基于HJ-CCD与TM影像的水稻LAI估测一致性分析

Analysis of consistency between HJ-CCD images and TM images in monitoring rice LAI

针对传感器参数与Landsat-5 TM较为相近的环境减灾小卫星HJ-CCD影像可否适用于水稻叶面积指数(LAI)估测的问题,该文利用同为2009-07-13获取的安徽省境内相同范围的HJ-CCD影像和TM影像,结合地面同步调查数据,分别从影像原始波段相关性、反演模型精度及LAI空间分布趋势3个方面,对两类数据在水稻LAI估测上的一致性进行分析和评价。结果表明,两类影像原始波段反射率相关性较高,基于两类影像的单变量、多变量模型在精度上一致性较高,估测所得的LAI在空间分布上也具有较高的一致性,数据集中分布的区间较为一致。因此,该研究初步证明将HJ-CCD影像用于水稻LAI估测在总体上是可行的。但同时,由于基于两类影像构建的模型在形式上存在一定程度的差异,据此得到的LAI填图结果的数据分布在某些区间内亦略有差异。因此,在应用中应注意针对不同数据源进行独立建模分析,不可相互套用模型。该研究一定程度上为有着重访周期短、覆盖范围宽等优势的国产中分辨率遥感数据在农业领域的应用提供依据。

This study aims to assess the capability of the Environment and Disaster Reduction Small Satellites（HJ-CCD） images in monitoring of rice leaf area index（LAI） in terms of comparing it with the widely used Landsat-5 TM images,which has the similar spatial resolution and band wavelength ranges.On July 13th,2009,a field investigation was conducted which exactly corresponded with the acquiring timing of a scene of TM image and a scene of HJ-CCD.The consistency of performance was evaluated in terms of the correlation of raw band reflectance,the accuracy of reversion model as well as the spatial distribution pattern of predicted LAI.From the results,a high level of correlation can be observed for raw band reflectance.The predicted accuracies of reversion models in both forms of single variable model and multi-variable model were rather approaching for HJ-CCD and TM images,which thus yielded a highly uniform spatial distribution and data distributed pattern of predicted rice LAI.Therefore,the conclusion may be safely drawn that the HJ-CCD image is feasible and suitable for rice LAI monitoring.Meanwhile,it should be noted that a certain degree of discrepancy was also existed in the model form as well as the data range of predicted rice LAI.For that reason,it is suggested that the reversion model of rice LAI should be built and applied specifically in the light of a certain type of image.In general,this study provides some important evidences that the indigenous remotely sensed data which owned the advantages as the higher revisit frequency and wider scene swath are able to satisfy several monitoring and assessing tasks in the field of agriculture.