The Umeda River is a small to medium-sized river located in Toyohashi City, Aichi Prefecture, which is prone to rising water levels during floods due to vegetation growth. This study utilized a combination of field survey and numerical modeling to assess the impact of in-channel vegetation on flood levels. Sensitivity experiments were conducted on vegetation height, considering vegetation conditions, and vegetation management plans were proposed. The sensitivity experiments revealed that it is crucial to maintain vegetation height below 2 meters to mitigate flood risks effectively. Furthermore, the study identified that vegetation on the right bank significantly contributes to the rise in water levels during flood events. Notably, the Umeda River experiences the highest frequency of floods in September. Consequently, clearing vegetation on both sides of the river by the end of August is considered to reduce the risk of flooding.

 In recent years, heavy rain disasters have become more frequent, and extensive damage to road bridges has been reported. A new concept of road risk assessment in Japan is presented for the purpose of constructing a highly reliable road network. Going forward, it is of utmost importance to clarify the damage risk presented by driftwood blockages of road bridges adjacent to rivers in the event of excessive flooding. In this study, we conducted basic experiments on bridges considered to be at high risk, as indicated in previous studies, and investigated how to prevent bridge outflow and flooding to the surrounding area due to driftwood blockage. In addition, the importance of the relationship between bridge girder position and water level, and the blocking mechanism of driftwood for bridges (girders and balustrade columns), are discussed. Finally, girder protection is proposed as a measure to reduce clogging around the girder under the floor slab, and its effectiveness was demonstrated.

 The aim of this study was to develop spatial imaging technology to assist in the development and management of small and medium sized watercourse infrastructure. The watercourse infrastructure should facilitate daily water usage, protect or restore riverine ecosystems, ensure safety in times of flooding, and incorporate community preferences. The study focused on the Chiribetsu River, an under-repair watercourse in Muroran, Hokkaido. A focus group was formed to ascertain the residents’ attitudes toward the river. Based on the findings, a virtual environment of the Chiribetsu River was created. The purpose was to use this virtual content as a means to remind residents about flood threats. The content incorporated information on river swelling and overflow scenarios, estimated using iRICNays2DFlood. In light of the results, the following recommendations are made: 1) Resident opinions should be considered when developing virtual watercourse environments; 2) The virtual environment should feature a seamless visual transition from normal conditions to an emergency scenario; and 3) The virtual-reality technology should be compatible with small and medium watercourses, for which information may be limited.

 Flooding experiments were carried out on a building arrangement that simulates an actual urban area and an urban area model that can consider inundation into buildings, and the overall behavior of the flood flow and the water depth at several points were measured. Based on this data, we verified three models that differ in how to handle inundation to buildings in the flood inundation analysis model for urban areas. Building Gap (BG) model in which the building boundary and the mesh boundary are matched and flooded through the slight gap in the building, Building Porosity (BP) model defined by the Ministry of Land, Infrastructure, Transport and Tourism, which evaluates by giving the porosity and drag coefficient in a uniform grid, Building in Mesh (BM) model that assumes a virtual building in a mesh. Of these, the BM model was able to express the inundation situation of flood water without requiring a large computational load. In addition, it was also shown that the application of an unstructured grid that differentiates between streets and city blocks greatly improves the calculation results.

 In recent years, the effects of global climate change have led to an increasing possibility of unprecedented large-scale torrential rainfall events. In Kanazawa, Yokohama, the target area of this study, both coastal and inland areas have been inundated by Typhoon No. 15 in 2019 and other occasions. In this study, S-uiPS, a sophisticated urban inundation forecasting method, is used to evaluate the inundation risk by uniformly applying the maximum expected rainfall to the target area. Because the target area includes a residential area developed on a steep slope, the S-uiPS was modified to newly account for water movement through steep slopes such as cliffs. Using this enhanced S-uiPS, flooding prediction results were obtained for areas including residential areas that could not be calculated before.

 In response to flood disasters, which are expected to become more severe due to climate change, it is becoming increasingly important to develop flood-resilient communities using urban planning methods. In this study, we estimated future changes of urban structure in the area around the Kakehashi River basin in Komatsu City, Ishikawa Prefecture, Japan by using a computable urban economic (CUE) model. The scenario assumed is the relocation and consolidation of elementary schools and the adjustment of the urbanization promotion area under the population decrease in 2045. Based on the results of inundation analysis using future rainfall based on a pseudo-global warming simulation as input, the flood economic losses before and after the urban structural change was calculated, and the flood risk mitigation effect of the assumed scenarios was evaluated. Relocation and consolidation of elementary schools had the largest decrease of 15.1% in the amount of flood economic loss. The combination of the relocation and consolidation of elementary schools and the change in the urbanization promotion area resulted in a 19.9% reduction in the flood economic loss, indicating a synergistic effect of the combination of measures.

