Microsoft Word - 陳昶憲 body.dot

Size: px

Start display at page:

Download "Microsoft Word - 陳昶憲 body.dot"

十伊
4 years ago
Views:

1 62 4 Journal of Taiwan Agricultural Engineering Vol. 62, No. 4, December Forecasting 24-Hour Rainfall Distribution of Typhoon by Heuristic Programming for Taichung City You-Da Jhong Yan-Po He Wei-Lun Tang Chang-Shian Chen ABSTRACT Located at the subtropical high pressure region of Pacific Ocean, Taiwan is easily attacked by typhoon which is influenced by the subtropical high pressure. Because the regional rainfall distributions are highly affected by the moving tracks of typhoon. The relationship between the rainfall distributions of rainfall stations and meteorological factors of typhoon is particularly complex during typhoon event, and this makes rainfall distributions difficult to forecast during typhoon period. * cschen@fcu.edu.tw 38

2 This study established the rainfall forecasting during typhoon event with different moving tracks for the rainfall stations of Taichung City. The research analysis could be divided into two stages. In stage I, the moving track-set of typhoons was categorized using self-organize maps based on historical data of typhoon tracks records. At Stage II, the Correlation Matrix was applied to identify the relationship between meteorological factors of typhoon announced by Central Weather Bureau and rainfall of rainfall station. After removing low correlation factors by decrement technique, the best model of rainfall forecast was set up base on genetic programming. Finally, the combination of these two stages could perform the forecasting of rainfall distributions for rainfall stations after comparing the most similar typhoon track. Keywords: Self-organize maps, Genetic programming, Rainfall forecasting. (SOM)Alvarez (2008) Chang (2014) Kalteh Berndtsson (2007) Hsu (2002) (ANN) DA Savic (1999) PA Whigham (2001) Kashid (2012) Sedki,A. (2009) -A Guven (2008) (2006) CHEN (2007)

3 Labovitz (1967) (Self-Organize maps, SOM) (Self-Organize maps SOM) T.Kohonen 1980 SOM 1 SOM 2 SOM SOM 2(1) R R R n n 1 _ rate,... (1) R _ rate ( 1.0) 40

(2) r ( X C ) ( Y C )...(2) 2 2 j j x j y XY j j j C x C y (3) R _ factor exp( r / R)...(3) j j r j = 0 R_ factor 1 rj = R_ factorj 0 SOM 1. 2. 3. 2.2 (Genetic Programing, GP) Kozay 1992 (Genetic Algorithm, GA) GA (0 1 ) j 3 3.

4 (2) r ( X C ) ( Y C )...(2) 2 2 j j x j y XY j j j C x C y (3) R _ factor exp( r / R)...(3) j j r j = 0 R_ factor 1 rj = R_ factorj 0 SOM (Genetic Programing, GP) Kozay 1992 (Genetic Algorithm, GA) GA (0 1 ) j SOM SOM :1 SOM

32 30 28 26 24 22 20 18 16 14 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 1 1 0 0 0 0 1 0000 00000000000111000011100001

5 SOM SOM SOM

6 Rˆt 1 MAE 2. (Root Mean Square Error, RMSE) M 2 p ( Tp Yp) RMSE... (5) M T p P Y p P M 3. (R-square, R 2 ) 6-2 GP GP (Mean Absolute Error, MAE) GP (Mean Absolute Error, MAE) M ˆ l1 Rt 1 Rt 1 MAE...(4) M M R t+1 2 SSR SSE R 1 SST SST SST SSR SSE ~9 1~

Hits Hits 3 2 2.5 1.5 2 14 13 17 1 5 6 1.5 1 10 9 10 0.5 0.5 0 11 4 0 23 15 14-0.5-0.5-1 -1-1 -0.5 0 0.

7 Hits Hits 圖8 東來與西來颱風拓樸圖(由左至右) 圖 11 SOM 分類結果為東來路徑 1 之路徑圖圖 12 SOM 分類結果為東來路徑 2 之路徑圖圖9 中央氣象局路徑分類圖(路徑 1 至 5) 圖 13 SOM 分類結果為東來路徑 3 之路徑圖個數並與中央氣象路徑分類圖(圖 9 與圖 10)比較本研究中東來颱風中路徑 1 2 5(圖 )為經過臺灣西南部之颱風路徑此三種路徑颱風起始位置與中央氣象局第 5 類較為接近但路線圖 10 中央氣象局路徑分類圖(路徑 6 至 9) 則與中央氣象局第 7 類路徑較為相似本研究東 44

8 圖 14 SOM 分類結果為東來路徑 4 之路徑圖圖 17 SOM 分類結果為東來路徑 7 之路徑圖圖 15 SOM 分類結果為東來路徑 5 之路徑圖圖 18 SOM 分類結果為東來路徑 8 之路徑圖圖 16 SOM 分類結果為東來路徑 6 之路徑圖圖 19 SOM 分類結果為東來路徑 9 之路徑圖來颱風路徑 3 6(圖 13 16)為經過臺灣東北部之研究東來颱風路徑 7 8 9(圖 17~圖 19)為經臺颱風路徑其中路徑 3 與中央氣象局第 1 類路徑灣本島之颱風路徑主要差別在於貫穿臺灣本島相似路徑 6 與中央氣象局第 1 2 類路徑相似之路線位於北中南部路徑 7 與中央氣象局本研究東來颱風路徑 4(圖 14)為沿中央山脈東邊之分類路徑則介於第 2 3 類路徑 8 與氣象局北上之路線與中央氣象局第 6 類路徑相似本之分類路徑則介於第 3 4 類路徑 9 與中央氣 45

