Hiromune Mitsuhashi

We would like to introduce our recently developed systems for taking images of herbarium specimens and for the automatic extraction of data from specimen labels at the Herbarium of the Museum of Nature and Human Activities, Hyogo, Japan (HYO). Firstly, we designed a low-cost, but high-quality specimen imaging system for non-professional photographers to obtain images rapidly (Takano et al. 2019). Our system uses a mass-produced, mirrorless single-lens reflex (SLR) camera (SONY ILCE6300) with a zoom lens (Samyang Optics SYIO35AF-E35 mm F/2.8). We made a photo stand by ourselves to reduce costs. In addition, we have adopted an LED (light-emitting diode) lighting system with high color rendering. This imaging system has been introduced, with some improvements or adjustments for available space, to various herbaria in Japan (e.g., University of Tokyo (TI), Kyoto University (KYO)), contributing to the digitization of herbarium specimens across Japan. Next, we developed a system to extract label information from specimen images. The specimen image was uploaded to Google OCR and data were extracted in the form of text. Uploading the whole specimen image decreased the reading accuracy of the software because the plant images behaved as OCR (Optical Character Reader) noise. Therefore, the label part was cut out from the whole specimen image by using D-Lib*1 and uploaded to tesseract OCR*2 for OCR extraction of the label information (Aoki 2019, Takano et al. 2020). When installing this system for HYO, we designed it as an application accessible externally via the internet, which proved very useful during the coronavirus pandemic: part-time workers checked and conducted label data input from home. Finally, we decided to develop a system that would automatically label the text data extracted by OCR and input them into the appropriate cells of the database. Even though the text data could be extracted from specimen images, it needed a human to input them into the database. Therefore, we adopted Named Entity Recognition (NER), a system that extracts named entities such as place names, identifying proper nouns from unstructured text data. It enables information recorded in herbarium specimens to be tagged as named entities. We tried text matching at first, but the result was not satisfactory, so we started to use machine learning instead. We compared three natural language libraries for Japanese: BERT (Bidirectional Encoder Representations from Transformers), Albert (A Lite version of BERT), and SpaCy. Despite BERT and SpaCy returning similarly high f-scores (indicating good performance), we decided to use SpaCy because it runs better on ordinary PCs or servers. With sufficient machine learning after the creation of a text corpus (a specialised dataset) specific to labels on herbarium specimens, we successfully developed the application. The project files are available on GitHub*3 (Takano et al. 2024). We then examined whether this system could be applied to non-plant specimen images, i.e., fishes or birds, and found that it could efficiently extract data. Therefore, we decided to publicize this system on the cloud server and share it with other natural history museums in Japan*4. Curators can obtain a unique ID and password and upload specimen images from their collection to extract label data. The digitization of natural history collections in Japan has been long behind other countries, and this system will help to accelerate it. The system mentioned above is specialized for the natural history collections of Japan, but we believe it is possible to build similar programs in other countries, and we hope our experience will contribute to the mobilization of the world’s natural history collections.

Major international herbaria, natural history museums and universities have recently begun to digitise their collections to facilitate studies and improve access to collections. In Japan, more than 10 million herbarium specimens are housed in various universities/museums; however, only 1% of these have been digitised. In this paper, we describe a new method for imaging herbarium specimens that is applicable to local/small herbaria. It is safe, fast, simple and inexpensive, but also satisfies usage guidelines for minimum image quality and can produce digital files suitable for long-term storage and future post production. During an eight-month trial at the Museum of Nature and Human Activities, Hyogo, with three part-time workers using a custom-made copy stand and a mirrorless interchangeable lens camera with a large LED light bank system, we were able to image 73,180 herbarium specimens (571 per day on average), obtaining two RAW and two JPEG files for each specimen.

In Japan, a yatsuda is a poorly drained traditional paddy field surrounded by a hilly forested landscape. Yatsudas are biodiversity hotspots, although urbanization is currently causing fragmentation and isolation of these habitats. The Japanese brown frog (Rana japonica) is a yatsuda indicator species that is on red lists in most prefectures in the Kanto region. The habitats and breeding ecology of this species have been well studied; however, little information about fine-scale (100 m to 5 km) isolation of breeding populations is available. We analysed the genetic structure and diversity of breeding-site populations in the suburban landscape of Chiba, Japan. Four major populations were genetically differentiated, to various extents, and we detected gene flow restriction between breeding sites upstream and downstream of the Sakatsuki River. Small breeding population sizes in suburban landscapes can cause unique genetic compositions, and isolation among populations might maintain such differences. Creating terrestrial vegetation corridors and small puddles beside rivers for use as stepping stones might be effective conservation measures.

兵庫県西部を流れる千種川は清流として知られる一方, しばしば大雨による河川氾濫が発生している. 2016年まで, 災害防止の目的で河川整備工事が行われたが, 工事による河川環境の変化や河川生態系への影響が懸念されている. 本研究では千種川の水循環や人為改変の環境影響評価の研究に必要な基礎データを得ることを目的として, 2015年8月9日に採取された河川水試料83点について, 水の水素・酸素同位体比を分析した．分析の結果は地理情報システム（GIS）により流域地図にマッピングした. 結果として, 千種川流域では, 水素・酸素同位体比ともに上流部で低い値を示し, 主要支流の佐用川や, 千種川下流に流入する支流では高い値を示した. また, 上流～中流では高度効果の影響を受けた降水の寄与がみられており, 千種川流域全体で蒸発の影響が大きいことが推測される.

