基于高光谱的草坪草光合参数模拟估算

Simulation and estimation of photosynthetic parameter of turfgrass based on hyperspectrum

摘要:

【目的】光合参数是衡量草坪草生长状况的重要生理指标，探索基于高光谱技术的草坪草光合参数的模拟估算对于草坪养护管理具有重要意义。【方法】以 3 个常用草坪草品种红象高羊茅（Festuca arundinacea cv. Hongxiang）、百灵鸟多年生黑麦草（Lolium perenne cv. Bailingniao）和肯塔基草地早熟禾（Poa pratensis cv. Kentucky）为试验材料，通过盆栽试验，在草坪草生长旺盛期，使用 SOC710VP 成像光谱仪和 CIRAS‐3 便携式光合仪分别测定了草坪草冠层光谱数据、净光合速率（P_n）和蒸腾速率 （T_r ），筛选与两种光合参数显著相关的原始光谱波段与植被指数，构建偏最小二乘法（PLS）估算模型， 并利用 PLS 模型中变量投影重要性（Variable Importance Projection，VIP）方法筛选 VIP 值＞1. 2 的重要波段与植被指数。【结果】1）共筛选与 P_n显著相关的 54 个原始光谱波段（435、450、460、475、490~ 550、560~565、590~725、990~1 000、1 015~1 030 nm）与 9 个植被指数（GI、NDVI、NDVI₆₇₀、CI、PSRI、 NRI、SIPI、PRI、SR），其中原始光谱 460 nm 与植被指数 CI 相关系数绝对值最高，分别为 0. 46 和 0. 77， 共筛选与 T_r 显著相关的 115 个原始光谱波段（435~440、450~1 010 nm）与 7 个植被指数（SIPI、SR、 NDVI、NDVI₆₇₀、MSR₇₀₅、CI、DVI），其中原始光谱 475 nm 与植被指数 SIPI 相关系数绝对值最高，分别为 0. 61 与 0. 54；2）P_n偏最小二乘法模型因变量方差解释率为 75. 24%，模型拟合精度 R² 为 0. 95，均方根误差 RMSE 为 0. 1，T_r偏最小二乘法模型因变量方差解释率为 73. 43%，模型拟合精度 R² 为 0. 73，均方根误差 RMSE 为 0. 5，可满足反演需求；3）根据偏最小二乘法中变量投影重要性 VIP 法筛选最优指标，得出反演 P_n的最优指标为 CI，T_r最优指标为 SR。【结论】草坪草净光合速率与蒸腾速率的偏最小二乘法光谱反演模型，为草坪草光合指标评估提供了更便利的方案。

Abstract:

【Objective】 The photosynthetic parameters are important physiological indicators for assessing the growth status of turfgrass. It is of great significance for turf maintenance management to explore the simulated estimation of turfgrass photosynthetic parameters based on hyperspectral technology.【Method】 In this experiment，three commonly used turfgrass species，“Hongxiang tall” fescue （Festuca arundinacea cv. Hongxiang），“Bailingniao”perennial ryegrass （Lolium perenne cv. Bailingniao），and “Kentucky” Kentucky bluegrass （Poa pratensis cv. Kentucky），were selected as experimental materials. During the vigorous growth period of turfgrass，spectral data of turfgrass canopy，net photosynthetic rate （P_n），and transpiration rate （T_r ） were measured using the SOC710VP imaging spectrometer and the CIRAS‐3 portable photosynthesis system. The original spectral bands and vegetation indices significantly correlated with the two photosynthetic parameters were selected through pot experiments. Partial least squares （PLS） regression models were constructed，and the Variable Importance Projection （VIP） method was used to screen important spectral bands and vegetation indices with VIP values ＞1. 2 in the PLS model.【Result】1） A total of 54 original spectral bands （435，450，460，475，490~550，560~565，590~725，990~1000 nm，1 015~ 1 030 nm） and 9 vegetation indices （GI，NDVI，NDVI₆₇₀，CI，PSRI，NRI，SIPI，PRI，SR） significantly correlated with Pn were selected. Among them，the absolute values of correlation coefficients between the original spectral band at 460 nm and the vegetation index CI were 0. 46 and 0. 77，respectively. A total of 115 original spectral bands （435~ 440 nm，450~1 010 nm） and 7 vegetation indices （SIPI，SR，NDVI，NDVI₆₇₀，MSR₇₀₅，CI，DVI） significantly correlated with T_r were selected. Among them，the original spectral band at 475nm and the vegetation index SIPI had the highest absolute correlation coefficients of 0. 61 and 0. 54，respectively. 2） The PLS regression model for P_n had a variance explanation rate of 75. 24%，a model fitting accuracy （R² ） of 0. 95，and a root mean square error （RMSE） of 0. 1，while the PLS regression model for T_r had a variance explanation rate of 73. 43%，an R² of 0. 73，and an RMSE of 0. 5，meeting the requirements of inversion. 3） According to the VIP method in the PLS regression，the optimal indicator for estimating P_n was CI，and the optimal indicator for estimating T_r was SR.【Conclusion】 The PLS regression spectral inversion models for the net photosynthetic rate and transpiration rate of turfgrass provide a more convenient solution for the assessment of turfgrass photosynthetic indicators.