It is a technical field of weather forecasting that uses computers to numerically solve dynamic equations, which are mathematical expressions of the laws of fluid mechanics and thermodynamics that govern the movement and changes in the state of the atmosphere, and dynamically predict future atmospheric conditions. The original translation is numerical weather prediction, but in Japan, numerical forecasting is the official term. [Hiroshi Matano] The emergence of numerical predictionIn principle, it is possible to forecast weather mechanically by solving the dynamic equations applied to the atmosphere under given boundary and initial conditions. However, the atmosphere is not a simple fluid like water, but a thermally sensitive gas that moves relative to the rotating Earth. Moreover, reflecting the complex nature of the fluid, including water vapor that undergoes phase changes, the mathematical equations are also complex, and the formulas are in the form of nonlinear partial differential equations, making it almost impossible to solve them mathematically. However, after World War II, the development of computers and the advancement of numerical calculation technology made it possible to numerically solve the dynamic equations applied to the atmosphere, and weather forecasting using dynamic methods reached a practical level. This is numerical forecasting, the most brilliant application of meteorological dynamics. In Japan, numerical forecasting was introduced to the Japan Meteorological Agency in March 1959, and from the late 1970s, traditional forecasting work also changed into a new form centered on numerical forecasting. [Hiroshi Matano] Principles of Numerical Weather ForecastingThe dynamic equations used in numerical forecasting consist of the equation of motion (laws of motion), equation of state (gas laws), continuity equation (law of conservation of mass), thermodynamic equation (law of conservation of heat), and water vapor equation (law of conservation of water vapor), but differential equations are converted to difference equations for actual numerical calculations. In numerical forecasting, the atmosphere is divided into many layers in the vertical direction, and each layer is covered with grid points at regular intervals like a checkerboard. The time changes in meteorological elements such as wind, temperature, pressure, and humidity are calculated for each grid point, and future grid point values are obtained by time integration (grid point method). However, as the performance of computers improved, a method was developed in which the spatial distribution of meteorological elements, which are dependent variables of the equations, is expanded with harmonic or trigonometric functions, and the time changes in the coefficients of the functions are calculated (spectral method), and this method was introduced in Japan in the 1980s. The spectral method has the advantage that it is not necessary to calculate the time changes in meteorological elements for each grid point, and the calculation results are output as grid point values. Of course, it is also output as a weather forecast in the form of a weather chart. [Hiroshi Matano] Target of Numerical PredictionPhenomena that can be identified on a weather chart, such as low and high pressure systems that are important for weather forecasting, are called synoptic-scale phenomena. Since the main dynamical properties of synoptic-scale phenomena were clarified after the Second World War, numerical forecasting for the next day to the next day (48- or 72-hour forecasts) was largely completed