Behavior

Behavior of individual mice is expected to be associated with disease dynamics in a number of ways. The timing and location of movements can influences exposure to pathogens and how an individual chooses to expend energy can influence its ability to fight off infection. We chose to assess 4 metrics of behavior for this project: 1) Activity timing, 2) range size, 3) exploration, and 4) learning. Since this is a field study and we weren’t able to follow each individual, we estimated proxies for these behavioral metrics

Many of these behavioral metrics rely on recaptures. From 2022 to 2024, there were 2,142 Peromyscus individuals that were recaptured at least once (44.9% of total individuals). On average, individuals that were captured more than once were captured 3.3 times on average.

Activity timing

Our indicator for activity timing in this study is capture time. From 2022 - 2024, our group added ibutton temperature sensor pairs to all traps at one of each plots within the eight NEON sites of interest. These sensors, which log temperatures every five minutes, allowed us to estimate the time at which an individual entered a trap⁴ and, thus, the time at which the individual is active.

This technique of deriving capture timing works because the air inside increases above the ambient air when an animal is in it. Figure 6 shows an example of these data. For most analyses, capture time is presented as the fraction of the night (sunset to sunrise) that had elapsed when the capture occurred. This relative capture time scaling allows for more equitable comparisons across latitudinal gradients and photoperiods (Figure 7).

Figure 8 shows the distribution of those relative capture times for Peromyscus, how they differ across sites and through time.

Figure 6: Capture time estimate example: ibutton data (green and blue lines) from a three-day trapping event at plot #2 from the UNDE NEON site in 2024. The dotted vertical lines correspond to the algorithmically-derived estimate of the time an individual entered the trap. Purple rectangles correspond to the segment of time during which the temperature difference between inner and outer buttons is above the trigger threshold (1°C). No data are shown for trapnights in which there was no capture.

Figure 7: Linear relationship between relative capture time (y) and latitude (x).

Figure 8: Peromyscus relative capture time density distributions across eight NEON sites over three years.

Range size

Our indicator of range size is the average distance between traps that an individual was captured between each consecutive capture. As it depends upon recaptures, this metric is biased towards individuals that have been captured more times times (Figure 9); individuals captured only once always have an average trap distance of 0. But for those individuals that were recaptured, it provides some information about the size of the space that the individual occupies. Figure 10 shows the distribution of movement distances among Peromyscus across NEON sites.

Figure 9: Distribution of the log-transformed distance (meters) between consecutive captures, averaged across all recaptures, for an individual Peromyscus.

Figure 10: Distribution of log-transformed distance (D, meters) between consecutive captures of recaptured individuals, seperated by site and sex.

Exploration

Our proxy for exploration is trap diversity: the number of unique traps an individual was captured in. We scaled trap diversity by the total number of captures (relative to the population) to alleviate some of the bias towards recaptures (Figures 11, 12). While this (weighted) trap diversity is somewhat associated with trap distance, they are not equivalent (Figure 13). An individual that travels far along a consistent path is probably functionally different from one that stays within a relatively small range, but uses more of that range. The former individual would have a higher trap distance and a lower trap diversity than the latter.

Figure 11: Relationship between individual capture count (x) and number of unique traps captured in (y).

Figure 12: Relationship between raw (x) and weighted (y) trap diversity. The regression line is a second-order polynomial.

Figure 13: Relationship between (log-transformed) trap distance and weighted trap diversity. Red points represent individuals that were only caught at at one plot and blue dot individuals were captured at 2 plots.

Learning

The metric of learning that we used is trapability: Individuals more prone to being captured - especially after the initial capture - have likely learned to identify traps as a source of food. This is demonstrated by the tendency of individuals to enter traps earlier in the night than on their first capture. This pattern is true of NEON populations (Figure 14) and has been demonstrated elsewhere⁵.

In our case, trapability is the proportion of trap nights on which an individual was captured between their first and last capture. By this method, individuals captured only once have a trapability of 1, but this is misleading since we have the least information about and confidence in these mice. To alleviate this bias, we weighted trapability by number of captures. A comparison between raw and weighted-trapability are shown by Figure 15. This figure shows that raw trapability of an individual captured once and an individual captured 15 times across 15 trap nights are equal (x = 1), whereas weighted trapability accounts for the higher degree of certainty in estimates for the mouse captured 15 times.

Figure 14: Distribution in the change in relative capture time for recaptured individuials after their first capture. Nevative values indicate that recaptures occurred earlier in the night relative to the initial capture. Positive shifts occured later. The red line represents the mean value.

Figure 15: Relationship between weighted (y) and raw (x) trapability and number of captures (color, size).

Outliers?

A handful of mice were captured in many traps (Table 10) at the eight sites and across all of NEON’s duration. One individual was trapped 38 times in 22 traps over 2 years! On average, these trap-happy individuals stuck to relatively small areas (average of 21.25483 meters), with a maximum distance of 120 meters and 17 traps. Compare that to the very furthest distance between observations of an individual in the larger Peromyscus population of 9 kilometers⁶!

We evaluated the validity of these observations to the best of our abilities and believe that the records for these individuals are corrected and not, as we initially suspected, erroneous grouping of multiple individuals that share the same tag number⁷

Table 10: Most trapped Peromyscus: top ten most captured individuals. Columns represent individual ID, sex, number of times captured, number of unique traps captured in (i.e., trap diversity), number of unique plots caught on, number of trap nights between first and last capture (across all plots visited), calendar days between first and last capture, trapability (weighted by times captured), average distance (meters) between consecutive captures, and maximum distance (meters) among observations of the individual.

iid	sex	caps	traps	plots	trap nights	days	trapability	avg D	max D
11123	M	38	22	1	47	771	0.81	19.91	98.49
10885	M	27	16	1	38	469	0.50	16.86	94.29
10928	F	27	8	1	35	435	0.55	11.94	31.59
11227	M	27	19	1	37	688	0.52	33.03	120.38
11375	F	27	21	1	35	534	0.55	21.29	82.49
5875	M	26	20	1	48	462	0.37	23.69	90.53
6124	F	25	14	1	63	760	0.26	18.12	78.28
10849	M	25	15	1	50	770	0.33	18.54	70.69
10919	M	24	16	1	31	407	0.49	25.60	72.03
14338	F	22	17	1	37	588	0.34	23.57	64.00

Tick contact

A primary way that we expect behavior to influence Borrelia infection in Peromyscus is through interactions with ticks (specifically Ixodes). For example, male mice are most likely to move around driven by mate-seeking behaviors. This extra movement makes them more likely to interact with ticks that are sit-and-wait predators. Figure 16 demonstrates that males are, indeed, more likely to accumulate ticks. Similarly, individuals that were captured in more traps - suggesting more diverse space use - acquired slightly more ticks on average, regardless of sex (Figure 17).

Figure 16: Comparison of average tick attachement (± 95% CI) between male and female Peromyscus

Figure 17: Logistic regression of tick attachment (y) by weighted trap diversity (x, log-transformed).

---

4. Citation to original Orrock paper needed↩︎

5. Citation Needed↩︎

6. this is not unusual for a Peromyscus dispersal event (Citation needed)↩︎

7. NEON’s tagID values are not entirely unique to individuals. Tags must be combined with site and recapture information to differentiate individual animals (i.e., iid).↩︎