Migratory routes are inherited primarily from mother in a terrestrial herbivore

For many migratory ungulates, cultural inheritance maintains migration through social transmission of information between individuals, while learning and memory inform movement within individuals. It often is assumed, but rarely tested, that offspring inherit migratory routes by learning from their mothers. Here, we evaluate whether daughters inherit migratory routes from their mothers by following 16 mother-daughter pairs of mule deer (Odocoileus hemionus) from each daughter’s first migration, through their yearling migration, and into adulthood. Adult routes for two-thirds of daughters overlapped with their mother’s route, suggesting that they inherited migratory routes from their mothers. The adult routes for the remaining daughters, however, bore little or no similarity to their mother’s routes, suggesting that these routes were instead shaped by individual experience or non-maternal social interactions. Regardless of whether routes were inherited or not, the strategy that daughters used was influenced by their yearling migratory route, which underscores the importance of this period for establishing life-long behaviors. For mule deer and other species where migration is informed by cultural inheritance, learning, and memory, the specific mechanisms that establish memory can have life-long and cross-generational ramifications. Our work emphasizes the role of social information and early-life experiences in establishing and maintaining migratory behavior, raises new lines of inquiry about what underpins variation in migratory behavior, and points to potential strategies for resilience of an often-imperiled behavior in a changing world.

Methods
Animal capture and handling: As part of a larger study that began in 2015, adult female mule deer were captured in western Wyoming, USA, via helicopter and fit with GPS collars (VERTEX PLUS, Vectronic Aerospace GmbH, Berlin, Germany, or G2110E2, Advanced Telemetry Systems Inc, Isanti, Minnesota, USA) that collected locations every 1 or 2 hours, depending on the individual and the season (Supplemental Materials Figure S1).28 In March, we assessed whether captured deer were pregnant using ultrasonography and, if they were pregnant, inserted a vaginal implant transmitter (Advanced Telemetry Systems or Vectronic Aerospace GmbH). Upon notification of a birth event, we found the neonates and, after taking a suite of measurements, fit them with either VHF collars that were connected to the mother’s collar (M4230BU, Advanced Telemetry Systems) or GPS collars that collected locations every 2 hours (VERTEX MINI, Vectronic Aerospace GmbH). We received an alert if the neonate and mother were separated or if the neonate died, which ensured that the neonates we collared belonged to that mother. We recaptured deer we collared as neonates when they were 6 or 9 months old, depending on the year. At that time, we fitted them with adult GPS collars (VERTEX LITE or VERTEX PLUS, Vectronic Aerospace GmbH, Berlin, Germany) that were programmed to collect locations every 1 or 2 hours, depending on the individual and the season. We continued to monitor daughters and their mothers until their death. Permits to capture adult and neonate mule deer were provided by the Wyoming Game and Fish Department and all protocols were approved by the Institutional Animal Care and Use Committee at the University of Wyoming (20131111KM00040, 20151204KM00135, 20170215KM00260, 20200305KM00412). All capture protocols meet the guidelines that are approved by the American Society of Mammalogists. Additional details about the capture and handling methods are explained elsewhere. 

Although over 200 neonates have been collared through this long-term study, mother-daughter pairs were included in this analysis if 1) daughters survived through their third (i.e., adult) autumn migration, and 2) mothers survived for the entirety of their daughters’ first (i.e., natal) autumn migration. We focused on the autumn migratory route because mother and daughter typically associate during this autumn route, and association is more variable after the first 6 months of life. Daughters completed their first autumn migratory routes between 2015 and 2021. Two mothers gave birth to 2 daughters that were included in this analysis. One individual was considered either a daughter or a mother depending on the analysis, because she was the daughter in a mother-daughter pair and then recruited a daughter that met the criteria above. 

Identifying migratory routes: Because of the variation in fix rates of collars across years, we rarified data to a 2-hour fix rate to create the migratory route. We identified migratory routes using net-squared displacement and visual inspection using Migration Mapper. We considered initiation of migration to occur when deer made directed movements away from their seasonal range, and considered migration to terminate when the directed movements ended. We determined route length by calculating the Euclidean distance between the beginning and end of the migratory route. Migratory routes were, on average, 71.7 km (range = 19.0 – 154.5 km) and they did not differ between daughters and mothers (two-tailed t-test, p-value = 0.4, df = 61.3; alpha = 0.05).  Finally, we created migratory routes using a line-buffer approach. This approach is robust to variation in parameters including fix rate and movement behaviors. Buffers were created by adding 380 m, which was the mean distance traveled in 2 hours, on either side of GPS points during migration. The line-buffer approach produced a footprint of the migratory route, which was used for subsequent analyses of spatial overlap.

Spatial overlap: We used measures of spatial overlap to test the hypothesis that mule deer inherit their migratory routes from their mothers. First, we identified an expected level of spatial overlap by determining the amount of overlap that mothers exhibited from year to year. We calculated the percentage of a mothers’ route in a given year that overlapped with her route in the previous year. We included mothers that completed more than 1 autumn migratory event during the study period; two mothers were excluded from this analysis because they did not meet this requirement. We made this comparison for all routes for remaining mothers through the study period (n =  29 comparisons across 12 individuals; mean number of routes per individual = 3.4; range = 2 – 6). Second, we identified the amount of overlap that daughters exhibited between their adult and natal routes by determining the percentage of a daughters’ adult route that overlapped with her natal route. We used a two-tailed t-test (alpha = 0.05) to determine if the overlap that mothers exhibited from year to year differed from the overlap daughters exhibited between their adult and natal routes. We then repeated this test while retaining only those daughters whose adult and natal routes overlapped more than 25%. Finally, we investigated whether a mother’s mean overlap was correlated with her daughter’s adult to natal overlap using Pearson’s correlation (alpha = 0.05). 

Association: We determined whether spatial overlap hinged on mothers and daughters migrating together. We first tested whether spatial overlap between the adult and natal route differed between daughters whose mother was alive during their adult migration and those whose mothers had died prior using a two-tailed t-test (alpha = 0.05). Next, we evaluated whether spatial overlap was correlated with association for the daughters whose mother was alive (n = 9) using Pearson’s correlation (alpha = 0.05). To determine association, we identified the Euclidean distance between simultaneous GPS points (within 1 hour), and averaged this distance across the migratory event. Mean distances < 1.5 km indicated that mothers and daughters moved together for most or all of the migratory event, whereas distances > 1.5 km indicated that pairs moved separately for most or all of the migratory event.

Timing: We evaluated whether temporal information about migratory behavior could be passed from mothers by investigating whether mothers and daughters differed in day of route initiation (day of the year), route duration (number of days migrating), and day of route termination (day of the year). We constrained our analysis to the 9 pairs where the mother was alive for her daughter’s adult route. For this analysis, we assessed similarity of initiation, duration, and termination using a paired t-test (alpha = 0.05). 

Stage of establishment: To evaluate the point at which migratory behaviors were established, we determined if there was a difference in the amount of overlap between 1) the adult and natal route and 2) the adult and yearling route using a paired t-test (alpha = 0.05). We used the measures of spatial overlap for the adult and natal route described above, and we followed the same procedure to identify the spatial overlap between the adult and yearling route. One daughter was excluded from this analysis because her collar malfunctioned during her yearling route. 

Software: All data manipulation and statistical analyses were conducted in R (version 4.4.1). We used the following packages to manipulate data and conduct analyses: sf, tidyverse, lubridate, wildlifeDI, amt, move2, and MoveTools. We used the following packages to visualize results: ggplot2, ggpubr, ggspatial, rosm, patchwork, and ggmap.