FCM-m

FCM-m: Optimized fuzzy c-means clustering for water optical classification

FCM-m optimizes the fuzzy c-means (FCM) algorithm by dynamically adjusting the fuzzifier parameter (m) to improve optical clustering and classification of inland and eutrophic waters based on spectral data.

Key Features:

• Optimized Fuzzifier Parameter (m): Dynamically estimates m instead of using the default value of 2, reducing erroneous membership assignments and improving classification accuracy.
• Adaptive Parameterization: Adjusts m according to dataset biogeochemical variability and dimensionality, with m approaching 1 as spectral band numbers increase.
• Enhanced Cluster Unitarity: Increases cluster compactness and minimizes membership degrees assigned to non-belonging water types.
• Sentinel-3 OLCI Integration: Classifies water optical clusters using Sentinel-3 Ocean and Land Colour Imager (OLCI) bands and atmospherically corrected OLCI imagery.
• Performance Validation: Evaluates clustering robustness using the Fuzzy Silhouette Index, achieving a score of 0.513 for optimized results.
• Chlorophyll-a Estimation: Improves Chlorophyll-a concentration estimation by reducing errors associated with non-belonging clusters.

Scientific Applications:

• Water Quality and Biogeochemical Analysis: Supports regional and global optical classification of inland and eutrophic waters for ecological and environmental studies.
• Remote Sensing Monitoring: Enables satellite-based water type discrimination using spectral reflectance data.

Methodology:

FCM-m applies fuzzy c-means clustering with dataset-specific optimization of the fuzzifier parameter (m). In situ spectral data with co-measured water quality parameters are used to determine optimal m values that maximize cluster unitarity. Clustering performance is validated using the Fuzzy Silhouette Index.