Abstract Fire regimes in North American forests are diverse and modern fire records are often too short to capture important patterns, trends, feedbacks, and drivers of variability. Tree-ring fire scars provide valuable perspectives on fire regimes, including centuries-long records of fire year, season, frequency, severity, and size. Here, we introduce the newly compiled North American tree-ring fire-scar network (NAFSN), which contains 2562 sites, >37,000 fire-scarred trees, and covers large parts of North America. We investigate the NAFSN in terms of geography, sample depth, vegetation, topography, climate, and human land use. Fire scars are found in most ecoregions, from boreal forests in northern Alaska and Canada to subtropical forests in southern Florida and Mexico. The network includes 91 tree species, but is dominated by gymnosperms in the genus Pinus. Fire scars are found from sea level to >4000-m elevation and across a range of topographic settings that vary by ecoregion. Multiple regions are densely sampled (e.g., >1000 fire-scarred trees), enabling new spatial analyses such as reconstructions of area burned. To demonstrate the potential of the network, we compared the climate space of the NAFSN to those of modern fires and forests; the NAFSN spans a climate space largely representative of the forested areas in North America, with notable gaps in warmer tropical climates. Modern fires are burning in similar climate spaces as historical fires, but disproportionately in warmer regions compared to the historical record, possibly related to under-sampling of warm subtropical forests or supporting observations of changing fire regimes. The historical influence of Indigenous and non-Indigenous human land use on fire regimes varies in space and time. A 20th century fire deficit associated with human activities is evident in many regions, yet fire regimes characterized by frequent surface fires are still active in some areas (e.g., Mexico and the southeastern United States). These analyses provide a foundation and framework for future studies using the hundreds of thousands of annually- to sub-annually-resolved tree-ring records of fire spanning centuries, which will further advance our understanding of the interactions among fire, climate, topography, vegetation, and humans across North America.

Top-down effects, like predation, are drivers of insect outbreaks, but bottom-up effects, like host nutritional quality, also influence outbreaks and could in turn be altered by insect-caused defoliation. We evaluated the prediction that herbivory leads to a positive feedback on outbreak severity as nutrient concentration in plant tissues increases through improved soil nutrient availability from frass and litter deposition. Over seven years of a spruce budworm outbreak, we quantified litter nutrient fluxes, soil nitrogen availability, and host tree foliar nutrient status along a forest susceptibility gradient. As the outbreak progressed, both soil nutrient fluxes and availability increased which, in turn, improved foliage quality in surviving host trees. This is consistent with boosted insect fitness and increased population density and defoliation as outbreaks grow. Our results suggest that a positive bottom-up feedback to forest ecosystems from defoliation may result in conditions favorable to self-amplifying population dynamics in insect herbivores that can contribute to driving broad-scale outbreaks.

 

Phenology has become a field of growing importance due to the increasingly apparent impacts of climate change. However, the time-consuming, subjective and tedious nature of traditional human field observations have hindered the development of large-scale phenology networks. Such networks are rare and rely on time-lapse cameras and simplistic color indexes to monitor phenology. To automatize rapid, detailed and repeatable analyzes, we propose an Artificial Intelligence (AI) framework based on machine learning and computer vision techniques. Our approach extracts multiple ecologically-relevant indicators from time-lapse digital photography datasets. The proposed framework consists of three main components: (i) a random forest model to automatically select relevant images based on color information; (ii) a convolutional neural network (CNN) to identify and localize open tree buds; and (iii) a density-based spatial clustering algorithm to cluster open bud detections across the time-series. We tested this framework on a dataset including thousands of black spruce and balsam fir tree images captured using our phenological camera network. The performed experiments showed the efficiency of the proposed approach under challenging perturbation factors, such as significant image noise. Our framework is exceedingly faster and more accurate than human analysts, reducing the time-series processing time from multiple days to under an hour. The proposed methodology is particularly appropriate for large-scale and long-term analyzes of ecological imagery datasets. Our work demonstrates that the use of computer vision and machine learning methods represents a promising direction for the implementation of national, continental, or even global plant phenology networks.