Wind Speed Data for TCs that Have Affected Hong Kong 影響香港氣旋之風速數據
RUSS (1994) made its CPA at its formation; tropical cyclone signals were issued when it was further away from HK. 1994 年熱帶氣旋羅士於形成時最接近香港，而熱帶氣旋警告信號則於其距離香港較遠時發出。
NURI (2008) did not give a direct hit on the Observatory, but instead skirted past it within 1 kilometre. 2008 年熱帶氣旋鸚鵡於香港天文台一公里範圍內掠過。
Tropical cyclone strength 熱帶氣旋強度:
The Observatory introduced two new categories, "severe typhoon" and "super typhoon", in 2009. For consistency, tropical cyclones before 2009 are also categorized based on this new system. 天文台於 2009 年加入 "強颱風" 和 "超強颱風" 兩個等級。為保持一致性，上表中 2009 年前的熱帶氣旋亦同樣按照新等級來作分類。
Full name 全寫
Central wind speed range 中心附近風速範圍
Colour scheme* 顏色*
Tropical depression 熱帶低氣壓
41 - 62 km/h 公里每小時
Tropical storm 熱帶風暴
63 - 87 km/h 公里每小時
Severe tropical storm 強烈熱帶風暴
88 - 117 km/h 公里每小時
118 - 149 km/h 公里每小時
Severe typhoon 強颱風
150 - 184 km/h 公里每小時
Super typhoon 超強颱風
185+ km/h 公里每小時
*Colour used in the column "CPA strength". 在 "最接近強度" 一欄中的顏色。
The percentile rank of a certain entry (loosely) shows the percentage of observations that have values smaller than that given entry. The values in the tables above are given in percentage terms, and a high value (close to 100[%]) means that the observation is among the highest ever observed. Here we adopt an algorithm that is similar to that used by Microsoft Excel, but with a modification to better deal with tied values. (Note that Excel does not calculate exactly according to the definition; more appropriately what we calculate is the cumulative distribution at that particular value, with the density describing the observations we have here. Linear interpolation is done if necessary.)
一個參數的百分等級顯示有多少項數據的數值比該參數為低 (以百分率表示)，比較大的百分等級代表該參數接近所有觀測數據的上限。這裡我們使用跟 Microsoft Excel 類似的計算方法，但稍作調整，從而令同值數據的計算更有代表性。(留意 Excel 的計算方法並不完全依照百分等級的定義；這個計算實為某一數值的累積分佈。有需要時我們會使用線性插值法。)
In the second table "Percentile Ranks Sorted by Signal Issued", data are divided into two groups, with group 1 containing all observations from wind speeds resulting from cyclones that necessitated #3 Signal (there are four subgroups as we have four station categories, excluding "All stations") and group 2 from cyclones that necessitated #8 Signal (again there are four subgroups here). Percentile rank for each cyclone (and each category) is calculated using these groups. A few notes on this are in order:
No percentile rank is calculated for storms that necessitated the #10 Hurricane Signal because of its rarity in the data we have included here.
For storms that necessitated the #9 Signal, the "#8 Signal" group is still used for calculation (i.e. group 2). This is due to two reasons:
The #9 Signal is relatively rare, and the final ranking may not reflect the true situation if we open a group 3 solely for those storms;
Storms that necessitated Signal #9 ("#9 cyclones", similarly defined below for storms that necessitated other signals) are excluded from group 2 because such data may affect another more important calculation: the percentile ranks for "#8 cyclones" - one usually of high importance because upgrading to Signal #8 usually means that most daily activities are affected; the difference between Signal #8 and Signal #9 is relatively smaller in terms of social cost.
By calculating in such a way, we can observe the strength of winds a particular "#9 cyclone" brings (in the "#8 cyclone" scale) without distorting the ranking for "#8 cyclones".
在表二 ("以發出信號為基準的百分等級")，我們把資料細分為兩組 : A 組為令天文台發出三號信號的氣旋 ("三號信號氣旋") 所帶給香港之風力數據；而 B 組為 "八號信號氣旋" 所帶給香港之風力數據。由於氣象站被分為四類，每組將再被分為四個小組。各氣旋各類氣象站的百分等級是基於這些組別計算而成，但請注意以下數點 :
For the third table "Overall Percentile Ranks", the calculation is similar to that for the second table, with the exception that the group considered now consists of all storms in the list, and is not subdivided according to the highest signal. Therefore one will normally expect the percentile ranks for "#8 cyclones" is higher in the third table than in the second, because in the third table the group also includes many "#3 cyclones".
表三 ("總百分等級") 的計算跟表二類似，但基準組別只得一個 : 表列所有氣旋均包括在內。由於計算基準組別不同，我們可以預計某一 "八號信號氣旋" 於表三的百分等級將較表二為高，原因是表三的基準組別包含 "三號信號氣旋"。
Due to the insignificance of the column "All stations" in the first table, it is not included in the calculation of percentile ranks.
