds083.2. For the lower boundary condition over water, the New

Generation Sea Surface Temperature for Open Ocean (NGSST-O, 0.05° × 0.05°, daily) provided by

Tohoku University (Sakaida et al. 2009) is used. The two horizontal wind components, temperature

and specific humidity simulated by WRF are nudged toward NCEP FNL with the FDDA option, and

the FDDA option is applied to the whole domain 1, the domain 2 and 3 excluding PBLs in this

simulation. Here, the PBL height in the simulation is diagnostically calculated from the vertical

gradient of turbulence kinetic energy. Primary physics options selected in this simulation are also

shown in Table 3-1.

Table 3-1. Configuration of WRF

PeriodStart: 00:00 UTC 1 Jan 2005End: 24:00 UTC 31 Dec 2005

Input dataNCEP FNL Analysis (6-houlry, 1° x 1°)NGSST-O (daily, 0.05° x 0.05°)

Nesting 2-way nesting

DomainDomain 1: 18 km (86 x 75 grids)Domain 2: 6 km (100 x 82 grids)Domain 3: 2 km (70 x 70 grids)

Vertical layer 28 levels (suface to 50 hPa)

Physics options

Dudhia short wave radiationRRTM long wave radiationWSM3 cloud micropysicsKain-Fritsch 2 cumulus parameterizationFive-layer soil modelMYJ PBL parameterization

FDDA option Enable excluding domain 2 and 3 PBL

Fig.3-1. Domains used in the simulation

Shirahma research platform

3-4

3.2.2 In-situ data used for evaluation of simulated wind speed

Accuracy of the surface wind speeds simulated by WRF is examined using in-situ

measurements from the offshore research platform of Shirahama Oceanographic Observatory,

Disaster Prevention Research Institute, Kyoto University. The research platform is located at the

mouth of Tanabe Bay (33°42’32”N, 135°19’58”E), 2 km away from the nearest coastline. The

research platform is 23 m in height, standing on the seabed at a depth of 10 m. 10-minute averaged

wind speed is measured by a propeller anemometer on the top of the research platform and the

measurements are used for the validation of WRF-simulated wind speeds. Additional details of the

research platform are given by Shimada et al. (2009). For the simulated wind speeds above the

surface layer, the accuracy is evaluated based on wind profiler measurements from the Japan

Meteorological Agency (JMA) observation station, Mihama, located approximately 40 km away

from the Shirahama offshore research platform. Further descriptions of the JMA wind profiler is

given by Ishihara et al. (2006). The location of the Shirahama offshore research platform and the

Mihama wind profiler station are shown in Fig. 3-2.

3.3 Results and discussion

3.3.1 Accuracy of simulated surface wind speed

Hereafter the accuracy and characteristics of wind speed simulated by WRF are examined from

a comparison with measurements, and in order to show the advantage of the use of a mesoscale

model for the offshore wind simulation than coarser resolution datasets, in parallel same analysis is

carried out with NCEP FNL, which is used as input for the WRF simulation.

Fig. 3-3 shows annual mean wind speed fields obtained from the NCEP FNL and WRF wind

speeds for the year 2005, and land grids represented in each data are also shown in Fig. 3-3. It is

Fig. 3-2. Location of the Shirahama offshore research platform and the Mihama wind profiler station

3-5

found that since NCEP FNL has a much coarser land-ocean distribution (1° × 1°) compared to WRF

(6 km × 6 km), the offshore wind speed field from NCEP FNL does not well correspond to indented

coastlines. In contrast, the WRF wind speed field obviously exhibits more realistic wind fields

including influences from complex topography and indented coastlines. Moreover it seems that the

NCEP FNL wind speeds are higher than the WRF wind speeds in open oceans.

In order to examine the accuracy of the NCEP FNL and WRF wind speeds, bias, RMSE and

correlation coefficient are calculated using the in-situ measurements from the Shirahama research

platform. The nearest grid point values from the research platform in NCEP FNL and WRF in

domain 3 are used for the validation. Since the NCEP FNL grid spacing is too coarse to specify the

nearest gird, the NCEP FNL data is firstly interpolated onto the same grids with WRF (2 km × 2 km)

using the WRF pre-processing system. Next the NCEP FNL wind speed is extrapolated to the

anemometer height of 23 m using the log profile assuming roughness length of 0.2 mm. Since wind

speed at the 10 m height is diagnostically calculated using the Monin-Obukhov similarity theory in

the WRF simulation, the WRF wind speed is interpolated into the 23 m height using upper and lower

grid values as shown in Eq. 3.1.

27 1023 23 10 10

27 10

ln( ) ln( )ln( ) ln( )

U UU z z U

z z

(3.1)

Here 23U shows the interpolated wind speed at the 23 m height, 10U and 27U show the WRF

grid point values at height of 10 and 27 m, and 10z and 27z show the 10 and 27 m heights,

respectively. Also in order to adjust temporal resolution between the NCEP FNL and WRF wind

speeds, only 6 hourly data is used for the validation.

3-6

Fig. 3-4 shows the monthly biases, RMSEs and correlation coefficients of the NCEP FNL and

WRF wind speeds against the measurements at Shirahama. Here, since the bias and RMSE generally

tend to increase with the mean wind speed, relative values to the measured mean wind speed are

mainly used in this study. The NCEP FNL and WRF wind speeds have annual biases of 0.7 and

15.3%, respectively. Compared to the NCEP FNL, the WRF wind speed is found to have a much

larger positive bias. In Fig. 3-4 it is moreover found that the WRF wind speed has a notable positive

bias every month. In contrast, annual correlation coefficients are 0.62 and 0.81 for NECP FNL and

WRF, respectively, and the WRF wind speed has much higher values than NCEP FNL. This implies

Fig. 3-3. Annual mean wind speeds obtained from (a) NCEP FNL (1° × 1°) and (b) WRF (6 km × 6 km) at the 23 m height in the year 2005. Black dots show land grids represented in each data, respectively

(a) NCEP FNL

(b) WRF

3-7

that WRF succeeds in reproducing local effects such as land-sea breezes by taking into account more

realistic land-ocean distribution in the model than NCEP FNL, and this leads to the high correlation

coefficients. As for RMSE, NCEP FNL and WRF have similar values (53.8 and 46.0%, respectively)

in terms of the annual mean.

Besides, hourly mean wind speeds obtained from measurements (black squares), NCEP FNL

(red triangles) and WRF (blue circles) at the Shirahama research platform in the year 2005 are

shown in Fig. 3-5. It is found that the hourly mean wind speed from measurements has an obvious

diurnal cycle. While the wind speed during nighttime shows an almost constant value of

approximately 5 m/s, it begins to increase at 9 LST, and then it reaches maximum at 14 LST. After

that the wind speed gradually decreases as time passes. On the other hand, the hourly mean wind

speed from WRF is found to well capture such diurnal cycle found in measurements compared to

NCEP FNL. However, the amplitude of diurnal cycle in the WRF wind speed is slightly smaller than

measurements. This is speculated to be attributable to the fact that the SST dataset used in the WRF

simulation does not take into account diurnal variations.

