Bayesian Inference of Human-Made Hazards in Networks of Island Wetlands: The Case of the Aegean Archipelago

Abstract

This research aims to provide guidelines using probabilistic methods to understand better the dominant characteristics of the 824 under-pressure wetlands on 75 islands within Greece and to inform future conservation efforts. Identifying the characteristics and types of anthropogenic pressures is crucial for developing effective conservation strategies. The study employs power-law modeling to validate the natural size distribution of wetlands, naïve Bayesian inference to model human impacts, and the epsilon statistic to assess wetland sensitivity to specific pressures, addressing potential sampling biases. Power-law modeling reveals a natural heavy-tailed distribution of wetland sizes, highlighting the ecological significance of larger, rarer systems. Naïve Bayesian inference indicates that agriculture and transportation are the predominant pressures affecting natural coastal wetlands. The epsilon statistic further differentiates wetland sensitivity, identifying estuaries, lagoons, and marshes as particularly vulnerable. By profiling the most vulnerable wetlands using these methods, the research provides a framework for assessing anthropogenic impacts and informing targeted conservation and management strategies to protect these vital ecosystems.