Modeling Estuarine Algal Bloom Dynamics with Satellite Data and Spectral Index-Based Classification
Macroalgae are an integral part of estuarine primary production; however, their excessive growth may have severe negative impacts on the ecosystem. Although it is generally believed that algal blooms may be caused by a combination of excessive nutrients and temperature, their occurrences are hard to...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-05-01
|
| Series: | Hydrology |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2306-5338/12/6/130 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Macroalgae are an integral part of estuarine primary production; however, their excessive growth may have severe negative impacts on the ecosystem. Although it is generally believed that algal blooms may be caused by a combination of excessive nutrients and temperature, their occurrences are hard to predict, and quantitative monitoring is a logistical challenge which requires the development of reliable and inexpensive techniques. This can be achieved by implementation of processing algorithms and indices on multi-spectral satellite images. Tuggerah Lakes estuary on the Central Coast of NSW was studied because of the regular occurrences of blooms, primarily of green filamentous algae. The detection of algal blooms based on the red-edge effect of the chlorophyll provided consistent results supported by direct observations. The Floating Algae Index (FAI) was identified as the most accurate index for detecting algal blooms in shallow areas, following a comparative analysis of six commonly used algae detection indices. Logistic regression was implemented where FAI was used as a predictor of two clusters, “bloom” and “non-bloom”. FAI was calculated for multi-spectral satellite images based on pixels of 20 × 20 m, covering the entire area of the Tuggerah Lakes. Seven sample points (pixels) were chosen, and the optimal threshold was found for each pixel to assign it to one of the two clusters. The logistic regression model was trained for each pixel; then the optimal parameters for its coefficients and the optimal classification threshold were obtained by cross-validation based on bootstrapping. Probabilities for classifying clusters as either “bloom” or “non-bloom” were predicted with respect to the optimal threshold. The resulting model can be used to estimate probability of macroalgal blooms in coastal estuaries, allowing quantitative monitoring through time and space. |
|---|---|
| ISSN: | 2306-5338 |