基于无人机数字表面模型的冬小麦生物量估算模型构建及迁移能力分析

Construction and transferability of winter wheat aboveground biomass estimation model:A digital surface model-based analysis

生物量(Aboveground Biomass,AGB)是表征作物长势状况的重要参数,快速精准地估算AGB对指导农田精准管理和挖掘生产潜力具有重要意义,而且AGB的估算也能够为粮食安全问题提供重要的数据参考。文中通过设置不同的水氮处理,采用无人机搭载多光谱成像光谱仪获取冬小麦关键生育期影像,提取数字表面模型,采用BP(Back Propagation)神经网络回归方法建立和改进了基于无人机影像提取株高(Height from Digital Surface Model,H_(dsm))的AGB估算模型。结果表明:H_(dsm)与实测株高(H)和AGB相关性较高,直接利用H_(dsm)构建的AGB估算模型R^(2)、RMSE(Root Mean Square Error)和RPD(Residual predictive deviation)分别为0.58、4528.23kg·hm^(-2)和1.25,估算平均值较实测值平均值小,且估算值较为分散。改进的基于比值的AGB估算模型R^(2)、RMSE和RPD分别为0.88、2291.90kg·hm^(-2)和2.75,估算平均值与实测平均值较为相近,模型较直接用H_(dsm)进行AGB估算,精度提高了51.72%。而且,改进的AGB估算模型在不同水处理和不同年份情景下表现出了较强迁移估算能力。数据特征的差异是导致AGB估算模型迁移估算能力不同的关键因子,文中为模型构建和迁移估算作物长势参数时提供了一种前置条件,如果不同的数据集具有相似的直方图特征,模型在新的情景下具有较强的迁移估算能力。

Aboveground biomass(AGB)is an important indicator for characterizing the crop growth conditions.Rapid and accurate estimation of AGB is critical for guiding the management of farmland and achieving the production potential,and can provide vital data for ensuring food security.In this study,by setting different water and nitrogen treatments,an unmanned aerial vehicle(UAV)which equipped with a multispectral imaging spectrometer was used to acquire images of winter wheat during critical growth stages.Then the plant height(H_(dsm))extracted from the digital surface model(DSM)information was used to establish and improve the estimation model of AGB by applying the backpropagation(BP)neural network and machine learning method.The results showed that the correlation between H_(dsm)and H and AGB is higher,the R^(2),root mean square error(RMSE),and residual predictive deviation(RPD)of the AGB estimation model constructed directly using H_(dsm)is 0.58,4528.23kg·hm^(-2),and 1.25,respectively.The estimated mean AGB(16,198.27kg·hm^(-2))is slightly smaller than the measured AGB(16,960.23kg·hm^(-2)).The R^(2),RMSE,and RPD of the improved AGB estimation model based on AGB/H_(dsm)is 0.88,2291.90kg·hm^(-2),and 2.75,respectively,and the estimated AGB(17478.21kg·hm^(-2))is closer to the measured AGB(17222.59kg·hm^(-2)).The accuracy of improved AGB estimation model is increased by 51.72%compared with the AGB estimation model directly using H_(dsm).Moreover,the improved AGB estimation model shows strong transferability under different water treatments and scenarios of different years,but there are differences in transferability under different N level scenarios.Differences in data characteristics are the key factors that lead to different transferability of the AGB estimation models.This study provides an antecedent for model construction and transferable estimation of AGB for winter wheat.And we confirmed that when different datasets have similar histogram characteristics,the model is applicable to the new scenarios.