The advancement of spring is a widespread biological response to climatechange observed across taxa and biomes. However, the species level responsesto warming are complex and the underlying mechanisms are difficult to disen-tangle. This is partly due to a lack of data, which are typically collected bydirect observations, and thus very time-consuming to obtain. Data deficiency isespecially pronounced in the Arctic where the warming is particularly severe.We present a method for automated monitoring of flowering phenology ofspecific plant species at very high temporal resolution through full growing sea-sons and across geographical regions. The method consists of image-basedmonitoring of field plots using near-surface time-lapse cameras and subsequentautomated detection and counting of flowers in the images using a convolu-tional neural network. We demonstrate the feasibility of collecting flower phe-nology data using automatic time-lapse cameras and show that the temporalresolution of the results surpasses what can be collected by traditional observa-tion methods. We focus on two Arctic species, the mountain avensDryas octo-petalaandDryas integrifoliain 20 image series from four sites. Our flowerdetection model proved capable of detecting flowers of the two species with aremarkable precision of 0.918 (adjusted to 0.966) and a recall of 0.907. Thus,the method can automatically quantify the seasonal dynamics of flower abun-dance at fine scale and return reliable estimates of traditional phenological vari-ables such as the timing of onset, peak, and end of flowering. We describe thesystem and compare manual and automatic extraction of flowering phenologydata from the images. Our method can be directly applied on sites containingmountain avens using our trained model, or the model could be fine-tuned toother species. We discuss the potential of automatic image-based monitoring offlower phenology and how the method can be improved and expanded forfuture studies.
Arctic, computer vision, convolutional neuralnetwork,Dryas integrifolia,Dryas octopetala,ecological monitoring, life-history variation,machine learning
