A new twist on a very old binary similarity coefficient.

Area of Science:

• Ecology
• Biodiversity Science
• Quantitative Ecology

Background:

• Pairwise similarity coefficients are often biased downwards in ecological studies when using presence-only data and partial sampling.
• Existing methods like the Dice and Simpson indices, while widely used, can be unreliable under such conditions.

Purpose of the Study:

• To re-evaluate and apply a historically overlooked similarity index proposed by Stephen Forbes in 1907.
• To introduce a heuristic correction to Forbes' index to account for unknown absent species in partial sampling scenarios.
• To demonstrate the superiority of the corrected index over commonly used similarity coefficients.

Main Methods:

• A heuristic correction was applied to Forbes' 1907 similarity index by ignoring the count of absent species.
• The corrected index was compared against the Dice and Simpson indices using simulations with varied species pool sizes, sampling intensities, and abundance distributions (uniform, log-normal, geometric).
• The corrected index was also tested on empirical datasets, including bat samples from Peninsular Malaysia and North American mammal assemblages.

Main Results:

• The corrected Forbes' index consistently outperformed the Dice and Simpson indices in simulations, providing similarity values closer to the true values across various conditions.
• The corrected index demonstrated greater robustness, particularly with unequal sampling intensities.
• Analyses of empirical data supported the simulation findings, showing improved community comparisons and revealing that local mammal assemblages appear to be random subsamples of larger species pools.

Conclusions:

• The corrected Forbes' index offers a more accurate and robust measure of ecological similarity, especially when dealing with incomplete presence-only data.
• This revised index has the potential to replace widely used alternatives like the Dice and Jaccard indices in many ecological applications.
• The index provides valuable insights into community structure, including the random subsampling of species pools and improved ordination of biome-scale data.