3-8

4.0

4.5

5.0

5.5

6.0

6.5

7.0

7.5

0 2 4 6 8 10 12 14 16 18 20 22

Hou

rly

mea

n w

ind

spee

d (m

/s)

Time (JST)

OBSNCEP FNLWRF

Fig. 3-5. Hourly mean wind speeds obtained from measurements (black), NCEP FNL (red) and WRF (blue) at the 23 m height at the Shirahama research platform in the year 2005

Fig. 3-4. Monthly bias, RMSE and correlation coefficient at Shirahama for NCEP FNL and WRF. The bias and RMSE are expressed as relative values to the monthly mean wind speed

-20

-10

0

10

20

30

40

50

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Annual

BIA

S (%

)

NCEP FNLWRF

2030405060708090

100

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Annual

RM

SE (%

)

NCEP FNLWRF

0.1

0.3

0.5

0.7

0.9

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Annual

Cor

rela

tion

coe

ffic

ient

(-) NCEP FNL WRF

3-9

3.3.2 Evaluation and improvement of wind speed errors

Three statistics (bias, RMSE and correlation coefficient) are examined in the previous section,

but it is noted that these three values are not independent to each other. According to Lange and

Focken (2006), RMSE can be decomposed into bias (BIAS), standard deviation bias (SDBIAS) and

dispersion (DISP). The decomposition of RMSE is given by

RMSE2=BIAS2+SDBIAS2+DISP2. (3.2)

Fig. 3-6 (a) shows the square values of RMSE, BIAS, SDBIAS and DISP calculated from the NCEP

FNL and the WRF wind speeds at Shirahama, and their contribution ratios to RMSE are shown in

Fig. 3-6 (b).

It is obvious in Fig. 3-6 that a large part of RMSE comes from DISP in the NCEP FNL and

WRF wind speeds. DISP is a function of correlation coefficient, and it generally tends to decrease as

correlation coefficient increases. In fact, the order of DISPs shown in Fig. 3-6 is completely opposite

to that of annual correlation coefficients shown in Fig. 3-4. Moreover, Fig. 3-6 (a) indicates that the

WRF wind speeds have significantly smaller DISP than the NCEP FNL. This difference between

WRF and NCEP FNL seems to be attributable to the improvement of reproducibility of local effects

due to the much finer land-ocean distribution in the WRF simulation compared to NCEP FNL. While

WRF exhibits smaller values in SDBIAS as well as DISP in Fig. 3-6 (a), it is only BIAS that is

larger in WRF than in NCEP FNL. After all, the results from Figs. 3-4 and 6 imply that improvement

of the large positive bias is a key to increasing the accuracy of the WRF wind speed at the research

platform.

This point of view can be also confirmed in terms of wind energy density (1/2 × air density × a

cube of wind speed), which is a more suitable parameter for wind resource assessment. Fig. 3-7

compares two kinds of errors in the mean wind energy density at Shirahama; One is calculated from

the raw WRF wind speeds and another is calculated from the WRF wind speeds after removing the

monthly biases. As found in Fig. 3-7, the wind energy density from the raw wind speed has an error

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

RMSE BIAS SDBIAS DISP

(m/s

)2

NCEP FNLWRF

BIASSDBIASDISP

(a)

(b)

Fig. 3-6. (a) Decomposition of RMSE in the NCEP FNL and WRF wind speeds at Shirahama, and (b) contribution ratio (%) of BIAS, SDBIAS and DISP to RMSE

0.06.4

94.9

NCEP FNL

11.21.0

87.9

WRF

3-10

of 47 % for the annual mean. In contrast, the wind energy density after removing the wind speed bias

has the error of 4 %. That is, the estimation error in wind energy density considerably decreases by

removing the annual bias of wind speed. In other words, reducing the bias, rather than reducing

RMSE and increasing correlation coefficient, is crucially important in terms of wind resource

assessment.

3.3.3 Accuracy of wind speeds above surface layer

Consequently, “where does the large positive bias of the WRF simulations come from?” is a

question that arises here. In order to investigate a cause of the bias, in this study, the wind profiler

measurements at Mihama, described in Section 3.2.2, are used to see if WRF has such a large

positive bias in and above the PBL as well as the surface layer. Fig. 3-8 is the result, showing the

annual bias, RMSE and correlation coefficient of the NCEP FNL and WRF wind speeds at heights

from 400 to 9,000 m. It is found that the accuracy of the NCEP FNL and WRF wind speeds

gradually become worse toward the surface. The differences between NCEP FNL and WRF become

pronounced under a height of 1,900 m, because in the WRF simulations the grid nudging is applied

for only above the PBL. The height of 1,900 m corresponds to the average height of the top of the

PBL simulated in the model.

As shown in Fig. 3-8, at the top of the PBL, NCEP FNL has an annual bias of -6.9 %, RMSE of

28.5 % and correlation coefficient of 0.87, while WRF has an annual bias of -6.0 %, RMSE of

26.6 % and correlation coefficient of 0.88. In the PBL, these statistics for WRF are all better than

those for NCEP FNL, indicating that the accuracy of wind speeds in the PBL is greatly improved by

using WRF, which has more realistic land-ocean distribution near coastlines. But, it is noteworthy

that the WRF wind speeds become to have a positive bias as the surface approaches, despite the

negative bias at the top of the PBL. The bias drastically increases under the height of 1,000 m, and

this tendency seems to lead to the large positive bias in the surface layer found at Shirahama. In

other words, the large positive bias is not limited in the surface layer but prevails in the lower PBL.

In our recent research (Shimada et al. 2011) shown in Chapter. 4, it is shown that the sea surface

temperature (SST) from NGSST-O, used for a lower boundary condition in the WRF simulation,

Fig. 3-7 Monthly and annual biases in wind energy densities at Shirahama calculated from WRF (blue bars) and from WRF after removing monthly biases of wind speed (yellow bars)

-20

0

20

40

60

80

100

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Annual

BIA

S (%

)

WRFWRF after removing bias

3-11

tends to have a positive bias (higher temperatures) in the area around Shirahama. This

overestimation of SST is then found to cause a positive bias in the simulated surface wind speed by

making stronger unstable conditions. The author thus made more accurate SST dataset based on the

MODIS satellite data to remove the positive bias of SST and applied it to the WRF simulation. This

attempt, however, could not completely remove the large positive bias of wind speed, and there still

remained an annual bias of 7.5 %.

Up to the present, the main cause of the large positive biases of wind found in the surface layer

and the lower PBL have not been identified, unfortunately. Thus, in order to improve the accuracy of

offshore wind speed in the lower PBL as well as surface wind speed, further studies are needed

especially for re-examining of the PBL scheme implemented in the WRF model．

3.4 Conclusions

In this chapter, in order to investigate the reliability of a mesoscale model for offshore wind

simulation, the accuracy and characteristics of offshore wind speeds simulated by the WRF model

are examined using in-situ measurements from the Shirahama offshore research platform and wind

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

-30.0 -20.0 -10.0 0.0 10.0 20.0 30.0

Hei

ght (

m)

Bias (%)

NCEP FNLWRF

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

0.0 10.0 20.0 30.0 40.0 50.0 60.0

Hei

ght (

m)

RMSE (%)

NCEP FNLWRF

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

0.6 0.7 0.8 0.9 1.0

Hei

ght (

m)

Correlation coefficient (-)

NCEP FNLWRF

Fig. 3-8. Annual bias, RMSE and correlation coefficient of the NCEP FNL and WRF wind speeds at heights from 400 to 9,000 m

3-12

profiler measurements from the adjacent JMA Mihama station. The main conclusions are

summarized as follows.

1. Compared with in-situ measurements at Shirahama, the WRF wind speeds have an annual

bias of 15.3%, RMSE of 46.0% and correlation coefficient of 0.81.

2. Decomposition of RMSE in the WRF wind speeds indicates that improvement of the large

positive bias is a key to increasing the accuracy of the WRF wind speed. In fact, if the bias

is removed, the accuracy of wind energy density estimation is greatly improved (from 47%

down to 4%).

