Feeding habits of the endangered Japanese diving beetle Hydaticus bowringii (Coleoptera: Dytiscidae) larvae in paddy fields and implications for its conservation

The diving beetle Hydaticus bowringii Clark, 1864 (Coleoptera: Dytiscidae) is on the Red List of Japan as ‘Near Threatened’. However, there is no quantitative information on the feeding habits of its larvae, which could be used to aid its conservation. In order to determine the prey that are important for the survival and growth of larvae of H. bowringii, we combined the results of fi eld surveys of paddy fi elds and rearing experiments. In the fi eld, H. bowringii larvae predominantly feed on tadpoles of fi ve species of frogs and occasionally also on insects, loaches and worms. The phenology of the tadpoles was similar to that of larvae of H. bowringii, as their abundances increased from May to June and decreased in July. Experimentally reared larvae of H. bowringii grew faster when fed tadpoles than when fed Sigara nymphs or a mixture of both prey, and more emerged as adults when tadpoles were included in their diet. Adults were larger in the tadpole treatment than in the Sigara treatment. Based on these results, we conclude that tadpoles are more suitable prey for the survival and growth of larvae of H. bowringii than insects. The decline in the abundance of frogs could be one of the factors determining the decrease in the local abundances of H. bowringii. In conclusion, we affi rm that in order to conserve populations of H. bowringii it is crucial to maintain paddy fi eld environments in which frogs are abundant.


INTRODUCTION
Diving beetles (Coleoptera: Dytiscidae) are top predators, especially in shallow and fi shless wetlands, such as paddy fi elds (Cobbaert et al., 2010;Ohba, 2011). Both the larvae and adults are carnivorous. The larvae are exclusively predatory, whereas adults also scavenge for food (Culler et al., 2014). They prey on zooplankton, insects, horsehair worms, gastropods, fi sh, amphibians and reptiles (Culler et al., 2014;Watanabe, 2019). They are often used as indicators of biodiversity in freshwater environments worldwide (Foster & Bilton, 2014) and also in Japanese paddy fi elds (MAFF & NARO, 2012). In Japan, approximately 46% (60/131 species) of the species of diving beetles inhabit paddy fi elds, which also include agricultural ditches and irrigation ponds (Saijo, 2001;Mitamura et al., 2017;Watanabe et al., 2019;Nakajima et al., 2020) and 45% (27/60 species) of them are on the Red List of Japan (Ministry of the Environment of Japan, 2019). The diversity of diving beetles is declining in paddy fi elds because of farmland City, Ibaraki Prefecture, Japan from 10 May to 31 August 2018 and 26 April to 31 August 2019 (Fig. 1). Mean ± SD water depth was 6.18 ± 2.38 cm (range 1.0-12.6 cm) in Paddy A and 6.04 ± 2.67 cm (1.3-14.0 cm) in Paddy B. The landscape surrounding the sites studied was dominated by paddy fi elds and forests, with few other wetlands (i.e. rivers, abandoned paddy fi elds and ponds). Paddy fi elds were fl ooded from late April to late August in both 2018 and 2019. Paddy A was connected to a traditional earth ditch, whereas Paddy B was not. No insecticides or herbicides were applied in either paddy fi eld during the study period. There was an abundance of water plants in addition to planted rice, including Eleocharis kuroguwai Ohwi, Monochoria vaginalis (Burm.f.) C. Presl ex Kunth, Murdannia keisak (Hassk.) Hand.-Mazz., Oenanthe javanica (Blume) DC. and Persicaria thunbergii (Siebold et Zucc.) H.Gross at both sites. On June 2, 29 and July 17, 2018 and on June 1, 16 and July 8, 2019, weeding was carried out in paddy fi elds using a hand-operated weeding machine. No surveys were conducted on these days as the water was turbid and unsuitable for observation.

