小波变换的潮滩沉积物含水量预测

潮滩沉积物水分的分布在空间和时间上会有很大的变化，含水量的变化会导致沉积物中生源要素含量的变化。因此，实时、准确、快速的监测潮滩沉积物含水量，对了解潮滩的各种特性，掌握潮滩生源要素信息，潮滩资源的开发有着重要意义。采集青岛市东大洋村潮间带的沉积物115份，分别测定新鲜样品、风干4周、风干8周样品的可见近红外光谱和含水量。以db10小波基和sym6小波基对原始光谱进行小波变换，采用偏最小二乘回归建立潮滩沉积物含水量模型。通过10阶小波变换获取原始光谱的低频信息A_n和高频信息D_n(n=1, 2, …，10)，通过原始光谱S分别与高频信息D_n做差值，得到S-D_n，对A_n，D_n和S-D_n建立潮滩沉积物含水量模型，并对模型结果进行分析。原始光谱建立模型的R2_p为0.841，RMSEP为2.767，RPD值为2.481。通过对db10小波基变换后的低频和高频信息分析，无用信息主要集中在D₃和D₄，去除D₃和D₄建立的含水量模型，相比于原始光谱模型精度有明显提高，R2_p为0.878，RMSEP为2.501，RPD值为2.749；通过sym6小波基变换后进行分析，无用信息主要集中在D₅和D₉，去除D5和D9建立含水量模型与原始光谱模型相比，精度也有一定提高，R2_p为0.87，RMSEP为2.475，RPD值为2.768。因此通过小波变换对原始光谱划分低频信息和高频信息进行分析，能够有效找到潮滩沉积物含水量的干扰信息，实现特征信息提取，从而建立准确度更高的潮滩沉积物含水量模型，为潮滩沉积物含水量实时、动态监测提供理论基础。

Prediction of Tidal Flat Sediment Moisture Content Based on Wavelet Transform

The distribution of water in flat tidal sediments will change greatly in space and time, and the changes will lead to the changes of biogenic elements in sediments. Therefore, the tidal flat sediment water content data are monitored in real time, accurately and quickly, which is of great significance to understanding the tidal flat characteristics, grasp the information of tidal flat biogenic elements, and develop tidal flat resources. This paper collected 115 samples of intertidal sediments from Dongdayang village, Qingdao city. The visible near-infrared spectra and moisture content of fresh samples, air-dried for 4 weeks and 8 weeks were measured. The db10 and sym6 wavelet basis were used to transform the original spectrum, and partial least squares regression was used to establish the tidal flat sediment moisture content model. The low-frequency information An and high-frequency information D_n (n=1, 2, …, 10) of the original spectrum were obtained by 10 order wavelet transform. S- D_n was calculated by the difference between the original spectrum S and D_n. The moisture content models were established using A_n, D_n and S- D_n, respectively, and the results were analyzed. The original spectrum model’s R2_P, RMSEP and RPD were 0.841, 2.767 and 2.481. By analysing low-frequency and high-frequency information, after db10 wavelet basis transforms, the useless information was mainly concentrated in D₃ and D₄, and the accuracy of the moisture content model established by removing D₃ and D₄ was significantly improved, R2_P was 0.878, RMSEP was 2.501, RPD was 2.749. Through the analysis of sym6 wavelet basis transform, the useless information was mainly concentrated in D₅ and D₉, the R2_P, RMSEP and RPD by removing D₅ and D₉ were 0.87, 2.475 and 2.768. Therefore, by analyzing the low-frequency and high-frequency information using wavelet transform, the interference information of sediment moisture content can be effectively found, and the feature information can be extracted. The more accurate the tidal flat sediment moisture content model is established, it provides a theoretical basis for real-time and dynamic monitoring of tidal flat sediment moisture content.