w» wz, 15«1y(2013) (ISSN 1229-5671) Korean Journal of Agricultural and Forest Meteorology, Vol. 15, No. 1, (2013), pp. 40~49 DOI: 10.5532/KJAFM.2013.15.1.040 Author(s) 2013. CC Attribution 3.0 License. x w y d ½ 1 *Á 2 Á 3 1 ( ) ƒ» l, 2» w»»zq, 3 w y œw (2013 3 19 ; 2013 3 23 ; 2013 3 23 ) Prediction of Blooming Dates of Spring Flowers by Using Digital Temperature Forecasts and Phenology Models Jin-Hee Kim 1 *, Eun-Jung Lee 2 and Jin I. Yun 3 1 National Center for Agro-Meteorology, Seoul National University, Seoul 151-742, Korea 2 Korean Peninsula Weather and Climate Division, KMA, Seoul 156-720, Korea 3 College of Life Sciences, Kyung Hee University, Yongin 446-701, Korea (Received March 19, 2013; Revised March 23, 2013; Accepted March 23, 2013) ABSTRACT Current service system of the Korea Meteorological Administration (KMA) for blooming date forecasting in spring depends on regression equations derived from long term observations in both temperature and phenology at a given station. This regression based system does not allow a timely correction or update of forecasts that are highly sensitive to fluctuating weather conditions. Furthermore, the system cannot afford plant responses to climate extremes which were not observed before. Most of all, this method may not be applicable to locations other than that which the regression equations were derived from. This note suggests a way to replace the location restricted regression equations with a thermal time based phenology model to complement the KMA blooming forecast system. Necessary parameters such as reference temperature, chilling requirement and heating requirement were derived from phenology data for forsythia, azaleas and Japanese cherry at 29 KMA stations for the 1951-1980 period to optimize spring phenology prediction model for each species. Best fit models for each species were used to predict blooming dates and the results were compared with the observed dates to produce a correction grid across the whole nation. The models were driven by the KMA s daily temperature data at a 5km grid spacing and subsequently adjusted by the correction grid to produce the blooming date maps. Validation with the 1971-2012 period data showed the RMSE of 2-3 days for Japanese cherry, showing a feasibility of operational service; whereas higher RMSE values were observed with forsythia and azaleas. Key words: Blooming date forecast, Spring phenology, Azaleas, Forsythia, Cherry blossom I. y y ƒ w»z y w ƒ wù, y w»z ƒ ƒ» d ƒ w w r (Yun, 2010; Chung et al., 2011).»z y ƒ w ù w» wš» w» ³e y w d * Corresponding Author : Jin-Hee Kim (7jhee@naver.com)
Jin-Hee Kim et al.: Prediction of Blooming Dates of Spring Flowers by Using Digital Temperature... 41 (phenology network) wš,»z y wsƒ» w» ù { y dw x ƒ š (De Melo-Abreu et al., 2004; Aono and Kazui, 2008). ü» w y d x, d y» w 2 ü y w (Jung et al., 2005), m»z ù ûw y s w (Yun, 2006). w l DC y s w (Chung et al., 2011). w» œw y w x w»»z y d l z w. z w ƒ ƒ w, s³ w»» y š, d»z w w w w ƒ. w w» w w ƒ x x ywš w. II. y 2.1. y» w tx tw ù,, ƒ y ƒ yw w yw t y. ü { (endodormancy) ƒ {» { kq š. (plant phenology) ü {,, y ew, m, ƒ y w x»ƒ» ƒ w (Oh, 2004). y» w w x ƒ Chill-day x y y-{ - - y w (thermal time) txw, ƒ š w y v w š» y w (Richardson et al., 1974; Aron, 1983; Cesaraccio et al., 2004; Jung et al., 2005). x w ù» w ü { ƒš þƒ (chill-days). þƒ ú» w w (chilling) w, ú þƒ (chill requirement, Cr) w ¾ { k w., { w k þƒ ( {, dormancy depth) w txw. { ƒ þƒ j š ¼ Á ù k, ü { w k. ü { w l. w ü { w ƒ û ww {, y { (environmental dormancy) y { (forced dormancy) š (Oh, 2004). ƒ» w ü { w z ú ƒ {» ú w w. þƒ ƒ (anti-chill days) w tx g. yƒ ƒ y š. ü { w þƒ» š, { w ƒ» (Fig. 1). z y ƒ w tx, yw v w ƒ. ƒ ¾ (growing degree days, GDD) w. z ƒ w w yw w š (heating requirement, Hr) š. x»