3. Comparison with the wind profiler measurements shows that the bias in the WRF wind

speed is negative at the top of the PBL, but it becomes positive under a height of 1,000 m,

drastically increasing toward the surface. That is, the positive bias is not limited in the

surface layer but prevails in the lower PBL.

4. Although the positive bias is found to be partly due to overestimation of SST, the main

cause has not been identified in this chapter. Further studies are needed especially for

re-examining of the PBL scheme in WRF to improve the accuracy of offshore wind speed

estimation.

3-13

References

Ishihara, M., Y. Kato, T. Abo, K. Kobayashi, and Y. Izumikawa, 2006: Characteristics and

performance of the operational wind profiler network of the Japan Meteorological Agency. J

Meteorol Soc Jpn, 84, 1085-1096.

Jimenez, B., F. Durante, B. Lange, T. Kreutzer, and J. Tambke, 2007: Offshore wind resource

assessment with WAsP and MM5: Comparative study for the German Bight. Wind Energy, 10,

121-134.

Lange, M., and U. Focken, Eds., 2006: Physical Approach to Short-Term Wind Power Prediction. Vol.

1-208, Springer.

Ohsawa, T., A. Hashimoto, S. Shimada, J. Yoshino, T. D. Paus, D. Heinemann, and B. Lange, 2007:

Evaluation of Offshore WindSimulations with MM5 in the Japanese and Danish Coastalwaters.

Proc. of EWEC 2007, Milan, Italy.

Sakaida, F., H. Kawamura, S. Takahashi, T. Shimada, Y. Kawai, K. Hosoda, and L. Guan, 2009:

Research and Development of the New Generation Sea Surface Temperature for Open Ocean

(NGSST-O) Product and Its Demonstration Operation. J Oceanogr, 65, 859-870.

Shimada, S., T. Ohsawa, S. Serizawa, and I. Yoneda, 2009: Characteristics of offshore winds at

Shirahama Oceanographic Observatory. Proc. of ISOPE 2009, 424-428.

Shimada, S., T. Ohsawa, D. Heinemann, and G. Steinfeld, 2011: Effects of Accuracy of Sea Surface

Temperature on Offshore Wind Resource Assessment using a Mesoscale Model. J Appl Meteorol

Clim, under review.

Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, and J. G. Powers,

2008: A description of theadvanced research WRF version 3. Tech. Note TN-475+STR, 1-96.

Suselj, K., and A. Sood, 2010: Improving the Mellor-Yamada-Janjic Parameterization for wind

conditions in the marine planetary boundary layer. Bound-Lay Meteorol, 136, 301-324.

5-1

Chapter 5

Accuracy of the wind speed profile in the lower PBL as simulated by the WRF model Abstract

Wind resource assessment for coastal areas requires accurate wind speed simulation using a

mesoscale model. Author’s previous study found that the annual mean wind speed simulated by the

advanced research Weather Research and Forecasting (WRF-ARW) model has a remarkable positive

bias in the lower part of the planetary boundary layer (PBL). This result was obtained from a

comparison with wind profiler measurements at Mihama, which is an observation station of the

WInd profiler Network and Data Acquisition System (WINDAS) operated by Japan Meteorological

Agency (JMA). In this study, the author examines whether such a positive bias can be seen at other

WINDAS stations from comparisons of the WRF-simulated wind speed profile using the

Mellor-Yamada-Janjic (MYJ) PBL scheme with wind profiler measurements at ten WINDAS

stations. The results show that the positive bias is found at all stations, and, moreover, that the

positive bias is unlikely to be caused by either error in wind profiler measurement or the objective

analysis data input into WRF. In addition, the author compares the wind speed profiles simulated by

WRF with seven different PBL schemes for a month. The result shows that the positive bias cannot

be simply reduced by using other PBL schemes.

5.1 Introduction

Facilities for offshore wind power generation have been expected to proliferate in the future as

attempts are made to reduce greenhouse gas emissions associated with energy production (EWEA

2009). However, since conducting feasibility studies using in situ meteorological masts for offshore

wind farms is generally more difficult than for those onshore due to the financial and technical

constraints as mentioned before, satellite and numerical model-based methods have recently been

developed to assess offshore wind energy potential (Sempreviva et al. 2008).

Offshore wind resource assessment using a mesoscale model is a useful method for regional scale

site assessment. In order to evaluate the reliability of mesoscale models, the author (Shimada and

Ohsawa 2011) have examined the accuracy of wind speeds simulated by the advanced research

Weather Research and Forecasting (WRF-ARW) model (Skamarock et al. 2008) (hereafter WRF),

which is a fully compressible, non-hydrostatic mesoscale model developed by the National Center

for Atmospheric Research (NCAR) and the National Center for Environmental Prediction (NCEP).

Using the WRF model, Shimada and Ohsawa (2011) calculated annual mean wind speeds in the

coastal area around Shirahama in Tanabe Bay, Japan, and showed that the simulated surface wind

5-2

speed has a large positive bias at the Shirahama offshore research platform. They also showed that

the positive bias is found not only at the surface but also in the lower part of the planetary boundary

layer (PBL), from comparison with wind profiler measurements at Mihama, which is an observation

station of the WInd profiler Network and Data Acquisition System (WINDAS) operated by the Japan

Meteorological Agency (JMA)(Ishihara et al. 2006).

The accuracy of wind speed profiles simulated by mesoscale models in the PBL have been

examined using sonde measurements and remote sensing techniques in previous studies (e.g.

Hayashi et al. 2008; Hu et al. 2010; Zhang and Zheng 2004). Hayashi et al. (2008) examined the

accuracy of wind speed profiles simulated by the JMA non-hydrostatic model (NHM) and WRF

using sonde measurements at several stations in eastern Asia. Their results identify a positive bias

similar to that found by Shimada and Ohsawa (2011) in the wind speeds simulated by both JMA

NHM and WRF. On the other hand, Hu et al. (2010) compared the accuracy of wind speed profiles

simulated by WRF using three different PBL schemes. They showed that the wind speeds simulated

by WRF using the Mellor-Yamada-Janjic (MYJ) PBL scheme (Janjic 1994) overestimates wind

speeds compared to measurements near the surface for both stable and unstable conditions.

These previous studies indicate that wind speeds simulated by WRF possibly have a positive bias

near the surface. Since the accuracy of wind speeds simulated by mesoscale models depends

strongly on the validation conditions (e.g., location, period, boundary conditions and model settings),

there is considerably uncertainty about whether this positive bias always occurs in the WRF

simulations or only under specific conditions. Since wind energy is proportional to the cube of wind

speed, the bias in annual mean wind speed is a crucial parameter for offshore wind resource

assessment (Shimada and Ohsawa 2011) and it is thus worthwhile to focus on this positive bias to

see if a solution to eliminate it from the WRF wind simulation can be found. Therefore, this chapter

aims to examine whether the large positive bias is peculiar to Mihama or if it can be seen at other

measurement sites from WINDAS as well. Moreover, possible causes of the large positive bias are

also discussed.

5.2 Data and methods

5.2.1 WRF model

The simulation is performed using WRF-ARW version 3.3 for one year from January to December

2005 with one domain covering the west part of Japan, where most WINDAS stations are located

near the coast. The model configuration and domain used in the WRF simulation are shown in Fig.

