Potential of web-based data for studying reproductive biology
We demonstrated the usefulness of web-based citizen data to investigate breeding season and the associated latitudinal cline, and nuptial color variations of a mutually ornamented fish, Japanese dace. It is now widely recognized that global climate change significantly affects organismal phenology, including breeding period (Walther et al. 2002). Web image searching can facilitate comparison between current and past phenologies in a wide range of species. In our case, the mean breeding period of Japanese dace in Chitose River was mid-July in the 1930s (Okada 1935), but was estimated to be mid-June based on photographs taken after 2009, which might be due to climate change or anthropogenic impacts.
Breeding period can directly influence population divergence via temporal reproductive isolation (reviewed in Ketterson et al. 2015) and candidate genes for phenology, such as clock genes, are known to vary with latitude (e.g., O’Malley et al. 2010). Empirical evidence on the latitudinal cline of breeding phenology, however, is limited (but see Berthold 1996; O’Malley et al. 2010). We successfully estimated the latitudinal pattern of breeding timing in Japanese dace over 1500 km, with only several days of web searching. Differences of 30–40 days in average breeding timing might contribute to intra-specific divergence. Future research can utilize this geographic variation to elucidate the temporal and spatial components of genetic variations within this species.
We also found large variation in nuptial coloration across Japan. The variation had some common trends, such as correlated changes within similar color patches/bands (i.e., red–yellow or black–gray) but not between the color patches/bands. Variation within populations was also large, although some geographic patterns still existed. Nuptial color variation has been occasionally mentioned in Japanese dace (Nakamura 1969; Sviridov et al. 2003), but this web-based approach is the first to quantitatively describe the variation in Japan.
Web-based data collection is useful for species that are frequently photographed by citizens and their photograph localities are easy to obtain (e.g., not conserved species). In particular, we were able to use web photographs with a relatively high proportion (i.e., 43.5%) that had available locality and date data compared with another similar study (26% in Leighton et al. 2016). This is probably because many photographs were uploaded on anglers’ blogs, which provide accurate date and approximate localities (e.g., names of rivers, cities, or villages) in many cases. Phenological traits may be able to be estimated for species that (1) change body colorations or shapes (e.g., plants, cyprinid, and salmonid fishes) or (2) migrate to spawning habitat during breeding seasons (e.g., migratory birds). In addition, variations in coloration could be revealed using citizen photographs, especially in organisms with discrete traits such as number of bands/spots, which can prevent observational errors. Although coloration and morphology seem easy to analyze using citizen photographs, various potential biases may exist because these characteristics are highly sensitive for to photography conditions (e.g., light conditions such as weather, and distance and angle to the targets) (Stevens et al. 2007; Sanz et al. 2013). Therefore, as a first step, categorical analysis of multiple spots/bands/patches, as used in this study, is preferable.
Although a web-based approach is cost-effective for describing large-scale patterns, it is important to recognize some limitations. First, inter-annual variation was not considered in this study, because we could not find enough photographs within years. To estimate the breeding period, photographs from before, during, and after the breeding period were needed: we could not estimate the breeding period in Chikuma River, mainly because of the lack of photographs in the spring. A balanced, relatively large dataset is needed to evaluate inter-annual variations (possibly 50–100 photographs per year based on the three rivers examined). If breeding period varies among years, and if the dace with nuptial coloration are more likely to be photographed than non-nuptial-colored dace, our analyses could overestimate the length of the breeding period. Second, non-random sampling is suggested, because the number of photographs was affected by human populations and relative abundance of Japanese dace. In addition, if rivers are long and the fish spawn in several places with different timing within rivers, estimation will be more difficult, because detailed localities are barely available from web images (only the names of rivers or towns, villages, and cities were available for most photographs). Another potential sampling bias is that uncommon phenotypes might be more likely to be photographed. This could be problematic if researchers are trying to evaluate the frequency of morphs or variations. However, it may be beneficial for describing overall variation if citizens tend to take photographs of rare phenotypes. Third, we subjectively categorized colorations, because qualitative evaluation of color hue from compressed photographs (e.g., jpeg and png) can have some biases (Stevens et al. 2007). Nonetheless, Leighton et al. (2016) subjectively evaluated color patterns of various animals using citizen photograph data on the web, and yielded results consistent with previous studies. We believe that evaluated patterns in this study were biologically meaningful, because the consistency with a previous study was high: over 80% of individuals were categorized with the same coloration as reported by Nakamura (1969). In addition, repeatability between persons was fairly high: consistency with another person was (1) 96% for judging the presence or absence of nuptial color, and (2) 80.5% in categorizing color patterns. Higher repeatability in judging presence or absence than scoring coloration could be due to evaluation simplicity.