 In the case of instituting river planning, it is important to determine the exceedance probabilities based on long-term observational data. However, there are many data-scarce basins in the world, such as in developing countries. Although reanalysis data has been used instead of observation data in these areas, it has biases between the observational data since the spatial resolution of reanalysis data is low. In this study, we developed a statistical correction method for reanalysis data (ERA5) with four different calibration periods (rewind to 5, 10, 15 and 20 years from the latest year of observation data) and created long-term pseudo hydro-meteorological data such as precipitation and temperature. In addition, we examined the type of probability distribution and the number of years for the calibration period that should be applied to create highly accurate corrected data. As a result, it is found that river flow indices based on ERA5 which was applied to the statistical correction method are the almost same relative to the observational data. Furthermore, the corrected data which applied the statistical correction method with the gamma distribution developed based on over 15 years of observational data in the calibration period revealed the high accuracy river flow indices and annual exceedance probability flow based on observational data.

 Agriculture is one of the industries most vulnerable to climate change. Moreover, the yields and appearance quality of rice are predicted to deteriorate under climate change as a result of higher summer temperatures. To reduce the negative impacts on rice cultivation, various adaptation measures, ranging from incremental to transformative, have been proposed. Among these adaptation measures, shifting the transplantation date is relatively inexpensive and easy to implement. Thus, it has been implemented widely by farmers. However, farmers’ requests to shift the transplantation date may be constrained by river administrators because river water in Japan is currently being used to nearly its maximum extent. In this study, we developed a framework to evaluate the effects of shifting the transplantation date of rice under climate change on the rice yield and water supply–demand balance of agricultural water. The framework consisted of two process-based models of crop science and hydrology. The rice yield and water balance were calculated by shifting the current transplantation date at increments of one week to five weeks before and after the current date. To evaluate the effects in different regions of Japan, we selected 77 areas based on reference gauging stations for river water uses and associated rice cultivation areas. The results showed that the transplantation date with higher yield improved the water balance (harmonious relationship) at 38 areas, but it worsened the water balance (tradeoff relationship) at 37 areas. The results imply that transplantation dates that increase the yield may be easily selected from the perspective of river administrators in the areas with harmonious relationships but may be hampered by deterioration in the water supply–demand balance in areas with tradeoff relationships.

 According to the paddy cropping in Japan, no previous studies assess the climate change impacts including the relationship between rice crop growth and water resource availability, nor analyze the change of trends. In this study, by using a combined model of a paddy rice growth model and an agricultural dam model, and climatic data calculated by a 150-year continuous experiment based on the RCP8.5 scenario, the long-term trend analysis in the number of crop growth days, crop yield, dam storage rate, and water stress was conducted in a district in Mie Prefecture. As a result, the increasing trend in crop yield stops around the 2070s, the period with on average small dam water storage becomes earlier as time goes on, and water stress becomes severer. In addition, when the transplanting date is changed as an adaptation measure, although an earlier transplanting date than the current is desirable from the aspect of average yield maximization, a later transplanting date is desirable from the aspect of stable irrigation supply.

 This study estimated the flood damage reduction effects by the order of riparian vegetation removal on class A river systems in Japan. River systems with a high potential for both adaptation and mitigation measures of riparian vegetation removal were evaluated. Flood inundation analyses which reflected the influence of riparian vegetation through roughness coefficients were conducted, and the damage costs were calculated based on inundation depths. The biomass potential was calculated as the normalized vegetation index multiplied by the vegetated area. The results showed that the flood damage reduction effect varied depending on the order of riparian vegetation removal. A relationship was also observed between the order of riparian vegetation removal and the population distribution. The river systems where both adaptation and mitigation measures by riparian vegetation removal were highly effective were estimated to be the Tokachi, Yubetsu and Iwaki River systems.

 In the future, it is said that the risk of sediment-related disasters will increase due to increased precipitation accompanying climate change due to global warming, and railways will also be affected. Therefore, in this study, we target three major railway lines with a high frequency of collapse in the Tohoku and Kanto regions, and use the output data of numerical experiments based on the RCP8.5 scenario by the climate model (NHRCM02) to evaluate the disaster. Regarding the future change in the number of occurrences, we evaluated the difference by line section. As a result, it was found that the rate of increase in the number of collapse occurrences at the end of the 21st century will be 3.4 times higher than the present on railway lines in the Kanto region, and 2.5 times on average on railway lines in the Tohoku region.

