Theory predicts that species engaged in intraguild predation (IGP) can only coexist under limited conditions, yet IGP is common in nature. Habitat complexity can promote co-existence by reducing encounter rates, but little is known about the contribution of differential habitat use. We hypothesized that differential use of alternative habitats promotes coexistence of an IG predator and prey. We evaluated predictions of this hypothesis with an experimental introduction of an IG predator fish into four natural stream communities that previously contained only the IG prey fish. We monitored the development of this IGP over the course of four years to determine how each species used alternative stream habitats. The introduced species prefers pool habitats while the resident species was more evenly distributed across pools and riffles. The density of the resident decreased in the pool habitat preferred by the invader, accompanied by a local increase in the mean of the resident size distribution. Selective predation by the invader on hatchling residents appears to impact the residents’ demographic response. The continued recruitment of resident juveniles in riffles, where the introduced species is rare, facilitated the persistence of the resident. This differential use of habitats was not accompanied by a change in the resident’s growth rates in either habitat. Our results show that differential habitat selection and recruitment promoted persistence during an invasion involving IGP, which helps to bridge the gap between theory and observation in explaining coexistence in IGP systems. 

Preliminary site preparation: We introduced guppies into four KO headwater streams in the upper Guanapo River watershed. These streams exhibit a pool-riffle structure where permanent pools alternate with rocky riffles. Other habitat types exist, such as seepages and flood pools. We combine these subcategories into a single, non-pool category, hereafter “riffles”. Barrier waterfalls separate fish populations and communities occupying this pool-riffle mosaic into defined sections of stream “reaches”. The Introduction and Control reaches each have long stretches of stream with good pool-riffle development and barrier waterfalls which confined guppies within the Introduction reach. In association with other goals of the transplant experiment, we thinned the tree canopy in the Control and Introduction portions of two streams (UPL and TAY) and in that way increased primary productivity.

Transplant and sampling procedures: We transplanted guppies into LOL and UPL in March 2008 and into CAI and TAY in March 2009. LOL and UPL each received 78 guppies and CAI and TAY each received 104 guppies with an even sex ratio in all four streams. Introduction reaches varied from 65 to 150 m in length. Killifish Control reaches were upstream of the guppy Introduction reaches, above barrier falls that prevented further upstream movement of introduced guppies. Control reaches varied from 26 to 62 m in length. 

We censused guppy populations every month during daylight hours. We sampled pools and riffles with dipnets with the goal of capturing all guppies > 14 mm standard length. Guppies from each pool or riffle were kept separate, as were males and females. All fish were measured, marked by subcutaneous injection of colored elastomer (Northwest Marine Technologies, Shaw Island, Washington, USA) then released at the point of capture. 

We sampled killifish with aquarium dipnets after dark with headlamps because they are more easily seen and caught at night. We also used baited minnow traps in deeper pools. Captured fish were separated by location of capture and transported to the laboratory where each fish was measured (total length), weighed, marked in the same fashion as guppies, then returned to the point of capture. We classified the habitat type, pool or riffle, at the time and place of each individual capture. We standardized sampling procedures, e.g. times and personnel, to minimize variation in effort across habitats and samples, so that the number of captures and resulting size distributions could be compared. We censused the killifish in each stream 6 times per year, beginning 1 year (6 samples) prior to guppy introduction in LOL and UPL and 3 months (3 samples) prior to guppy introduction in CAI and TAY. During the first two years of the study (2007-2008), we captured fish every other month. Beginning in 2009, we altered the sampling schedule to better resolve recruitment and growth of the smaller size classes of killifish. We sampled the streams once each month for three months, then skipped three months and sampled again, once each month for three months. The three-month sampling periods were chosen to represent the dry and wet seasons. Logistical constraints near the end of our study limited sampling in late 2011 and 2012 to a single sample.

mark-recapture dataset

site = stream identity

reach = reach identity (Control or guppy Introduction)

fish_id = curated individual id code

capture_history = a string of 0 and 1 values. The length of the string indicates the number of sampling occasions. Values of 1 indicate an individual was captured on a particular sampling occasion and values of 0 indicate no capture on that occasion. 

mark-recapture lags dataset

site = stream identity
reach = reach identity (Control or guppy Introduction)
date = date on which a sample took place
lag30 = number of months (30 days) until the following sample took place

habitat-use dataset

site = stream identity

stream_pair = guppy introduction performed in 2008 (coded: 0) or 2009 (coded:1)

canopy_treatment = intact or thinned canopy

reach = reach identity (Control or guppy Introduction)

habitat = pool or riffle habitat

species = Killifish or Guppy

time_since_introduciton = time since guppy introduction in months (30 days)

stream_length = length of stream of each site-reach-habitat locality

no_captured = number of individuals captured by mark-recapture sampling

capture_density = no_captured/stream_length in individuals per meter

log_capture_density = natural logarithm of capture_density

recruit dataset

site = stream identity

stream_pair = guppy introduction performed in 2008 (coded: 0) or 2009 (coded:1)

canopy_treatment = intact or thinned canopy

reach = reach identity (Control or guppy Introduction)

habitat = pool or riffle habitat

time_since_introduciton = time since guppy introduction in months (30 days)

season  = season identifier (dry: d; wet: w)

search_period = indicator for change in killifish sampling protocol pre-2009 (0) and post-2009 (1)

stream_length = length of stream of each site-reach-habitat locality

adult_n = number of adult killifish (>28 mm) captured in sampling event

recruit_n = number of immature killifish (<28 mm) captured in sampling event

recruit_density = recruit_n/stream_length in individuals/m

log_recruit_density = natural logarithm of recruit_density

growth dataset

site = stream identity

stream_pair = guppy introduction performed in 2008 (coded: 0) or 2009 (coded:1)

canopy_treatment = intact or thinned canopy

reach = reach identity (Control or guppy Introduction)

habitat = pool or riffle habitat

fish_id = curated individual id code

time_since_introduction_centered = time since guppy introduction in months (30 days) centered at the time series mid-point

total_length_initial = initial total length for the growth interval in mm

total_length_final = final total length for the growth interval in mm

time_interval = length of growth interval in months (30 days)

growth_rate = (total_length_final - total_length_initial)/time_interval in mm/mo.

log_growth_rate = natural logarithm of growth_rate

size distribution dataset

site = stream identity

stream_pair = guppy introduction performed in 2008 (coded: 0) or 2009 (coded:1)

canopy_treatment = intact or thinned canopy

reach = reach identity (Control or guppy Introduction)

habitat = pool or riffle habitat

time_since_introduciton = time since guppy introduction in months (30 days)

season  = season identifier (dry: d; wet: w)

no_captured = number of individuals captured by mark-recapture sampling

mean_total_length = mean total length of individuals captured in mm