5-1 and Table 5-1, respectively. Horizontally, the domain consists of 190 × 159 grid cells with a 3

km cell resolution, and 40 vertical levels are configured between the surface and the 50 hPa pressure

level. The vertical grid-size decreases towards the surface, and the heights of the lowest three levels

are about 12, 39 and 74 m, respectively. The JMA Meso Analysis (MANAL) data, which is 6-hourly,

5-3

10 km × 10 km, 20-level objective analysis data, is used for the four-dimensional data assimilation

(FDDA) as well as initial and lateral boundary conditions. Besides, soil temperature and moisture

data obtained from the NCEP Final Analysis (1° × 1°, 6-hourly, http://dss.ucar.edu/datasets/ds083.2)

(NCEP FNL) is used for the Noah land surface model. The FDDA is performed using the grid

nudging technique in the WRF model. In this simulation, the FDDA is applied to the levels above a

height of 2 000 m, where the two horizontal wind components, temperature and specific humidity

simulated by WRF are nudged toward MANAL. As for Sea Surface Temperature (SST), the daily

MODerate resolution Imaging Spectroradiometer (MODIS)-based SST composite with 0.02° × 0.02°

grid cells is used as the lower boundary condition. The primary physics options, selected by

reference to Suselj and Sood (2010), are also shown in Table 5-1. The MYJ PBL scheme, which is

widely used for wind resource assessment with mesoscale models (e.g. Hasager et al. 2011; Jimenez

et al. 2007; Suselj and Sood 2010), is at first selected for the validation. The performance of other

PBL schemes implemented in the WRF-ARW is then tested.

5-4

Fig. 5-1 Simulation domain and observation

Table 5-1. Model configurations

PeriodStart: 00:00 UTC 1st Jan 2005End: 24:00 UTC 31st Dec 2005

Input data

JMA MANAL(6-hourly, 10 km × 10 km)NCEP FNL (6-hourly, 1° × 1°)MODIS-based SST (daily, 0.02° × 0.02°)

Domain 3 km (190 × 159 grids)

Vertical layer 40 levels (suface to 50 hPa)

Physics options

Dudhia short wave radiationRRTM long wave radiationEta microphysicsBetts-Miller-Janjic cumulusparameterizationNoah land surface modelMYJ PBL parameterization

FDDA option Enabled above 2000 m height

5-5

5.2.2 Data used in the validation

Vertical profiles of wind speed simulated by WRF are compared with measurements from

WINDAS at ten stations. The WINDAS measures wind speed at the heights from 400 to 9 000 m

AGL at intervals of 300 m, based on the Doppler effect from electromagnetic radiation beams in five

directions. The ten-minute averaged measurements are used in this validation. WINDAS is presently

operated at 31 observation stations in Japan, and this study uses the data from ten stations: Nagoya,

Owase, Mihama, Tottori, Hamada, Takamatsu, Kochi, Shimizu, Oita and Nobeoka. The locations of

these stations are shown in Fig. 5-1. In addition, wind speeds measured from the JMA rawinsonde at

Yonago and Shionomisaki are also used in this validation.

5.3 Results and discussion

5.3.1 Wind speed bias at 10 wind profiler stations

Firstly, the author examines whether the large positive bias in the lower PBL found at Mihama in

our previous study (Shimada and Ohsawa 2011) is evident at other WINDAS stations. The hourly

model output, which is linearly interpolated to the wind profiler measurement height, is compared

with the ten-minute averaged WINDAS wind speed at each height. Here, the WRF-simulated wind

speed profile taken from the land grid cell and vertical height corresponding best to the location of

the WINDAS observation station was compared to the measurements. Although the surface level

height of each WRF grid was not necessarily identical to the observation station heights, since the

maximum difference of height between the WRF grids and the WINDAS stations is 45 m, the

correction for the difference is not performed in this validation.

Fig. 5-2 shows vertical profiles of annual mean wind speed from WINDAS and the bias in the

WRF wind speed at all stations. Here, the word “bias” means the value of an annual mean

“simulated” wind speed minus an annual mean “measured” wind speed. In order to reduce the

difference originating from on-site conditions, the relative value normalized by the measured wind

speed is frequently used for the evaluation of the accuracy of wind speed simulated by a numerical

model in wind resource assessment (Lange and Focken 2006). Thus relative values (%), defined as

the bias divided by the mean measured wind speed at each height, are shown as well as absolute

bias.

In Fig. 5-2, the absolute biases at most stations are found to have slightly negative values at levels

higher than 4 000 m, and the biases exhibit negative peaks at a height of approximately 2 500 m.

Between 1 300 m and 2 500 m, the bias changes the sign to positive and increases toward the surface

at all stations. At the lowest level (400 m), the positive bias at each station reaches up to +1.0 to +2.7

m/s (+12 to +83 % of mean measured wind speed). In short, the large positive bias in the lower PBL

can be clearly seen at not only Mihama but also at all nine other stations.

Since mesoscale model simulations with a few kilometers spatial grid resolution cannot entirely

5-6

reproduce undulations in actual topography, the simulated wind speeds near the surface can be

expected to have a large negative or positive bias due to the coarse topographic representation in the

model. Nevertheless, the biases at all ten stations in Fig. 5-2 exhibit positive values as well as results

shown in previous studies (Shimada and Ohsawa 2011). Thus these results suggest that the large

positive bias found in the WRF wind speed is more likely to be caused by systematic error in the

WRF model itself rather than the model settings. It is also found that the positive bias in the lower

PBL tends to be larger when the simulated wind speed has a more remarkable negative peak in bias

in the upper PBL. If this result is associated with an overestimation of vertical mixing within the

PBL, this would suggest the possibility of systematic error in the eddy viscosity coefficient in the

MYJ scheme.

5-7

As mentioned in Sec. 5-1, Hayashi et al. (2008) and Hu et al. (2010) also found a similar positive

bias in the WRF-simulated wind speed profile near the surface from a comparison with

measurements for a few summer months. This finding raises the question whether the positive bias

in the annual mean wind speed is mainly attributable to the bias in the summer season. Thus, the bias

in each month is examined. Fig. 5-3 shows vertical profiles of the monthly bias averaged for the ten

stations. From Fig. 5-3, the positive bias in the lower PBL can be seen in all months. At the lowest

level, the bias ranges from +1.4 to +2.3 m/s (from +32 to +46 % of monthly mean wind speed). Any

tendency that summer months particularly exhibit a large positive bias cannot be clearly identified

from Fig. 5-3.

Fig. 5-2 Vertical profiles of (a) measured mean wind speed, (b) absolute bias, and (c) relative bias in the WRF wind speed at ten wind profiler stations in the year 2005.

-100 -80 -60 -40 -20 0 20 40 60 80 1000

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [%]

Hei

ght [

m A

GL

]

(c) Relative bias

NagoyaOwaseTottoriHamadaMihamaOitaNobeokaTakamatsuKochiShimizu

0 5 10 15 20 25 30 350

1000

2000

3000

4000

5000

6000

7000

8000

9000

WPR wind speed [ms-1]

Hei

ght [

m A

GL

]

(a) Mean wind speed

Nagoya

Owase

Tottori

Hamada

Mihama

Oita

Nobeoka

Takamatsu

Kochi

Shimizu

-5 -4 -3 -2 -1 0 1 2 3 4 50

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [ms-1]H

eigh

t [m

AG

L]

(b) Absolute bias

NagoyaOwaseTottoriHamadaMihamaOitaNobeokaTakamatsuKochiShimizu

5-8

5.3.2 Possible causes of the positive bias

The JMA WINDAS measures wind speed in a unit of 1 m/s, which is comparable to the value of

the bias between the WRF-simulated and the measured annual mean wind speeds. Moreover, there

might be some influences from ground or artificial and terrestrial objects around the wind profiler,

which can affect the wind speed measurement based on remote sensing observation and lead to

positive bias. Thus next, the WRF wind speed profile is compared with the profile measured with a

rawinsonde, which is based on in situ observation. 12-hourly data for the year 2005 is used for the

comparison. Fig. 5-4 shows vertical profiles of the bias in the WRF wind speed at two rawinsonde

observation stations: Yonago and Shionomisaki. Both stations exhibit positive biases in the lower

PBL, similar to the bias found using WINDAS. This result indicates that the positive biases found in

Fig. 5-2 are not likely to be caused by some error in the wind profiler measurements.

