North African coastal lagoons are unique ecosystems that often suffer degradation due to human activities. Therefore, monitoring methods are required to identify stressors and assist with the management of these valuable and often understudied ecosystems. A synthetic indicator of water ecological quality would be desirable for regular monitoring of these ecosystems under pressure. In 2008 an optical procedure was developed and applied in Ghar El Melh, a Tunisian lagoon which has been increasingly impacted by pollutant loading, especially from agriculture. In situ hyperspectral irradiance was measured at several stations, from which the apparent optical properties (AOPs), namely the irradiance attenuation coefficient K(λ) and the reflectance ratio R(λ), were obtained in order to relate them to water composition, in terms of light-attenuating substances (LASs). The significant relationships observed between R and LAS values enabled the application of a hyperspectral optical classification, which effectively highlighted threatened sectors of the lagoon. The pattern of differing water quality across the lagoon system that was derived from the hyperspectral classification agreed well with that obtained from a conventional optical classification that included AOPs and LASs. We suggest that hyperspectral analysis and classification is a useful monitoring tool for the assessment of change in coastal lagoons, and perhaps also in other shallow-water ecosystems.