玉米光合有效辐射分量高光谱估算的初步研究

 【目的】光合有效辐射分量（FPAR）是各种生产力模型、作物估产模型等的重要参数，本文将对高光谱估算FPAR效果作初步探讨分析，为提高FPAR估算及遥感产品验证精度和各种生态模型模拟精度提供科学支持。【方法】本文基于地面实测玉米数据，详细分析了光合有效辐射分量与光谱反射率及其一阶导数之间的相关关系及FPAR估算机理，并利用反射率、一阶导数、植被指数方法研究了玉米冠层FPAR估算效果。【结果】玉米FPAR与整个可见光波段反射率相关性都相对较好，明显好于近红外波段；FPAR与一阶导数的相关关系曲线波动较反射率大，仅520、570、670、805、950和1 010 nm几个波长处一阶导数与FPAR相关性较好。FPAR与典型单波段反射率和一阶导数具有较好拟合关系，确定性系数分别高达0.791和0.882。总的来说，一阶导数法和植被指数法估算FPAR效果较反射率法好些，其中一阶导数的多波段逐步回归分析取得最优的估算效果，R2高达0.944。除红、绿和常用的近红外波段外，375nm附近紫色波段和950nm附近的近红外波段用于估算FPAR也可以得到较理想的结果，特别是水分强吸收的波段具有较好的提高FPAR估算精度的潜力。【结论】一阶导数法和植被指数法估算FPAR效果稍好，充分挖掘高光谱数据估算FPAR潜力，选择最佳波段能够较好得提高FPAR估算精度。

Hyperspectral Estimation of Corn Fraction of Photosynthetically Active Radiation

【OBJECTIVE】Fraction of Absorbed Photosynthetically Active Radiation (FPAR) is one of important variables in many productivity and biomass estimation models, therefore, it is significant to retrieve FPAR accurately for the improvement of model precision. 【METHOD】Based on the field experiment of corn, this paper analyzed the correlations between FPAR and spectral reflectance or the differential coefficient, and discussed the regression of FPAR and the typical spectrum bands reflectance or differential coefficient，which was compared with the regression of NDVI, RVI and FPAR. 【RESULTS】The reflectance of visible bands shows much better correlations with FPAR than near-infrared bands. The correlation curve between FPAR and differential coefficient varies more frequently and greatly than the curve of FPAR and reflectance. Reflectance and differential coefficient both have good regressions with FPAR of the typical single band, with the maximum R2 of 0.873 and 0.882, and have better stepwise regressions of multiple bands (R2 is 0.906 and 0.944, individually). In a word, differential coefficient is a little more effective than reflectance for FPAR estimation. However, normalized difference vegetation index (NDVI) and ratio vegetation index (RVI) show the best regression results, compared to reflectance and differential coefficient. 【CONCLUSION】On the whole, the reflectance and differential coefficient have good relationships with FPAR , and could be used for FAPR estimation effectively.