基于作物生长监测诊断仪的双季稻叶面积指数监测模型

为探索作物生长监测诊断仪(Crop Growth Monitoring and Diagnosis Apparatus,CGMD)在不同株型双季稻长势指标监测应用的准确性和适用性,该研究开展了不同株型品种和施氮量的田间试验,采用CGMD获取冠层差值植被指数(Differential Vegetation Index,DVI)、归一化植被指数(Normalized Difference Vegetation Index,NDVI)和比值植被指数(RatioVegetationIndex,RVI),并同步采用高光谱仪(AnalyticalSpectralDevices,ASD)获取冠层光谱反射率,构建DVI、NDVI和RVI;通过比较2种光谱仪获取的植被指数变化特征及相互定量关系,评价CGMD的监测精度,建立基于CGMD的不同株型双季稻叶面积指数(Leaf Area Index,LAI)监测模型,并用独立数据对模型进行检验。结果表明:不同株型品种的LAI、DVI、NDVI和RVI随施氮量增加而增大,随生育进程推进呈"低—高—低"的变化趋势;基于CGMD与ASD的DVI、NDVI和RVI间的决定系数(Determination Coefficient,R2)分别为0.959～0.968、0.961～0.966和0.957～0.959,表明CGMD具有较高监测精度,可替代价格昂贵的ASD获取DVI、NDVI和RVI。基于CGMD植被指数的单生育期LAI监测模型的预测效果优于全生育期,基于CGMD植被指数的松散型品种LAI监测模型的预测效果优于紧凑型品种;基于DVICGMD的线性方程可较好地预测LAI,模型R2为0.857～0.903,模型检验的相关系数(Correlation Coefficient,r)、均方根误差(Root Mean Square Error,RMSE)和相对均方根误差(Relative Root Mean Square Error,RRMSE)分别为0.950～0.984、0.18～0.43和3.95%～9.40%;基于NDVICGMD的指数方程可较好地预测LAI,模型R2为0.831～0.884,模型检验的r、RMSE和RRMSE分别为0.906～0.967、0.24～0.38和5.73%～9.16%;基于RVICGMD的幂函数方程可较好地预测LAI,模型R2为0.830～0.881,模型检验的r、RMSE和RRMSE分别为0.905～0.954、0.25～0.56和7.37%～9.99%。与传统人工取样测定LAI法相比,利用CGMD可实时无损监测双季稻LAI动态变化,可替代SunScan植物冠层分析仪获取双季稻LAI,在双季稻生产中具有推广应用价值。

Model for monitoring leaf area index of double cropping rice based on crop growth monitoring and diagnosis apparatus

The real-time, fast, non-destructive and quantitative monitoring of leaf area index (LAI) is critical for precise regulation population quality of double cropping rice production. The objective of this study was to test the accuracy and adaptability of crop growth monitoring and diagnosis apparatus (CGMD) in double cropping rice of different plant types growth index monitoring and application, and to establish the leaf area index (LAI) monitoring model of double cropping rice based on CGMD. Field experiments were conducted in Jiangxi China in 2016 and 2017, including different plant type cultivars and nitrogen application rates. The differential vegetation index (DVI), normalized difference vegetation index (NDVI) and ratio vegetation index (RVI) were measured at tillering stage, jointing stage, booting stage, heading stage and filling stage with two spectrometers, i.e., CGMD (a passive multispectral spectrometer containing 810 and 720 nm wavelengths) and analytical spectral devices (ASD, a passive hyper-spectral spectrometer containing 325 to 1 075 nm wavelengths). Vegetation indexes change characteristics were compared between CGMD and ASD, and their quantitative relationships were analyzed. The LAI monitoring models for compact and loose plant type cultivars of double cropping rice were established based on CGMD from field experimental dataset in 2016 and then validated using field experimental dataset in 2017. The results showed that the LAI, DVI, NDVI and RVI of different plant type cultivars were increased with increasing nitrogen application rate at different growth stages. All of them showed a "low-high-low" trend with double cropping rice development progress. The determination coefficient (R2) of DVI, NDVI and RVI based on CGMD and ASD were 0.959-0.968, 0.961-0.966 and 0.957-0.959, respectively. This indicated that vegetation indexes based on CGMD and ASD was highly consistent, and the CGMD could be used to replace expensive ASD to measure NDVI, DVI and RVI. The prediction effect of LAI monitoring model at single growth stage based on CGMD vegetation indexes was better than that in the whole stage, and the prediction effect of LAI monitoring model in the loose plant type cultivar based on CGMD vegetation indexes was better than that in the compact plant type cultivar. The linear equation based on DVICGMD could be used to estimate LAI with the R2 in the range of 0.857-0.903, and the correlation coefficient (r), root mean square error (RMSE) and relation root mean square error (RRMSE) of model validation in the range of 0.950-0.984, 0.18-0.43 and 3.95%-9.40%, respectively. The exponential equation based on NDVICGMD could be used to estimate LAI with the R2 in the range of 0.831-0.884, and the r, RMSE and RRMSE of model validation in the range of 0.906-0.967, 0.24-0.38 and 5.73%-9.16%, respectively. The power function equation based on RVICGMD could be used to estimate LAI with the R2 in the range of 0.830-0.881, and the r, RMSE and RRMSE of model validation in the range of 0.905-0.954, 0.25-0.56 and 7.37%-9.99%, respectively. Compared with the normal manual sampling method, using the CGMD can real-time and non-destructive monitoring the LAI dynamic change of double cropping rice. The CGMD could be used to replace SunScan (an expensive plant canopy analyzer can measure LAI) to measure LAI of double cropping rice, which has a potential to be widely applied for precise regulation of LAI and high yield cultivation in double cropping rice production.