Lynette’s review on measuring habitat complexity published in Ecology Letters

Our review on measuring habitat complexity in ecology has just been published in Ecology Letters. The review is led by Lynette Loke, a post-doctoral fellow at Macquarie University and previous collaborator of our lab. Ecologists have theorised that habitats with higher complexity have higher diversity, and there is some empirical evidence to support this. But generalisations are difficult, partly because complexity is not measured in a standardised way. We review frequently used metrics of habitat complexity and identify qualities that an ideal metric of complexity should possess.

We find that fractal dimension, one of the most commonly used metrics of complexity, is fraught with problems: fractal dimension is hard to measure accurately (see second figure below); most real ecological habitats may not have fractal properties; and fractal dimension is often poorly correlated with diversity. Rugosity, or surface roughness, is another commonly used metric that is easier to measure and better correlated with diversity, but it may not capture important aspects of habitat complexity. We see promise in information-based metrics of complexity, such as entropy, which are more holistic.

Loke, L.H.L. & Chisholm, R.A. (2022) Measuring habitat complexity and spatial heterogeneity in ecology. Ecology Letters (in press)

Examples of complex surfaces. (a) Three simulated surfaces with different fractal dimensions (D). (b) Three real surfaces at different spatial scales: a rock surface; a rocky shore; and a forest landscape.
Estimates of fractal dimension (vertical axis) can be inaccurate, even under the idealised conditions here where the simulated object being measured is generated from a truly fractal process with known fractal dimension (horizontal axis). The best-performing method is the box-counting method at intermediate scales, but this is rarely if ever applied in ecology (the standard box-counting and variation methods are more prevalent but exhibit strong biases).