基于偏最小二乘法的玉米FPAR高光谱反演模型研究

以ASD FR便携式光谱仪与LI-191SA光量子仪对吉林中西部的玉米田进行多次观测,采集到123组有效数据,基于偏最小二乘法(PLS)对玉米FPAR进行高光谱反演。对可见光与近红外光谱(400～1 500nm)进行分析并建立反演模型,对FPAR预测效果进行验证,验证模型的R2为0.785,RMSE为0.117;同时进行了玉米FPAR与光谱反射率、反射率一阶导数之间的关系分析及植被指数与玉米FPAR之间的回归分析。研究结果表明,PLS方法建立的模型可有效地从玉米高光谱反射率数据反演出FPAR含量,反演结果精度较植被指数高。

Hyperspectral Remote Sensing Modeling of FPAR for Corn Based on Partial Least Squares (PLS) Regression Analysis

Fraction of incident Photosynthetically Active Radiation(FPAR) absorbed by vegetation is an important variable in many environmental processes,i.e.,net primary production,vegetative substance accumulation.Thus accurate estimation of FPAR is critically importance for carbon circulation and biogeochemical modeling.Spectral data over corn canopy were collected with ASD portable spectrometer in the Mid-west of Jilin Province during three year field surveys(2006,2007 and 2010),concurrently FPAR were collected with LI-191 Linear Quantum Sensor.The Partial Least Squares(PLS) regression analysis was applied for hyperspectral inversion of FPAR using in situ collected corn canopy spectra.In the PLS regression analysis,123 samples of spectra data in the visible and near-infrared spectral region(400~1 500 nm) were resembled with an interval of 10 nm,of which 84 samples were used for model calibration while the rest of 39 samples were used to validate the model accuracy.In the PLS regression model,5 latent variables were selected for model calibration and validation,yielding a determination coefficient(R2) of 0.785 and the RMSE of 0.117,respectively.As a comparison,regression models based on first derivative and vegetation index also established for FPAR estimation.The result indicates that the model using PLS regression is effective approach for inversion of corn FPAR through in situ collected canopy hyperspectral reflectance spectra,and the model accuracy is better than that from models based on vegetation index.