42 Korean Journal of Agricultural and Forest Meteorology, Vol. 15, No. 1, š» w ú þƒ ù ƒ 0. x w š»,»,»» y 6 ƒ w ƒƒ (Table 1). Fig. 1. Concept of the two step phonological model to predict flowering date in temperate zone deciduous trees. Floral buds must be exposed sequentially to sufficiently long periods of chilling temperature (R c ) and heating temperature (R h ) for flowering (from Jung et al., 2005). w, p(, 4 5 )ƒ w ú, txw ú ú w. þƒ ƒ š» w v w» (temperature threshold, Tc)ƒ. w þƒ ú š»,» w ÿ w 2.2. y d x x ù,, ù t y d w» w ù,, ƒƒ w» (Tc), (Cr) š (Hr) x (parameters) w. yw ü» w w w y x mw l {,, y w» w, x» w» d w.» d e» w t w - y, t w d» wš.»z y t wù ƒ 1950 15 d w» 1970 50 š, x ¾ ã» š.» e d ƒ d yw w» Table 1. Equations to calculate chill days (C d ) and anti-chill days (C a ) for the five cases that relate the daily maximum (T x ) and minimum (T n ) termperature to the threshold temperature ( ) and 0 o C, where T m is the mean daily temperature (Cesaraccio et al., 2004) Case Temperature Chill Days (C d ) Anti-Chill Days (C a ) 1 0 T c T n T x C = 0 d C = a T m 2 T x T x 0 T n T c T x C = ( d T m T ) -------------- n C = -------------- 2 a 2 3 0 T n T x C = ( d T m T ) n C = 0 a 4 T x T x T < 0 < n T x C = -------------- ---- d C = 0 T x T 2 a n 5 T x T T x x T x T < 0 < n T < c T x C = -------------- ---- -------------- d C = -------------- T x T 2 2 a 2 n 6 T < 0 x C = 0 d C = 0 a C d : Chill days C a : Anti-chill days T x : Daily maximum temperature T n : Daily minimum temperature T m : Daily mean temperature : Threshold temperature
Jin-Hee Kim et al.: Prediction of Blooming Dates of Spring Flowers by Using Digital Temperature... 43 Fig. 2. Relationship between the model estimated- and the observed- flowering dates for forsythia, azaleas and cherry trees at the 29 phenology observation stations of the Korea Meteorological Administration during 1981-2012. w v, y d y» d x q w» w y. ù,, w» w» w. w -50 l -120 ¾ 10 8 wš,» 5, 6, 7 o C 3 8 3=24 w. w w þƒ w dwš ƒ d ƒ w w kw. 1951-1980» wš 29» w 24 ƒƒ w d» w z d w ƒ w» - w ù,, w ƒƒ wš, w ü t w. 3 w ƒƒ t» - d w 1951-1980» d y wš d y ƒ ƒ w š w. w ù» 6 o C, -90, š 128.5 š(ÿ ),» 6 o C, -80, š 94 ( ),» 7 o C, -100, š 158 ( ). y d x w 1981-2012» 29»» w wš w» d y w. ù Fig. 3. Performance of the best fit models in prediction of flowering dates of forsythiag(top), azaleasg(middle), and cherry (bottom) at 3 representative geographical locations in South Korea during 1981-2012.
44 Korean Journal of Agricultural and Forest Meteorology, Vol. 15, No. 1 걸쳐 29개 지점 생물계절 관측 기상대를 대상으로 조사한 바로는, 표준목의 개 화일이 실제로는 그 지역 대표성이 없다는 점이 전체 답변 중 28.2%로 가장 큰 비중을 차지했고, 코스모스, 개나리에서 자주 관찰되는 불규칙한 개화시기가 22.6%, 군락단지로 지정된 지점과의 거리가 멀어서 생 기는 개화 상시 관측의 어려움이 14%로 그 뒤를 이 었다. 그 밖에도 관측목 표준관리(수령, 품종, 예비목 등)의 어려움과 발아-개화-만개 단계의 불명확성, 주관 개화일예측모형의 현업서비스 적용 적인 관측기준 등이 거론되었다. 이처럼 모형에 반영 될 수 없는 관측업무 자체에 기인한 오류를 보정하기 전에는 생물계절모형만으로 예측된 개화일 분포는 신 지역간 오차보정 생물계절관측의 현황과 문제점을 파악하기 위해 뢰성이 낮을 수 밖에 없다. 더욱이 이 모형을 관측자 측 개화일에 비해 모형에 의한 추정 개화일이 지연되 는 경향이며 진달래는 반대로 실측 개화일에 비해 모 형 예측 개화일이 더 빠른 것으로 나타났다(Fig. 2). 29개 지점 가운데 서울, 대구, 광주에서의 개화예측 결과만 추출해 보면, 모형은 지구온난화로 인해 앞당 겨진 개화시기의 경향성과 이상기후에 따른 극심한 연 차변이를 RMSE (root mean square error) 3~4일 수준에서 재현하고 있다(Fig. 3). 2012년 6월부터 3개월에 III. 3.1. Fig. 4. Classification of 29 phenology stations into 6 geographical zones for spatial interpolation of the correction factor.