Another possible cause of the positive bias is some influence from JMA MANAL, which is the

objective analysis data used for FDDA and initial and boundary conditions in the WRF simulation. If

MANAL originally has a positive bias, WRF could also have a positive bias. The author thus

examines the profiles from MANAL in the same way as Fig. 5-2. Vertical profiles of the bias in the

MANAL wind speed at ten stations are plotted in Fig. 5-5. Here, in order to easily handle the

MANAL data, 6-hourly MANAL wind speeds are horizontally and vertically interpolated onto the

WRF grid cells using the WRF pre-processing packages. In Fig. 5-5, the MANAL wind speed does

not have a large bias at most stations except for Owase and Kochi. Even at these two stations, the

absolute biases are relatively small and the relative biases do not increase simply with a decrease of

Fig. 5-3 Vertical profiles of (a) absolute and (b) relative bias averaged for ten stations for each month.

-5 -4 -3 -2 -1 0 1 2 3 4 50

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [ms-1]

Hei

ght [

m A

GL

]

(a) Absolute bias

JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDEC

-50 -40 -30 -20 -10 0 10 20 30 40 500

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [%]

Hei

ght [

m A

GL

]

(b) Relative bias

JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDEC

5-9

height, unlike the WRF biases. Since JMA operationally assimilates the wind profiler data to

MANAL, this result seems reasonable to some extent. Thus, the positive bias found in the WRF

wind speed is speculated not to be caused by the FDDA with MANAL.

Fig. 5-5 Vertical profiles of bias in MANAL wind speed at ten wind profiler stations in the year 2005.

-5 -4 -3 -2 -1 0 1 2 3 4 50

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [ms-1]

Hei

ght [

m A

GL

]

(a) Absolute bias

NagoyaOwaseTottoriHamadaMihamaOitaNobeokaTakamatsuKochiShimizu

-80 -60 -40 -20 0 20 40 60 800

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [%]

Hei

ght [

m A

GL

]

(b) Relative bias

NagoyaOwaseTottoriHamadaMihamaOitaNobeokaTakamatsuKochiShimizu

Fig. 5-4 Vertical profiles of bias in the WRF wind speed compared to rawinsonde observations at Yonago and Shionomisaki in the year 2005.

-5 -4 -3 -2 -1 0 1 2 3 4 50

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [ms-1]

Hei

ght [

m A

GL

]

(a) Absolute bias

YonagoShionomisaki

-50 -40 -30 -20 -10 0 10 20 30 40 500

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [%]

Hei

ght [

m A

GL

]

(b) Relative bias

YonagoShionomisaki

5-10

5.3.3 Effects of the PBL scheme on the vertical wind profiles

From a comparison of wind speed profiles simulated by WRF using three different PBL schemes,

Hu et al. (2010) reported that the positive bias is clearly found only when using the MYJ PBL

scheme. In contrast, from a comparison of wind profiles simulated by MM5 (Grell et al. 1994) using

different five PBL schemes, including the MYJ scheme, Zhang and Zheng (2004) concluded that the

simulated wind speed profile have little dependence on the PBL scheme except near the ground. In

fact, in Hayashi et al (2008), a positive bias can be seen in the result obtained from the WRF

simulation using the Yonsei University (YSU) PBL scheme. If the result shown in Hu et al. (2010)

can be entirely applied to the WRF simulation in Japanese coastal water areas, the positive bias

found in Fig. 5-2 could be removed by the use of another PBL scheme in WRF. In the latest version

of WRF-ARW (version 3.3 updated in April 2011), ten PBL schemes based on different PBL

parameterizations are available. The features of these PBL schemes are summarized in Table 5-2, by

reference to the WRF-ARW users’ guide (available at http://www.mmm.ucar.edu).

Thus, the author finally examine whether the positive bias can be reduced by the use of other PBL

schemes in the WRF model. These additional simulations use the model configuration identical to

that shown in Table 5-1 except for the PBL scheme and its associated surface layer scheme. Here,

the BouLac scheme developed for an urban canopy model and the MRF scheme, which is the older

version of the YSU scheme, are excluded from the validation. In addition, since the simulation using

the MYNN3 scheme did not run correctly due to unexpected error on our cluster system, the

MYNN3 scheme was also excluded from the validation. The seven simulations are performed for the

month of January 2005. Fig. 5-6 shows the comparison of the vertical bias profiles averaged for ten

stations from the WRF simulations with the YSU, MYJ, QNSE, MYNN2, ACM2, UW and TEMF

Table 5-2. PBL schemes available in WRF-ARW ver. 3.3.

SchemeTurbulenceclosure order

Summary

YSU 1.0Non-local-K scheme with explicit entrainment layer andparabolic K profile in unstable mixed layer

MRF 1.0 Older version of the YSU scheme

ACM2 1.0Asymmetric Convective Model with non-local upwardmixing and local downward mixing

BouLac 1.5Designed for use with BEP (Building EnvironmentParameterization) urban model

MYJ 1.5One-dimensional prognostic turbulent kinetic energyscheme with local vertical mixing

QNSE 1.5A TKE-prediction option that uses a new theory forstably stratified regions

TEMF 1.5Sub-grid total energy prognostic variable, plus mass-fluxtype shallow convection

UW 1.5TKE scheme from NCAR CESM (Community EarthSystem Model) climate model

MYNN2 1.5 Predicts sub-grid TKE termsMYNN3 2.0 Predicts TKE and other second-moment terms

5-11

PBL schemes. The positive bias in the lower PBL can be seen in the wind speed simulated with other

PBL schemes as well as the MYJ scheme. From a one-month comparison, it was found that the

positive bias in the WRF wind speed using the MYJ scheme cannot be simply removed by using

other PBL schemes.

In additional studies, the author has already examined the impacts of the accuracy of SST,

horizontal grid resolution and the surface roughness length on the accuracy of the WRF-simulated

wind profiles. However, the author has not yet found the evidence that these settings directly lead to

the positive bias in the lower PBL (figures not shown). At the moment, the author therefore

concludes that the positive bias is most likely caused by a deficiency in the WRF model itself, such

as the deficiency in the formulation of PBL or surface layer parameterization rather than the

accuracy of external data input into WRF and the model settings.

5.4 Conclusions

The accuracy of vertical wind speed profiles simulated by the WRF model was investigated in

comparison with wind profiler measurements taken from the JMA WINDAS. The main conclusions

obtained in this chapter are summarized as follows:

1. From the comparison with measurements at ten stations, the WRF- simulated wind speed

profile has positive biases in the lower part of PBL not only at Mihama, as shown in our

previous study, but also at all other stations.

2. Comparison with rawinsonde measurements at two stations also shows similar positive

biases in the lower PBL. It is thus unlikely that the positive bias is due to error in the wind

Fig. 5-6 Vertical profiles of bias averaged for ten stations from the WRF simulations with seven PBL schemes for January, 2005.