in the 1980s. However, for small- to medium-scale phenomena smaller than synoptic scale and forecasts of 10 days or more, numerical forecasting is still far from being completed even in the 21st century. The main reasons for this are (1) the dynamical properties of phenomena smaller than synoptic scale have not been fully clarified to be able to handle them in numerical forecasting, and (2) meteorological noise (hereafter simply referred to as noise), which is a high-frequency phenomenon such as sound waves and small-scale, short-lived phenomena such as street whirlwinds, develop unnaturally during the numerical calculation process and disrupt the appropriate spatial distribution of meteorological elements. For this reason, suppression measures are built into the calculation process to prevent the noise from growing, but as the forecast period becomes longer, the noise gradually grows beyond the suppression measures. This is the unavoidable fate of numerical calculation (numerical integration). [Hiroshi Matano] Constraints on Numerical PredictionPhenomena smaller than the grid spacing are not the direct target of numerical forecasts, but for meteorologically important phenomena, their collective effects on synoptic-scale phenomena are indirectly incorporated into the calculation process. To be the direct target of numerical forecasts, a scale must include at least five grid points. Therefore, narrowing the grid spacing can directly target medium-scale phenomena (meso-scale phenomena) close to the synoptic scale, improving the accuracy of numerical forecasts. However, halving the grid spacing increases the amount of calculations by an order of magnitude. Therefore, in order to improve the accuracy of numerical forecasts, it is essential to clarify the mechanical properties of the phenomena as well as to improve the performance of computers. [Hiroshi Matano] Numerical forecast in practiceIt generally takes several hours to collect observation results from around the world at a certain time (main times are 9:00 and 21:00 Japan time). Moreover, ground and high-altitude meteorological observation points are distributed irregularly, so in order to perform numerical calculations, it is necessary to mathematically convert the observation values from each location into regularly-spaced grid point values. This is called objective analysis. In objective analysis, noise is not removed from the original observation values, so the noise is also allocated to the grid point values. Although measures to suppress noise have been taken, in order to avoid forcing unnecessary suppression operations in the calculation process, it is necessary to optimize the grid point values given in objective analysis as initial values for numerical calculations. This is called setting the initial state or initializing. Based on the initial state, forecast values for about 5 minutes of the forecast period are first calculated for each grid point, and then these distributions are used as new initial states to calculate forecast values for the 5 minutes of the forecast period for each grid point, and this process is repeated. In this way, for example, grid point values (forecast values) for a forecast up to the day after tomorrow (72-hour forecast) and their distribution map (projected weather map) are output about three hours after the calculation begins. For weekly and one-month forecasts, which have longer forecast periods and tend to be more noisy, a technique called ensemble forecasting is used in which several small deviations are artificially added to the initial