Jin-Hee Kim et al.: Prediction of Blooming Dates of Spring Flowers by Using Digital Temperature... 45 ƒ w y d ¾ w» w x v w w w» w 29 w 1981-2012» l s w. 29 Fig. 4 w w s x p š w 6 «wš «œ s³ ƒ (inverse distance weighting, IDW) w 5km w. w y x w 29 ù,, y w w x s w j w. x s p, ù û eƒ š w f w, û w f w. ù x d y, x d y û (Fig. 5). wr 3 y w x d y j x ùkù, ³ w» w x ƒ ƒ v w. 3.2. d y ù,, y ƒƒ Ÿ,, d w txw x w 29 1 y w, s» w w 1971-2012» y dwš d y w. d y ƒ d y RMSE txw, ù 29 ƒ ƒ yw (3.5 ), Ÿ ƒ (3.7 ). ƒƒ 4.0, 2.3 w ƒ. 3 û ù w w œ ù ü w y» y ƒ w. w d y» w ƒš p ù ù w RMSE 2~3 ü x q (Fig. 6-8). IV. š x ù ù,, w» w yw w Fig. 5. Correction factor grids for flowering date prediction in forsythia (left), azaleas (center), and cherry trees (right).
46 Korean Journal of Agricultural and Forest Meteorology, Vol. 15, No. 1 Fig. 6. Performance of the phenology model based method for flowering date estimation in forsythia. Comparison of the predicted and the observed flowering dates at various geographical locations in Korea for 1971-2012. w w y x l w y w w,» w» d w.» 29» w d ¼ w, d t, d ³, dy p, w y, d ü ƒvw. w d ù d w w» w x ƒ. x w d ü w w. w w, x
Jin-Hee Kim et al.: Prediction of Blooming Dates of Spring Flowers by Using Digital Temperature... 47 Fig. 7. Performance of the phenology model based method for flowering date estimation in azaleas. Comparison of the predicted and the observed flowering dates at various geographical locations in Korea for 1971-2012. ùy w y» w,»z w RMSE 3 ~4 xwš, x k w. p ùy» wq 3 d» w,» x x w x» w k w w x d w w. w wz x» w, û ù w w ƒ f ³ wš w w w, ù ƒ ù» w š w
48 Korean Journal of Agricultural and Forest Meteorology, Vol. 15, No. 1 Fig. 8. Performance of the phenology model based method for flowering date estimation in cherry trees. Comparison of the predicted and the observed flowering dates at various geographical locations in Korea for 1971-2012. x ƒ ƒ v w.» œw y w»»z y d l z w» y š, d»z w w w,» w y ƒ ƒ w. w w» w» x x yw ww. ûw 29» l 1951-1980» ù,,
Jin-Hee Kim et al.: Prediction of Blooming Dates of Spring Flowers by Using Digital Temperature... 49 d t y d» w x (»,, š ) wš w y d x w. d y s txw x ww 3 y d y w. w 1971-2012» 29 y dw d y w RMSEƒ 2~3 y w. œ ( y: PJ009292) w. REFERENCES Aono, Y. and K. Kazui, 2008: Phenological data series of cherry tree flowering in Kyoto, Japan, and its application to reconstruction of springtime temperatures since the 9th century. International Journal of Climatology 28(7), 905 914. doi:10.1002/joc.1594 Aron, R. H., 1983: Availability of chilling temperatures in California. Agricultural Meteorology 28, 351-363. Cesaraccio, C., D. Spano, R. L. Snyder, and P. Duce, 2004: Chilling and forcing model to predict bud-burst of crop and forest species. Agricultural and Forest Meteorology 126, 1-13. Chung, U., L. Mack, J. I. Yun, and S. H. Kim, 2011: Predicting the timing of cherry blossoms in Washington, DC and mid-atlantic States in response to climate change. PLoS ONE 6(11), e27439. De Melo-Abreu, J. P., D. Barranco, A. M. Cordiero, J. Tous, B. M. Rogado, and F. J. Villalobos, 2004: Modelling olive flowering date using chilling for dormancy release and thermal time. Agricultural and Forest Meteorology 125, 117-127. Jung, J. E., E. Y. Kwon, U. Chung, and J. I. Yun, 2005: Predicting cherry flowering date using a plant phenology model. Korean Journal of Agricultural and Forest Meteorology 7, 148-155. (In Korean with English abstract) Oh, S. D., S. M. Kang, D. I. Kim, M. S. Kim, and W. S. Kim, 2004: Fruit Tree Physiology in Relation to Temperature. Gilmogeum, Seoul, Korea, 364pp. (In Korean) Richardson, E. A., S. D. Seeley, and D. R. Walker, 1974: A model for estimating the completion of rest for Redhaven and Elberta peach trees. HortScience 9, 331-332. Yun, J. I., 2006: Climate change impact on the flowering season of Japanese cherry (prunus serrulata var. spontanea) in Korea during 1941-2100. Korean Journal of Agricultural and Forest Meteorology 8, 68-76. (In Korean with English abstract) Yun, J. I., 2010: Agroclimatic maps augmented by a GIS technology. Korean Journal of Agricultural and Forest Meteorology 12, 63-73. (In Korean with English abstract) doi:10.5532/kjafm.2010.12.1.063