Abundância de pirarucus (Arapaima gigas) em áreas de manejo comunitário: integrando os componentes da paisagem e o comportamento socioambiental das comunidades ribeirinhas

Abstract

The abundance of pirarucu is shaped by a variety of ecological and anthropogenic factors. Understanding how these factors interact with population abundance is essential for improving fisheries management strategies. From this perspective, the overarching objective of this study was to examine the relationship between pirarucu abundance, the landscape components of floodplain lakes in the Amazonian inundated plains, and the socioenvironmental behavior of riverine communities within fishing agreement areas in the municipality of Coari, Amazonas, Brazil. In Chapter I, we sought to identify the key elements present in the ethnoichthyological knowledge of fishery managers, agricultural suitability, and the participatory governance of associations engaged in pirarucu management. Our findings revealed a high level of local knowledge among managers, which was consistent with existing scientific literature. Agricultural suitability was found to be primarily composed of low-tech family-based production systems, supplemented by management activities and multispecies commercial fishing. The evaluation of participatory governance revealed disparities among the associations, with notable weaknesses in the dimensions of monitoring, surveillance systems, stock assessments, and productive capacity. In Chapter II, we assessed three supervised classification algorithms—Maximum Likelihood, Spectral Angle Mapper, and Random Forest—for identifying landscape components (open water, macrophytes, flooded forest, and non-flooded forest) in Amazonian floodplain environments. We used Landsat-8 satellite imagery, high-resolution aerial imagery obtained via drone, and conducted digital image processing using QGIS software. The Random Forest algorithm achieved the highest overall accuracy (93.7%) and Kappa index (0.92), followed by Maximum Likelihood (74.8% and 0.92) and Spectral Angle Mapper (69.9% and 0.60), respectively. Random Forest proved to be the most robust classification method, even in a heterogeneous landscape with high spectral variability such as the Amazonian floodplain. In Chapter III, we tested the hypothesis that landscape components and socioenvironmental variables significantly influence pirarucu abundance. Data were analyzed using a generalized linear mixed model (GLMM) with a negative binomial distribution, which explained 96.8% of the variance in pirarucu abundance. This variation was positively associated with landscape components (open water, macrophytes, and flooded forest) and socioenvironmental factors (participatory governance and ethnoichthyological knowledge). The integration of floodplain ecosystem conservation with the active participation of riverine communities is essential to ensure the sustainability of managed fisheries and the long-term preservation of the species.