-5 -4 -3 -2 -1 0 1 2 3 4 50

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [ms-1]

Hei

ght [

m A

GL

]

(a) Absolute bias

YSUMYJQNSEMYNN2ACM2UWTEMF

-60 -40 -20 0 20 40 600

1000

2000

3000

4000

5000

6000

7000

8000

9000

Bias [%]

Hei

ght [

m A

GL

]

(b) Relative bias

YSUMYJQNSEMYNN2ACM2UWTEMF

5-12

profiler measurements.

3. It is moreover found that JMA MANAL, used as input for WRF, has much smaller biases

in the lower PBL at most stations than the WRF wind speed, and therefore the positive bias

is likely not caused by the four dimensional data assimilation with MANAL.

4. From the comparison of vertical profiles of monthly bias at ten stations, the bias in the

lower PBL is found in every month through the year.

5. From the comparison of vertical profiles of monthly bias from the WRF simulations with

different PBL schemes, the positive bias can be seen in the WRF wind speed using other

PBL schemes as well as the MYJ scheme.

In summary, the causes of the positive bias are not clearly identified in this study. However, the

above facts suggest that the positive bias is attributed to the WRF model itself and is recognized as a

systematic error in the model. The author therefore suggests that further study is required for

identifying the cause of and removing the positive bias in the WRF wind simulation, because this

bias in wind speed can greatly affect the accuracy of wind resource assessment.

5-13

References EWEA, 2009: Oceans of Opportunity. http://www.ewea.org/offshore.

Grell, G. A., Dudhia J., and D. R. Stauffer, 1994: A description of the fifth-generation Penn

State/NCAR mesoscale model (MM5). NCAR Technical Note, NCAR/TN-398+STR, 117 pp.

Hasager, C. B., and Coauthors, 2011: Combining satellite wind maps and mesoscale modelling for a

wind atlas of the South Baltic Sea. Proc. of EWEC 2011, Brussels, Belgium.

Hayashi, S., K. Aranami, and K. Saito, 2008: Statistical Verification of Short Term NWP by NHM and

WRF-ARW with 20 km Horizontal Resolution around Japan and Southeast Asia. Sola, 4, 133-136.

Hu, X. M., J. W. Nielsen-Gammon, and F. Q. Zhang, 2010: Evaluation of Three Planetary Boundary

Layer Schemes in the WRF Model. J Appl Meteorol Clim, 49, 1831-1844.

Ishihara, M., Y. Kato, T. Abo, K. Kobayashi, and Y. Izumikawa, 2006: Characteristics and

performance of the operational wind profiler network of the Japan Meteorological Agency. J

Meteorol Soc Jpn, 84, 1085-1096.

Janjic, Z. I., 1994: The Step-Mountain Eta Coordinate Model - Further Developments of the

Convection, Viscous Sublayer, and Turbulence Closure Schemes. Mon Weather Rev, 122, 927-945.

Jimenez, B., F. Durante, B. Lange, T. Kreutzer, and J. Tambke, 2007: Offshore wind resource

assessment with WAsP and MM5: Comparative study for the German Bight. Wind Energy, 10,

121-134.

Lange, M., and U. Focken, 2006: Physical approach to short-term wind power prediction. Springer,

xii, 208 pp.

Sempreviva, A. M., R. J. Barthelmie, and S. C. Pryor, 2008: Review of Methodologies for Offshore

Wind Resource Assessment in European Seas. Surv Geophys, 29, 471-497.

Shimada, S., and T. Ohsawa, 2011: Accuracy and Characteristics of Offshore Wind Speeds Simulated

by WRF. Sola, 7, 21-24.

Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, and J. G. Powers,

2008: A description of theadvanced research WRF version 3. Tech. Note TN-475+STR, 1-96.

Suselj, K., and A. Sood, 2010: Improving the Mellor-Yamada-Janjic Parameterization for wind

conditions in the marine planetary boundary layer. Bound-Lay Meteorol, 136, 301-324.

Zhang, D. L., and W. Z. Zheng, 2004: Diurnal cycles of surface winds and temperatures as simulated

by five boundary layer parameterizations. J Appl Meteorol, 43, 157-169.

6-1

Chapter 6

Conclusions and summary 6.1 Conclusions obtained in each chapter

In this Ph.d thesis, for the purpose of accurate offshore wind resource assessment, the accuracy

and characteristics of offshore winds simulated by the mesoscale model WRF (Weather Research

and Forecasting) model were examined using in situ measurements taken from offshore research

platforms in Japanese coastal waters. Moreover, an attempt at increasing the accuracy of the

WRF-simulated wind speed was also established in this study. General conclusions obtained in each

chapter are summarized as follows.

In Chapter 1, the background and purpose of this Ph.d study were stated.

In Chapter 2, long-term offshore wind measurements taken at the Shirahama offshore research

platform, which is located 2 km off the nearest coastline of Tanabe Bay, Japan, were analyzed for the

purpose of understanding the characteristics of offshore winds in the waters of interest. Since the

characteristics of offshore winds based on offshore wind measurements have hardly been reported to

date, the obtained data shown in Chapter 2 are valuable in themselves as a reference for discussing

the characteristics of offshore winds in Japanese coastal waters. Moreover, the measurements

analyzed in this chapter are widely able to be used not only for the examination of the accuracy and

characteristics of WRF in the following chapters in this study, but also for the examination of the

accuracy of synthetic aperture radar (SAR)-estimated wind speeds by our research group.

In Chapter 3, the accuracy and characteristics of wind speed simulated by WRF were examined

using in situ measurements obtained at the Shirahama offshore research platform. Annual WRF

simulations on a 2 km resolution grid were performed, and the three statistics of bias, Root Mean

Square Error (RMSE) and correlation coefficients were calculated in comparison with in situ wind

speeds. As results, the wind speeds simulated by WRF were found to have a large positive bias of

more than +10 % of the measured wind speed and a RMSE of 40 %. Moreover, decomposition of

RMSE and the investigation of accuracy of wind energy density estimated by the WRF wind speed

indicated that improvement of the large positive bias is a key for the accurate offshore wind resource

assessment using mesoscale model.

In Chapter 4, the effects of the Sea Surface Temperature dataset used as a lower boundary

condition in mesoscale models on the simulated wind speed were investigated in order to increase

the accuracy of the WRF wind speed. Firstly, NGSST-O (New Generation Sea Surface Temperature

for Open Ocean), used in the previous mesoscale model simulations, was found to have a positive

bias in SST in coastal waters, and the positive bias in NGSST-O was speculated to be a factor

contributing to the positive bias in the WRF wind speed. In order to reduce the bias associated with

6-2

the inaccuracy of NGSST-O in the WRF wind speed, a new high spatial resolution SST dataset based

on the MODIS (MODerate resolution Imaging Spectroradiometer) product was developed. The

MODIS-based SST developed in this study was confirmed not to have unreasonable errors in coastal

waters as in NGSST-O. Secondly, annual WRF simulations were performed with both NGSST-O and

the MODIS-based SST and the effect of accuracy of SST on the WRF wind speed was examined

with measurements from five offshore platforms. As a result, the use of the MODIS-based SST was

found to improve the bias on average by 9. 3% and RMSE by 4.3 % in the WRF wind speed.

However, the positive bias in the WRF wind speed was found not to be entirely improved by the use

of the MODIS-based SST instead of NGSST-O.

