由植被高解析反射光譜模式化稻株之氮素含量

Abstract

本研究量測的水稻植被反射光譜具有一般綠色植物光譜曲線的類似特性，而不等量氮肥的施用將改變水稻的植被光譜曲線；稻株葉片及稈的含氮量百分比在生育期問的改變，無論一、二期稻作，均隨著生育的進展出現下降的趨勢。本研究發現，葉片氮素含量與稈氮素含量的相關強度曲線變化雷同，顯示各窄波段反射比對於葉片及稈的氮素含量具有相似的敏感反應，然而一期稻作與二期稻作的相關強度表現卻不一致。整體來說，一期稻作的反射光譜各窄波段反射比與氮素含量的相關性不高，相關係數介於-0.5至0.5之間；二期稻作則在近紅外光區至中紅外光區具有較大的相關性，尤其在1607 nm 窄波段位置的相關係數之絕對值最大，可獲得近75 ％的變異估測準確度。經比較各種特徵波段之反射比值組合及光譜特徵與氮素含量的關係，發現仍以二期稻作才出現顯著相關，在red edge slope 。為二次曲線關係，在red edge span / red edge location 則為一次直線關係。又根據研究結果，一二期稻作均可利用多元線性複迥歸模式估測葉片及稈的氮素狀態，其中適用於一期稻作全生育期的植株葉片及稈氮素含量估測的最佳三元（403 nm 、505 nm 及710 nm ）直線複迴歸模式決定係數（R2）分別可達0.734及0.717 (P<0.0001)，二期稻作相對的最佳三元（403 nm 、558 nm 及1607 nm ）直線複迴歸模式決定係數則分別可達0.814及0.812 (P<0.0001）。經比較這些三元直線複迴歸模式的估測值與實測值後，發現相關係數大於0.93 ，顯示無論葉片或稈的氮素含量，模式估測可達近於1:1的驗證結果。 Results indicated that reflectance spectra of rice canopy were similar to other green-like plants. Applications of different amounts of nitrogen fertilizer changed the spectral patterns, but did not changed the decreasing trend of nitrogen contents in leaves and shoots of rice following plant development in both cropping seasons. Correlation intensity curve between spectral reflectance and nitrogen content in 340-2400 nm was similar in leaves and shoots, suggesting that responses of reflectance of rice spectrum to nitrogen contents in leaves and shoots were quite alike. However, the correlation intensity curve of First Crops was not conformed to that of Second Crops. In general, change of correlation coefficient was in between ¡X0.5 and 0.5 in First Crops, yet the values were much higher in Second Crops. At 1607 nm for Second Crops, spectral reflectance may estimate more than 75% variability of nitrogen contents in leaves and shoots. Among the tested spectral indices and spectral characteristics, only red edge slope and red edge span/red edge location ratio had close correlation with nitrogen content in the second cropping season. The red edge slope was fitted to a quadratic function while red edge span/red edge location ratio was fitted to a linear function. Results also indicated that estimation of nitrogen content in leaves and shoots of rice plants may be improved by using multiple linear regression (MLR) models. The correlation coefficients for the relationships between the estimated and the measured values of nitrogen contents were greater than 0.93. The spectral wavebands suitable for the best-3 MLR models were 403 nm, 505 nm and 710 nm for First Crops and were 403 nm, 558 nm and 1607 nm for Second Crops, respectively.