A workflow to optimize spatial sampling in ecoacoustic studies

Context Landscape monitoring through sounds relies on spatial sampling to characterize soundscape variation across a heterogeneous space. This non-invasive approach complements traditional landscape metrics by capturing ecological dynamics not visible in spatial structure alone. However, the effectiveness of different sampling strategies remains poorly understood. Objectives We examined how spatial sampling design affects the representativeness of soundscape information across heterogeneous ecosystems. Our goal was to identify efficient acoustic Sub-sampling strategies that integrate landscape heterogeneity while balancing effort and performance. Specifically, we (1) assessed how sampling structure and effort affect interpolated soundscape representativeness; (2) compared random, thematic designs based on land-cover classes, and tessellation-based designs; (3) examined interactions among landscape, acoustic, and temporal variables; (4) evaluated the influence of landscape heterogeneity on sampling performance; and (5) developed a workflow to guide ecoacoustic Sub-sampling. Methods Our study was conducted across three Colombian ecosystems. We constructed discrete and continuous landscape proxies using Sentinel-2 imagery, calculated acoustic indices from field recordings, and then implemented various spatial sampling designs from sub samples based on image tessellation techniques. Using kriging, a geostatistical method for interpolation,we compared sub samples to full spatial scenarios and assessed their performance using structural and distributional metrics. We took representativeness as the spatial fidelity of interpolated soundscape surfaces relative to full sampling, capturing both acoustic patterns and underlying landscape structure. Results We found that tessellation-based methods, especially those based on watershed segmentation, significantly outperformed random and thematic designs, showing higher sample performance, stronger correlations with sample size, and greater representativeness of landscape heterogeneity. Multivariate analyses showed that tessellation-based designs exhibited consistent patterns across acoustic, temporal, and landscape variables. Our results demonstrate that the spatial arrangement of sampling locations, rather than sample size alone, critically determines sampling performance. We propose a landscape-based workflow to support the design of efficient ecoacoustic sampling strategies. This approach advances scalable and spatially informed biodiversity monitoring using ecoacoustics methods.