研究目的
1) prove the adequacy of remote sensing-based inputs for algal bloom data-driven models as a practical alternative for extensive field data-based inputs that are unavailable for many of the vulnerable coastal and inland water bodies across the globe; 2) apply two data-driven modeling techniques (i.e., multivariate regression [MR] and an artificial neural network [ANN]) for algal bloom occurrence probability modeling and compare the findings from the two models; 3) propose a new modeling approach that relies on the construction of a coupled temporal and spatial model for significantly improved accuracy; and 4) provide a replicable model using Kuwait Bay and surrounding waters as a test site.
研究成果
The developed methodologies are advantageous for their reliance on readily available global remote sensing datasets, cost-effectiveness, and ability to predict algal bloom occurrences. They are particularly useful in inaccessible areas and those lacking adequate monitoring systems, offering local, global, and technical implications for algal bloom forecasting.
研究不足
The main constraints are related to the spatial and temporal resolution of the remote sensing digital datasets used, which may not detect algal blooms smaller than the picture elements of digital products or those that occur and vanish within the acquisition period of two consecutive satellite images. Adverse weather conditions can also degrade the quality of remote sensing datasets.
