多背景条件下铁矿区降尘量的光谱反演实验研究

我国一些矿区的粉尘污染比较严重，利用遥感技术可实现全面、快速的降尘量监测.在铁矿区开展四类典型背景下的级差降尘光谱测量实验，研究不同背景下降尘量的反演精度差异.结果显示，对于高光谱模型(350~2500nm)，基于900nm光谱吸收指数(SAI)建模时植物背景反演精度最高(误差为4.92g/m²)；基于统计方法获得优势波段进行反演，发现植被和油毡楼顶的反演精度较高(误差分别为6.02和7.35g/m²).对于多光谱模型(按Landsat-8 OLI波段设置)，亦是植物和油毡楼顶背景在第7波段反演精度较高(误差分别为6.19和7.93g/m²).综合来看，植物背景下的降尘量反演精度最高，可作为降尘遥感监测的首选；若受限于生长季，可优先选择油毡楼顶背景.在无高光谱数据时，则以多光谱的第7波段为首选反演波段.

Experimental Study on Spectral Retrieval of Dustfall in Iron Ore Areas Under Multi-background Conditions

Dust pollution is serious in some mining areas. Remote sensing could monitor dustfall comprehensively and rapidly. Four kinds of typical backgrounds were selected to carry out iron dustfall spectroscopy measurement to study the accuracy difference of dustfall retrieval. The results show that the retrieval accuracy of plant background was the highest(error is 4.92g/m²) based on 900nm spectral absorption index(SAI)in the hyperspectral model(350~2500nm). By retrieving from the dominant band obtained by the statistical method， the retrieval accuracy on the plant and linoleum roof was relatively high(error is 6.02 and 7.35g/m²， respectively). For the multi-spectral model(according to Landsat-8 OLI bands)， the retrieving accuracy on the plant and linoleum roof was also high in the 7th band(error is 6.19 and 7.93g/m²， respectively). In summary， the dustfall retrieval accuracy under the plants background is the highest， which can be used as the first choice for dust fall remote sensing monitoring; if it is limited by the growing season， the background of linoleum roof can be the first choice. If there is no hyperspectral data， the 7th band of multispectral should be the first choice.