摘要
水分是影响冬小麦生长发育的重要指标,目前可以利用高光谱数据建模来对其进行预测诊断.但在冬小麦生育初期,这类预测模型精度较低,为解决这个问题,利用高光谱数据分析方法,结合叶面积指数,对基于高光谱的冬小麦水分状况预测模型进行优化.结果显示,冬小麦叶面积指数会随着灌水处理的不同产生显著差异;优化后,光谱与冬小麦水分状况相关的敏感波段范围在450-500 nm、620-690 nm和780 nm左右;相对于土壤含水率而言,优化模型对植株含水率有更好的预测精度;引入叶面积指数进行优化提高了冬小麦在生育前期的预测模型精度,使模型精度从0.2提升到0.4以上,并且还提高了基于原始光谱反射率模型的精度;最终获得的模型中,基于原始光谱反射率R780的植株含水率预测模型拟合精度最高,为0.862,基于光谱指数R(810,460)的植株含水率诊断模型验证效果最好,均方根误差(RMSE)为4.341,平均绝对误差(MAE)为2.361;基于光谱指数VARI700的土壤含水率预测模型验证效果最好,均方根误差(RMSE)为4.506,平均绝对误差(MAE)为6.293.本研究表明利用叶面积指数优化模型可以很好地提高模型精度,在土壤含水率预测模型方面优化尤其显著,这为基于高光谱的水分状况预测模型构建与实际应用提供了新思路.(图6表5参31)
Model accuracy is low when using hyperspectral imaging to predict the water conditions of winter wheat.In this study,hyperspectral data analysis and the leaf area index are used to optimize the prediction model of winter wheat water conditions based on hyperspectral imaging.The leaf area index of winter wheat differs with different irrigation treatments,and the sensitive band range of the winter wheat water conditions are approximately 450-500 nm,620-690 nm,and 780 nm.Compared to the soil moisture content,the optimization model has better prediction accuracy for the plant moisture content.Introduction of the leaf area index improved the accuracy of the model from 0.2 to more than 0.4,and the accuracy based on the original spectral reflectance model also improved.In the final model,the fitting accuracy of the model based on the original spectral reflectance at 780 nm was the highest(0.862).The model based on the spectral index R(810,460)had the best validation effect,with a root mean square error(RMSE)of 4.341 and a mean absolute error(MAE)of 2.361.The model based on the spectral index VARI700 had the best validation effect,with a RMSE of 4.506 and a MAE of 6.293.This study demonstrates model optimization with leaf area index that improved the model accuracy,especially in the predictive model of soil moisture content.This study provides new information for the construction and practical application of hyperspectral water condition prediction models.
作者
白青蒙
韩玉国
彭致功
刘露
林少喆
BAI Qingmeng;HAN Yuguo;PENG Zhigong;LIU Lu;LIN Shaozhe(Key Laboratory of Water and Soil Conservation Bureau of Beijing Forestry University,Beijing 100083,China;National Key Laboratory for Water Circulation Simulation and Regulation of China Institute of Water Resources and Hydropower Research,Beijing 100038,China)
出处
《应用与环境生物学报》
CAS
CSCD
北大核心
2020年第4期943-950,共8页
Chinese Journal of Applied and Environmental Biology
基金
国家重点研发计划项目(2018YFC047703)
中国水利水电科学研究院基本科研业务费专项(ID0145B082017,ID0145B742017,ID0145B492017)资助。
关键词
高光谱
冬小麦
水分状况
叶面积指数
预测模型
hyperspectral
winter wheat
water condition
leaf area index
prediction model