 We conducted a field survey of ice jams that occurred in Hokkaido in March 2023, and attempted to obtain quantitative knowledge of the ice jam phenomenon. The scale of ice jam was evaluated based on the amount of water level rise (∆𝐻) when ice jam occurred. By comparing the past occurrences with those in March 2023 using ∆𝐻, it was found that the water level change ranged from 0.57 m to 4.22 m. There is a positive correlation (r=0.70) between cumulative rainfall before ice jams and ∆𝐻. We tried to apply the method of estimating the time of ice break-up. It was confirmed that the ice thickness variation acceleration 𝑃𝐵 appears before the ice jam occurrence time, and that the time of ice break-up can be estimated.

 We developed a 5 node human heat balance model, which consists of core, muscle, fat, skin, and central blood pool node. The data from the outdoor thermal physiological experiments with 36 subjects were used to parameterize the human thermoregulation process and tune the model parameters. Our model was compared with one of the multisegment models, JOS-3 as well as a standard 2 node model and outperformed them in the prediction of core body temperatures and average skin surface temperatures.

 Large-scale flood analysis using ensemble climate change data requires a hydrological model that can simulate runoff and inundation with high speed computation. The objective of this study is to develop a 1K-RRI model that can solve runoff and inundation at the macro scale at 1 km resolution. The model was constructed by upscaling a nation-wide RRI model at 150 m resolution. Applying the RRI model to the macro scale without particular care will cause problems such as the inability to distinguish between different runoff processes of mountainous and plain land uses within the same grid cell. Therefore, we improved the RRI model by upscaling the average gradient and area of the mountainous and plain land uses and calculating each runoff component. The 1K-RRI model was compared with the original 150 m resolution model and a model upscaled directly to the 1 km resolution. We found that the 1K-RRI model can reproduce the runoff responses better compared to the one just upscaled without any particular attention.

 Since the required topographic resolution for runoff analysis is unknown, we used a runoff analysis method using a quadtree grid to compare runoff characteristics at different topographic resolutions and analyze their effects. Comparisons were also made with the RRI model and with methods that treated the river channel in one dimension. It was shown that improving the topographic resolution of the river channel by means of a quadtree grid improves the reproducibility of the water level hydro, and that the topographic resolution of the river channel contributes significantly to the runoff volume estimation. It was suggested that differences in topographic resolution in the mountains as well as in the river channel affect the amount of runoff.

 Due to global climate change, there are concerns about an increase in extreme torrential rains, and some places are suffering from disasters due to short-term extreme rains. Soil has its own infiltration capacity, and in the range where rainfall intensity is sufficiently high, the infiltration rate is almost equal to the infiltration capacity of the soil and the excess part becomes runoff. In this study, field sprinkling tests were conducted at multiple locations in Japan to investigate the relationship between steadystate infiltration rate and rainfall intensity, as well as the accuracy of estimating infiltration capacity. It shows that the steadystate infiltration rate equation using hyperbolic tangent function is rather promising in estimating infiltration capacity. It also introduces the comparison with the curve number method used in the United States.

 In this study, we developed a method to analyze a high-resolution geospatial dataset to easily and automatically detect river levees in a wide area. Specifically, we analyzed the high-resolution DEM in Google Earth Engine in conjunction with a Python program via API, calculated four topographic parameters (relative elevation, slope, aspect difference and curvature), constructed a system to combine these parameters, and developed a method to extract river levees. In the Kinu River and Mississippi River basins, the levee extension detected by the developed method was more than 75% of the validation data, and the levee height was within 1 m of the average error. Characteristic levees such as continuous levees in urban areas and “Kasumitei" were also sufficiently detected, confirming that the 10m resolution DEM used in this method is able to detect levees for medium or larger sized river.

 Accurate information pertaining to river levees holds significant importance in fulfilling the demand for comprehensive and precise inundation data. However, particularly concerning small and medium-sized rivers, the insufficiency of data regarding the height and location of levees can be attributed to the inadequate allocation of river maintenance funds by local governing bodies. Additionally, there exists the possibility of the presence of unrecognized old levees or those in tributaries, which have not been acknowledged by river administrators. This study endeavored to automatically detect levees in Shizuoka City, Shizuoka Prefecture, Japan. The algorithm successfully identified not only the primary levees of first-class rivers but also those of small and medium-sized rivers, ambiguous levees, and dual-line levees, in cases where the levees possessed trapezoidal cross-sections. Conversely, the algorithm encountered difficulties in detecting levees exhibiting shapes that deviated from those defined within the algorithm, while simultaneously incorrectly identifying structures resembling the shapes stipulated by the algorithm.

