In this study, we developed a vegetation discrimination method based on satellite remote sensing, topographical information, and machine learning for creating vegetation maps along the Arakawa River. First, satellite images were processed by object-based classification. Then spectral information from the satellite imagery, the vegetation index, and topographical information were added to each object. Vegetation discrimination models were created by machine learning algorithm (Random Forests, Support Vector Machine), and individual objects were adapted to the vegetation types. Comparison of the results with existing vegetation maps confirmed that the use of topographical information improved the accuracy of discrimination. When the amount of training data for machine learning was reduced to 10%, the classification accuracy of the Support Vector Machine algorithm was higher than that of the Random Forests algorithm. In addition, the accuracy of vegetation discrimination was reduced when vegetation types with a small number of data were generated in the training data. When using field survey data to classify vegetation with a small sampling rate, it was shown that the number of survey sites per vegetation type needs to be controlled so that the number of training data does not become too small.

The Genji firefly (Luciola cruciata) is an important tourism resource, especially in Japan. Nevertheless, its habitat has decreased, mainly due to human activity. Until now, its population size has been studied primarily by visual observation of adults, which fly and flash during the night. Surveys of larvae have been avoided not to disturb its populations and habitats. In this study, we performed environmental DNA (eDNA) analysis to quantify firefly larvae by showing a correlation between larval numbers and eDNA concentration in the field. We then compared the eDNA concentrations and larval numbers to the adult numbers of two generations to evaluate the suitability of this analysis for exploring bottleneck events in firefly populations. We designed "primer set I" and "primer-probe set II" to specifically detect and quantify Genji firefly DNA fragments, respectively. The species specificities of the primer sets were validated by PCR using tissue-derived DNA samples from the target species and its most closely related species (Heike firefly, Luciola lateralis), which were captured in the field. The quantifiability of primer-probe set II was confirmed by quantitative PCR using artificially synthesized DNA of the two species. Quantitative PCR using water samples collected from areas where firefly larvae were captured (15 locations) showed a positive correlation with the number of firefly larvae. The relationship between the number of adults in the previous and focal generations and the number of captured larvae and eDNA concentration suggested that repeated measurements of eDNA concentration during the larval stage are necessary. In conclusion, we showed a positive relationshipbetween eDNA concentration and firefly larvae number in the field for the first time. It may be possible to identify causes or seasons of Genji firefly reduction by repeated measurements of eDNA concentration during the larval stage.

Natural spawning of Chum salmon has recovered in Toyohira River, an urban river in Sapporo, Japan, from the 1980s, a time when this species had almost disappeared from this river because of water pollution. Channelization has altered river geomorphology, which may have influenced spawning habitat. We divide salmon spawning habitat in Toyohira River into three sections and measure changes in riverbed morphology using transverse cross-sectional profiles from 1988-2015. The relationships between 8 topographic parameters, number of spawning redds and their decadal changes are examined. Principal component analysis reveals the strengths of relationships between these topographic parameters and numbers of spawning redds per kilometer post. Upper-and middle riverbed sections were significantly deeper, but spawning redds occurred in localized groundwater springs where water levels were higher than the riverbed. Significant lower river section changes included increased elevation gaps between lowest and highest riverbed of the low-water channel, narrowing of the water surface during spawning season, and decreased thalweg migration distance. Of two principal components, PC1 correlated with changes in riverbed elevation, elevation gaps between the lowest and highest riverbed in the low-water channel, and water surface width during spawning season, which represent "susceptibility to flood disturbance." PC2 correlated with groundwater level and streambed particle size, possibly indicative of upwelling water volume. The number of chum salmon spawning redds was significantly, positively correlated with PC1 and PC2. Because the coefficient of PC1 is decreasing with time, we surmise that appropriate habitat for spawning condition in Toyohira River has deteriorated over 27 years. As Chum salmon is a symbolic species in Toyohira River, it is important to keep dynamic feature of thalweg and gravel bed for natural spawning.

