Potential for Using FORMOSAT-Ⅱ Data Simulated by Hyperspectral Reflectance to Estimate Growth and Predict Yield of Rice

以高解析反射光譜模擬福衛二號衛星资料估測水稻生長及預測產量之潛力

楊純明(Chwen-Ming Yang) ； 陳榮坤(Rong-Kuen Chen)

《台灣農業研究》 54卷1期 (2005/03) Pp. 54-69

https://doi.org/10.6156/JTAR/2005.05401.06

摘要

本研究旨在利用近地面高解析植被反射光譜模擬福衛二號衛星(FORMOSAT-Ⅱ)光譜測值以評估應用於水稻(Oryza sativa L. cv. TNG 67)生長估測與產量評估之潛力。田間試驗係在臺中縣霧峰鄉農委會農業試驗所農場進行，以量測2000-2002年兩期稻作之植被光譜、葉面積指數及產量等。福爾摩沙二號衛星承載之紅光(630-690 nm)及近紅外光(760-900 nm)寬頻波段測值由高解析光譜之相對累積窄波段測值平均值模擬，並以計算窄波段及寬波段之標準差植被指數（NDVI(下标 NB)及NDVI(下标 FORMOSAT-Ⅱ)）。結果顯示NDVI(下标 NB)及NDVI(下标 FORMOSAT-Ⅱ)在水稻生育期間呈現曲線分佈，而NDVI(下标 FORMOSAT-Ⅱ)概低於NDVI(下标 NB)。實際量測之葉面積指數(LAI(下标 measured))與NDVI(下标 NB)之間為一指數函數關係，並據以輸入NDVI(下标 NB)及NDVI(下标 FORMOSAT-Ⅱ)而估算出LAI(下标 NB)及LAI(下标 FORMOSAT-Ⅱ)。又發現生育期間估測之LAI(下标 FORMOSAT-Ⅱ)均小於LAI(下标 NB)，且這些估值自移植後55天起皆低於LAI(下标 measured)。結果也顯示，產量可由生育某段期間LAI(下标 measured)累加值預測，當由LAI(下标 NB)及LAI(下标 FORMOSAT-Ⅱ)取代時，產量預測值將低於實測值。顯然地，模擬福爾摩沙二號衛星寬波段測值將獲得小於地真(ground truth)資料之生長估測與產量預測。

關鍵字

福衛二號衛星 ； 高解析反射比 ； 水稻生長估測 ； 產量預測 ； 模擬

並列摘要

This study assessed the potential for using FORMOSAT-Ⅱ data, simulated by ground-based remotely sensed high resolution reflectance spectra, to estimate growth and predict yield of rice (Oryza sativa L. cv. TNG 67) crops. Field experiments were conducted in the Taiwan Agricultural Research Institute Experimental Farm at Wufeng to measure canopy reflectance spectra, leaf area indices (LAI(subscript measured)), and yields of rice grown in the cropping seasons from 2000 to 2002. Reflectance values of broadband sensors equipped in the satellite FORMOSAT-Ⅱ were simulated by using the mean values of reflectance in the respective band regions obtained from hyperspectral data. The values of normalized difference vegetation index (NDVI) were calculated using the mean reflectance in red (630-690 nm) and near-infrared (760-900 nm) regions defined in FORMOSAT-Ⅱ. Both NDVI(subscript NB), which was computed from narrow bands of hyperspectral data, and NDVI(subscript FORMOSAT-Ⅱ), which was computed from simulated FORMOSAT-Ⅱ data, showed a quadratic time trend during rice growth. Changes in NDVI(subscript FORMOSAT-Ⅱ) nearly paralleled the changes in NDVI(subscript NB), but to a lesser extent. The relationship between LAI(subscript measured) and NDVI(subscript NB) was fitted well by an exponential function, through which the estimates of satellite LAI (LAI(subscript FORMOSAT-Ⅱ)) were obtained. Results further showed that values of LAI(subscript FORMOSAT-Ⅱ) were lower than values of LAI(subscript NB) along plant development, while both of these estimated LAIs were lower than LAI(subscript measured) after approximately 55 days after transplanting. Results indicated that yield may be predicted by the accumulated values of LAI(subscript measured) during the growing periods. By substituting LAI(subscript measured) with the values of LAI(subscript NB) and LAI(subscript FORMOSAT-Ⅱ), it was found that the predicted yields were lower than the measured yields. Results suggest that the lower predicted yield by the FORMOSAT-Ⅱ satellite data is attributable to lower level of LAI(subscript FORMOSAT-Ⅱ) as aforementioned.

並列關鍵字

FORMOSAT Ⅱ ； Hyperspectral reflectance ； Rice growth estimation ； Yield prediction ； Simulation