Some limitations will be rapidly overcome by methodological advances. In particular, by asking photographers to include a color reference and scale bars in high-resolution photographs (e.g., Casanovas et al. 2014), citizen-based science might allow us to quantitatively analyze color hue and investigate the effect of body size on coloration. Therefore, we started gathering photographs for scientific use on the web site “Ugui nuptial coloration project” (http://ugui-guigui.wix.com/ugui-guigui), in which we ask citizens to take pictures with some forms of scale information, such as tobacco cases and postcards. Finally, the reproductive behavior of Japanese dace can also be studied based on data from the web: over 40 videos on the spawning of Japanese dace and closely related species (T. brandtii and T. sachalinensis) were uploaded to YouTube as of 31 December 2016, which will facilitate comparative studies.
Potential factors that affect nuptial coloration
Although our nuptial coloration analyses were preliminary, several important findings were yielded: (1) color co-varied within black bands and red bands/patches but not between black and red bands/patches, (2) nearly 20% of individuals lacked a few red bands/patches with some exceptional phenotypes (see Fig. 2f–h), and (3) large color variations existed within rivers. We also found that variation in color pattern was greater in allopatric than sympatric areas with a closely related species, T. brandtii, which can potentially hybridize with T. hakonensis (Fig. 2j). Potential factors that affect nuptial coloration include the conditions of individuals, such as nutritional status (Craig and Foote 2001); local environment, such as predation risk (Andersson 1994) and light conditions (Reimchen 1989), and timing (Kodric-Brown 1998).
Nuptial coloration of Japanese dace varied at a small spatial scale (i.e., within rivers). Red/orange coloration in Japanese dace is based on carotenoids (Matsuno and Katsuyama 1976), which animals can only acquire from diet (Goodwin 1984). Carotenoids are plant-synthesized pigments; thus, their abundance and coloration should co-vary with primary productivity (Leavitt 1993). Japanese dace migrate to the ocean or reside in rivers (e.g., Sakai and Imai 2005), which can shape substantial variation in body size within rivers (Nakamura 1969). Therefore, such tactics might result in a nuptial color variation via nutritional status, as is known in salmonids (Craig and Foote 2001). In addition, because variations within spawning schools were small (we only found one or two phenotypes within each photograph), such variations may also be due to temporal change. For example, Nakamura (1969) stated that Rb3 occasionally existed in the non-breeding season, which indicates that each coloration component appears/disappears at different timing.
Together with variations within rivers, geographic variations in nuptial coloration may also exist, because unique phenotypes were found in some areas (Fig. 2f–h: central Honshu, northern Kyushu, and Tama River, respectively). Hybridization with T. brandtii might also change nuptial coloration, because the two species differ in the coloration: T. brandtii has one dark grey/black band on the lateral line and one red (not orange) band below the lateral line (Nakamura 1969; Sviridov et al. 2002; Sakai and Amano 2014). We found three samples with an intermediate phenotype in nuptial coloration: two obvious orange bands above and below the lateral line, and one dark grey band on the lateral line. These samples were photographed in Tama River (Fig. 2h), where both species are abundant. The number of pre-dorsal scales (33–35) was also intermediate between the two species: it was within the range of Japanese dace (29–36: Amano and Sakai 2014), but also the range of T. brandtii (34–41: the range of the subspecies T. b. maruta, Sakai and Amano 2014). Because cranial morphology of the three individuals resembled T. brandtii (cheek wider than eye diameter), these individuals may be T. brandtii. However, the presence of an orange band above the lateral line, which is a characteristic of Japanese dace, has never been reported in T. brandtii (Nakamura 1969; Sviridov et al. 2002; Sakai and Amano 2014). These individuals are potentially hybrids, although hybridization between the daces is rare in northeastern Honshu (Tohoku district: Hanzawa et al. 1984; Sakai et al. 2007). Further study utilizing genetic markers in unexplored regions, including Tama River, is required. If coloration in daces prevents heterospecific matings, as previously suggested (Gritsenko 1974), species-specific coloration should be more conserved in sympatric than allopatric areas, because there is no risk of hybridization in allopatric regions (i.e., reinforcement: e.g., Higgie and Blows 2008).
We also suggest that the nuptial coloration of Japanese dace may differ between Japan and Primorye, Russia: in Japan, red/orange spots were absent from black/grey bands and Rb2 rarely reached the caudal fin; alternatively, in Primorye, Sviridov et al. (2003) reported that spots tended to be present (approximately 40% of individuals surveyed) and Rb2 reached the caudal fin (approximately 90% of individuals). This could be due to either the conditional or environmental factors discussed above, or genetic differences; Japanese dace in the eastern Sea of Japan, including Primorye, are genetically different from those in the Japanese archipelago (Sakai et al. 2002; Polyakova et al. 2015).
Mutual ornamentation is taxonomically widespread, but its function is still debated (Kraaijeveld et al. 2007). Promising approaches include comparative studies that analyze inter- or intra-specific variations in mutual ornamentation or degree of sexual difference, and other ecological or social factors that affect natural/sexual/social selection. Our findings regarding nuptial color variation within/between populations provides a foundation for investigating ecological/social factors that affect mutual ornamentation. In addition, some uploaders in western Japan suggested that nuptial coloration is more elaborate in males than females. Citizen-based science could facilitate studies that focus on intra-specific variation in strength of sexual difference, which is important for revealing the function of mutual ornamentation.