Remote sensing chlorophyll-a in the Strait of Georgia.

摘要

The main objective of this thesis was to evaluate the use of Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua imagery to estimate chlorophyll-a (chl) concentrations in the surface waters of the Strait of Georgia, located off the southwest coast of Canada. To meet this objective two components were addressed: (1) evaluate chl algorithms using ship-based radiometric and biophysical measurements, (2) evaluate atmospheric correction methods in conjunction with chl algorithms using MODIS Aqua imagery.;Using unattended fluorometric chl measurements, different atmospheric correction approaches in conjunction with the standard OC3M chl algorithm and modified GSM01 algorithm from Chapter 2 were evaluated in Chapter 3. Atmospheric correction methods that were evaluated included: the standard near infrared (NIR) correction, a shortwave infrared (SWIR) correction, and an adapted version of the correction developed at the Management Unit of the North Sea Mathematical Models (MUMM). The NIR correction with the OC3M algorithm was statistically significant with an adjusted R 2 = 0.759 outside the most turbid portions of the Strait, but had a relatively large RMSE of 0.523, was limited to chl < 21 mg m–3, and only estimated concentrations for 63% of the pixels. Effectiveness of the SWIR correction was limited with the OC3M algorithm because of the low percentage of estimated chl concentrations (21% of the pixels) and a lower adjusted R 2 = 0.572 outside the more turbid portion of the Strait. The adapted MUMM correction was the most effective, using the spatially averaged aerosol properties from the SWIR correction with the fixed NIR water-reflectance ratio defined by the MUMM method. The MUMM correction with the OC3M algorithm, when limited to pixels outside the Fraser River influence, had an adjusted R2 = 0.720 and provided chl estimates for 84% of the pixels. The modified GSM01 algorithm was ineffective with all three atmospheric corrections due to the overcorrection of the 412 nm band. Several possible reasons for this overcorrection were identified, including the presence of absorbing aerosols and absorption from atmospheric NO 2. When comparing spatial-temporal patterns in the MUMM corrected OC3M image with chl measurements, general spatial patterns and temporal trends match, with some explainable exceptions. Firstly, turbidity typically found near Fraser River plume makes the OC3M algorithm ineffective. Secondly, under lower aerosol reflectance conditions, the signal-to-noise ratio of the SWIR bands can make the spatial identification of the aerosol properties difficult.;In Chapter 2, biophysical and above-water reflectance measurements collected in 2006 were used to evaluate the OC3M, standard Garver-Siegel-Maritorena version 1 (GSM01), and a modified version of the GSM01 algorithms for estimating chl concentrations in the Strait. The Strait was generally classified as a case 2 water body, transitioning from chromophoric dissolved organic matter (CDOM) dominant in the central region to possibly particulate dominant in the Fraser River plume region. From these biophysical measurements, results showed that the OC3M algorithm was somewhat effective (r 2 = 0.552) outside the most turbid areas of the Fraser River plume. However, a systematic overestimation of lower chl concentrations was found, which may have been related to the higher CDOM absorption observed throughout the Strait. The standard GSM01 algorithm had moderately good agreement with measured CDOM absorption (r2 = 0.584) and total suspended solids (TSS) concentrations (r 2 = 0.866), but was ineffective at estimating chl concentrations. Localized characterization of the CDOM absorption, through a hyperbolic CDOM model, improved the modified GSM01 results by providing better agreement with measured CDOM absorption (r 2 = 0.620) and TSS concentrations (r2 = 0.935). By limiting the GSM01 algorithm to regions with lower combined CDOM and non-algal particulate absorption, the statistical relationship between measured and estimated chl improved (r 2 = 0.690). The further re-interpretation of phytoplankton absorption from the modified GSM01 algorithm with a two-component phytoplankton model resulted in a chl relationship with an R 2 = 0.702 and a linear slope closer to one. However, due to the nature of the GSM01 algorithm, its effectiveness is dependent the accurate characterization of the absorption and backscattering of the optically significant water constituents, which is not always available. Furthermore, the GSM01 algorithm effectiveness is contingent upon the accurate atmospheric correction of the shorter blue wavelengths within satellite imagery.