 Elevation and surface flow direction data are useful for the analysis of hydrologic processes in a catchment. Surface flow direction is generally determined by steepest downward slope based on depression-filled elevation data. However, since the steepest downward slope is affected by the elevation correction and errors in the input elevation data, resulted surface flow direction does not reflect the actual river geometry. Here, a method to form catchment dataset was developed based on actual river information to minimize elevation correction and improve reproducibility of river geometry. As a result, the reproducibility of river geometry was significantly improved and the elevation correction was reduced at multiple spatial resolutions from 30 m to 1000 m. In addition, the complex river geometry such as meandering and tributary confluence was reproduced while it was difficult to do in conventional approaches. This method contributes to catchment-scale flood control, wide-area river environment management, and disaster prevention measures using the digital twin.

 Lake Tataranuma in Gunma has been suffering from a deteriorating environment for aquatic plants and animals due to the development of the area and the increase in polluted wastewater since around the period of rapid economic growth. In many cases, water pollution in lakes is caused by a variety of factors, and it is necessary to make sure of the relationship between the phenomena and the factors. In this study, we investigated the spatiotemporal characteristics of water quality in Lake Tataranuma based on field surveys conducted four times a year. The results showed that thermal stratification in the central part of Lake Tataranuma, which has a large water depth, during the summer season. The vertical profiles of DO, pH and chlorophyll fluorescence in summer were related to thermal stratification.

 We investigated the differences in the effects of the previous double plume model and the revised version on reservoir hydraulics and prevention of harmful cyanobacteria. Using the two-dimensional reservoir simulation model developed by JWA as a platform, the differences between the two double-plume models were evaluated in terms of their effectiveness in combating bubble circulation. As reported recently, the previous double plume model was developed focusing only on strong effects of density stratification compared to diffused bubble buoyancy. This paper presents the quantitative characteristics of overestimation of the far field surface horizontal flow quantity estimated by the previous model. Furthermore, due to these characteristics, numerical simulations under actual reservoir conditions revealed that the previous double plume model underestimated the surface layer retention time, an important habitat condition of harmful cyanobacteria and overestimated the preventive effects of the bubble circulation countermeasure. Finally, the lessons from the numerical simulation for the design and application of a countermeasure and future research topics for the bubble plume are discussed.

 Understanding the spatio-temporal variation of water temperature is important for water quality management in headwater reservoirs. Although field observations provide relatively reliable data, a numerical fluid dynamics model is indispensable to predict spatio-temporal changes. In this study, we tried to improve the prediction accuracy by implementing ensemble Kalman filter in a three-dimensional environmental fluid model, Fantom-Refined. As a result, sequential assimilation continued to reduce the model error, and we can confirm the robustness of the data assimilation system. 3-week average MAE and RMSE were reduced to 48% and 74%, respectively. Also, in a stratified reservoir during the summer season, the accuracy of spatio-temporal water temperature prediction was improved even with limited observation data.

 Recently, the impact of climate change has a serious effect on water environments. On dam reservoirs, there are concerns about increasing water temperature and decreasing dissolved oxygen in the bottom layer. Therefore, this study evaluated the impacts of climate change on the thermal structure and dissolved oxygen in the bottom layer of the dam reservoirs.

 As a result, the difference in the temperature between the surface layer and the bottom layer tended to increase in the cases of RCP_4.5 and RCP_8.5 but the difference in the temperature between August and September tended to decrease in the case of RCP_6.0. Moreover, water temperature gradients tended to increase in the cases of future climate scenarios. The results suggested that the thermocline will become stronger in the future climate scenarios and the change in thermal structure will cause to decrease in the dissolved oxygen in the bottom layer.

 To clarify climate change impacts on hypoxia in Tokyo Bay and Ise–Mikawa Bay, we performed numerical predictions under the present (the end of the 20th century) and the RCP8.5 future (the end of the 21st century) climate conditions using an integrated river basin – coastal ocean hydro-environment assessment model. These results indicated that the sea surface temperature in RCP8.5 increases by 3–4 degrees C from that in the present, and that the decrease in primary production due to high temperature from summer to autumn was larger in Tokyo Bay, Ise Bay, and Mikawa Bay, in that order, depending on the water temperature in the sea area. In Mikawa Bay where the primary production decreased significantly, the hypoxic volume decreased in the future. Furthermore, the effects of nutrient management were evaluated. It was suggested that there were some areas where it was effective to control hypoxia with DO < 3 mg/L, but difficult to control hypoxia with DO < 4 mg/L only by loading reduction.

