Trees often experience migration lags in their response to rapidly changing climate. Preferential granivory by nocturnal small mammals has the potential to create lags by reducing germination success beyond range edges. To determine how granivory may limit establishment of trees beyond their range margins, we conducted a seed choice experiment which offered seeds of five regionally dominant tree species to small mammals within distinct forest communities across a 400m elevational gradient on four mountains in the northeastern United States. Multinomial logistic mixed effects models were used to (a) quantify seed preference of each species across the elevational gradient and (b) assess relationships between seed preference and abiotic variables. A separate seed dispersal experiment was used to compare the probability of seed consumption versus seed caching. The low-elevation temperate tree species Fagus grandifolia and Acer saccharum had an equally high probability of granivory within and beyond their range margins (~40% and ~20%, respectively). Generally, seed preference was positively correlated with seed mass and nutrient content regardless of elevation. Our seed dispersal experiment revealed that seeds were 3× more likely to be consumed than cached, suggesting that small mammals can potentially decrease germination success. Overall, temperate tree species with either high seed mass or nutritional value may experience substantial granivory beyond their range margin, partially explaining the observed lag between tree dispersal and climate change. Thus, granivory is vital to consider when modeling future tree species distributions under various climate change scenarios.

Seed Selection Experiment

Seeds of five regionally dominant tree species were offered in a cafeteria-style experiment in the fall of 2020 (mid-September through mid-October), coinciding with the timing of natural seed release and dispersal of our focal tree species (Burns and Honkala, 1990). The focal tree species included American beech (Fagus grandifoila), sugar maple (Acer saccharum), yellow birch (Betula alleghaniensis), balsam fir (Abies balsamea), and red spruce (Picea rubens). All seeds were purchased from either Sheffield’s Seed Company (Locke, NY) or F.W. Schumacher Seed Company (Sandwich, MA). Seed species were offered in equal amounts by mass (3 g per station) in order to standardize across species with various seed sizes (Boone & Mortelliti, 2019). Seeds were offered to mammals as they would naturally be found post-dispersal on the forest floor (conifer seeds removed from cones and sugar maple wings attached).

At each seed selection station, five Petri dishes (6 cm diameter) were affixed to one wooden plank with Velcro strips (see Figure 2). The planks were fixed to the ground with two metal lawn stakes to ensure stability. We opted not to enclose each station with a wire cage. While a cage would fully ensure exclusion of any non-target species (i.e., birds, large mammals), it also has the potential to discourage visitation by target species (Mortelliti, personal communication). We believed non-target visitation to seed stations would be low given the dense understory of the sites. In each Petri dish, seeds of one species were placed. Species were randomly assigned to a dish at each site. All stations were checked daily for damage, cleaned of debris, and removed after four nights. Seeds were not replaced after each daily check. At noon on the last day of deployment, any remaining seeds were collected in order to compare the final mass of seeds remaining for each species. Stations were then moved and redeployed on each of the remaining mountains for a total of 60 deployed stations at 12 distinct sites (elevations) across 16 total days.

Small mammal seed choice was determined using infrared game cameras (see SI for examples of videos). At each station, a Stealth Cam G45NG Pro camera was attached to a tree within 1 m of the dishes. The camera was angled down between 45 and 60 degrees to provide a full view of the station. All cameras were set up to record 30 s videos at 1080 HP, and with a 5 s trigger delay between videos when triggered using a near-field setting. In addition to seed choice and mammal species identification and behavior, the cameras were also used to record temperature, time, and the presence of rain at the time of each seed selection event. Sample videos were taken in the field prior to full deployment in order to ensure camera alignment with seed stations. 

Seed Dispersal Experiment

In each of the 12 sites, 3 seed dispersal stations were established to track dispersal distances of seeds and their ultimate fate (consumption vs. caching). Only American beech seeds (beechnuts) were used in this experiment as retrieval tags could not be fixed to the smaller seed species. In order to track beechnut dispersal, a length (1 inch) of yellow yarn was attached to the side of each seed using solvent-free glue, as detailed by Kempter et al. (2018). At each seed dispersal station, 30 beechnuts were placed on a small wooden platform secured to the ground with lawn stakes. The seeds were left uncaged as in the seed selection experiment. An infrared camera was affixed to a nearby tree within 1 m of the station to allow for small mammal identification and quantification of behaviors at the time of seed removal. After three nights, a 2-hour, gridded systematic search up to 30 meters away from the station was conducted for the tagged beechnuts. When one was found, we recorded distance away from the station, whether it was consumed or not, and the substrate in which the seed was found (either found on the surface of the forest floor, buried under the litter mat, or found on a tree). Across all sites, mean recovery rate for tagged beechnuts was 76%.

Data Collection and Processing

All 30-second field videos were processed to determine seed preferences by the resident small mammal community. Seed choice events were defined as the discrete times a small mammal handled a seed in any way. For each observed seed choice event ,we recorded time of day, the small mammal species, the seed species chosen, the number of seeds consumed per event, seed availability (the number of seeds remaining visible in the video) for all seed species, temperature, whether it was raining or not, illumination level, and handling time for the seed species during the event. The high quality of the video taken allowed us to count the number of seeds that were taken or consumed during each event. Running counts of remaining seed number at each station allowed for easy calculation of seed availability for each seed species. This number was divided by the total number of starting seeds to provide seed availability as a proportion. Time, rain, and temperature were taken directly from the video and camera metadata. As most seed predation events occurred during the night, illumination was characterized by the moon phase at the time of the event (new, crescent, quarter, gibbous, full, and a category for daytime light when events occurred during daylight hours). Local weather reports and the time and day of the seed predation event were used to determine the moon phase during each event. Seed handling time was measured with a stopwatch and was defined as the time between when a seed was first picked up and when it was fully consumed or dropped. Events where the endpoint could not be determined because the video ended before it occurred, or the mammal carried the seed outside the video frame were not used for handling time calculations.

(see manuscript for more details)