應用遙測技術推估水稻產量與品質之研究

Abstract

本文主要探討五種不同程度缺或無葉綠素b突變種水稻於生育期間其葉片之葉綠素(chlorophyll，Chl)、生合成中間物包括protoporphyrin IX (PPIX)、magnesium protoporphyrin IX (MGPP)、protochlorophyllide (Pchlide)與崩解代謝物包括chlorophyllide (Chlide)、pheophytin (Phe)、pheophorbide (Pho)等含量變化與崩解途徑及類胡蘿蔔素(carotenoid，Car)含量變化。結果顯示，農林8號Chl及其代謝物、Car之含量皆明顯高於突變種水稻。Chl、Phe及LP Car等極性較小之代謝物，其含量隨著生育日數增加而下降，而口比口林(PPIX、MGPP、Pchlide)、Chlide及MP Car等極性較大之代謝物，其含量亦隨著生育日數增加而下降。五種水稻在生育階段前期，也就是營養生長期至生殖生長期間，葉綠素之崩解途徑傾向以Chl→Phe→Pho為主要崩解途徑(major route)，而以Chl→Chlide→Pho為次要途徑(minor route)；而隨著葉片逐漸成熟與老化，其葉綠素崩解之途徑卻有明顯的不同，亦即在生育階段後期，也就是生殖生長期至成熟期間，葉綠素之崩解途徑傾向以Chl→Chlide→Pho為主要崩解途徑，而以Chl→Phe→Pho為次要途徑。 五種不同葉色水稻進行葉片反射光譜測定，並以反射率計算植生指數(vegetation index)後，估算其葉片之色素含量。由反射光譜分析顯示，五種不同葉色水稻葉片在705nm之反射率對色素含量變化具有最大敏感度。而葉綠素紅光吸收波段675nm之反射率對於色素含量變化之敏感度小，故此一波段不適用於估算葉片色素含量。波長750 nm以上之近紅外光波段對於葉片色素含量敏感度亦小，雖不適用於直接估算色素含量，卻可用於計算植生指數來估算色素含量。以波段705nm及750 nm反射率計算植生指數SRVI與NDVI後，與葉片色素含量進行迴歸分析，可得到最佳的估算模式(R2>0.78)。另以多波段平均反射率計算植生指數後，與色素含量進行迴歸分析亦可得到較佳的估算模式(R2>0.70)。由本研究結果顯示利用反射光譜計算植生指數，非破壞性地估算葉片色素含量之變化進而監測作物生長狀況，可能為有效可行的方法。 另於2006年在花蓮縣玉里鎮(台稉2號)、高雄縣美濃鎮(高雄145號)、彰化縣竹塘鄉(台稉11號)以及桃園縣新屋鄉(台稉14號)，設置水稻產量地真樣區，以不同氮肥等級施用建立產量以及稻米品質差異。利用福衛二號以及SPOT衛星影像資料，建立衛星遙測NDVI植生指數對水稻產量與米質之推估模式。水稻生殖生長期間孕穗期之衛星影像NDVI指數與稻穀產量具有最高相關性，建立衛星遙測NDVI指數推估稻穀產量之最佳預測模式，玉里試區為移植後第56天之Yield (kg/ha) = 7889.7 × NDVI ＋ 710.94；美濃試區為移植後第65天之Yield (kg/ha) = 14654 × NDVI － 364.14；竹塘試區為移植後第59天之Yield (kg/ha) = 14063 × NDVI ＋ 1669.4；新屋試區為移植後第65天之Yield (kg/ha) = 13233 × NDVI ＋ 1768.6。西半部美濃、竹塘及新屋試區合併資料建立之產量最佳預測模式為Yield (kg/ha) = 8602.2 × NDVI ＋ 2845.3，以此時期建立之產量預測模式，經驗證後顯示具有高準確度。西半部美濃、竹塘及新屋試區合併資料建立之產量估算模式，亦具有高準確度，顯示預測模式應用於不同品種及不同地區之可行性。 稻米品質食味值與產量具有負相關，與衛星遙測NDVI指數亦呈現負相關，玉里試區台稉2號，在進入生殖生長期後，其衛星遙測NDVI植生指數與稻米品質食味值之關係，相當密切，水稻生殖生長期間孕穗期之衛星影像NDVI指數建立稻米食味值之推估模式為y=-30.84x + 79.313，r2 = 0.83。在生育期最後階段，成熟期後期，衛星遙測NDVI與稻米食味值之關係亦非常高，其r2達0.9。竹塘試區台稉11號水稻生殖生長期間孕穗期之衛星影像NDVI指數建立稻米食味值之推估模式為y = -88.737x + 79.23，r2 = 0.89，以竹塘試區台稉11號水稻生殖生長期間孕穗期之衛星影像NDVI指數建立稻米食味值之推估模式，經驗證後其平均誤差為11.52，顯示具有高準確度。

We examined the contents of chlorophyll (Chl), biosynthetic intermediates (protoporphyrin IX, PPIX; magnesium protoporphyrin IX, MGPP; protochlorophyllide, Pchlide), degradative intermediates (chlorophyllide, Chlide; pheophytin, Phe; pheophorbide, Pho), and carotenoid in the leaves of rice with five different types of chlorophyll b-deficient or -lacking mutant during their growth and development. The levels of less polar (LP) intermediates such as Chl, Phe and LP Car decreased with increasing growth stage, while the levels of more polar (MP) intermediates such as porphyrins (PPIX, MGPP, Pchlide), Chlide and MP Car were also decreased. The biosynthetic and degradative rate of Chl in rice variety Norin No. 8 was higher than rice mutant type due to smaller amounts of Chl, intermediates and Car. Chl→Phe→Pho was the major route of Chl degradation at vegetative stage in five rice, while Chl→Chlide→Pho was the minor route. When leaves were aging and senescent, Chl→Chlide→Pho was the major route and Chl→Phe→Pho became the minor route of Chl degradation. The reflectance spectra of five rice leaves were recorded. By using of reflectance spectra, vegetation indices were calculated to remotely estimate the pigment content. The signature analysis of reflectance spectra indicated that in the leaves of five rice the maximum sensitivity to pigment concentration was found to be at 705 nm. The minimal sensitivity to pigment concentration coincided with the red absorption maximum of chlorophyll at 675 nm. Therefore, it seemed inappropriate to use this spectral band for pigment estimation. The near-infrared band ranging above 750 nm was not sensitive to pigment concentration, as found for 675 nm. The reflectance at near-infrared band could be used as reference in the calculation of vegetation indices. Vegetation indices calculated using reflectance at 705 nm and 750 nm correlated very well with pigment concentration (correlation R2>0.78). Vegetation indices calculated using broad-band reflectance also correlated well with pigment concentration (correlation R2>0.70). Thus, it appears possible to create indices using reflectance spectra for non-destructive estimation pigment content and monitoring crop growth. The satellite remote sensing normalization difference vegetation index (NDVI) of rice field at booting stage correlated very well with rice yield and rice quality in first period rice. The developed prediction models for rice yield and taste meter value were: yield (kg/ha) = 8602.2 × NDVI + 2845.3; taste meter value = －30.84 × NDVI＋79.313, r2 = 0.83; taste meter value = －88.737 × NDVI＋79.23, r2 = 0.89. The developed algorithms predicting rice yield and rice quality from satellite remote sensing data were validated in Chutang, Shinwu, Meinong and Yuli in 2007. Results indicated that estimation models were high accuracy.