9 20 SOM 1 23 SOM 4 21 SOM 2 22 SOM 3 2 1( 20) 9 23( 21 22) 9 4( 23) SOM SOM ( 9) 2 ( 4) MAE mm 46

10 1 (5 ) 2 (4 ) 3 ( 10 ) 3 (9 ) 4 (2 ) 5 (3 ) 6 (5 ) 7 (6 ) 8 (7 ) 9 (10 ) MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare ( 4) 47

11 MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare MAE RMSE Rsquare ( 9) R-square 4 R-square 0.76 R-square 0.91() 0.05( ) 48

24 4 () 25 (2007)- 4 () 26 4 () 24 27 2 9 R-square 0.

12 24 4 () 25 (2007)- 4 () 26 4 () R-square 0.51() 0.17() ( 4) MAE 26.87mm ( 9) mm

13 27 (2007)- 4 () 28 9 () 29 (2012)- 9 () ( 17 19) 50

14 30 9 () 31 (2012)- 9 () 32 (1991) ( 1)( 4)( 9)( 36~ 38)

15 33 (2007) (2012) (2008)- 800mm 495mm 52

16 圖 36 艾美(1991)颱風觀測(左)與預測(右)雨量值分布圖(第 5 報) 圖 37 帕布(2007) 颱風觀測(左)與預測(右)雨量值分布圖(第 17 報) 圖 38 蘇拉(2012) 颱風觀測(左)與預測(右)雨量值分布圖(第 35 報) 自組織映射圖網路分類的準確度受聚類個皆有其預測之準確性數(拓樸大小)的設定影響頗大太多將產生過度擬合與空白分類群過少則會失去其分類意義本研究之結果對於降雨之時間及空間分布 53 建議 1. 本研究以 61 場颱風建立 13 條颱風降雨預測

17 39 9 () 40 (2012)- 9 () 41 9 () 2. 54

18 42 (2012)- 9 () 1. Alvarez-Guerra, M., González-Piñuela, C., Andrés, A., Galán, B. and Viguri, J. R., Assessment of Self-Organizing Map artificial neural networks for the classification of sediment quality, Environment International, Vol. 34, Issue 6, pp , August, Chang, L. C., Shen, H. Y. and Chang, F. Y., Regional flood inundation nowcast using hybrid SOM and dynamic neural networks, Journal of Hydrology, Vol. 519, pp , November, Kalteh, A.M. and Berndtsson, R., Interpolating monthly precipitation by self-organizing map (SOM) and multilayer perceptron (MLP), Hydrological Sciences Journal, Vol. 52, Issue 2, pp , April, Hsu, K., Gupta, H.V., Gao, X., Sorooshian, S. and Imam, B., Self-organizing linear output map (SOLO): an artificial neural network suitable for hydrologic modeling and analysis, Water Resources Research, Vol. 38, Issue 12, pp , December, D.A. Savic, G.A. Walters and J.W. Davidson, A Genetic programming approach to rainfallrunoff modeling, Water Resour Manag, Vol. 13, Issue 3, pp , June, Whigham, P. A. and Crapper, P. F., Modelling Rainfall-Runoff using Genetic Programming, Mathematical and Computer Modelling, Vol. 33, Issue 6-7, pp , March-April, Kashid, S. S. and Maity, R., Prediction of monthly rainfall on homogeneous monsoon regions of India based on large scale circulation patterns using Genetic, Journal of Hydrology, Vol , pp , August, Sedki, A., Ouazar, D. and El Mazoudi, E., Evolving neural network using real coded genetic algorithm for daily rainfall runoff forecasting, Expert Systems with Applications, Vol. 36, Issue 3-Part1, April, Guven A. and Gunal M., Genetic programming approach for prediction of local Scour downstream of hydraulic structures, Journal of Irrigation and Drainage Engineering, Vol. 134, Issue 2, pp , April, Ching-Sen Chen, Yi-Leng Chen, Che-Ling Liu, Pay-Liam Lin and Wan-Chin Chen., Statistics of Heavy Rainfall Occurrences in Taiwan, Weather & Forecasting, Vol. 22, Issue 5, pp. 981, October, Labovitz, S., Some observations on measurement and statistics, Social Forces, Vol. 46, No. 2, pp , December,

19 38(1)

4 115,,. : p { ( x ( t), y ( t) ) x R m, y R n, t = 1,2,, p} (1),, x ( t), y ( t),,: F : R m R n.,m, n, u.,, Sigmoid. :,f Sigmoid,f ( x) = ^y k ( t) =

$4 115,,. : p { ( x ( t), y ( t) ) x R m, y R n, t = 1,2,, p} (1),, x ( t), y ( t),,: F : R m R n.,m, n, u.,, Sigmoid. :,f Sigmoid,f ( x) = ^y k ( t) =$ 2007 4 4 :100026788 (2007) 0420114206, (, 430074) :,,,,,,GIS.,,. : ; ; ; ; : TP391 ;P338 : A Development of Combinatorial Intelligentized Decision2Making Support System and Its Utilization in Runoff Forecasting