|Title:||由植被高解析反射光譜模式化水稻之產量||Other Titles:||Modeling Rice Yield Using Hyperspectral Reflectance Data||Authors:||陳榮坤
|Keywords:||遙測;標準差植被指數;葉面積指數;產量;模式化;Remote sensing;Normalized difference vegetation index;Leaf area index;Yield;Modeling||Issue Date:||Sep-2003||Publisher:||農業試驗所||Start page/Pages:||25-34||Source:||農業試驗所特刊第105號||Conference:||水稻精準農業體系||Abstract:||
為利用光譜遙測技術具有之非破壞性而即時大面積快速偵測優點，必先瞭解被測物反射光譜特性及其外觀物理徵狀與光譜特徵之關係，同時建立對應之光譜特徵模式，才能據以進行監測而提供被測物的現況資訊。本研究以不等氮肥施用量處理來建構不同植被狀態及產量表現之水稻族群，以探討標準差植被指數（normalized difference vegetation index , NDVI ；或稱正規差植生指數）、葉面積指數（leaf area index , LAI）及產量間的關係，試以利用植被高解析反射光譜模式化水稻的產量。試驗發現水稻全生育期的NDVI 及LAI 在不同年度和期作均呈現凸形曲線分佈，而NDVI 與LAI 之間則呈現指數曲線關係（RZ = 0 . 601 , P < 0 . 0001 ) ，顯示由NDVI 及LAI 相互推估的可行性。又比較抽穗期問量測的LAI 與水稻產量之關係，發現水稻產量可利用抽穗期問的LAI 予以預測，其中一期稻作為抽穗前後15 天之累加LAI 或平均LAI ，二期作則為抽穗前15 天與抽穗後10 天之累加LAI 或平均LAI 。綜合本研究結果，顯示可藉由水稻生育期間光譜遙測資料計算的NDVI 推估當時的LAI ，再以抽穗期間LAI 進行對產量的預測，達到利用光譜遙測技術在追蹤水稻族群生長狀態及預測產量的目的。惟試驗亦發現，由於NDVI 與LAI 的指數關係，限制了在高NDVI 臨界值的使用。
The optical properties of target canopy and the relationships between spectral characteristics and biophysical characters of crop should be clarified and well defined before using the advantages of spectral remote sensing to nondestructively detect and monitoring the timely information of growth status over large area. This study used different nitrogen fertilizer application rates to establish varied canopy status and yield performance so as to investigate the relations among normalized difference vegetation index (NDVI), leaf area index (LAI) and rice yield, and to modeling rice yield from ground-based remotely sensed hyperspectral reflectance data. It showed that changes of MDVI and LAI during the growing periods of rice, both in the first cropping and the second cropping seasons, were a sigmoidal pattem with the maximum observed near heading. The relationship between NDVI and LAI was fitted to an exponential function (R2=0.60 1, P<0.000 1), indicating that NDVI and LAI may be mutually estimated. By comparing the values of LAI before and after heading with yields at harvest, it was found that rice yield may be assessed from LAI near heading. For the first crops. the mean and the accumulated LAI from 15 days before heading to 15 days after heading were closely correlated to yield. It was the mean and the accumulated LAI from 15 days before heading to 10 days after heading showing close relation to yield for the second crops. Results from this study suggest that NDVI calculated from reflectance spectra measured during rice growth may be used to estimate LAI, which collect near heading may then be used to predict yield at harvest. However, the exponential function of NDVI and LAI limits the application of these two parameters in the upper end of the curve where the 'saturation effects' occur and NDVI was insensitive to change of LAI.
|Appears in Collections:||作物組|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.