由於表一中 "所有站平均" 缺乏代表性，於計算百分等級時不予考慮。
"Average Percentile Rank" takes the average of the four values that follow. It loosely represents the relative wind strength (brought to Hong Kong) of that cyclone.
The resulting percentile ranks are not absolute; instead they should be used in a relative manner. Therefore, in the second table for a storm that necessitated the #8 Signal and has an average value of 20 (i.e. 20%), it is correct to say that the Signal #8 for this storm is relatively weak; however we do not have information on whether this #8 is unjustified, because such signals are issued based on absolute scales in terms of wind speeds. (But in practice, for a storm that has an average percentile rank of 20%, winds are generally light and even if the signal is justified it probably falls near the margin.)
計算出的百分等級為相對指標，而不是一個絕對指標。例如某一 "八號信號氣旋" 於表二計算出的數字為 20 [%]，那我們可以說該次的八號信號相對較弱，但由於信號基於風速等絕對指標，我們並沒有足夠數據來判斷該次八號信號是否誤發。(但一般應用上，如該次的百分等級只有 20%，它的風速應該相對頗弱，即使達標也不會離開達標邊沿太遠。)
The percentile ranks can change as new storm data are added, because the underlying densities (i.e. groups 1 and 2 we used above) have changed.
The careful reader may discover that some stations are counted more than once (across categories) and different stations may carry different weights in the calculation (e.g. only 5 stations' data are used in calculating Victoria Harbour Average while 13 are used in calculating Urban Average). However, as long as the grouped average is representative, this will not cause a large deviation from the theoretically perfect outcome. [Note: generally stations that are close to the Victoria Harbour and/or belong to the reference stations will be counted more than once due to their significance]
讀者或會發覺某些站點會被計算多於一次 (於不同組別出現)，而由於不同的平均值利用不同數量的站點資料計出，每站的權重亦不同。但如果最終得出的每組平均具有代表性，這應不會令計算偏離至不可使用的程度。[註 : 由於靠近維港的站點和參考站的重要性較高，它們的數據通常會被重複計算]
From experience, while referring to the average ranks in table 2, a value of above 70 normally implies a strong Signal #3 / #8, while a value of below 30 implies a weak #3 / #8. Sometimes weak signals are issued not because of meeting wind speed standard; instead they may be issued whenever a storm is very close to Hong Kong.
根據經驗，如表二中的平均等級高於 70，通常該次為一較強的三號 / 八號信號；而如數值低於 30，該次則為一較弱的三號 / 八號信號。有時三或八號信號會在氣旋非常接近香港時發出，而該些情況風速或未達標。
Information about the data 數據資料:
All distances are in kilometres (km), and all average speeds are in kilometres per hour (km/h). 所有距離和風速的單位分別為公里和公里每小時。
Maximum hourly average wind speeds are used in calculating the averages. 所有平均由最高每小時平均風速計算而成。
The following stations are included in each of the categories appeared above (figures in brackets indicate elevation of anemometer): 上表各組所包括的氣象站如下 (括號內為風速計高度) :
Lamma Island 南丫島 (17 m) (Data available from 2019 數據於 2019 年開始)
Lau Fau Shan 流浮山 (50 m)
Peng Chau 坪洲 (47 m)
Ping Chau (東)平洲 (39 m)
Sai Kung 西貢 (31 m)
Sha Chau 沙洲 (31 m)
Sha Lo Wan 沙螺灣 (71 m)
Tai Mei Tuk 大美督 (71 m)
Tap Mun 塔門 (48 m) (35 m for the old station decomissioned in July 2017 於 2017 年 7 月停用的舊站高度為 35 米)
Tsing Yi (Ching Pak House) 青衣青柏樓 (136 m)
Waglan Island 橫瀾島 (83 m)
All Stations that appeared in HKO's TC Report 所有於天文台的熱帶氣旋報告中出現過的氣象站
"Obs" columns indicate the number of valid observations in computing the average wind speed. "Obs" 一欄顯示計算該平均之有效站數。
No altitude correction is made, and therefore stations at very high altitudes are not used to prevent distortion of data (except for the "all stations" column). 由於本表不設高地風速調整，為免令數據失準，在非常高的地點錄得的風力並不包含在本列表 ("全部" 站平均除外)。
Locations of the stations can be found here. 氣象站地點可於這裡查閱。
2013/05/27: Following an announcement from the Observatory, Wetland Park is replaced by Lau Fau Shan as one of the eight reference stations. Percentile ranks are re-calculated accordingly. Some data errors are also fixed. The resulting percentile rank changes are reflected in this table. 天文台較早前宣布將更改參考站，以流浮山替代濕地公園，因此本站重新計算所有百分等級，與此同時一些資料錯誤亦已修正。是次更新所引致的百分等級改變詳載於此表。
Last Accessed 最近訪問日期: Tue Jan 31 2023 14:32:15 HKT
Last Modified 最近修訂日期: Fri Jun 03 2022