 The long-term characteristics of the tides in the Ariake Sea and the data of nine years of continuous multi-point observations of current velocities in Isahaya Bay, which have been carried out by this research group, were reviewed and analysed with a focus on the changes. The results showed that the M₂ tidal amplitude has continued to decrease since the 2000s, and that the current tidal current velocity caused by the incoming and outgoing tides is at its lowest in the present Ariake Sea, when considered together with the decrease in the bay area due to past reclamation, the 18.6-year cycle variation of tides and the rise in mean sea level. Analysis of long-term continuous tidal current observation data for nine years, about half of the 18.6-year cycle in Isahaya Bay, showed that there was no perfect correspondence with tidal changes, suggesting the influence of density stratification and other factors.

 We calculated the threat score, false alarms rate, and misses rate for the upstream basins of dams in Japan, and comparisons between the guidance predicted rainfall which used in the preliminary release operation guidelines and observed rainfall. The ensemble predicted rainfall was also compared with observed rainfall in order to examine the possibility of utilizing the preliminary release operation.

 As a result, we confirmed that the guidance predicted rainfall tended to be underpredicted for dams with short rainfall duration and that the ensemble predicted rainfall by GEPS (Global Ensemble Prediction Systems) tended to be underpredicted and needed to be corrected.

 This work investigated effects of numerical weather model upgrading on prior release of a multi-purpose reservoir considering long-range ensemble forecast of precipitation. Change in error characteristics of two kinds of operational weekly ensemble forecast before and after upgrading was analyzed from the viewpoint of application to decision making for prior release. Through the simulations on long-term reservoir operation considering each forecast, it was found that forecast after upgrading allowed to estimate greater water amount that can be released in prior release, which contributed to more chance to secure greater flood control capacity in large-scale floods. On the other hand, miss-false alarm relationship was not improved by model upgrading, which resulted in greater risk of water supply or power generation.

 In this study, we proposed a reservoir operation method for two major reservoirs in the Chi River Basin in northeastern Thailand, where specific reservoir operation rules for flood control purposes are not stipulate and agriculture is a major industry, to contribute to both flood control and water utilization in the basin.First, based on the water levels correlation between upstream and downstream, the flood trvel time from the dam reservoir to the downstream area, which is the flood-prone zone, was calculated, and the lead time for reservoir operation was calculated. Next, we attempted to reduce the water level in the downstream area of the reservoir by giving the increase or decrease of the water level in the downstream area, which is a point of concern, as an explanatory variable for the reservoir operation. The proposed reservoir operation method was applied to the 2017 flood, when dam inflows and downstream flood inundation were very large, and resulted in a 0.4 m reduction in the maximum inundation depth downstream while ensuring sufficient water storage for water use by the beginning of the dry season at both of the dams.

 We studied the effectiveness of a water level prediction methods to support the operation of check gates in low-lying small stream, where both flood control and water utilization are required. From interview surveys with facility managers and on-site water level observations, it was found that the current decision to remove the gate weirs depends on managers' forecasts and visual observations of water level, and that water level predictions based on remote monitoring data are useful for facility management. In one rainfall event, it was estimated that the peak water level was kept from rising approximately 1 m by removing the weirs of the check gate. This indicates the importance of gate operation by the facility managers. The Nearest-Neighbor (NN) method, one of the pattern recognition methods that are easy to construct models, was applied to the study area as a water level prediction method. It was found that water levels could be predicted with an RMSE of 0.056 m to 0.13 m when the lead time was 4 h.

 We proposed a novel evaluation index on the rainwater storage capacity of irrigation ponds based on the total increase in stored water volume accompanying rainfall during peak inflow. On irrigation ponds in Miyagi Prefecture, we calculated the flood inflow and the quotient (drainage-area ratio) obtained by dividing the drainage area by the reservoir area, and investigated rainwater storage capacity using 14 model irrigation ponds prepared based on the drainage-area ratio distribution. As a result, it was revealed that: 1) rainwater storage capacity was recognized even in irrigation ponds of large drainage-area ratio and minor flood peak-cutting effect, which has conventionally been used to evaluate the flood control performance of irrigation ponds; 2) conversely, rainwater storage capacity was relatively small in ponds having high peakcutting effect and small drainage-area ratio; and 3) both kinds of pond can be enhanced in terms of rainwater storage capacity by lowering the water level before rainfall. In using irrigation ponds for flood control of rivers, we observed that release of water below the maximum storage level before rainfall is effective for enhancing the functions and effects of both kinds of pond.

