| 摘要: |
| 在淤泥质海岸，了解潮滩表层沉积物的粒度空间分布特征与粒径运移趋势，是认识潮滩水沙过程、冲淤演变和地貌演化的重要手段。针对传统粒径趋势分析空间范围有限、现有遥感反演模型形式简单且精度难以提高的问题，论文研究并实现了一种基于遥感粒度参数驱动的潮滩沉积物粒径运移趋势分析方法，首先利用PCA（Principal Component Analysis）主成分分析法从HJ-1A多光谱遥感影像中提取反演因子，采用小波神经网络（Wavelet Neural Network，WNN）模型结合野外采样数据进行参数训练与建模，反演沉积物粒度参数的空间分布；然后以遥感粒度参数驱动GSTA（Grain Size Trend Analysis）沉积物粒径趋势分析模型，实现了淤泥质潮滩表层沉积物的粒径运移趋势模拟。该方法在江苏中部淤泥质海岸的精度验证结果表明：平均粒径、分选系数、偏态的模型检验组数据10次运行结果平均绝对误差分别为0.22φ、0.15、0.42，平均相对误差分别为5.32%、12.47%、14.59%；三个粒度参数的变异系数值变化范围较稳定。与已有的遥感模型相比，平均粒径反演精度接近，而且分选系数、偏态的反演精度有较大提高。遥感反演与实测粒度参数模拟的粒径运移趋势矢量相似性系数为0.67，矢量长度差小于0.4的矢量占80.74%，角度差小于90°的矢量占84.31%，两者有较高的相似性。在潮滩不同位置，沉积物粒径运移趋势总体呈现不同的规律性特征，与当地水动力条件较为吻合。该方法基于遥感技术实现，为大范围的潮滩沉积物粒度特征分析与粒径运移趋势研究提供了一种快速且有效的途径。 |
| 关键词: 潮滩 沉积物 遥感 粒度参数 粒径运移趋势 小波神经网络 主成分分析 |
| DOI：10.11693/hyhz20190500092 |
| 分类号:P7 |
| 基金项目:国家自然科学基金项目，41771447号；江苏省海洋科技创新专项项目，HY2018-3号。 |
|
| REMOTE SENSING STUDY ON SEDIMENT GRAIN SIZE DISTRIBUTION AND ITS MIGRATION TREND ANALYSIS IN TIDAL FLAT BASED ON PCA AND WNN MODEL |
|
LIU Xing-Xing1, ZHANG Dong2,3, HAN Fei1
|
|
1.Department of Geography, Nanjing Normal University, Nanjing 210023, China;2.College of Marine Science and Engineering, Nanjing Normal University, Nanjing 210023, China;3.Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing 210023, China
|
| Abstract: |
| Understanding the spatial characteristics and particles movement in the surface sediments of tidal flats is important to recognize tidal flat water and sedimentation processes, erosion-deposition evolution, and geomorphological evolution. Traditional analysis of particle size trend is limited in spatial range; and the current remote sensing inversion model is simple in form and low in precision. In this paper, a trend analysis method of sediment particle size migration based on remote sensing particle size parameters is studied. First, inversion factors are extracted from HJ-1A multispectral remote sensing images using principal component analysis (PCA) with wavelet neural network in combination with field data for parameter training and modeling, by which the spatial distribution of sediment particle size parameters is inverted. Second, the remote-sensing particle-size parameter drives the trend analysis model to simulate the particle size migration trend in the muddy tidal flat. The results of accuracy verification for muddy coast in central Jiangsu show that the mean absolute errors of average grain size, sorting coefficient, and skewed model of 10 test runs are 0.22φ, 0.15, and 0.42, and the average percentage errors are 5.32%, 12.47%, and 14.59%, respectively; and the variation coefficient values of the three parameters are relatively stable. Compared with the results of previous models, the inversion accuracy of average particle size is very close, and those of the sorting coefficient and skewness have been increased greatly. The vector similarity coefficient between inverted particle size transport trend and measured one is 0.67, vectors with length difference of <0.4 account for 80.74%, and those with angular difference of <90° for 84.31%, indicating a high similarity. Moreover, in different locations of the tidal flat, the migration trend of sediment particle size shows different but regular characteristics, which is consistent with the local hydrodynamic conditions. This approach provides a fast and effective tool for studying grain size characteristics and particle size migration trend of tidal beach sediments in a large scope. |
| Key words: tidal flat sediment remote sensing particle size parameter particle size trend wavelet neural network principal component analysis |
