Study species
Pelophylax porosus brevipodus is found exclusively in rice paddy areas (Okochi et al. 1997), where it remains during both the breeding and nonbreeding seasons and rarely leaves the water’s edge (Sawahata 2002). Breeding occurs from April to July and metamorphosis takes place thereafter. The snout–vent length (SVL) is 35–62 mm in mature males and 37–73 mm in females (Matsui and Maeda 2018). Pelophylax porosus brevipodus is an opportunistic predator; its diet includes spiders, terrestrial and aquatic insects, and small frogs (Hirai and Matsui 2001). Its estimated longevity in the wild is 3–4 years (Khonsue et al. 2002).
Study site
Our study site was located in a lowland rice paddy area in Takashima, Shiga Prefecture (Fig. 1). Over the 10-year period from 2002 to 2012, the mean annual precipitation was 2,011 mm, the mean maximum snow depth in winter was 44.6 cm, and the mean temperature was 14.1°C (Japan Meteorological Agency 2013). We selected three rice paddy areas (areas A–C) and two adjacent biotopes (biotopes 1 and 2) as the study site (Fig. 1). Land consolidation and improvement projects have been undertaken around the study site: drainage channels were deepened and lined with concrete and irrigation channels were replaced with culverts. The Japanese tree frog (Dryophytes japonicus Günther, 1859), P. p. brevipodus, and the American Bullfrog (Lithobates catesbeianus Shaw, 1802) occur in the area (Naito et al. 2012a).
At the end of April, the rice paddies were yearly flooded and young rice plants were transplanted. Water depths were maintained at 4–10 cm until intermittent drainage called “naka-boshi” (drying of the soil to release harmful gases and supply oxygen to the rice roots for 1–2 weeks) in June or July. After naka-boshi, the rice paddies were flooded repeatedly and drained until approximately 1 week before harvesting in September or October. They remained drained until spring, when the next cultivation period started. Each rice paddy was surrounded by a 1–2-m-wide, 30–50-cm-high earthen vegetated levee; this vegetation was mowed to a height of a few centimeters two or three times a year during the cultivation season.
Biotopes 1 and 2 were constructed at the edge of the rice paddy area (Fig. 1) in 2005 and 2006, respectively. Biotope 1 measured 10 × 50 m and biotope 2 was 17 × 108 m. The maximum depth in both biotopes was about 50 cm. The flashboards between the drainage channel and biotopes were open from March to September and drainage water from upstream rice paddies ran through the biotopes. Even when the flashboards were closed, precipitation created water pools in the biotopes year-round. The biotopes were originally created to conserve fish species that swim up the drainage channel from Lake Biwa and spawn in the rice paddies. Reeds started growing in April, reaching heights of 4–5 m in summer. The open water surface decreased to about 10% in biotope 1 and 40% in biotope 2. In late fall, the aerial portions of the reeds decreased, and the open water surfaces fell to a few percent in biotope 1 and 35% in biotope 2. At the margins and banks of the biotopes, we found plant species that are common in rice paddy areas. The bank was mowed a few times each year in spring and summer. In March 2009 and 2010 and January 2012, the vegetation in the ponds, which was mainly reeds, was cut down and removed to prevent succession to a dry environment.
Survey
We walked a circuit of the levee and captured any P. p. brevipodus seen within 30 cm on either side of the edge of the levee in the rice paddy and biotope area (Osawa et al. 2005). For the biotope area, we captured frogs both in the water and on levees. After catching the frogs, we measured their SVL to the nearest 1 mm with plastic calipers and classified individuals as adults (SVL ≥ 37 mm; including subadults) (Matsui and Maeda 2018) or juveniles (SVL < 37 mm). Although individuals with 35- and 36-mm SVL could potentially include male ones, we confirmed all the individuals lacked secondary sexual characters of males and were categorized as juvenile. For adult frogs, we also recorded the sex. Males were identified using secondary sexual characters (nuptial pads and vocal sacs) and females by the lack of these characters (Matsui and Maeda 2018). If we could not identify presence/absence of the sexual characters from adults, the individuals were categorized as “unknown”. Frogs were released at the capture site. At both sites, we conducted surveys at 800–1200 h or 1400–1800 h, for 3 h on average, and conducted 22 surveys from May to November 2010 on a weekly basis, 11 surveys from April to October 2011 on a biweekly basis, and no survey in September 2011.