 Because of the recent increase in heavy rainfall, it is expected for the hydropower dams to not only increase a power generation amount but also to expand the water storage capacity for flood management by discharging the stored water earlier than the flooding using the information of long-term ensemble rainfall predictions. However, there are not many studies discussing the practical use of ensemble rainfall forecast considering the operational situations of the dam and power generation, nor studies evaluating the effect on the whole water basin including the downstream. In this study, we discussed the practical method of using the ensemble rainfall forecast for the effective dam operation, and aimed to suggest the method to control the stored water for power generation in advance to the flooding and to expand the water storage capacity. We tried to predict the flooding 7–13 days before by using the ensemble forecast for June and July, and evaluated the effect on the dam operation of Kurobe and the downstream two dams when using the current prediction accuracy. As a result, in 2020 when ineffective discharge actually occurred, we were able to confirm the increase in the power generation due to the reduction in the amount of ineffective discharge.

 This study demonstrates that two-type pretrained models built with artificial neural network (ANN) for reasonable predictions of unprecedented floods in major rivers in Kyushu region, Japan that has similar hydrological features using a small amount of flood datasets. One pretrained model, created by the flood datasets observed in dam-affected rivers, provided better predictions than those of the convention ANN for short lead times and small amounts of the flood datasets. Another pretrained model was created by all flood datasets, available in the major rivers and even no-dam rivers. The difference between the pretrained models shows that the latter model performed better predictions than the former model doing only in the limited condition such as longer lead time and relatively large amount of the datasets.

 In the event of flooding, damage can be reduced by pre-release water. In Thailand, however, the rainy and dry seasons are separated, and dams need to store enough water for the dry season. This results in a high risk of pre-discharge flood. Therefore, it is important to accurately predict dam inflows to make decisions on pre-releases. In this study, we used the Transformer model, which has been attracting attention recently and is becoming a trend in machine learning. A large amount of data is necessary to improve the accuracy of machine learning, but there is only a limited number of monthly inflow data. This study uses the data from Sirikit Dam for pretraining to predict the monthly inflows at Bhumipol Dam and Srinakarin Dam. As a result, the Nash-Sutcliffe efficiency is successfully improved from 0.17 to 0.75 by using incremental learning when the training period was only 5 years.

 Short-term rainfall prediction using deep learning has gained attention recently. However, the evaluation of forecast accuracy has been limited, and comparisons with operational models are rare. In this study, we improved an existing forecasting method for rainfall prediction in Japan up to 6 hours ahead and comprehensively evaluated the results. The proposed method outperforms operational forecasts for precipitation exceeding 50 mm h^-1 and 5 mm h^-1, and the results indicated that the possibility that the model learn the physical phenomena, which the operational model could not consider. However, even during the winter season, there are events caused by low-pressure-induced rainfall and other similar characteristics to summer rainfall, which can be predicted. These findings suggest the need for further discussions on constructing an accurate model and building the dataset appropriately.

 The decreasing production of cultured nori seaweed in Okayama Prefecture has been partly attributed to the feeding damage of Acanthopagrus schlegelii. We applied a deep learning-based detection model (YOLOv5) to underwater camera images of the target species to understand their feeding behaviors efficiently. It is commonly noticed that both the quality and quantity of training data greatly affects the detection accuracy of the model. However, it was challenging to establish objective conditions for labeling because some images were taken unclearly in the natural environment. In this study, we used the Segment Anything Model (SAM), which was recently developed to segment image regions based on huge amount of training dataset, to create a labeling criteria efficiently for unclear images. We analyzed the characteristics of images that were difficult for the SAM model to detect when counting the target species in actual sea areas.We also considered which unclear images should be selectively added to the training data based on the judge from fisheries professionals. Results showed that including both SAM-detectable and non-detectable unclear images of the target species in the training data would be effective in improving accuracy.

 The increase in flood damage due to climate change in recent years has created a need for quantitative assessment of flood risk. The purpose of this study is to adapt the failure probability to flood risk assessment. We have developed a method for evaluating the failure probability and a method for integrating the failure probabilities of multiple evaluation points. The developed method uses a Monte Carlo approach to represent a levee failure due to overtopping and takes into account uncertainties in levee height, water level, and erosion resistance of the slopes. The results of applying this method to a real river levee showed that it is possible to set conditions that prevent levee failure even when the water level exceeds the levee height. We also focused on the dependency relationships between upstream and downstream breach points and proposed a method for integrating the failure probabilities based on the assumption of these relationships. This method reveals that there is a significant difference in the failure probability for the entire section depending on whether or not the dependency relationship is taken into account.

