Font Size: a A A

Retrieval Of Water Quality In The Pearl River Estuary Using Hyperspetral Technique

Posted on:2008-09-26Degree:DoctorType:Dissertation
Country:ChinaCandidate:Q G XingFull Text:PDF
GTID:1100360242455421Subject:Physical oceanography
Abstract/Summary:PDF Full Text Request
Several cruises of investigation were carried out in the waters of Pearl River Estuary. The techniques of boat-based in-situ spectrum measuring, relationships among water quality parameters and high-resolution spectra, and fresnel effects of sky light on the interface of air and water, were discussed. Based on the work of air radiation simulation using MODTRAN, the possibility to remove atmospheric effects was analyzed, and the concept on atmospheric windows of derivative spectrum was introduced, too. Finally, the retrieval models were established to estimate the water quality parameters of Pear River Estuary waters by high-resolution derivative spectrums. The key conclusions are as below.For the in-situ work of spectrum-measuring in the Case-II waters, especially in the Pearl River Estuary, the above-water measuring technique is recommended. When on a boat, the height of sensor can vary in a scope of 1-5m above the sea surface, and the atmospheric effects can be ignored.When high-resolution spectrum (1nm) is resampled to low-resolution spectrum (10nm), the shape of original spectrum does not change much, and there is no obvious change for the coefficient between band value and water quality index. But it is not the same to its derivatives. Generally, the correlation relationship between the water quality parameters and high/low resolution derivative bands can not be applied to each other, although their central wavelengths are the same. The original spectrums should be resampled to lower resolution first, and then processed to derivative spectrum.For the water quality parameters of chlorophyll, TSM, turbidity, COD, DO, nitrate, phosphate, the largest coefficients between derivative bands and them are higher than that between original spectral bands and them. Generally, derivative spectrum is more sensitive to water quality changes than original spectrum, especially for the optically sensitive parameters, such as, chlorophyll, TSM, and turbidity. The coefficient is low between original or derivative spectrum and DO, nitrate, and phosphate which are usually regarded as optically non-sensitive factors. COD is a comprehensive index, and it has the characteristics of optically sensitive and non-sensitive water quality parameters.For the water quality parameters of Pearl River Estuary waters, simple linear /non-linear retrieval models were established upon the in-situ derivative spectrum. The model for chlorophyll can be used to estimate the chlorophyll concentration when the concentration ranges from a few to tens of units (μg/L), R2>0.7. But relatively large errors will be found when chlorophyll concentration is low. The R2 value of regression model for COD and phosphate ranges from 0.5 to 0.7, and the averaged relative error of estimation is lower than 30%. For the models of DO and nitrate, the R2 value is under 0.5, and the averaged relative error is also less than 30%.There is a high linear correlation relationship between TSM and turbidity, and their retrieval model is similar to each other. The averaged relative error is under 30%, and they can indicate the water quality parameters in the range from a few to a hundred above units (mg/L, NTU). These models are expected to be used in actual monitoring.
Keywords/Search Tags:water quality, retrieval, hyperspectrum, derivative spectrum, Pearl River Estuary
PDF Full Text Request
Related items