Development and application of a remote sensing-based Chlorophyll-a concentration prediction model for complex coastal waters of Hong Kong

• Chl-a estimation using coincident Landsat and in situ datasets over a 13-year period.
• Phytoplankton species important in selection of relevant spectral bands.
• Error of 1% in surface reflectance can produce an error of 6% in Chl-a estimation.
• Model can detect and quantify higher Chl-a concentrations including red tides.
• Model can be adopted for routine monitoring of Chl-a in Hong Kong’s coastal region.

SummaryDespite recent advances in estimation of water quality parameters using satellite remote sensing, the estimation of Chlorophyll-a (Chl-a) has remained problematic due to optical complexity of coastal waters and imprecise atmospheric correction of imagery. Local environmental agencies require frequent measurement and monitoring of Chl-a over coastal regions at detailed level, for water quality assessment and control. To monitor Chl-a around the complex coastal waters of Hong Kong using remote sensing, 27 Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) images over a 13-year period from January 2000 to December 2012, were used, along with 120 in situ Chl-a samples. Atmospherically corrected Landsat TM/ETM+ bands 1–4 along with in situ Chl-a data were used to develop and validate regression models for a Chl-a concentration range of 0.3–13.0 μg/l. Validation results indicated that the ratio of band 3 (red, 0.63–0.69 μm) and the square of band 1 (blue, 0.45–0.52 μm), with correlation coefficient (R) of 0.89, Root Mean Square Error (RMSE) of 2.53 μg/l and Mean Absolute Error (MAE) of 1.02 μg/l was most capable of representing actual Chl-a concentrations. This is attributed to the differential response of the red and blue wavebands to the Chl-a signal. The study is considered more robust than previous studies of Chl-a retrieval, due to the much larger number of images and in situ samples used for model development and validation, as well as the different times of year, water quality zones, and wide range of Chl-a concentrations which were investigated. The robustness of the developed model was also tested by its application to monitoring an extensive red tide event. The results indicate that the developed model is capable of routine monitoring of such algal blooms which frequently occur from late summer to early autumn in Hong Kong and its adjacent coastal waters.