Spatial patterns of flower colour variation in native and introduced ranges of Convolvulus arvensis (Convolvulaceae) revealed by citizen-science data and machine learning.

Abstract

Flower colour polymorphism refers to the presence of multiple colour variants in plant populations. Investigation of this phenomenon led to multiple discoveries, including the principles of heredity and the foundations of population genetics. I examined flower colour variation across native and introduced ranges of Convolvulus arvensis, which exhibits flower colour polymorphism (individuals have white or pink petals). To study flower colour variation of this species throughout large geographic scale, I used observations gathered from the iNaturalist platform. To handle a large amount of data, I trained a neural network to classify the plants' morphs based on photographs. After which I performed spatial analyses to examine the patterns of the colour frequency, also in relation to environmental factors. The results show that flower colours are polymorphic across the whole species range, but the frequency of pink versus white flowers varies. In the Palearctic, I observed geographic clines of colour morph frequencies: a higher frequency of the pink morph in populations from Northwest Europe, whereas in South and East Europe, towards the eastern edge of the range, the white morph was dominant. In contrast, pattern of colour distribution in North America (where the species is invasive) seems random, but the model indicates a link between higher proportions of pink morphs in mild and humid climates. The mechanisms behind the observed patterns remain largely unknown, as changes in a morphs' frequency are not strongly linked to abiotic factors. To understand the spatial pattern, a detailed investigation, accounting for the species' phylogeography is needed. This study provides another example of how the general public may collect data relevant to ecological studies, even when the data are not collected for a specific project.