Weathering the storm for love? The mate-searching behaviour of wild male Sydney funnel-web spiders (Atrax robustus)
Data files
Sep 19, 2024 version files 48.29 KB
-
allspiders1.csv
26.45 KB
-
Displacement.csv
1.38 KB
-
README.md
5.17 KB
-
weathering_the_storm2.qmd
15.29 KB
Abstract
The risky business of mate-searching often leaves the actively searching sex facing threats and rapidly changing conditions. Yet, active mate-searching behaviour is rarely studied in invertebrates, and we have limited understanding of how mate-searching strategies have evolved to cope with risks posed by harsh weather. We investigated how mate-searching males move through their habitat and how their movement is affected by weather conditions in the Sydney funnel-web spider (Atrax robustus), one of the world’s most venomous spiders. As is common in mygalomorphs spiders, females are functionally sessile, and are thought to spend their whole lives in a single burrow, whereas males must permanently abandon their burrows to mate during the breeding season. Nineteen male spiders were fitted with micro-radio transmitters and tracked during their mating seasons in 2020 (n = 2), 2021 (n = 8) and 2022 (n = 9) in Lane Cove National Park, in Sydney, Australia. Males moved at night, typically in a zig-zag pattern, and were found in new locations on approximately 50% of daily resighting’s. Males often spent several days in a female’s burrow, and some female burrows were visited by multiple males. When outside a female’s burrow, males constructed and occupied temporary shelters (‘temporacula’). Males were most likely to move and/or moved furthest when there was no rain, and on warm nights after cool days. Our findings suggest that mate-searching A. robustus males prefer to search for females in less risky conditions, revealing novel risk-minimizing strategies, especially in response to rainfall and temperature.
README: Weathering the storm for love? The mate searching behaviour of wild males of (the Sydney funnel-web spider (Atracidae: Atrax robustus)
https://doi.org/10.5061/dryad.mw6m90656
Authors: Caitlin N. Creak, Hugo Muirhead, Michael Kasumovic, Russell Bonduriansky, Bruno A. Buzatto
Summary of Project:
We investigated how mate-searching males move through their habitat and how their movement is affected by weather conditions in the Sydney funnel-web spider (Atrax robustus). Nineteen male spiders were fitted with micro-radio transmitters and tracked one at a time every day during their mating seasons in Sydney, Australia. We then compared their movement with temperature and rainfall data collected from the Australian Bureau of Meteorology.
List of files
weathering_the_storm1.qmd\
Displacement.csv\
allspiders1.csv
File descriptions:
weathering_the_storm1.qmd - The only file needed to see all analyses in manuscript. Loads in summary dataset and full dataset and runs the correlations, glmmTMB analyses, then runs the dredge and model selection. This can be opened in R studio as a Quardo Doc.
Displacement.csv- A summary dataset for each individual spider
Variables:
ID: Identification code for each individual spider including the year
Total: The sum of all meters travelled
Displacement: Linear projection of each spider’s movement = distance between the first location of capture and the last location recorded.
Ratio: Quantifying the linearity of movement, = displacement divided by total
Moved: Number of times that individual moved throughout the study
Days tracked: Number of days the individual had a transmitter attached
Name: Field name of each individual
Weight: Weight of individual in grams
Cephwid: The cephalothorax width of each individual, measured at widest point, in cm
Cephleng: The cephalothorax length of each individual, measured down centre of cephalothorax, in cm
Mean_disp: Mean nightly displacement = displacement divided by Days tracked
Mean_meters: Mean nightly movement + total distance divided by Moved
Missing values:
Displacement M3.20, potentially inaccurate GPS coordinates so omitted from the displacement analysis
Ratio M3.20, as above
Weight Scales only procured for the 2022 season, M4.22 had measurement error so omitted weight
Cephwid Only individuals that had a quality photo taken were measured
Mean_disp M3.20 as above for displacement.
allspiders1.csv - Full raw dataset of tracking and observations
Variables:
ID: Identification code for each individual spider including the year
day: Day of the month the observation was recorded
month: Month the observation was made
year: Year the observation was recorded
name: Field name of individual
order: Order number of the day the individual was tracked
moved: Whether the individual moved (1) or didn't move (0) for that observation
meters: Distance (meters) between previous and new location
mintemp: Minimum temperature (degrees Celsius) recorded of the previous night
maxtemp: Maximum temperature (degrees Celsius)recorded of the previous day
avtemp: Average temperature (degrees Celsius) of the previous 24 hours = mintemp + maxtemp/2
rain: Total precipitation measured for the previous day (mm)
burrow: Whether an individual was confirmed inside the burrow of a female (Y/N)
plantcom: Plant community type, has four categories
caldate: Number of day since tracking began, starting at 1 at the beginning of each year
caldatecat: Number of day since tracking began + year code A = 2020, B = 2021, C = 2022
latdecimal: Decimal of the latitude the observation was recorded at
londecimal: Decimal of the longitude the observation was recorded at
Missing values:
meters 19 missing values. Missing measurements due to park closure during floods or unsafe conditions. Addressed in the gaussian glmmTMB by removing all observations that have missing values.
