不同地域冬小麦叶片SPAD高光谱估算研究

HYPERSPECTRAL-BASED ESTIMATION OF WINTER WHEAT SPAD IN TWO DIFFERENT REGIONS

关键词:  高光谱 冬小麦 SPAD 反演模型 不同地域

Key words:  hyperspectral  winter wheat  SPAD  estimation model  different regions

摘要:

叶绿素（SPAD）是重要的植被生物参量。针对传统的SPAD监测方法效率低的问题，该研究探讨 了利用高光谱遥感技术构建SPAD估算模型，实现大范围的冬小麦叶绿素快速无损检测和预测。首先，选 取我国江汉平原的湖北省潜江市后湖管理区和黄淮海平原的山东省济南市长清区作为研究区域，分别采集 冬小麦生育期的冠层反射光谱和倒二叶片SPAD数据。方差分析发现，两个地区冬小麦SPAD在各生育期 存在显著差异。之后，选取和计算RVI、DVI、NDVI和GRVI四种植被指数，利用回归模型建立了不同地 区的冬小麦叶片SPAD测算模型。模型精度检验表明，NDVI估算模型对两个样区的冬小麦叶片SPAD的总 体估算效果较好，满足精度要求，可以应用于大区域的SPAD遥感估测。该研究结果可以为不同区域大面 积的作物SPAD遥感监测提供技术和方法。

Abstract:

Vegetation SPAD（Soil and Plant Analyzer Development）is one of the most important agronomic pa- rameters for assessing vegetation growth status and health condition. Yet， traditional method for vegetation SPAD monitoring is inefficiency due to chemical methods to determine chlorophyll content for a long time and a complex process. This paper aimed to propose a method of using hyperspectral remote sensing to estimate the winter wheat SPAM at the regional level. Two study areas， located in Jianghan Plain and Huanghuaihai Plain of China， were se- lected for this study. Firstly， canopy spectral reflectance and SPAD of winter wheat at different growing periods were measured by using the ASD FieldSpec 3 and SPAD-502.Secondly，the optimal hyperspectral bands were selected to calculate four hyperspectral vegetation indices（GRVI、RVI、NDVI and DVI）， and then these four vegetation indices were used for regression so as to build the SPAD estimation models for winter wheat. The two re- gions' winter wheat leaf SPAD and four hyperspectral vegetation indices correlation coefficient was between 0.686 and 0.901， and the correlations were overall good. For example， the Houhu region's coefficient was 0.901， and Changqing region's coefficient was 0.873. The model validation using observed SPAD showed that NDVI-based estimation model generally had a higher accuracy in winter wheat SPAD estimation in both study areas. Houhu re- gion optimal estimation model was NDVI index model， and Changqing area was NDVI linear model. It was conclu- ded that the proposed method had the potentials of being used for SPAD estimation for crop monitoring.