values beforehand, and the forecast value is calculated as the average of the ensemble of several scattered calculation results. [Hiroshi Matano] Numerical weather prediction modelsThe dynamic equations used in numerical forecasts are called numerical forecast models because they model the complex processes that occur in the atmosphere. The improvement of numerical forecast models is inextricably linked to the improvement of computer performance, so numerical forecast models are updated every time the computers are updated. For an overview of the latest numerical forecast models, it is useful to refer to Meteorological Operations Now (published annually) edited by the Japan Meteorological Agency. There are six types of numerical forecast models in Japan (as of March 2003): (1) Global model (area is the whole Earth, grid spacing is 55 km, forecast period is 90 to 216 hours = about 4 to 9 days. The following contents are in the same order) "Numerical Forecasting Theory" by Kishiho Kanzaburo (1955, Chijin Shokan)" ▽ "New Lectures on Numerical Forecasting" by Kishiho Kanzaburo (1968, Chijin Shokan)" ▽ "One Hundred Years of Meteorology - Exploring the Modern History of Meteorology" by Takahashi Koichiro, Uchida Eiji, and Nitta Hisashi (1987, Tokyodo Publishing)" ▽ "Numerical Forecasting of Floods - The First Steps" by Hino Mikio, Ota Takehiko, Sunada Kengo, and Watanabe Kunio (1989, Morikita Publishing)" ▽ "Numerical Simulation of Meteorology" edited by Asai Tomio and Matsuno Taro, Tokioka Tatsushi, Yamasaki Masaki, and Sato Nobuo (1993, University of Tokyo Press)" ▽ "Numerical Forecasting - Its Theory and Practice" 5th Edition by Masuda Yoshinobu (1993, Tokyodo Publishing)" ▽ "Numerical Forecasting - New Weather Forecasting Using Supercomputers" by Iwasaki Toshiki (1993, Kyoritsu Shuppan)" ▽ "Easy Weather Classroom" by Shimada Moriie (1994, Tokai University Press)" ▽ "Basic Knowledge of Numerical Forecasting - Numerical Forecasting in Practice" edited by the Japan Meteorological Agency (1995, Meteorological Operations Support Center)" ▽ "Atmospheric Movement and Dynamics for Weather Forecasting" by Matano Hiroshi (1997, Tokyodo Shuppan)" ▽ "How to Read Weather Charts for Weather Forecasting" by Shimoyama Norio (1998, Tokyodo Shuppan)" ▽ "New Numerical Analysis Forecasting System" edited and published by the Forecasting Department of the Japan Meteorological Agency (2000)" ▽ "Verification of the New Numerical Analysis Forecasting System" edited and published by the Forecasting Department of the Japan Meteorological Agency (2001)" ▽ "Weather Science for Weather Forecasters" by Nishimoto Hiroaki (2002, Seizando Shoten)" ▽ "Verification and Improvement of Numerical Analysis Forecast Systems" compiled and published by the Forecast Department of the Japan Meteorological Agency (2002)" ▽ "Current Meteorological Operations" compiled by the Japan Meteorological Agency, various annual editions (Fuji Micro) [References] | | | | | | |Source: Shogakukan Encyclopedia Nipponica About Encyclopedia Nipponica Information | Legend |