latdecimal 23 missing values due to GPS error, these values were either removed or manually adjusted in QGIS but are not used in the statistical analysis.
londecimal Same as latdecimal
Environmental data (Minimum temperature, Maximum temperature and rainfall) was obtained from the Australian Government Bureau of Meteorology
Software
R version 4.4.0 (2024-04-24 ucrt)
R packages used:
cowplot (version 1.1.3): Streamlined Plot Theme and Plot Annotations for 'ggplot2'.\
MuMIn (version 1.47.5): A package that contains functions carry out model averaging based on information criteria.\
tidyverse (version 2.0.0): A collection of packages for data manipulation, exploration, and visualization in R.\
ggpubr (version 0.6.0): A system that for creating and customizing 'ggplot2'- based publication ready plots.\
report (version 0.5.8): Automatic reporting of R objects.\
glmmTMB (version 1.1.9): Fits a generalized linear mixed model (GLMM) using Template Model Builder (TMB).\
DHARMa (version 0.4.6): Uses a simulation-based approach to create readily interpretable scaled (quantile) residuals for fitted (generalized) linear mixed models.
Methods
Collection of wild spiders
This study was conducted on Eora and Kuring-gai (also Guringai) Nations’ lands. Individual male Atrax robustus specimens (n = 19) were collected between February and April of 2020 (n = 2), 2021 (n = 8), and 2022 (n = 9), from Lane Cove National Park (-33.761, 151.106), Sydney, Australia. To collect males, trips on foot after 8pm AEST were made along multiple fire trails within the national park. Males were spotted on and alongside the fire trail in undergrowth and upon elevated roadside banks. Collecting individuals involved placing an open 500ml plastic container over the top of the spider, then encouraging the spider to walk into the container using the lid. Once collected, we labelled the location using flagging tape and saved the GPS location of each individual using a Garmin GPSMAP 62sc.
Transmitter attachment
We used model T15 micro-radio transmitters (Advanced Telemetry Systems, Gold Coast, Qld., Australia) to tag spiders (Figure 1). The T15 transmitters weigh 0.15 grams and are 3.4mm x 11mm for the main body (Figure 1 A) plus a 150mm flexible antenna. Transmitters had a battery life of approximately 30 days. Two ATS R410 scanning receivers with a 3-element folding yagi directional antenna were used to locate individuals with transmitters.
At first capture, individuals were taken back to the lab at UNSW Sydney and photographed inside their container. In 2022, males were also weighed using a HT-120 compact balance (A&D Ltd., Thebarton, SA, Australia). Individuals were given a minimum of 12 hours to settle and were provided with water ad libitum before the transmitter was attached. Each ATS T15 transmitter was switched on using the ATS T15 controller app immediately prior to attachment to avoid battery life wastage. To attach the transmitters, each individual spider was placed into a 500ml plastic container with holes drilled into the sides and lid (Figure 1 B). A damp sponge was placed in the bottom of the container to ensure individuals did not desiccate. Once the individual was transferred to the container, the container was placed into a sealed plastic bag, which was then filled with CO2 gas and sealed off for three minutes or until the individual was lethargic and easily manoeuvrable.
The individual was them moved using forceps to the centre of the sponge (location marked with an X in Figure 1 B), and a second sponge (cut into a ring shape with a Perspex plate around the outside and two dowel rods attached to it) was placed over the top of the individual to expose the cephalothorax but restrain the legs and abdomen (Figure 1 B). A blunt size 28 tapestry needle was then used to smear a small amount of 430 Loctite super glue (Henkel Corp., Ohio, USA) onto the cephalothorax and then onto the transmitter itself. Using fine point forceps, the transmitter was placed onto the cephalothorax and held there for approximately 15 seconds until the glue set. The cephalothorax was chosen as the site of transmitter attachment because it is a non-porous sclerite, whereas the abdomen is a membranous structure that expands and retracts and thus attempting to attach the transmitter on it would injure the spider.
The top sponge was then removed, and the spider remained in the container for 30 minutes to allow the glue to cure further. The spider was then moved into its original container and allowed to recover for a minimum of 12 hours. During CO2 exposure, the spiders did not elicit any signs of discomfort (Dombrowski et al., 2013). Each spider recovered normally in a dimly lit, quiet environment, and subsequent behaviour during the recovery period appeared normal. Only one fatality occurred using this procedure due to excess glue being applied. This individual was humanely euthanised using CO2. All other individuals were released in the location where they were found 12 – 48 hours after the transmitter was attached in the lab. In three cases, individuals were kept in the lab for seven days as they were collected immediately prior to the field site flooding and to ensure a safe release and safe field practices, we allowed extra time after the flood had passed.