 The material properties inside river embankments are uncertain, depending on the embankment construction history. With the recent intensification of rainfall, there's a need to evaluate the resilience of embankments during overtopping. In this study, we propose an evaluation method for material uncertainty in overtopping erosion analysis. We've realized an approximate multi-solution analysis for different grain-size distributions in river embankments, and assessed the impact of overtopping erosion due to material uncertainty. In this paper, we enhance the analysis method through an extension of the scouring sand and suspended sand models, considering the grain-size distribution effect on erosion resistance, using principal component analysis for dimension reduction, and employing Gaussian process regression for an alternative analysis. We demonstrate the effectiveness of the proposed method through an analysis of the grain-size distribution scatter in real river embankments.

 We have succeeded in visualizing density distribution internal of river embankment, which is necessary for maintenance and quantitative evaluation of embankment failure risk, by using cosmic ray muon. However, conventional method can only evaluate density qualitatively because solution depends on initial value. In this study, we propose the new exploration method that introduces the Moore-Penrose inverse matrix, which uniquely determines the solution of underdetermined system of simultaneous equations, Tikhonov's regularization method, which suppresses effect of observation errors, and Morozov's contradiction principle, which determines regularization parameter. The results show good agreement between the analytical and actual ground densities. It is also found that when space of wave rays is less than 1/4 scale of analytical grid scale, it is possible to estimate location of structures and cavities in ground.

 Based on the Kuma River flood caused by the torrential rainfall in July 2020, this study developed a three-dimensional inundation analysis model that can evaluate the hydrodynamic forces of river bridges and examined the effects of them on flood flows from various angles. As a result, it was confirmed that the our model can take into account changes in the vertical velocity distribution of river flow due to its girder height and shape. The effects of them were suggested that the rapid rise in water level caused by multiple bridges contributes to damage to houses and human casualties in the flooded area.

 This study proposes real-time inundation inside levee prediction method evaluating inundation depth distribution by inundation water volume and wave arrival time. This method saves the depth distributions pre-calculated by flood analysis with multiple hyetographs in preparation phase, and calls these depth distributions according to inundation water volume and wave arrival time by the arbitrary hyetograph in realtime operation phase. This reduces the computational load. At the time for verification of the proposed method, flood analysis results are compared between conventional two-dimensional indefinite flow calculation and proposal method. In addition, it is examined the display of the proposed method by color scheme according to Guide to Creating Flood Hazard Maps in Japan. The proposed method can evaluate such as inundation water volume against elapsed time, maximum inundation depth distribution, and occurrence time and location of inundation above the floor. As a result, the proposed method can evaluate inundation above the floor.

 In the Ichinomiya River basin, a flood control project has been promoted, but observed data such as water level and flow discharge in rivers are not enough to understand hydraulic and hydrological features in the basin. In this study, we observed water level at multiple locations in the Ichinomiya River basin and showed the relationship between rainfall and water level. As a result, we indicated that the water level hydrograph in the main river depend on the confluence of tributaries. In addition, the HWL rainfall amount equivalent to the designed water level in the main river was generally higher than that in tributaries and was different even in the main river, suggesting the differences of flow capacity at each locations.

 Although studies have been conducted on the overtopping and loss of vegetation during floods, focusing on the hydrodynamic forces and scavenging forces acting on vegetation, it is still difficult to fully understand the phenomena. In this study, a three-dimensional vegetation model was created using aerial laser data for the Asahi River branch in Okayama Prefecture, and a three-dimensional flood flow analysis was conducted to construct a model of vegetation overtopping and loss due to moments by considering the vertical distribution of fluid forces acting on vegetation, which could not be considered in the conventional two-dimensional analysis. The model was constructed by considering the vertical distribution of fluid forces acting on vegetation, which could not be considered in conventional two-dimensional analyses. First, correct data for the overtopping and loss of trees were generated from the difference values of ALB data before and after the flooding, and compared with the analysis. As a result, the validity of the model was confirmed. Furthermore, although the effect of vegetation on the diversion flow rate was considered, the analysis of the vegetation overtopping and runoff model at the planned high water flow rate showed that the effect could be negligible.

 This study examined the following (A) and (B): (A) the reproducibility of subsurface and surface flows in a rainfall-runoff and flood inundation model; (B) the usefulness of parameter estimation methods using Bayesian optimization. We also conducted experiments on subsurface and surface flows. The following (1)-(4) were confirmed through the examinations: (1) Under conditions of small suction, the model can reproduce the water level and flow rate fluctuation process; (2) the model parameters can be estimated when surface flow occurs, and suction is small; (3) the model can reproduce the experimental results using the optimal parameters in (2); ( 4) parameter estimation for actual soils may be possible under small suction and surface flow conditions. In other cases, soil tests are necessary to determine the parameters.