Feeding habits and phenology of larvae of H. bowringii
Nocturnal observations of the paddy fi elds were carried out at 1-9 day intervals from May to August 2018 and 2019 (47 times in 2018 and 42 times in 2019). Following previous studies, (Ohba, 2009a, b) these observations lasted a few hours between 20:00-2:00 h, but not on rainy days. We searched for larvae of H. bowringii using a 400 lm fl ashlight (MG-186R, GENTOS, Tokyo) while walking at a constant speed (3 m/min) in the same direction on the levee around the paddy fi elds. The number of larvae within 1.5 m of the levee edge during one round of each paddy fi eld was recorded. When we saw a hunting event, both the larva and the prey were captured using a D-frame net (30 cm width, 1 mm mesh size). Each larva of H. bowringii was placed in a plastic container (7.5 cm width, 10.5 cm long, 3.5 cm deep) attached to a ruler and photographed. Head widths of larvae were measured using ImageJ software (Abràmoff et al., 2004) was used as a measure of body length (Johansson & Nilsson, 1992;Ohba, 2009a). Then, the larva was released back into the paddy fi eld from which it was captured. The prey was preserved in 80% ethanol and brought to the laboratory for identifi cation and measurement. The prey species were identifi ed based on Mukai a result of which, the abundance of D. sharpi increases (Nishihara, 2012).
Hydaticus bowringii Clark, 1864 (Coleoptera: Dytiscidae) is another endangered diving beetle in Japan, which on the Red List of Japan is listed as 'Near Threatened' (Ministry of the Environment of Japan, 2019). It is distributed widely in Japan, China, Korea and Taiwan (Mori & Kitayama, 2002). This species is considered to be univoltine (Tsuzuki et al., 2003), although there is no quantitative fi eld survey for this species. From spring to early autumn (i.e. late April to September), adults occur in paddy fi elds and irrigation ponds, and their larvae can be found in paddy fi elds (Saijo, 2001;Watanabe, 2017a). Adults leave water bodies from late autumn to early spring (i.e. October to early April) and probably overwinter on land (Mori & Kitayama, 2002) as overwintering adults have been found resting on mud at the base of water plants growing along the shores of drained irrigation ponds (Watanabe, 2017b).
Very little is known about the ecology of the larvae of H. bowringii. They have been observed feeding on tadpoles in the fi eld and on chironomid (Diptera) larvae in the laboratory (Tsuzuki et al., 2003;Watanabe, 2017a), which may be an opportunistic option for survival in the laboratory and not necessarily a common prey in nature. However, there is no quantitative information about the larval diet of this species to guide its conservation. In the present study, we identify the important species of prey for the survival and growth of larvae of H. bowringii by using fi eld surveys and rearing this species in the laboratory. We use the results to discuss its habitat requirements in paddy fi elds necessary for maintaining thriving populations of H. bowringii.

Quantifi cation of potential prey
To determine the phenology and abundance of potential prey, we conducted fi eld surveys once a week from May to August 2018 and late April to August 2019 at both paddy fi elds. We placed a quadrat (25 cm vertical, 20 cm horizontal, 20 cm height) at fi ve random locations near the edge of the levee and swept fi ve times within the quadrat using a square-framed net (15 cm by 12 cm with 0.5 mm mesh size) (bl-g1, Mitani-Turigyogu, Saitama).
We swept ins ide the quadrat, including the surface of the mud. The samples were transferred to a white plastic tray and all the suffi ciently large animals (> approximately 3 mm) were sorted, counted, identifi ed to the lowest possible taxon and released at the site. Prey that could not be identifi ed in the fi eld were preserved in 80% ethanol and brought to the laboratory for identifi cation using the keys of Mukai (2014), Matsui & Maeda (2018) and Kawai & Tanida (2018).

Rearing experiment: Effect of diet
To compare the effect of potential prey on the survival and growth of larvae of H. bowringii, we reared them in a laboratory from May to July 2019. The fi eld survey in 2018 revealed that H. bowringii mainly feed on tadpoles of the Japanese tree frog, Dryophytes japonica (Günther, 1859) (see Table 2). Among insects, larvae of H. bowringii attacked Sigara nymphs in the fi eld ( Table 2) that were abundant at both sites when the larvae of H. bowringii were present (Fig. 2). Therefore, we used D. japonica tadpoles and Sigara nymphs; mostly those of S. septemlineata (Paiva, 1918), but also some S. nigroventalis (Matsumura, 1905) as prey. Although chironomid larvae were also abundant and attacked by larvae of H. bowringii, we did not use them as prey because they are usually buried in mud. Tadpoles and Sigara nymphs were collected from paddy fi elds at the Tsukuba-Plant Innovation Research Centre, in Tsukuba City and Ishioka City.
