By Jennifer Donnini, PhD student at Concordia University

For over a decade, two simple ideas have captivated ecologists and remote sensing scientists alike: the Spectral Variation Hypothesis (SVH) and the Acoustic Variation Hypothesis (AVH). By measuring the variability of the signals ecosystems produce (Figure 1), both promise a way to monitor biodiversity that doesn’t require identifying every species on the ground. In theory, a forest that reflects more diverse colors or emits a richer soundscape should also host a more diverse community of species.

It’s an elegant and optimistic concept, but as the fields of remote sensing and ecoacoustics mature, it may be time to admit that these hypotheses have reached the limits of what they can tell us.  The problem isn’t that they’re wrong, but that their simplicity conceals the messy ecological realities that make biodiversity so challenging and so interesting to measure.

A seductive shortcut

When the Spectral Variation Hypothesis was proposed, it provided a unifying framework for linking satellite data to biodiversity. More spectral heterogeneity in a satellite image (that is, more variation in pixel reflectance patterns across space) reflects more variation in vegetation structure and chemistry, which should in turn indicate more plant species (Palmer et al., 2000). The Acoustic Variation Hypothesis (AVH) followed a similar idea for soundscapes, suggesting that greater acoustic variability indicates more vocally active species and therefore higher biodiversity. It draws on two related concepts: the Acoustic Niche Hypothesis (ANH), which predicts that species partition time and frequency to reduce overlap, so more species should fill more of the acoustic space and the Acoustic Adaptation Hypothesis (AAH), which proposes that signal characteristics evolve in response to habitat structure, meaning that communities across diverse habitats are expected to produce more varied acoustic outputs (Alcocer et al., 2022).

Both ideas were appealing because they offered a potential shortcut, a way to bypass the time-consuming, costly, and often spatially limited nature of field surveys. Satellites and autonomous recorders could, in principle, detect biodiversity from space or from sound. These hypotheses motivated hundreds of studies, encouraged cross-disciplinary collaborations, and inspired new tools for global biodiversity monitoring.

The cracks in the logic

Yet after years of testing, results are mixed. Some studies find strong correlations between spectral or acoustic variability and biodiversity; others find none (Alcocer et al., 2022; Torresani et al., 2024). The issue isn’t that these relationships never exist, but that they often depend on context.

Variation in reflectance or sound doesn’t always stem from variation in species. In optical data, shadows, soil exposure, and illumination geometry can create apparent “spectral diversity” unrelated to biological diversity. Moreover, species identity alone is not always a reliable determinant of spectral uniqueness. Different plant species can produce remarkably similar spectral signatures, while individuals of the same species may vary widely depending on stress, phenology, or water availability. Similarly, in soundscapes, rainstorms, wind, and human noise can inflate acoustic variability without reflecting true faunal diversity. Even within species, vocalizations may shift with context, age, or motivation, while calls from different species may overlap in frequency and pattern, blurring the very distinctions acoustic indices aim to capture.

Why correlation isn’t enough

The persistence of these hypotheses speaks to our human desire for simplicity. But the assumption that complexity in a signal automatically mirrors ecological complexity can lead to circular reasoning: we look for relationships because we expect them, not because the mechanisms are well understood. 

In practice, spectral and acoustic diversity indices capture a mix of biological and non-biological variation. While approaches such as trait-based analyses, spectral unmixing, or multi-index acoustic frameworks can help separate ecological signals from background noise, their success depends heavily on context. We should be cautious about treating the Spectral and Acoustic Variation Hypotheses as universally valid rather than as situational tools that sometimes, but not always, reflect true biodiversity patterns.

A cautious path forward

While the Spectral and Acoustic Variation Hypotheses have inspired valuable research, it is worth asking how useful a metric can be when it is influenced by so many external factors and can vary widely across sensors, seasons, and spatial scales. A measure that shifts with illumination, noise, or data resolution may tell us more about conditions of observation than about the ecosystems themselves.  These hypotheses helped open the door to new ways of linking environmental signals with biodiversity, but their continued use as stand-alone indicators should be approached with caution. As new technologies and analytical methods emerge, it will be important to reassess whether these relationships can be made more stable and ecologically meaningful, or whether it is time to move on to frameworks better grounded in measurable biological processes.

About the author:

Jennifer Donnini is PhD student in the Earth Observation Lab at Concordia University in Montreal. Her research interests include remote sensing of biodiversity, forests, and wildlife ecology.

Palmer, M. W., Thomas, W., Peter, E., Jose Ramon, A., & Steven, T. (2000). Opportunities for Long-Term Ecological Research at the Tallgrass Prairie Preserve, Oklahoma. ILTER Regional Workshop, 123–128.