Effects of Sensor Noise in Spectral Measurements on Chlorophyll-a Retrieval in Nanhu Lake of Changchun, China

Abstract

We present an algorithm from spectral measurements of upwelling water light to estimate the concentration of optically active material (OAM) such as chlorophyll-a (chl-a). The problem of chla concentration retrieval is modelled using a simulated hyperspectral sensor based on spectral measurements. Sensor noise is taken into account during the algorithm development for the best linear estimation of concentration mean and residual variance. In addition, it requires the joint distribution of radiance spectra and chl-a concentration in the algorithm development. This distribution can be obtained by simulation using water-leaving radiance models. The result shows that sensor noise greatly decreases the accuracy of chla concentration retrieval. But the accuracy can also be improved if a priori information concerning observation conditions is known and included in the algorithm. The effects of sensor noise in spectral measurements on retrieval of chl-a concentration are analysed and assessed in the study.This method is applied to the optical properties of the eutrophic water in Nanhu Lake of Changchun, China. The eutrophic water of Nanhu Lake is mainly impacted by the input from rivers and drainage systems from the area of Changchun, which discharge a high concentration of nutrients, mineral suspended solids, and dissolved organic matter (DOM or referred to yellow substance). We measured reflectance spectra (400–900 nm) of Nanhu Lake using the ASD-field spectroradiometer from April to October in 2004. The application of this algorithm shows that neglecting the sensor noise greatly underestimates the accuracy of chl-a retrieval in the study.