To obtain fi rst instar larvae, we collected adult females of H. bowringii from paddy fi elds in Ishioka City from May to June 2019 and kept them individually in plastic cups (12.9 cm diameter, 6.5 cm deep) containing 200 ml dechlorinated tap water. The cups containing larvae were kept at 25°C and under a 14L: 10D photo cycle in an incubator (Bio Multi Incubator, Nippon Medical & Chemical Instruments, Osaka). The temperatur e used was based on the average temperature from June to A ugust 2018 in the study area (i.e. 25.3°C, Japan Meteorological Agency (2020)). Plastic mesh (16.5 cm by 5 cm with 3 mm mesh) in the plastic cups served as an oviposition substrate and was checked for the presence of eggs daily. When eggs were found, the mesh with eggs was moved to another plastic cup containing 200 ml of water. We checked the eggs daily and transferred hatchlings to individual rearing cages (88 mm × 88 mm × 52 mm).
The experimental design followed that of Ohba (2009a, b). Rearing cages were placed in a plastic container (91.1 cm × 61.1 cm × 20.3 cm) with two fi ltration devices (Suisaku Eight Flower S, Suisaku, Tokyo). To prevent the water in the rearing cages from fouling, a 3 × 3 cm hole was made in the bottom of each rearing cage, and a 1-mm polyethylene mesh covered the hole to allow the water to move between the plastic container and rearing cages. We placed a 1-cm layer of river gravel (2-3 mm in size; Happiness Chuume, Stone Dealer Shinsei, Osaka) on the mesh. Water depth in rearing cages was kept at 2 cm above the gravel layer (3 cm for third instar larvae). In addition, a 5-cm long piece of an aquatic plant, Ceratophyllum demersum L., was placed in each rearing cage as a perch for the larvae.
Tadpole, Sigara and tadpole-Sigara mixed diet were the three experimental diets. In the tadpole treatment (n = 21), six small tadpoles (1.4-2.0 cm snout-vent length, SVL) were provided daily for the fi rst instar larvae, six medium tadpoles (2.1-2.5 cm SVL) for second instar larvae and six large tadpoles (2.6-3.2 cm SVL) for third instar larvae. In the Sigara treatment (n = 19), 10 individuals of 4-5th instar nymphs (4.0-5.0 mm total body length) were provided daily for fi rst instar larvae, 10 adults (5.0-6.0 mm) for second instar larvae and 20 adults for third instar larvae. In the tadpole-Sigara mixed diet treatment (n = 12), six small tadpoles or six Sigara nymphs were provided daily for fi rst instar larvae, six medium tadpoles or six Sigara adults fore second instar larva and six large tadpoles or 10 Sigara adults for third instar larvae, and the species of prey alternated every 2 days beginning with tadpoles. Some of the fi rst instar larvae moulted to the second instar within 2 days and were not fed Sigara nymphs.
Based on the preliminary tests, the number of prey provided daily in each treatment was suffi cient for the larvae to be satiated. We checked the rearing cages daily and recorded both the survival and development of the larvae of H. bowringii and number of prey consumed. Dead prey were removed and replaced with new prey daily to maintain a constant prey density. Third instar larvae just b efore pupation often swam hurriedly and tried to climb the walls of the rearing cage (Watanabe, pers. observ.). Larvae that showed such behaviour were checked again 2 h after feeding and those that did not eat any prey were moved to a plastic cup (88 mm × 88 mm × 52 mm) with wet peat moss for pupation (Ohba 2009a, b). The day when the larva made the pupal chamber in the moss was defi ned as the fi nal day of the larval period. We also recorded the pupal period for each individual, i.e. interval between the construction of the pupal chamber and adult emergence.
