A new conceptual and quantitative approach to exploring and defining potential open-access olfactory information

Abstract

All organisms emit odour, providing 'open-access' olfactory information for any receiver with the right sensory apparatus. Characterizing open-access information emitted by groups of organisms, such as plant species, provides the means to answer significant questions about ecological interactions and their evolution. We present a new conceptual framework defining information reliability and a practical method to characterize and recover information from amongst olfactory noise. We quantified odour emissions from two tree species, one focal group and one outgroup, to demonstrate our approach using two new R statistical functions. We explore the consequences of relaxing or tightening criteria defining information and, from thousands of odour combinations, we identify and quantify those few likely to be informative. Our method uses core general principles characterizing information while incorporating knowledge of how receivers detect and discriminate odours. We can now map information in consistency-precision reliability space, explore the concept of information, and test information-noise boundaries, and between cues and signals.

Keywords: cue; ecological interactions; information recovery; signal; volatile organic compounds.

Fig. 1

Conceptual diagram of information flow between an emitter (a plant) and various receivers (such as vertebrate and invertebrate pollinators, herbivores, and other plants). (a) By studying one specific receiver, we likely see only a fraction of both the available information and the possible ecological interactions. (b) By studying the emitter, we can identify all the potential ‘open‐access’ information it produces (step 1 – this study) and then test it on different receivers to detect possible interactions (step 2). Dashed coloured lines indicate the different components of information used by particular receivers, and signs indicate resulting beneficial (+) or harmful (−) interactions for the emitter.

Fig. 2

Conceptual framework showing how consistency and precision, two within‐group criteria, relate to noise vs information; and within information, cue vs signal. Reliability of odour emissions increases with increasing consistency (occurring in more members of a group) and/or precision (tight relative amounts of the volatile organic compounds (VOCs) among members of a group). Odour emissions with both low consistency and low precision are completely unreliable, uninformative, and hence noise. Cues, such as (a) odours emitted from undamaged palatable seedlings, are predicted to have VOC combinations that are either highly consistent or highly precise but rarely both. Signals, such as plant odour emissions aimed to (b) attract natural enemies of damaging insect herbivores or (c) attract pollinators should have both highly consistent and highly precise VOC combinations.

Fig. 3

Matrix of results from the focal group of 38 Eucalyptus punctata seedlings using the heuristic method to identify number of potentially informative volatile organic compound (VOC) pairs by varying values for the two within‐group criteria, consistency and precision. The numbers in the matrix indicate the number of informative VOC pairs for each consistency and precision combination. The colour of the matrix changes from purple, for a low number of informative VOC pairs, to orange, for a high number of informative VOC pairs. Green arrows indicate the specific values chosen for further investigation in aim 2, in which we applied all three criteria (consistency, precision, and accuracy) to the focal group, E. punctata, by including an outgroup, Corymbia gummifera.

Fig. 4

Results from the focal group of 38 Eucalyptus punctata seedlings using the heuristic method to identify pairs of volatile organic compounds (VOCs) in consistency–precision reliability space. Here, values for the two within‐group criteria vary: consistency (low 0.5 to high 0.8) and precision (low 0.5 to high 0.17). Specific VOC pairs, presented in a different colour, are shown when one to four pairs were detected. Where a cell contains the number 0, no pairs were identified. Where more than four pairs were identified, just the number of pairs is shown (as in Fig. 2). The method and R function ‘Odour Information Definition Algorithm’ also provide output on the relative proportion of the two component VOCs in a given pair, but this detailed information is not presented here.

Fig. 5

Matrix generated from the heuristic method showing percentages of the focal group of 38 Eucalyptus punctata seedlings that have no informative volatile organic compound (VOC) pairs, for each combination of the two within‐group criteria, consistency and precision. The colour of the matrix changes from orange, for a low percentage of replicates without informative VOC pairs, to purple, for a high percentage of replicates without informative VOC pairs. Green arrows indicate the specific values chosen for further investigation in aim 2, in which we applied all three criteria (consistency, precision, and accuracy) to the focal group, E. punctata, by including an outgroup, Corymbia gummifera.

Fig. 6

Results for the focal group of 38 Eucalyptus punctata seedlings, using the heuristic method and applying the two within‐group criteria. (a) Criterion 1, consistency (0.67 of all seedlings): abundance (as total ion count per million) of the six most frequent volatile organic compounds (VOCs) (when present). Left and right plots show the same information but at a different scale. (b) Criterion 2, precision (variation 0.21): proportion of the first listed compound in each VOC pair (calculated after fourth‐root transformation, see the Materials and Methods section) for all 15 possible combinations of the six most frequent VOCs identified in (a) criterion 1. Blue dashed line indicates the space where the paired proportion is 0.5 (i.e. both VOCs in the pair have the same relative abundance). Green arrows indicate the final four precise VOC pairs identified using these specific values for consistency (0.67) and precision (0.21). Boxplots indicate the median, the first and third quartiles, and the maximum and minimum values. Empty circles indicate outliers. Numbers inside the figures indicate the number of seedlings (out of 38) in which (a) each VOC or (b) a pair of VOCs was detected.

Fig. 7

Results from the random forest algorithm. The left‐hand column lists each replicate taken for each batch (B) for our focal group (Ep, Eucalyptus punctata) and outgroup (Cg, Corymbia gummifera). The top row lists the six volatile organic compounds (VOCs) selected by the algorithm in the interpretation step. Of these six, only three remained after the prediction step (the three first columns from the left in the top row). The amount of emitted VOC (quantified as ion count) for each replicate is presented by colour from blue (zero) to green (highly abundant); see the key. For example, 1,8‐cineole is mid to highly abundant in most E. punctata seedling replicates but is absent or near absent from C. gummifera seedling replicates.