In recent years, invasive alien plant species, Amorpha fruticosa has been spreading across the shoreline area especially in the drawdown zone, where the water level fluctuates (8 m) between the flood and non-flood seasons, in the Miharu dam reservoir. In this study, the target was to establish native plant species, Salix subfragilis and Salix jessoensis growing naturally on the shoreline of the Miharu dam reservoir, after logging A. fruticosa. Frequency of logging A. fruticosa and timing of willow planting were changed to seek optimal conditions which the two willow species became dominant over A. fruticosa. Two willow species were planted with cuttings in June and December. The survival rate in June was higher than one in December. Due to the periodical inundation, optimal time to transplant willows in the drawdown area was limited to June in the Miharu dam reservoir. A. fruticosa was cut multiple times but logging three times in 13 months was less effective on establishing the willow species. By cutting A. fruticosa five times in two years (twice a year or more during growing period), the two willow species became higher and dominant over A. fruticosa until the third year after logging. The slope of the dam reservoir in this study was submerged for more than 100 days per year, which was different from the case in the previous study. However, the number of logging required to suppress, A. fruticosa in the drawdown area was the same as one required for slopes without inundation. This indicated that inundation did not suppress growth of A. fruticosa. Logging effects was disappeared in 4^th year (no significant difference with control area) and it is desirable to cut down A. fruticosa at least before reaching 4^th year.

The Pfankuch stability index (PI) is known as a simple and suitable measure to assess streambed disturbance for benthic organisms. Although this method has been widely applied to unimpacted mountain streams, the applicability for human-altered lowland streams has not been clarified yet. In this study, we conducted surveys on stream invertebrates and bed-stability using the PI in ten streams flowing the Dogo Plain in Ehime Prefecture, southwestern Japan. We aimed to evaluate the applicability of the PI to lowland streams and to develop a better index (modified PI) that is specialized to assess streambed disturbance in lowland streams by considering the effect of water quality degradation. As a result, the PI varied markedly among the sites. The modified PI also showed high variation but the rank of the score did not change for each site when compared with the original PI. Our analyses demonstrated that the invertebrate abundance decreased with increasing PI. A negative relationship with abundance was also found for the modified PI but there was little improvement in the relationship compared with the original PI. Overall, this study suggests that the Pfankuch method can be applied to assess the streambed disturbance for benthic macroinvertebrates in human-altered lowland streams. Our attempt to modify the pplicability of score was not successful, implying that further trials are recommended to increase the applicability of the PI to lowland streams. We suggest that developing methodlogy for measuring flood disturbance, as in this study, is key to boosting disturbance studies in streams and managing stream ecosystems from the viewpoint of disturbance ecology.

We studied the distribution and physical environmental characteristics of chum (Oncorhynchus keta) and masu salmon （O. masoui） redds within the alluvial fan of the Koonihassam River. After the spawning period, we evaluated the differences between surface and subsurface water temperatures during the winter to determine the presence or abence of underground water discharge. Our results indicated that the redds of masu salmon were mainly distributed in the upstream section and those of chum salmon were only distributed in the downstream section of the river. Chum salmon redds were characterized by significantly larger grain sizes than those of masu salmon. There were no significant differences between surface and subsurface water temperatures of masu salmon redds in the up-stream section. However, the subsurface water temperature of chum salmon in the down-stream section was significantly higher than the surface water temperature, suggesting the presence of underground water discharge. Chum salmon redds were largely distributed at the lower edge of the alluvial fan, where underground water tends to discharge from the river bed. In contrast, masu salmon redds were mainly found in the upper area of the alluvial fan, and underground water discharge from the river bed was not confirmed. Our results suggested that the spawning site selection of masu salmon is independent of the presence or absence of underground water discharge. These results suggest that suitable spawning habitats for these species are likely affected by multiple environmental conditions, including geomorphological features, underground water discharge, and riverbed composition.