When an adult emerged, its sex, total body length and wet weight were recorded. Total body length and wet weight were measured using a calliper (0.1 mm precision, Pocket Vernier Caliper, Shinwa Rules, Niigata) and an electronic balance (0.01 mg precision, HR-202i , A & D Company, Limited, Tokyo), respectively. To compare the effect of different prey on beetle development, we also recorded the total body length of adults of H. bowringii (53 males and 46 females) collected from the study sites between May and July 2019. We visually searched for adults at night and collected them using a D-frame net. After the adults were measured, they were released.

Statistical analysis
The statistical software R version 3.6.0 (R Core Team, 2019) was used for all analyses. To compare the diets of the larvae of different instars of H. bowringii, their diet was classifi ed into three prey categories: tadpole (including frog eggs), loach and invertebrate. Then, we used multinomial logit models (MLM) with prey category as a response variable, and the larval instar of H. bowringii, the site (i.e. Paddy A or Paddy B), and year as explanatory variables ('brglm2' package, Kosmidis et al., 2020). We created models with all pos sible combinations of explanatory variables and regarded the model with lowest Akaike's information criterion (AIC) as the best model.
To determine the predator-prey size relationship in the fi eld (n = 136), we used multiple linear regression model s (MLRM) with log 10 -transformed prey body width as the response variable and the log 10 -transformed head width of lar vae of H. bowringii, the site and prey category as explanatory variables. We created models with all possible combinations of explanatory variables and regarded the model with lowest value of AIC as a best model ('MASS' package, Ripley et al., 2020). Cases in which several larvae were feedi ng on a single item of prey were not included in these analyses. We also used a one-way analysis of variance (ANOVA) to compare the log 10 -transformed head widths of larvae of H. bowringii at the different sites.
To assess the prey preferences of larvae of H. bowringii for three categories of prey, we calculated the Manly's alpha (α) (Chesson, 1978): where r i is the proportion of prey categor y i in the diet, p i is the proportion of prey category i in the environment, and m is the total number of categories of prey in the environment (paddy A: m = 3, paddy B: m = 2). In Paddy B, since there was no predation on lo ach, we could not calculate α for loach. We summarized the fi eld predation events and the prey abundance in terms of periods of high (May to June) and low (July to August) tadpole abundance in each year and paddy fi eld to calculate α i ('electivity' package, Quintans, 2019). Sin ce the proportions of different categories of prey in the diet of larvae of H. bowringii did not differ among instars (see below), we pooled the fi eld predation events for all instars. Values of α i range between 0 (avoidance) and 1 (preference). For visualization, values of α i were converted i nto electivity indices according to Chesson (1983). When the value of the electivity index equals 0, the prey category is consumed in proportion to the abundance in the environment, positive values indicate preference and negative values indicate avoidance of a prey category.
To assess the correlation between population density of larvae of H. bowringii (instars pooled) and that of their potential prey at each site, we calculated the Spearman's rank correlation coeffi cient using the data on population density summarized for each month; number of larvae of H. bowringii recorded per night; total number of each prey taxa recorded in four surveys per month.
We used a two-way ANOVA to determine differences in the developmental period from hatching to emergence of the adults of both sexes and in the different treatments. In addition, the Tukey-Kramer HSD test was used to compare the developmental periods between treatments. To determine if emergence success differed between treatments and sexes, we used Firth's bias-reduced penalized likelihood logistic regression ('logistf' package, Heinze et al., 2018). Using this method, it is possible to estimate the parameters in a logistic regression even for small samples in which separation occurs (Heinze & Schemper, 2002). In the model, 'normally emerged or not' was the response variable, and sex and treatment (tadpole, Sigara, or tadpole-Sigara mixture) were included as explanatory variables. Model signifi cance was assessed using a penalized likelihood ratio test. To compare the total body length and wet weight of newly emerged adults by sex and treatment, we used a two-way ANOVA followed by the Tukey-Kramer HSD test. To determine the association between type of prey and the adult size of H. bowringii, total body length was also compared with that of wild-caught individuals. Since the wet weight of wild-caught individuals varied because of differences in the development of their reproductive organs and level of satiation, we did not compare the wet weights of the reared and wild-caught individuals.