In Chapter 5, the accuracy and characteristics of the WRF-simulated wind speed profiles was

examined in comparison to JMA (Japan Meteorological Agency) wind profiler measurements to

examine causes of the positive bias in the WRF wind speed apart from the use of inaccurate SST.

Comparisons of the WRF-simulated wind speed profile with measurements from ten JMA wind

profiler stations showed that the large positive bias found in the WRF wind speed near the sea

surface can be also seen in the lower part of the PBL at all stations, and the bias tends to increase

towards the surface. Moreover, the positive bias is unlikely to be caused by either error in the wind

profiler measurements or the objective analysis data input into WRF. In addition, the author

compared the wind speed profiles simulated by WRF with seven different PBL schemes for a month.

The results show that the positive bias cannot be reduced simply by using alternative PBL schemes.

Consequently, these results indicate that the positive bias is most likely caused by a deficiency in the

WRF model, such as an overestimation of the eddy viscosity coefficient or non-local mixing in the

PBL scheme.

6.2 Summary

Offshore wind resource assessment using a mesoscale model is expected as a new method which

is less dependent on in situ measurements taken from offshore meteorological masts. However, since

mesoscale meteorological simulations with a high spatial resolution need much computational

resource, and moreover measurements taken offshore is very few, the accuracy and characteristics of

offshore winds simulated by mesoscale models were yet to be examined except for a few case

studies. Hence it was still uncertain whether mesoscale models are truly an effective application for

offshore wind resource assessment or not.

In this study, annual WRF simulations on a 2 km grid resolution were performed with parallelized

computer cluster systems, and the accuracy and characteristics of the WRF wind speed were

examined using valuable measurements taken at an offshore research platform in Japanese coastal

waters.

A comparison of the WRF-simulated offshore wind speed with winds obtained from a regional

6-3

scale objective analysis showed that the WRF wind speed can reproduce more realistic wind

distributions along the indented coastlines in Japanese coastal waters, compared to the objective

analysis, and moreover it can take into account the diurnal cycle of wind speed induced by a local

circulation. Consequently, it was found that the WRF model can be expected as a promising method

for offshore wind resource assessment.

On the other hand, a comparison with in situ measurements exhibited that the WRF wind speed

near the sea surface in Japanese coastal waters has an annual bias of more than +10% to the annual

measured wind speed. Moreover it was also found that if this over-predicted WRF wind speed is

used for calculating wind energy density, the large positive bias in the WRF wind speed will lead a

bias of more than +45% in the estimated wind energy density. These results indicate that the

improvement of accuracy of simulated wind speed is necessary for evaluating the expected wind

energy with the WRF wind speed directly.

Through the validation of the WRF wind speed at an offshore research platform using in situ

measurements, the SST dataset used in our previous mesoscale model simulations was found to have

an unreasonable error in coastal waters. At the same time, from a review of previous works, it was

found that offshore winds near the sea surface are strongly affected by atmospheric stability which is

mainly determined by the temperature difference between air and seawater. Thus the author

developed a more accurate SST dataset with a higher spatial resolution in coastal waters. Moreover it

was found that one third of the positive bias in the WRF wind speed can be reduced by the use of the

developed accurate SST.

The WRF simulation with a more accurate SST was partly successful to reduce the positive bias in

the simulated wind speed near the sea surface, but the positive bias was not entirely removed by the

use of the accurate SST. In order to examine the causes of the remaining positive bias, the accuracy

and characteristics of the WRF-simulated wind speed profiles in the lower PBL was investigated in

comparison with wind profiler measurement from JMA WINDAS. It was found that the large

positive bias in the WRF wind speed can be seen not only near the sea surface but also the whole of

lower PBL at several wind profiler measurement stations. This examination showed that the positive

bias in the WRF wind speed is most likely caused by a discrepancy of the surface and PBL

parameterizations in the WRF model than the problems of external data input into the WRF model

and model configurations.

The major academic findings obtained from this research are finally summarized as follows.

i. Accuracy and characteristics of the WRF wind speed offshore

ii. Impact of accuracy of SST on the WRF wind speed in coastal waters

iii. Feature of the WRF-simulated wind speed profiles

The validation of simulated wind speed with measurements will be an informative index for those

who are conducting offshore wind resource assessment using mesoscale models such as wind

6-4

analysis engineers. Regarding the impact of SST on the simulated wind speed, the result that the

accuracy of SST plays an important role to coastal wind climates is an interesting result from a point

of view of not only wind power engineering, but also atmospheric science. Moreover the feature of

the WRF-simulated wind speed profile in the lower PBL exhibits a possibility of discrepancy in the

WRF model, and it shows a hint that will lead to a considerable improvement of the accuracy of

offshore winds simulated by the WRF model.

Unfortunately, there are still unknowns about the causes of the positive bias in the WRF wind

speed, and room for improvement in the accuracy also remains so far. On the other hand, the need

for the accurate offshore wind simulation increases with expansion of facilities of offshore wind

power generation. Thus, further studies are necessary to pinpoint the causes of the positive bias in

the WRF model and improve the accuracy of the WRF wind speed. The author has recently prepared

for the development of a PBL scheme, which enables us to offshore wind speeds without

unreasonable errors. As a first step towards re-examination of the PBL scheme in the WRF model in

order to select the most suitable PBL scheme implemented in the latest version WRF, the

performance of each PBL scheme has been already compared using measurements from an offshore

meteorological mast in the North Sea. After comparison of PBL schemes, a modification of the PBL

scheme will be performed to achieve more accurate offshore wind simulations using WRF.

7-1

List of abbreviations

ACM2 Asymmetric Convective Model, version 2

AMSR-E Advanced Microwave Scanning Radiometer for Earth Observing System

AVHRR Advanced Very High Resolution Radiometer

BouLac Bougeault and Lacarrere

DISP DISPersion

EEZ Exclusive Economic Zone

EWEA European Wind Energy Association

FDDA Four Dimensional Data Assimilation

IBL Internal Boundary Layer

IEC International Electrotechnical Commision

JAXA Japan Aerospace Exploration Agency

JMA Japan Meteorological Agency

MANAL JMA Meso Analysis

MLIT Ministry of Land, Infrastructure, Transport and Tourism of Japan

MM5 Mesoscale Model fifth generation

MODIS MODerate resolution Imaging Spectroradiometer

MRF Medium Range Forecast

MYJ Mellor-Yamada-Janjic

MYNN2 Mellor-Yamada-Nakanishi-Niino Level 2.5

MYNN3 Mellor-Yamada-Nakanishi-Niino Level 3

NCAR National Center for Atmospheric Research

NCEP National Center for Environmental Prediction

NCEP FNL National Center for Environmental Prediction Final Analysis data

NEDO New Energy and Industrial Technology Development Organization

NGSST-O New Generation Sea Surface Temperature for Open Ocean

PBL Planetary Boundary Layer

QNSE Quasi-Normal Scale Elimination

RMSE Root Mean Square Error

SDBAIS Standard Deviation bias

SST Sea Surface Temperature

TEMF Total Energy Mass Flux

UW University of Wisconsin

WAsP Wind Atlas Analysis and Application Program

7-2

WINDAS WInd profiler Network and Data Acquisition System

WRF SCM WRF Single Column Model

WRF-ARW Advance Research Weather Research and Forecasting

YSU Yonsei University

7-3

Curriculum Vitae

Aug. 22, 2011

PERSONAL DATA

Name: Susumu SHIMADA

Place and date of birth: Ishikawa Prefecture, December 15, 1980

Nationality: Japan

EDUCATION

April 2009 - March 2011 Ph. D. Eng. in Graduate school of Maritime Sciences, Kobe

University in Japan

April 2003 - March 2005 M. Eng. in Environmental and Renewable Energy System, Gifu

University in Japan

April 1999 - March 2003 B. Eng. in Civil Engineering, Gifu University in Japan

WORKING EXPERIENCE

April 2011 - present Research Associate in Graduate School of Engineering, Gifu

University in Japan

April 2010 - March 2011 Research Fellow of the Japan Society for the Promotion of Science

April 2005 - May 2009 Newjec Inc.