The degradation of wetlands caused by the overgrowth of aquatic plants is a problem in many areas; therefore, low-cost, labor-saving vegetation management using robot boats is under development. Herein, we propose a method to classify aquatic plants and obstacles by real-time image processing to enable the full-autonomous operation of a robot boat. We adopted a semantic segmentation method using deep learning for the image processing and conducted teaching and testing on our own dataset. Regarding classification by segmentation, it was possible to classify the lotus (Nelumbo nucifera), which is the target of harvesting, and a rare species of fringed water lily (Nymphoides peltata), with an accuracy rate of about 90%. The deep neural network could learn the color, shape, and position information in the image. This reduced the influence of external environmental disturbances such as changes in illuminance and harsh sunlight reflections on the water surface, and the inference was robust. Regarding the obstacles, it was possible to classify with an accuracy rate of 80% or more within the depth-obtainable range of an RGB-D camera. It was also confirmed that classification was possible even when the obstacle was located 4.0 m or more away.

In Lake Izunuma, Miyagi Prefecture, there is concern that the feeding habitats in the lakes of egrets is affected by the decreaseof open water and the change of fish composition. In this study, we examined the availability of UAV-based monitoring for egrets. The survey was conducted in Lake Izunuma (369 ha), and a total of 7,852 images were obtained using UAV in September 2018. Using the UAV imagery, the ability to discriminate egrets was examined. As a result, egrets were confirmed in 855 images, and UAV was considered available for monitoring egrets in lakes. To improve the efficiency of UAV monitoring, the availability of image analysis methods using binarization processing was evaluated. As a result, it was found that the binarization process is useful for screening UAV images. However it is difficult to extract only egrets, and further consideration of the method is necessary. UAV and image processing technologies were able to gather new insights of egrets. It is recommended that consider better protocols for UAV ecosystem monitoring.

Mires are regarded as one of the most fragile ecosystems, and active vegetation monitoring is required for their conservation. By using unmanned aerial vehicles (UAVs) equipped with omnidirectional video cameras, it may be possible to obtain information inside vegetation easily through close-up super-wide-angle aerial views of vegetation communities. Although this approach may allow for non-invasive vegetation monitoring in mires, there is little knowledge about the advantages and disadvantages of vegetation surveying using such UAVs. In this study, we set 27 quadrats in the Bekanbeushi mire and conducted surveys using UAV aerial views from an omnidirectional video camera on the community surface, still cameras 2 and 5 m above the community, as well as using conventional field vegetation survey methods. Using these views, we examined species appearances and coverages on a PC monitor and evaluated species detection rates, coverage errors, and statistical vegetation classification via two-way indicator species analysis (TWINSPAN). From this data, we discussed the prospect of vegetation surveying by UAV aerial views. The mean values in species detection rate and coverage error were not clearly different between the omnidirectional movie and the still images. In the omnidirectional movie, the detection rates of small to medium species growing inside the community were higher than those in the still images, which contributed to more accurate vegetation classification in TWINSPAN. In contrast, in the omnidirectional movie, the detection rates of narrow-leaf and low-coverage species decreased owing to the wind generated by the UAV and the narrowing of the cameraʼs field of view upon approaching the community surface. Furthermore, species with plant heights of less than 10 cm were hardly detected, even in the omnidirectional movie. Therefore, to perform vegetation surveys using UAV aerial views, it will be necessary to first further improve the detection rate of small species.

We inspected rivers where mangrove forests were distributed on the main island of Okinawa, and surveyed four rivers (Okukubi River, Ooura River, Gesashi River, and Waku River) with existing relatively large mangrove forest areas. Based on the information that can be obtained with UAV, we examined a simple and versatile evaluation method that enables quantitative evaluation. The method was a combination of vegetation classificaion using ortho-mosaic image and mangrove degradation evaluation using vegetation index. We could classify vegetation by supervised object-based image classification. As a result of totaling the area for each classification class in 4 rivers, the area ratio for each class was different among rivers. The histograms of the polygon areas classified as gaps were compared in four rivers. The average value of NDVI in the polygons classified as Bruguiera gymnorhiza tended to be less than 0.5 polygons in the Okukubi River. In the Okukubi River, polygons with low NDVI tended to increase as the distance from the forest edge decreased, and the maximum NDVI tended to decrease as the distance from the forest edge increased. It was suggested that the increase in NDVI to be close to the forest edge and the decrease in NDVI inside the patch could be used as an indicator of the degradation of the mangrove forest. From the comparison of the four rivers, the characteristics of the mangrove forests of each river can be grasped, and in particular, the mangrove forests of the Okukubi River are declining.