Phenology of larvae of Hydaticus bowringii and their potential prey
Larvae of Hydaticus bowringii were recorded from May to August 2018 and 2019 (Fig. 2a). The number of larvae peaked in May and decreased rapidly in June in both years, with a more or less visible second peak in August. Seasonal dynamics of tadpoles (D. japonica, Rana japonica Boulenger, 1879, Pelophylax porosus (Cope, 1868), Glandirana rugosa (Temminck & Schlegel, 1838), and Zhangixalus schlegelii (Günther, 1858)) was similar to that of larvae of H. bowringii (Fig. 2b). Population densities of larvae of H. bowringii and tadpoles were positively correlated at both sites (  Ris, 1916 and Anax parthenope (Sélys, 1839)), larvae of Trichoptera and Diptera (Syrphidae, Tabanidae and Tipulidae) peaked in density from July to August in both years ( Fig. 2b-f). Chironomid larvae were always abundant during the period studie d with a peak in May (Fig. 2c). There were two peaks in the abundance of larvae of Culicidae, o ne in May and the other in August 2019, whereas there was only one peak in May 2018 (Fig. 2d). Abundance of loaches (Misgurnus anguillicaudatus (Cantor, 1842) and Lefua echigonia Jordan & Richardson, 1907) increased from May to June and August, and that of Tubifi cidae peaked in May 2018 and June 2019 (Fig. 2e, f).

Larval diet in the fi eld
The null model was selected as the best model of the diet composition of larvae of H. bowringii and revealed that it is not associated with larval instar, site or year. All larvae mainly consumed tadpoles of the fi ve frog species and loaches, and insects and Tubifi cidae made up < 6% of their diet (Table 2, Fig. 3). Larvae of H. bowringii other than the fi rst instar were the most frequently recorded eating tadpoles of Dryophytes japonica, followed by those of Zhangixalus schlegelii. In a few cases, several beetle larvae were recorded feeding on the same prey individual (Fig. 3e). The second and third instar larvae also fed on frog eggs laid on the stems of water plants (Fig. 3f). We observed a larva thrusting its head into a jelly-covered frog egg mass and grasping an egg with its mandibles. In Paddy A, larvae of H. bowringii consistently preferred tadpoles along with loach in July to August 2018 and May to June 2019, whereas they avoided invertebrates (Fig. S1a). In Paddy B, larvae of H. bowringii consistently preferred tadpoles and avoided invertebrates (Fig. S1b).
The model with log 10 -transformed head width of larvae of H. bowringii, site and category of prey was selected as the best model for predicting the log 10 -transformed body width of prey. Large larvae killed signifi cantly larger prey than small larvae (MLRM, p < 0.0001, Fig. 4) and the relationships differed signifi cantly for the different categories of prey, such that body width of invertebrate prey was smaller than that of the tadpoles (MLRM, p = 0.014, Fig.   S2) and in Paddy B was bigger than in Paddy A (MLRM, p = 0.003). In addition, the log 10 -transformed head width of larvae of H. bowringii was bigger in Paddy B than Paddy A (ANOVA, F 1, 134 = 8.85, p = 0.003).

DISCUSSION
This study revealed that tadpoles are an important source of food for larvae of H. bowringii. Larvae of several diving beetles, such as, Acillius, Colymbetes, Cybister, Dytiscus, Hydaticus, Ilybius and Rhantus are known to consume tadpoles (Wells, 2007). Hydaticus larvae are reported feeding on tadpoles in the fi eld and laboratory (Nishida, 2000;Borzée, 2019;Gould et al., 2019). We showed that larvae of H. bowringii predominantly feed on frog tadpoles in the fi eld and the inclusion of tadpoles in their diet greatly enhanced larval survival and development in the laboratory. The larvae may eat more insects and other invertebrates than recorded in the fi eld. However, predation events involving small insects may be less detectable because of the shorter handling times associated with small prey. In addition, small prey may not have been detected by us during the night sur veys due to visual limitations. For further elucidation, it is necessary to clarify the diet of larvae of H. bowringii using isotope analysis and DNA barcoding.