Stream classification at catchment scale is important in river ecosystem management; in recent years, it has often been used in conservation planning. Stream classification at catchment scale is often based on biological data obtained at discrete survey locations. Obtaining continuous biological data as vegetation maps on rivers is labor intensive. In this study, we divided the river into many segments of equal length (500 m) and recorded vegetation types that appeared in each segment through fieldwork. Obtained data that recorded the presence and absence of vegetation type for each segments were ordinated into two-dimensional space using a non-metric multidimensional scaling method, and segments were classified through non-hierarchical cluster analysis using the k-means method based on the scores obtained by the non-metric multidimensional scaling. Using index indicators (IndVal), we obtained indicator vegetation types specific to each cluster. We discussed the characteristics of each cluster on the basis of the river topography and composition of vegetation types. Clusters obtained using the present method represent a good classification that reflects characteristics of river segments. The labor effective stream classification at catchment scale become possible by this method and it is a useful tool in river ecosystem management. Our approach should be applied and evaluated in other river systems with various topography, surrounding land uses, and climates regime.

Stream grazers have a major impact on food web structure and the productivity of stream ecosystems; however, studies on the longitudinal (upstream versus downstream) and temporal changes in their drift dynamics and resulting distributions remain limited. Here, we investigated the longitudinal and temporal distributions and drift propensity of a trichopteran grazer, the caddisfly, <italic>Micrasema quadriloba</italic> , during its life cycle in a Japanese stream. The distribution of larvae significantly shifted downstream during the fifth instar larval stage during late winter; with periphyton abundance (i.e. their food source) showing similar shifts downstream. Therefore, our results show that the drift dispersal the caddisfly occurs in response to decline in available food resources (i.e. food-resource scarcity) and an increase in food requirements by growing individuals. Furthermore, our results show that this observed longitudinal shift in larval distribution varies through their life cycle, because the drift dispersal of fifth instar larvae was greater than that of immature larvae. The correlation between periphyton abundance and drift propensity of fourth instar larvae was not statistically significant, whereas that of fifth instar larvae was significantly negative. In conclusion, we detected an ontogenetic shift in drift propensity, which might explain the longitudinal and temporal distributions of this species.

With rapid declines in global riverine biodiversity, conservation is needed on a nationwide scale. Since 1990, the National Census on River Environments (NCRE) has collected meaningful data on riverine organisms and environments in Japan. However, there are several issues with using the NCRE database for scientific studies. In this study, we summarised issues related to the NCRE database, and used the 3^rd NCRE dataset (2001-2005) to evaluate nationwide trends in species richness and the rarity of fishes (primarily freshwater and diadromous fishes) and taxon richness and the rarity of benthic animals (aquatic insects and shellfishes). First, we summarised data quality (e.g. monitoring sites and season) and data formatting issues (e.g. site name and species name). Second, we mitigated these potential issues as thoroughly as possible and tested the relationships of species (or taxon) richness and rarity with latitude using generalised linear models. We found that species and taxon richness showed different latitudinal distributions between primarily freshwater and diadromous fishes and between aquatic insects and shellfishes. In contrast, rarities showed spatial congruence with species diversity, suggesting that more species and rare species could be conserved in the same regions.

To manage the expansion of bamboo forests efficiently, it is necessary to predict where expansion is likely. However, predicting bamboo distribution using statistical models is challenging due to biased distributions, heterogeneous invasion histories, and strong neighbor effects. Furthermore, clonal and invasive species such as bamboo show a non-linear response to environmental factors. Here, we construct a spatial model to predict locations of high penetration probability after controlling for these problems, using data for bamboo distribution in Toyooka, Hyogo Prefecture, Japan. First, we used MaxEnt to reveal the full range of possible bamboo survival and to control for the effects of clonal growth. Second, we constructed a predictive model of invasion using environmental factors. The first model showed that distance from the source patch strongly affected expansion and that constraints from environmental factors were not significant. The final model explained about 70% of actual expansion. Scenario analyses showed that bamboo expansion is best controlled by removing smaller patches. Based on our results, we discuss issues in bamboo forest management and planning.

Stream classifications and indicators that are capable of summarizing the complex information of ecological communities provide useful tools in river ecosystem management. However, these tools have not yet achieved practical application here in Japan due to the fact that in many studies, the criteria for selecting stream classifications and indicators are neither objective nor quantitative. Combining model-based clustering and indicator values (IndVal) makes it possible to assess the appropriateness of stream classifications based on the indexability from multiple classifications. Additionally, by establishing threshold values, indicators can be objectively selected. Using these techniques, we examined appropriate data acquisition methods for determining a stream classification system for vegetation at the catchment scale. Additionally, we studied the suitability of methods that were derived solely from either biological or environmental data as well as those that combined the two. We prepared three sets of data that utilized different research zone and scope-setting methods and then created three types of stream classifications from each, ultimately evaluating the appropriateness by the quality of selected indicators and the level of indexability. The most appropriate methods involved those derived from biological and environmental data exhibiting consistently high indexability as well as those that created vegetation maps in certain areas of the distribution over the entire catchment area. Furthermore, the quality of the selected indicators was consistent with the indexability results.