SUMMER SCHOOLS

June 2009 Ph. D. Summer School: Remote sensing for wind energy (Risø DTU, Denmark)

PROFICIENCY OF FOREIGN LANGUAGES

Japanese: Mother tongue

English: good

CUMPUTER SKILLS

Numerical models: - Mesoscale model fifth-generation (MM5)

- Weather Research and Forecasting (WRF)

Programming Languages - Fortan

- Matlab

- Unix Shell programming

Operating systems - UNIX/Linux

- Windows

7-4

PUBLICATIONS

Proceedings on conference

1. Kozai, K., T. Ohsawa, S. Shimada, Y. Takeyama, C. Hasager, and M. Badger, 2009: Comparison of Envisat/ASAR-estimated Offshore Wind Resource Maps around Shirahama with those from Mesoscale Models MM5 and WRF. Proc. of EOW 2009, PO. 131, Stockholm, Sweden.

2. Kozai, K., T. Ohsawa, Y. Takeyama, S. Shimada, R. Niwa, C. Hasager, and M. Badger, 2010: Comparison of SAR Wind Speed Retrieval Algorithms for Evaluating Offshore Wind Energy Resources. Proc. of Techno Ocean 2010, Kobe, Japan, 2010.

3. Ohsawa, T., S. Shimada, J. Tambke, and B. Lange, 2009: A study on effective usage of mesoscale model for accurate offshore wind simulation. Proc. of EOW 2009, PO. 11, Stockholm, Sweden.

4. Ohsawa, T., S. Shimada, N. Tsubouchi, and K. Kozai, 2009: Offshore Wind Resource Assessment in Japanese Coastal Waters. Proc. of EWEC 2009, PO. 65, Mareseille, France.

5. Ohsawa, T., N. Tsubouchi, S. Shimada, and K. Kozai, 2010: Impact of atmospheric stability on QuikSCAT wind speeds in Northeast Asian Region. Proc. of Techno Ocean 2010, Kobe, Japan, 2010.

6. Ohsawa, T., M. Tanaka, S. Shimada, N. Tsubouchi, and K. Kozai, 2009: Assessment of Offshore Wind Resources within Japan's EEZ using QuikSCAT Data. Proc. of World Wind Energy Conference 2009, S25, Jeju Is., South Korea.

7. Ohsawa, T., A. Hashimoto, S. Shimada, J. Yoshino, T. D. Paus, D. Heinemann, and B. Lange, 2007: Evaluation of Offshore WindSimulations with MM5 in the Japanese and Danish Coastalwaters. Proc. of EWEC 2007, BL3. 103, Milan, Italy.

8. Shimada, S., and T. Ohsawa, 2009: Investigation of causes of inaccurate wind speeds in WRF simulation for an offshore site in Japan. Proc. of EOW 2009, PO. 131, Stockholm, Sweden.

9. Shimada, S., T. Ohsawa, and K. Yatsu, 2009: A study on the ability of mesoscale model MM5 for offshore wind resource assessment in Japanese coastal waters. Proc. of EWEC 2009, PO. 62, Marseille, France.

10. 香西克俊, 大澤輝夫, 嶋田進, 竹山優子, C. Hasager, M. Badger, 2009: 白浜周辺海域におけるメソスケールモデルWRF及び現場観測によるEnvisat/ASAR洋上風力資源パラメータの検証. 日本リモートセンシング学会 第47回学術講演会論文集, A. 10.

11. 大澤輝夫, 壷内伸樹, 嶋田進, 香西克俊, 2008: 日本周辺海域の洋上風況マップに関する研究. 第30回風力エネルギー利用シンポジウム予稿集, 30, 225-228.

12. 壷内伸樹, 大澤輝夫, 嶋田進, 香西克俊, 2010: QuikSCAT海上風データの大気安定度補正と推定精度の改善評価. 第32回風力エネルギー利用シンポジウム予稿集, 355-358.

13. 嶋田進, 大澤輝夫, 武藤裕則, 鈴木崇之, 久保輝広, 2009: 白浜海洋鉄塔におけるメソ気象モデルWRFの海上風況計算精度. 第31回風力エネルギー利用シンポジウム予稿集, 31, 153-156.

14. 嶋田進, 大澤輝夫, 芹澤重厚, 米田格, 2008: 白浜海象観測所における洋上風況特性について. 第30回風力エネルギー利用シンポジウム予稿集, 30, 233-236.

15. 壷内伸樹, 嶋田進, 大澤輝夫, 香西克俊, 2010: QuikSCAT衛星データに基づく洋上風況マップの作成及び精度検証. 海洋気象学会・日本気象学会関西支部近畿地区2009年度例会要旨集, 35-38.

7-5

16. 丹羽亮介, 嶋田進, 大澤輝夫, 香西克俊, 2010: 合成開口レーダー画像を用いた沿岸海域におけるWRFの海上風精度検証. 海洋気象学会・日本気象学会関西支部近畿地区2009年度例会要旨集, 31-34.

17. 谷津健, 嶋田進, 大澤輝夫, 芹澤重厚, 米田格, 2009: メソ気象モデルによる海上風計算精度向上に関する検討. 海洋気象学会・日本気象学会関西支部近畿地区2008年度例会要旨集, 39-42.

Articles in peer-reviewed journals

1. Ohsawa, T., M. Tanaka, S. Shimada, N. Tsubouchi, and K. Kozai, 2009: Assessment of Offshore Wind Resources within Japan's EEZ using QuikSCAT Data. Journal of the Environmental Sciences, 18, 841-845.

2. Shimada, S., and T. Ohsawa, 2011: Accuracy and Characteristics of Offshore Wind Speeds Simulated by WRF. Sola, 7, 21-24.

3. Shimada, S., T. Ohsawa, S. Serizawa, and I. Yoneda, 2009: Characteristics of offshore winds at Shirahama Oceanographic Observatory. Proc. of ISOPE 2009, 424-428.

4. Shimada, S., T. Ohsawa, S. Chikaoka, and K. Kozai, 2011: Accuracy of the wind speed profile in the lower PBL as simulated by the WRF model. Sola, 7, 109-112.

5. Shimada, S., T. Ohsawa, D. Heinemann, and G. Steinfeld, 2011: Effects of Accuracy of Sea Surface Temperature on Offshore Wind Resource Assessment using a Mesoscale Model. J Appl Meteorol Clim, under review.

6. 大澤輝夫, 壷内伸樹, 嶋田進, 香西克俊, 2009: 日本周辺海域の洋上風況マップに関する研究. 風力エネルギー, 33, 92-97.

7. 嶋田進, 大澤輝夫, 橋本篤, 深尾一仁, 安田孝志, 2005: 伊勢湾における洋上風力発電の可能性に関する検討. 風力エネルギー, 29, 92-97.

8. 丹羽亮介, 大澤輝夫, 嶋田進, 香西克俊, 竹山優子, 2010: 合成開口レーダー画像を用いたメソ気象モデルWRFによる沿岸海上風速分布の検証. 第21回風工学シンポジウム論文集, 203-208.