大気の運動や状態の変化を支配する流体力学と熱力学の法則を数学的に表現した力学方程式を、コンピュータを使って数値的に解き、将来の大気の状態を力学的に予測する天気予報の技術的分野をいう。原語の訳は数値天気予測だが、日本では数値予報が公式用語である。 [股野宏志] 数値予報の登場与えられた境界条件と初期条件の下に、大気に適用された力学方程式を解くことによって天気予報を力学的に行うことは原理的に可能である。しかし、大気は水のような単純な流体ではなく、自転する地球に相対的な運動をしている熱的に敏感な気体である。しかも、相変化を伴う水蒸気を含むなど複雑な性質の流体であることを反映し、数式も複雑であるうえに、数式の形式が非線形偏微分方程式であることから、これを数理的に解くことはほとんど不可能である。しかし、第二次世界大戦後、コンピュータの発達と数値計算技術の進歩により、大気に適用される力学方程式を数値的に解くことが可能となり、力学的方法による天気予報が実用の域に達した。これが数値予報で、気象力学のもっとも輝かしい応用成果である。日本では1959年(昭和34)3月に数値予報が気象庁に導入され、1970年代後半からは伝統的な予報作業も数値予報を軸とする新しい形態に変貌(へんぼう)した。 [股野宏志] 数値予報の原理数値予報に用いられる力学方程式は、運動方程式(運動の法則)、状態方程式(気体の法則)、連続方程式(質量保存の法則)、熱力学方程式(熱量保存の法則)および水蒸気の式(水蒸気量保存の法則)からなるが、実際の数値計算には微分方程式が差分方程式に変換される。数値予報では、大気を鉛直方向に何層にも分け、各層を碁盤の目のような一定間隔の格子点で覆い、格子点ごとに風、気温、気圧、湿度などの気象要素の時間変化量を計算し、将来の格子点値を時間積分して求める方法(格子点法)が基本である。しかし、コンピュータの性能が向上し、方程式の従属変数である気象要素の空間分布を調和関数または三角関数で展開し、関数にかかる係数の時間変化量を計算する方法(スペクトル法)が実用化されたので、日本では1980年代からこの方法が導入された。スペクトル法では、格子点ごとに気象要素の時間変化量を計算する必要がなく、しかも計算結果は格子点値として出力される利点がある。もちろん、天気図形式の予想図としても出力される。 [股野宏志] 数値予報の対象天気予報に重要な低気圧や高気圧のように天気図上で識別できる規模の現象は総観規模現象とよばれる。とくに第二次世界大戦後は総観規模現象のおもな力学的性質も解明されたので、総観規模現象の明日から明後日までの予報(48時間ないし72時間予報)に関していえば、数値予報は1980年代におおむね完成の域に達した。しかし、総観規模より小さい中小規模現象と10日以上の予報に関していえば、数値予報は21世紀を迎えてもいまだ完成の域にはほど遠い。そのおもな理由は、(1)総観規模より小さい現象の力学的性質が数値予報で扱いうるほど十分に解明されていないこと、(2)気象学的雑音(以下、単に雑音)とよばれる音波のような高周波の現象や、街頭のつむじ風のような規模が小さく短命の現象が数値計算の過程で不自然に発達して気象要素の適切な空間分布を乱すこと、である。そのため、雑音が発達しないように抑制措置が計算過程に組み込まれているが、予報期間が長くなると抑制措置を超えて雑音がしだいに成長する。これは数値計算(数値積分)の不可避的な宿命である。 [股野宏志] 数値予報の制約格子間隔より小さい規模の現象は数値予報の直接の対象とはならないが、気象学的に重要な現象については、それらが総観規模現象に集団的に及ぼす効果を間接的な形で計算過程に取り入れている。数値予報で直接の対象となるには、少なくとも5個以上の格子点を含む規模でなければならない。したがって、格子間隔を狭めると、総観規模に近い中規模の現象(メソ現象)も直接の対象となりうるので、数値予報の精度をあげることができる。しかし、格子間隔を半分に縮めると計算量は1桁多くなる。そのため、数値予報の精度向上には現象の力学的性質の解明とともにコンピュータの高性能化が不可欠である。 [股野宏志] 数値予報の実際世界各地で一定の時刻(おもな時刻は日本時で9時と21時)に観測された結果を収集するには、おおむね数時間を要する。しかも、地上と高層の気象観測点は不規則に分布しているので、数値計算を行うために各地の観測値を規則正しく並んだ格子点の値に数理的に置き換える必要がある。これを客観解析という。客観解析では元の観測値に雑音が含まれていても除去されないので、雑音は格子点値にも配分されることになる。雑音の抑制措置が講じられているとはいえ、計算過程でむだな抑制操作を強いることを避けるため、客観解析で与えられた格子点値を数値計算の初期値として適正化する必要がある。これを初期状態の設定または初期値化という。初期状態に基づいて、まず予報期間約5分の予報値が格子点ごとに計算され、ついでこれらの分布を新たな初期状態として予報期間5分の予報値が格子点ごとに計算される、という繰り返しが行われる。こうして、たとえば明後日までの予報(72時間予報)の格子点値(予報値)とその分布図(予想天気図)が計算開始約3時間で出力される。予報期間が長くなって雑音が卓越する週間予報と1か月予報には、あらかじめ初期値に人為的に幾通りかの微小な偏差を与えて計算し、幾通りかの散らばった計算結果の集まり(アンサンブル)の平均を予報値とする技法(アンサンブル予報)が用いられる。 [股野宏志] 数値予報のモデル数値予報に用いられる力学方程式は実際の大気の複雑な過程をモデル化しているので数値予報モデルとよばれる。数値予報モデルの精密化はコンピュータの高性能化と一体なので、コンピュータの更新ごとに数値予報モデルも更新される。最新の数値予報モデルの概要は気象庁編集の『気象業務はいま』(毎年発行)を参照するのが有用である。日本の数値予報モデル(2003年3月時点)は、以下の6種類である。 (1)全球モデル(範囲は全地球、格子間隔は55キロメートル、予報期間は90~216時間=約4~9日。以下内容は同順) 『岸保勘三郎著『数値予報論』(1955・地人書館)』▽『岸保勘三郎著『数値予報新講』(1968・地人書館)』▽『高橋浩一郎・内田英治・新田尚著『気象学百年史――気象学の近代史を探究する』(1987・東京堂出版)』▽『日野幹雄・太田猛彦・砂田憲吾・渡辺邦夫著『洪水の数値予報――その第一歩』(1989・森北出版)』▽『浅井富雄・松野太郎編、時岡達志・山岬正紀・佐藤信夫著『気象の数値シミュレーション』(1993・東京大学出版会)』▽『増田善信著『数値予報――その理論と実際』第5版(1993・東京堂出版)』▽『岩崎俊樹著『数値予報――スーパーコンピュータを利用した新しい天気予報』(1993・共立出版)』▽『島田守家著『やさしい気象教室』(1994・東海大学出版会)』▽『気象庁編『数値予報の基礎知識――数値予報の実際』(1995・気象業務支援センター)』▽『股野宏志著『天気予報のための大気の運動と力学』(1997・東京堂出版)』▽『下山紀夫著『気象予報のための天気図のみかた』(1998・東京堂出版)』▽『気象庁予報部編・刊『新しい数値解析予報システム』(2000)』▽『気象庁予報部編・刊『新しい数値解析予報システムの検証』(2001)』▽『西本洋相著『気象予報士の天気学』(2002・成山堂書店)』▽『気象庁予報部編・刊『数値解析予報システムの検証と改良』(2002)』▽『気象庁編『気象業務はいま』各年版(富士マイクロ)』 [参照項目] | | | | | | |出典 小学館 日本大百科全書(ニッポニカ)日本大百科全書(ニッポニカ)について 情報 | 凡例 |
>>: Numerical integration - Numerical integration
…Agricultural pests include the seed fly, onion f...
It belongs to group 14 of the periodic table and ...
1837‐76 A gunfighter in the American West. His rea...
… [Ezo Development] The development of Ezochi in ...
…The town was founded by the Slavs in the 11th ce...
…Area = 13,782.48 km2 (3rd in Japan)Population (1...
[1] 〘Noun〙 (Formerly "hassho")① A perenn...
French painter. Born in Paris. After the death of...
The original title is "Gofukugaki no Tsume Na...
Year of death: August 27, 6th year of Ansei (Septe...
〘noun〙 (service)① To provide services that do not ...
《 low-density polyethylene 》 Low-density polyethyl...
...The 17th Dynasty, established in Thebes under ...
This refers to currents near the bottom of the oce...
A summer-green fern of the Dryopteris family. It ...