頁籤選單縮合
| 題 名 | 利用植被反射光譜特徵估測稻株氮素含量=Estimation of Nitrogen Content in Rice Plants from Spectral Characteristics of Canopy Reflectance Spectrum |
|---|---|
| 作 者 | 張芳銘; 楊純明; 張愛華; | 書刊名 | 中華農業研究 |
| 卷 期 | 53:1 2004.03[民93.03] |
| 頁 次 | 頁63-74 |
| 分類號 | 434.113 |
| 關鍵詞 | 植被反射比光譜; 光譜特徵; 氮素含量; 相關係數; 多元線性複迴歸模式; Canopy reflectance spectrum; Spectral characteristics; Nitrogen content in rice; Correlation coefficient; Multiple linear regression model; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | 本研究旨在量測近地面水稻植被高解析反射光譜,篩檢與稻株氮素濃度(重量百分比)有關的光譜特徵,並建立估測氮素濃度的光譜遙測模式。試驗結果發現水稻植被反射光譜類似於一般綠色植物的典型光譜,而施用不等量氮肥將改變水稻光譜曲線。無論一、二期稻作,大致上稻株葉片及全株的含氮百分比均隨著生育的進展而出現下降的趨勢。經反射光譜之各窄波段反射比與含氮百分比進行簡單直線相關分析,顯示各窄波段反射比對於葉片與全株的氮素濃度具有相似的敏感反應,惟各窄波段之反應強度不一,且一、二期稻作之間存在差異。整體來說,植被反射光譜各窄波段反射比與氮素濃度的相關性在一期稻作低於二期稻作,然兩期作皆以近紅外光波段(740-1300 nm)至短波紅外光波段(1300-1800 nm)的相關性高於其他波段;其中一期稻作以1240 nm窄波段位置之相關係數絕對值最高,然決定係數(R²)低於0.18,顯示存在相當變異,二期稻作以1208 nm窄波段位置之相關係數絕對值最高,R²可達0.70以上。續以多元線性複迴歸分析篩檢出與葉片及全株含氮百分比有關的多個特徵窄波段,發現兩期稻作均可利用403 nm、558 nm及1607 nm之反射比組成最佳的三元直線複迴歸模式,進行對葉片及全株氮素濃度的估測。模式估測值對一期稻作之稻株葉片及全株含氮百分比的估測準確度分別提升至R²=0.763(P<0.0001)及R²=0.738(P<0.0001),對二期稻作則分別提升至R²=0.790(P<0.0001)及R²=0.803(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,發現機差均方根(RMSE)均小於0.548,斜率則均大於0.94,顯示無論葉片獲全株的氮素濃度,模式估測可達近於1:1的比較結果。 |
| 英文摘要 | This study was to measure near-ground remotely sensed canopy hyper-spectral reflectance data during the growing periods under varied treatments of nitrogen (N) fertilizer and to identify spectral characteristics in related to N concentration so as to establish spectral models for estimating N status in rice plants. Results indicated that reflectance spectra of rice canopy were similar to other green plants. Applications of different amounts of N fertilizer changed the spectral patterns, but did not change the decreasing trend of N concentrations in leaves and shoot following plant development in both cropping seasons. Correlation intensity curve between spectral reflectance in range of 340-2400 nm and N concentration was similar in leaves and shoot, suggesting that responses of reflectance along the spectral domain to N concentrations in leaves and shoot were quite alike. However, the correlation intensity curve of the first crops was not conformed to that of the second crops, and varied correlation coefficients existed in different narrow bands. In general, correlation coefficients from near-infrared (740-1300 nm) to short-wave infrared region (1300-1800 nm) had larger values than the other bands. Change of correlation coefficient was in between -0.5 and 0.5 in the first crops, with the maximum absoulte value of correlation coefficient located at 1240 nm (R²<0.18). The values of correlation coefficient in the second crops had greater variability, with the maximum absolute value found at 1028 mn (R²>0.70). Results also indicated that estimation of N concentrations in leaves and shoot of rice plant may be improved by using multiple linear regression (MLR) models. The spectral wavebands suitable for the best three variables MLR models were 403 nm, 558 nm and 1607 nm for both cropping seasons. The slopes for the relationship between the estimated and the measured values of N concentrations were greater than 0.940, with the root mean square error (RMSE) less than 0.548. |
本系統中英文摘要資訊取自各篇刊載內容。