Proportions of different categories of prey in the diet of larvae of H. bowringii did not differ among instars as all instars fed mainly on tadpoles. This contrasts with ontogenetic diet shifts reported for Cybister chinensis, in which the fi rst and second instar feed mainly on insects, whereas the third instar feeds equally on insects and vertebrates (tadpoles and fi sh) (Ohba, 2009b). In general, predator and prey body sizes covary, especially in freshwater ge neralist predators, which each capture prey of a particular body size (Brose et al., 2006;Nakazawa et al., 2013). The mode of feeding greatly infl uences the predator-prey body mass allometry in aquatic insects, with sucking predators, such as diving beetle larvae and aquatic bugs, can feed on larger prey than chewing predators of the same size (Klecka & Boukal, 2013). In addition, the predator-prey body size relationsh ips are weaker in specialist predators that focus on a particular species of prey (Nakazawa et al., 2013). The prey body size increased with increase in the size of the larvae of H. bowringii as previously reported for Cybister brevis and C. chinensis (Ohba, 2009a, b). The prey body width was larger in Paddy B than Paddy A, because there were many cases of predation by large larvae in Paddy B than in Paddy A. Although we did not weigh the prey, tadpoles have a greater wet weight than aquatic insects of the same body length (Nakazawa et al., 2013). Therefore, the larvae of the tadpole-eater, H. bowringii, can consume larger prey relative to own body size than the larvae of Cybister chinensis and C. brevis, which are insectivores. Thus, larvae of H. bowringii may specialize on tadpoles from the fi rst instar as does the giant water bug, Kirkaldyia deyrolli (Vuillefroy, 1864) (Ohba et al., 2008a;Nakazawa et al., 2013;Ohba & Tatsuta, 2016). Functional morphology and behavioural adaptations of H. bowringii for catching tadpoles could be an interesting future study.
Interestingly, we also observed a larva of H. bowringii attacking frog eggs in the fi eld. Previous studies report that the larvae of Ilybius and adults of Agabus bipustulatus (Linnaeus, 1767) consume amphibian eggs in laboratory experiments (Resetarits, 1998;Kurdíková et al., 2011) but our observation may be the fi rst example of it occurring in the fi eld. Diving beetle larvae detect prey using visual, tactile, or chemical cues (Culler et al., 2014), which may enable larvae of H. bowringii to fi nd frog eggs, although we cannot exclude that such events occur simply at random. In addition , it is reported that Hydaticus parallelus Clark, 1864 and Rhantus pacifi cus (Boisduval, 1835) oviposit on the surface of frog egg masses and their larvae fed on newly hatched tadpoles (Williams, 1936;Gould et al., 2019). At the sites studied, all larval instars of H. bowringii were often observed on frog egg masses (Watanabe, personal observation), which may indicate that adults of H. bowringii lay their eggs on or near frog egg masses in order to provide their larvae with an easily available source of prey.
Our results confi rm that tadpoles are a suitable prey for the survival and growth of larvae of H. bowringii. Ohba et al. (2012) report that the protein content of frogs is greater than that of odonate nymphs. This may affect the survival and growth of larvae of H. bowringii. Culler & Lamp (2009) report that larvae of Agabus disintegratus (Crotch, 1873) are bigger when fed only mosquito larvae or copepods than when fed ostracods and suggest that mosquito larvae and copepods are of poorer nutritional value for A. disintegratus than ostracods. On the other hand, the Japanese species of Cybister (C. brevis, C. chinensis, and C. tripunctatus lateralis) are known to prey mainly on insects and odonate nymphs, which are a better prey for the survival and growth of these species than tadpoles (Ohba, 2009a, b;Ohba & Inatani, 2012). That the diet of different species differs might be due the differences in body size, phenology, microhabitat use (Yee et al., 2013), digestive enzymes (Walker et al., 2016) and prey selectivity related to prey nutritional values (Dudová et al., 2019). Further studies are needed to quantify interspecifi c dietary overlaps of co-occurring diving beetle larvae in paddy fi elds.