We used CMR methods to investigate movement patterns and GRs, and to estimate population parameters. Frogs with SVL ≥ 37 mm (in 2009) and SVL ≥ 45 mm (2010–2012) were marked with unique passive integrated transponder (PIT) tags (from 2009 to May 2011: TX148511B [2.12 × 8.5 mm, 0.067 g]; Destron Fearing, Dallas, TX, USA); after May 2011: BIO 9.HG.01 [2.1 × 8 mm, 0.08 g]; Biomark, Boise, ID, USA). We increased the minimum SVL size for marking because the PIT tags sometimes slipped off frogs with SVL < 45 mm before release. We defined the sexes as described above. We chose PIT tagging as the marking method because it has little effect on reproduction or long-term survival (Pyke 2005) and is suitable for long-term studies.
At the bottom left dorsal side of a frog, we made a skin incision a few millimeters long using ophthalmic surgery scissors, and inserted a PIT tag between the skin and muscle. The PIT tags were soaked in 100% ethanol for > 24 h and the scissors were dipped in 100% ethanol before use. Frogs were kept for 48–72 h to heal and then released. In the rice paddies, when a frog was captured in area A, we released it at point A; the same concept was applied to areas B and C (Fig. 1). We released the frogs in the same biotope from which they were captured. CMR was conducted simultaneously in the biotopes and rice paddies 14 times from May to November 2009 on a biweekly basis, 8 times from May to December 2010 on a monthly basis, 6 times from April to October 2011 on a monthly basis, and once each in January and May 2012.
During the survey period, each captured frog was scanned with a portable microchip reader (EXHS6103L3; Biomark) to determine if it had been tagged. The recapture effort occurred at least 4 days after the day on which the frogs were released. One assumption of the Jolly–Seber model is that sampling and tagging are instantaneous, but because we needed to treat the insertion point in the skin, there was a 48–72 h time difference. For our analysis, the sampling date was the date on which newly tagged frogs were released. As part of the movement behavior and GR survey, we occasionally looked for marked individuals and either did not measure SVL or mark newly captured frogs (34 individuals in 2009), or measured SVL but did not mark new frogs (6 individuals in 2010).
Data analysis
We compared the body size distributions between the biotope and rice paddy areas using pooled data from the 2010 and 2011 surveys, as we were primarily interested in the difference between the habitat categories rather than variability within a category. To examine the differences in population variance between habitat categories, a Levene’s test was performed using the lawstat package (Gastwirth et al. 2023) of R 3.6.3 (R Core Team 2021).
Based on the CMR data, we investigated seasonal movement patterns, and recorded the direction and distance of frog movement between recapture locations. We also calculated individual GRs (mm/day) as the change in SVL divided by the number of days between captures. Only one GR (the last measurement minus the first measurement divided by the interval in days) was used for each individual. We used linear regression analysis to estimate the relationship between GR and length (Galatti 1992). Analysis of variance was used to assess the significance of this relationship (Sokal and Rohlf 1995).
We estimated population parameters using the Jolly–Seber population analysis (POPAN) model (Arnason and Schwarz 1999) implemented in the program MARK ver. 6.2 (White and Burnham 1999). This model has been successfully applied to birds, butterflies, marine mammals, reptiles, and amphibians (Arnason and Schwarz 1999). The POPAN model postulates the existence of a super-population (N) that includes all of the animals present in the study area at any time between the first and last sampling occasions, and estimates the probabilities of survival between occasion i and i + 1 (φi), capture at occasion i (pi), and an animal from the super-population entering the population between occasion i and i + 1 (bi) using the maximum likelihood method. Net population entrance (including immigration and births) (Bi) and population size (Ni) at occasion i were derived from the estimated parameters.
We used the POPAN model for two reasons. First, it considers imperfect detection, which can seriously bias inferences from surveys and indicate negative population trends (Laurance et al. 1996). The second reason was that the POPAN model estimates the size of the super-population. In this study, because one of our primary interests was to document the movement patterns of P. p. brevipodus, we defined the super-population as the frogs in all three rice paddies and both biotopes (Wagner et al. 2011).
We created a global model that depended fully on sex and time, and performed a goodness-of fit test using the RELEASE program (TESTs 2 and 3) within MARK to validate the model assumptions of capture homogeneity and survival homogeneity between marked and non-marked individuals. After confirming that the global model satisfied the assumptions, we created models in which each parameter depended on sex, time, or neither, and selected the variant with the smallest corrected Akaike information criterion (AICc) value as the best model (Hurvich and Tsai 1989; Lebreton et al. 1992; Burnham and Anderson 2002). A sine function was used for φ and p because these parameters are constrained between 0 and 1, whereas we used a log link function for N to ensure that it was positive (i.e., N > 0). To specify that a set of b summed to 1, a multinomial logit link function was used for each group.