Invasive species threaten ecosystems globally, but their impacts can be cryptic when they occur indirectly. Invader phenology can also differ from that of native species, potentially causing seasonality in invader impacts. Yet, it is unclear if invader phenology can drive seasonal patterns in indirect effects. We used a field experiment to test if an invasive grass (Imperata cylindrica) caused seasonal indirect effects by altering rodent foraging and seed predation patterns through time. Using seeds from native longleaf pine (Pinus palustris), we found seed predation was 25% greater, on average, in invaded than control plots, but this effect varied by season. Seed predation was 24% - 157% greater in invaded plots during spring and fall months, but invasion had no effect on seed predation in other months. One of the largest effects occurred in October when longleaf pine seeds are dispersed, suggesting potential effects on tree regeneration. Thus, seasonal patterns in indirect effects from invaders may cause underappreciated impacts on ecological communities.

Study site & experimental design

To evaluate seasonal patterns in indirect effects of an invasive species, we used a field experiment in Gainesville Florida, USA that manipulated cogongrass, an invader from southeastern Asia. The experiment consists of forty 4m x 4m plots with established longleaf pine and understory plant communities, treated with a factorial combination of cogongrass invasion and simulated drought. We used 10 invaded and 10 uninvaded control plots. Native plant communities were initially established in 2012 with cogongrass added in half the plots in 2013.

Seed predation experiment

To test if invasion altered seed predation by rodents, we used 26 cm x 26 cm x 6 cm (L x W x H) plastic trays with ~100 dry longleaf pine seeds (estimated by weight) per tray mixed with 1.5 liters of fine sand. Each sampling night, one seed tray was placed in the center of invaded and control plots (20 trays per night, 10 per treatment) and removed the following morning. After collecting trays, sand was sifted and undamaged seeds were counted by hand, to provide a precise count of fully intact seeds. The number of seeds consumed by rodents per tray was calculated by subtracting the number of fully intact seeds from 100. We sampled four consecutive nights and repeated four-night sampling sessions every two months from February 2019 to April 2020, totaling eight sessions and 32 sample nights. Each sampling session was conducted within four days of full moon to control for variation in moonlight, which can alter rodent behavior. To account for any influence of cloud cover on foraging behavior, we used data from a regional weather station (www.timeanddate.com/weather/usa/gainesville/historic) and scored relative cloud cover per night (hourly average 7:00 pm – 8:00 am) corresponding to: clear (0), passing clouds (1), partly cloudy (2), mostly cloudy (3), overcast (4), and fog (5).

Percent vegetation cover (henceforth “cover”) was measured every two months from February 2019 to April 2020, within two weeks of seed tray sampling nights. Cover was estimated by randomly selecting four 1 m2 quadrats within each plot, estimating percent cover for each, and taking the average [24]. Cover included dead and live plants 10 cm or taller, a height sufficient to conceal foraging rodents.

To identify seed predators, we used seven camera traps per treatment (due to limited camera availability) suspended 1m above trays. We also examined trays for scat, tracks, and chewed seeds prior to sifting sand to better assess species that visited trays. To examine differences in rodent foraging time, we assessed camera data from April and August 2019 when average differences in cover between treatments was relatively large (18%) and small (4%).

Data was collected throughout the experiement and recorded in excel sheets. Data was subsequently cleaned and combined in R to produce the final dataset uploaded here. 

This dataset contains data for seed predation and vegetation conditions of each plot during each sampling. Each row represents a single tray placed in one experimental plot for 1 of the sample nights. 

Here is a description of each column:

MonthYear: The month and year for the sample

SamplingPeriod: each sample month we conducted a 4 night sampling session. This column specifies which night the row comes from within the 4 nights of sampling during that particular month.

PlotNumber: the plot ID number

SampleMonth: the month 

Year: the year

TrayID: specific tray number placed in that plot 

InitialSeedWght(g): initial weight of seeds put into the tray prior to sampling

InitialCount: Initial seed count based on the weight having been callibrated each sampling session with the specific batch of seeds.

FinalCount: The count of fully intact seeds

Eaten: Initial count - final count = the total seeds consumed by rodents

Tracks: yes/no for if animal tracks were in the tray when collected

TrackType: specifies type of tracks observed if any

SeedsChewed: yes/no for if there was visual evidence of seeds having been chewed by rodents

Scat/Pee: NA if none, otherwise specifies the species for which scat and pee was found in the trays

notes: general notes from sampling night

PlotTreatment: A = "ambient" or control plot with no cogongrass, And  AC = "ambient cogon" or cogongrass invaded plot

RowNumber: row number from which the plot is from

EdgeOrCore: specifies if the plot was found on the edge of core of the experiment

Lat: latitude

Long: Longitude

LowCover: % vegetation cover less than 100cm tall and taller than 10cm tall

TallCover: % vegetation cover taller than 100cm tall

Bare: % area that had no vegetation cover taller than 10cm

TotalCover: total area with cover greater than 10cm tall

AverageSky: average nightly cloud cover score based on a 1-5 score specified in the methods

SumSkyScore: nightly sum of cloud cover score

RemainingSeeds: total intact seeds remaining in each tray