Tracking of tagged spiders in the wild
Individuals were tracked one at a time every day using the receiver and antenna, unless extreme weather events prevented entry into the National Park. We recorded details of each individual on every day that they were tracked once the location of an individual was narrowed down to < 1 square meter or the individual was sighted. In each case, we recorded a GPS coordinate and secured labelled flagging tape to foliage where the spider was located. We next measured the distance between the previous and current location (“nightly distance moved”) using a retractable 60 m measuring tape. If individuals were tracked to the same location and had not moved, this was recorded as a nightly distance moved of 0 m. If an individual had not moved and had not been physically sighted after three days, the site was thoroughly examined to ensure the transmitter had not dislodged or that the individual had not died. All males were tracked until either their transmitters dislodged (n=15), the transmitter ran out of battery (n=2), or the spider was found dead (n=2). The number of days of tracking and number of observations for each individual spider are shown in Table 1. If males were tracked into a burrow, the GPS coordinates of the burrow were recorded, and the individuals were left undisturbed. All males that were tracked into burrows subsequently re-emerged from the burrows.
Statistical analysis
All statistical analyses were conducted using R version 4.3.1 (2023). Information on plant community types was obtained from the NSW State Vegetation Type Map (State Government of NSW and NSW Department of Climate Change, Energy, the Environment and Water, 2022) then layered over maps (configured in QGIS 3.30.3) containing data on GPS locations for individual spiders. Temperature and rainfall data were obtained from The Australian Government, Bureau of Meteorology (Australian Government, Bureau of Meteorology, 2023). The data consist of minimum and maximum temperature and rainfall in all forms of precipitation recorded at 9am AEST, and therefore represent weather conditions for the preceding 24 hours.
Total distance travelled over the course of the season was calculated for each spider as the sum of its nightly movement distances, and the mean nightly movement distance was calculated by dividing the total distance by the number of nights when the spider moved to a new location. In addition, we estimated the linear projection (“displacement”) of each spider’s movement over the course of the season by measuring the distance between the first location of capture and the last location recorded in QGIS. To calculate how far on average each male moved per night away from the initial place of capture, we divided displacement by the number of nights between the date of release and the date of final observation (“mean nightly displacement”, Table 3). To quantify the linearity of movement, we then divided displacement by the total distance travelled (“displacement ratio”, Figure 2, Table 3). The displacement ratio would equal 1 if the spider moved in a straight line away from its location of capture, and 0 if the spider circled back around to its initial capture location.
Because we had data on cephalothorax length and weight for different sub-sets of males, we tested for effects of both spider body weight and cephalothorax length (measured once at first capture) on mean nightly movement and mean nightly displacement using Pearson correlation analyses. Effects of environmental factors on nightly movement were analysed in two steps. First, we modelled a binomial variable (moved/not moved) to investigate the factors that affected whether spiders moved or stayed in the previous night’s location. Next, for spiders that had moved since the previous observation, we modelled a continuous (Gaussian) variable (nightly distance moved in m) to investigate the factors that affected how far spiders travelled while they were searching for mates.
The binomial and Gaussian models both contained the following environmental parameters as fixed effects: daily rainfall (mm), nightly minimum temperature (ºC), daily maximum temperature (from the previous day; also, in ºC), and plant community type. The models also included fixed effects of year, and order of days tracked for each individual (each day tracked for each spider was given a number starting at 1 on day 1 and increasing in numerical order). Both models included individual identification and a code representing each day when tracking occurred as random effects. The correlation between minimum daily temperature and maximum daily temperature was weak (R = 0.16), presumably because there was little seasonal change in average temperature during the time that spiders were tracked. We therefore included both the minimum and maximum daily temperature as fixed effects in our models. The plant community type (PCT) data obtained from NSW State Vegetation Type Map describes three broad habitat types present in Lane Cove National Park. These models were fitted using the TMB package (Kristensen, 2016).
For both the binomial and Gaussian response variables, we carried out model selection based on Akaike’s information criterion corrected for small sample size (AICc) using the MuMIn package (Barton, 2022). We compared models that included all combinations of fixed effects (without interactions) and intercept-only models (Kristensen, 2016). All models included the two random effects, and diagnostics were assessed visually. Models with ΔAICc < 2 relative to the best-supported model (i.e., the model with the lowest AICc value) were considered to have some support, whereas models with ΔAICc > 2 were considered to have very low support.