基于光谱指数和偏最小二乘的棉花类胡萝卜素/叶绿素a比值估算

类胡萝卜素(Car)与叶绿素a含量比值(Car/Chla)的变化与植被生长发育变化、环境胁迫及叶片衰老特征等密切相关,可作为植被生理生态及物候的监测指标。不同植被类型和植被品种其色素变化随植被生长发育呈现出不同的变化特征。为了探究适用于干旱区棉花Car/Chla比值估算的光谱指数和估算方法,本研究通过2011年和2012年连续2年的大面积田间试验,获取了棉花不同生育期的叶片及冠层尺度光谱反射率及色素含量信息,对多种光谱指数及偏最小二乘回归(Partial Least Square Regression, PLSR)用于Car/Chla比值和Car估算进行了探讨。对比表明,基于光化学指数(PhotochemicalReflectanceIndex,PRI)的线性和一元二次模型对Car/Chla比值和Car的估算精度最高,由PRI-Car/Chla线性模型得到的叶片和冠层尺度的Car/Chla比值估算值与实测值之间的决定系数R2大于0.6, PRI-Car的R2大于0.36;基于PLSR模型得到的Car/Chla比值估算值与实测值之间的拟合关系略优于基于PRI的估算模型,由其得到的叶片及冠...

Estimation of cotton Car/Chla ratio by hyperspectral vegetation indices and partial least square regression

Estimating the ratio between carotenoid to chlorophyll a (Car/Chla) provides an additional avenue for the assessment of physiology and phenology of plant growth and development. With the aim of assessing cotton Car/Chla ratio from hyperspectral reflectance, a wide range of carotenoid (Car) and chlorophyll a concentrations, and leaf and canopy reflectance at cotton different growth stages were measured. The performance of a variety of Car/Chla ratio related vegetation indices and partial least square regression (PLSR) for Car/Chla ratio and Car estimation were tested. Among all tested vegetation indices, PRI (Photochemical Reflectance Index) and linear PRI models had the most significant correlations with Car/Chla ratio and Car, and could accurately estimate, Car/Chla ratio (R²_{leaf level} = 0.69 and R²_{canopy level} = 0.67) and Car concentration (R²_{leaf level} = 0.44 and R²_{canopy level} = 0.36). The best estimation of the Car/Chla ratio and Car was provided by PLSR models with R² > 0.80 between the estimated and measured value for Car/Chla ratio and R²= 0.74 for Car. Both reflectance indices and PLSR method were more successful for the estimation of Car/Chla ratio than for that of Car concentration, indicating the promising potential of Car/Chla ratio as a powerful indicator using for plant status monitoring by remote sensing. Besides, accuracy test of models using validation dataset highlighted the remarkable performance of PLSR for Car/Chla (R²_{leaf level} = 0.87 and R²_{canopy level} = 0.84) and Car (R²_{leaf level} = 0.73 and R²_{canopy level} = 0.74) estimated by hyperspectral reflectance at both the leaf and canopy levels. The results further prove the remarkable performance of hyperspectral reflectance for the estimation of Car/Chla ratio, and enrich the parameters for monitoring high temperature stress, water deficit stress, and nutrient stress and pest diseases by remote sensing in cotton.