Spatio-temporal dynamics of giant kelp forests in the Humboldt current system revealed by satellite data and hierarchical Bayesian modeling

Understanding the spatio-temporal dynamics of foundation species is critical for assessing ecosystem variability and plan its management accordingly. The giant kelp ( Macrocystis pyrifera ) forms extensive forests in the Humboldt Current System (HCS), a region of pronounced oceanographic heterogeneity, yet its basin-scale population dynamics remain poorly quantified. Here, we integrate Sentinel-2 satellite imagery (2016–2023) with hierarchical Bayesian modeling to resolve the floating canopy cover dynamics of M. pyrifera across three ecoregions of the HCS. Our model explicitly decomposes spatial and spatio-temporal dependencies through Gaussian Markov Random Fields, revealing multi-scale patterns in kelp forest variability. We found a strong latitudinal gradient, with the highest mean canopy cover in the Araucanian ecoregion (south of 33°S), intermediate values in the Humboldtian ecoregion (18°S-25°S), and reduced cover in Central Chile (25°S-33°S). The model identified two distinct spatial scales of organization: fine-scale spatial structure (range ∼ 7 km) indicative of local processes, and broad-scale spatio-temporal coherence (range ∼ 225 km) reflecting regional synchrony. Positive spatial effects were concentrated near major coastal upwelling centers, highlighting their role in enhancing kelp abundance. Temporally, seasonal cycles were most pronounced in the south, while northern ecoregions exhibited higher interannual variability, yet overall canopy cover remained stable throughout the study period. This study establishes the first comprehensive baseline of M. pyrifera dynamics across the HCS, and opens the door to future mechanistic studies of the determinants of its structure and variability.