Diving beetle larvae gain most weight and increase in body size during the third instar (Kingsley, 1985;Scholten et al., 2018). Given their smaller size, the Sigara nymphs are of lower nutritional value per individual than are tadpoles for larvae of H. bowringii, especially during the third instar, which is indicated by the higher number consumed and longer development in the Sigara treatment than in the tadpole treatment. The relatively minor retardation of growth and lower adult weight in the tadpole-Sigara mixture treatment might have been greater if the fi rst instar larvae were fi rst fed with Sigara nymphs rather than tadpoles. In addition, our results might be at least partly driven by the amount of food provided for the larvae than the type of prey.

Implications for the conservation of Hydaticus bowringii
Seasonally occurring pre dators should synchronize their phenology with the occurrence of prey (Evans, 1982;Sota, 1985;Ohba et al., 2008a). The development of H. bowringii larvae matched tadpole presence from May to June in 2018 and 2019 at our fi eld sites. In temporary wetlands, such as paddy fi elds, larval period of aquatic insects needs to be short enough for them to complete th eir development before the wetland dries out (Williams, 2006). Given their large size, tadpoles are a suitable prey for fast growth and shortening of the larval period. Chironomid larvae were also abundant when larvae of H. bowringii were present, however, their availability was low because they are usually buried in the mud.
Recently, frog populations have been declining in Japanese paddy fi elds because of modern improvements in drainage systems by consolidating fi elds and habitat fragmentation by urbanization (Natuhara, 2013). In Japan, many paddy fi elds are fl ooded between late April and early June and drained for 2 weeks in mid-July to stop the tillering of rice plants (Nisikawa & Miyashita, 2014). This midseason drainage is considered detrimental for large-sized diving beetles because they cannot pupate before this event (Ichikawa, 2002;Nishihara et al., 2006). In our laboratory experiment, larvae of H. bowringii became adults in approximately 19 days in the tadpole treatment and most third instar larvae disappeared in the fi eld in early July. These results indicate that H. bowringii may be less affected by the mid-season drainage when tadpoles are available because the adults can escape the drainage by fl ying to other water bodies. On the other han d, mid-season drainage can negatively affect the tadpoles before they metamorphose (Nisikawa & Miyashita, 2014), which may indirectly affect H. bowringii through prey availability. We conclude that in order to maintain a suffi cient abundance of tadpoles it is important to adjust the timing of the mid-season drainage of paddy fi elds.
Our study indicates that proper maintenance of paddy fi elds is essential for sustaining populations of H. bowringii. We recommend the following practices for the conservation of H. bowringii. To maintain amphibian populations in paddy fi elds, create a traditional earth ditch connected to the paddy fi elds (Uruma et al., 2012;Taqumori et al., 2020) or dig a V-shaped groove to keep some water in the paddy fi eld (Yabu, 2005). The earth ditch would act as a refuge for tadpoles during the mid-season drainage and a habitat and refuge from predators for frogs (Takeuchi et al., 2019). Aquatic insects such as odonate nymphs can also utilize the earth ditch as a habitat and a refuge during drainage (Yabu, 2005), which may further support other diving beetle species such as Cybister. In fact, Watanabe (2016) reports that the abundance of aquatic beetles and bugs was higher in paddy fi elds with an earth ditch than those without. The presence of a diversity prey in paddy fi elds may also prevent intraguild predation occurring in diving beetles.
Forests surrounding paddy fi elds could also be important for maintaining populations of H. bowringii, because the abundance of some frogs (Rana ornativentris and Zhangixalus schlegelii) in paddy fi elds increases with forest cover (Uruma et al., 2012;Zheng & Natuhara, 2020). Maintaining forests surrounding paddy fi elds could also contribute to the conservation of Dytiscus sharpi, which can only complete their development when feeding on the tadpoles of species of Rana (Inoda et al., 2009;Nishihara, 2012).