Measurement of light absorption by aquatic particles: improvement of the quantitative filter technique by use of an integrating sphere approach.

Determination of particulate absorption in natural waters is often made by measuring the transmittance of samples on glass-fiber filters with the so-called quantitative filter technique (QFT). The accuracy of this technique is limited due to variations in the optical properties of the sample/filter composite, and due to uncertainties in the path-length amplification induced by multiple scattering inside the filter. Some variations in the optical properties of the sample/filter composite can be compensated by additional measurements of the filter's reflectance (transmittance-reflectance method [T-R] [S. Tassan and G. M. Ferrari, Limnol. Oceanogr. 40, 1358 (1995)]). We propose a different, rarely used approach, namely to measure the filter's absorptance in the center of a large integrating sphere, to avoid problems with light losses due to scattering. A comparison with other QFTs includes a sensitivity study for different error sources and determination of path-length amplification factors for each measurement technique. Measurements with a point-source integrating-cavity absorption meter were therefore used to determine the true absorption. Filter to filter variability induced a much lower error in absorptance compared to a measured transmittance. This reduced error permits more accurate determination of the usually low absorption coefficient in the near IR spectral region. The error of the T-R method was lower than that of the transmittance measurement but slightly higher than that of an absorptance measurement. The mean path-length amplification was much higher for the absorptance measurement compared to the T-R method (4.50 versus 2.45) but was found to be largely independent of wavelength and optical density. With natural samples the path-length amplification was less variable for the absorptance measurement, reducing the overall error for absorption to less than ±14%, compared to ±25% for the T-R method.