基于CASI高光谱数据的作物叶面积指数估算

叶面积指数(LAI)的快速估算对于及时了解作物长势、病虫害监测以及产量评估具有重要意义。利用2012年7月7日在黑河流域张掖市获取的CASI高光谱数据，精确提取出了不同作物的光谱反射率，同时结合地面实测数据，对比分析了宽波段和“红边”植被指数在估算作物LAI方面的潜力，在此基础上，基于波段组合算法，筛选出作物LAI估算的敏感波段，并构建了两个新型光谱指数NDSI和RSI，最后对研究区域作物LAI的空间分布进行了分析。结果表明，在植被覆盖度较低的情况下，宽波段植被指数NDVI对LAI具有较好的估算效果，模型的精度R2与RMSE分别为0.52，0.45(p<0.01);对于“红边”植被指数，由于CI_{red edge}充分考虑了不同的作物类型，其对LAI的估算精度与NDVI一致;利用波段组合算法构建的光谱指数NDSI_{(569.00, 654.80)}和RSI_{(597.6, 654.80)}对LAI估算的效果要优于NDVI与CI_{red edge}，其中，NDSI_{(569.00, 654.80)}主要利用了植被光谱“绿峰”和“红谷”附近的波段，模型估算的精度R2可达0.77(p<0.000 1);根据LAI与NDSI_{(569.00, 654.80)}之间的函数关系，绘制作物LAI的空间分布图，经分析，研究区域的西北部LAI值偏低，需增施肥料。研究结果，可为农业管理部门及时掌握作物长势信息、制定施肥策略提供技术支持。

Estimating Leaf Area Index of Crops Based on Hyperspectral Compact Airborne Spectrographic Imager (CASI) Data

The fast estimation of leaf area index (LAI) is significant for learning the crops growth, monitoring the disease and insect, and assessing the yield of crops. This study used the hyperspectral compact airborne spectrographic imager (CASI) data of Zhangye city, in Heihe River basin, on July 7, 2012, and extracted the spectral reflectance accurately. The potential of broadband and red-edge vegetation index for estimating the LAI of crops was comparatively investigated by combined with the field measured data. On this basis, the sensitive wavebands for estimating the LAI of crops were selected and two new spectral indexes (NDSI and RSI) were constructed, subsequently, the spatial distribution of LAI in study area was analyzed. The result showed that broadband vegetation index NDVI had good effect for estimating the LAI when the vegetation coverage is relatively lower, the R2 and RMSE of estimation model were 0.52, 0.45 (p<0.01), respectively. For red-edge vegetation index, CI_{red edge} took the different crop types into account fully, thus it gained the same estimation accuracy with NDVI. NDSI_{(569.00, 654.80)} and RSI_{(597.60, 654.80)} were constructed by using waveband combination algorithm, which has superior estimation results than NDVI and CI_{red edge}. The R2 of estimation model used NDSI_{(569.00, 654.80)} was 0.77(p<0.000 1), it mainly used the wavebands near the green peak and red valley of vegetation spectrum. The spatial distribution map of LAI was made according to the functional relationship between the NDSI_{(569.00, 654.80)} and LAI. After analyzing this map, the LAI values were lower in the northwest of study area, this indicated that more fertilizer should be increased in this area. This study can provide technical support for the agricultural administrative department to learn the growth of crops quickly and make a suitable fertilization strategy.