 In this study, correlation between the surface instantaneous velocity distribution characteristics and internal structure of turbulent flow in a smooth open channel were investigated using flow velocity measurements and flow visualization methods. As a result, it was clarified that a parallel cycloidal flow was formed in the instantaneous flow velocity distribution even when the bottom wall surface was smooth. Simultaneous visualization of instantaneous water surface velocity and internal flow regime revealed that the longitudinal vortex structure directly contributed to the velocity distribution on the water surface. Furthermore, as a result of examining the correction factor of the water surface float from the instantaneous flow velocity distribution, the ratio of the average flow velocity in the relatively high-speed region corresponding to the parallel spiral flow to the cross-sectional average flow velocity was 0.77. The correction factor of the surface float is 0.85, which is smaller than the value of this study, about 0.08, but it was found to be in line with the results of previous investigations.

 Wind influences hydrodynamics in rivers. Therefore, it is necessary to consider such wind effects when we estimate the depth-averaged velocity and the river discharge from the float velocity. However, in practice, the river discharge is generally estimated by using the calibration of the float without considering wind effects, and the accuracy of estimated discharge is questionable. In this study, we measured the water flow velocity using the float in wind-effected flows, and then this result was compared with PIV data. Based on these experiments, we revealed the influence of wind on the float velocity and evaluated the correction coefficient in wind-induced flows.

 In conventional river engineering, wind stress has been assumed to affect only the temporary increases in water level. However, as climate change progresses and super typhoons are predicted, it is necessary to review the treatment of wind in river calculations. In this study, numerical experiments were conducted on the Edo River during the flood caused by typhoon Hagibis to evaluate the effect of wind on flood flow in an actual river. In addition, long-term observations of tide levels, river water levels, and meteorological data around Tokyo Bay were also analyzed. The results indicate that not only the tide level but also the river water level rises with the increase of wind velocity. Furthermore, the numerical experiments suggest that intense wind stress causes flow velocity and water level response, and that the rise in water level during the water rising period can also affect the peak period due to the propagation of water surface waves.

 This study examined the applicability of several flow discharge estimation methods using observed water levels to flow in a diversion channel. We used the following methods. (A) a ID estimation method using temporal changes of the water-surface profile. (B)a 2D and (C) a quasi-3D numerical model with the water level as boundary conditions. We conducted experiments of steady and unsteady flow in channels with diversion. In the experiments, we observed temporal changes in water level and discharge. The computed results were verified against the observed data to compare the accuracy of the methods. The comparison showed the following. (I) Analysis method (A) can reproduce the flow rates upstream and downstream of the diversion and the diversion with sufficient accuracy except for unsteady flow, (2) Analysis methods (B) and (C) can reproduce the flow rates when the diversion is not included, and (3) Analysis methods (B) and (C) can reproduce the flow rates when the diversion is included because strong 3-D flow is generated in case of weir backwater. ( 4) On the other hand, in the case of a lowered backwater, both analysis methods (B) and (C) can reproduce the main flow rate.

 In this study, the characteristics of the flow field associated with the suction operation of the longitudinal ridge element in an open channel with the longitudinal ridge element installed on the bottom wall were investigated by flow velocity measurement and flow visualization. From the flow velocity measurement results, it was clarified that the remarkable upward flow and paired swirling flow disappeared by the suction operation, and the primary velocity distribution on the longitudinal ridge element became uniform in the transverse direction. In addition, the horizontal Reynolds stress (-uw) is reduced by the suction operation, and the positive and negative phases of the distribution are reversed. Furthermore, in the horizontal turbulence energy generation term (-uw(dU/dy)) distribution, it was clarified that a suction operation produces a negative value on the longitudinal ridge element. It is speculated that the characteristics of the above flow velocity measurement results are caused by the intermittent and small-scale longutudinal vortex structure on the longitudinal ridge element due to the suction operation.

 Although channel width change longitudinally in river, detailed studies of hydraulic characteristics in areas where the river width changes, such as concave areas, have not focused on the longitudinal effects as fundamental research. In this study, experiments have been conducted to investigate the basic changes in the water surface profile and flow characteristics on the condition that steady and unsteady water supply in flume with and without widening area. As a result, it was possible to show the flow and sediment deposition trends peculiar to the widening area under steady flow, as well as the flood propagation characteristics under non-steady flow. In particular, under the condition that the widening area is more wide, the effect of the rise in water level in the upstream of narrow section is widely propagated upstream. In addition, it was shown that the flood wave propagated downstream before the water level rose sufficiently and also reduced the peak water level downstream although the short rise time of the flood wave delayed the rise of the water level within the widening area.

 In numerical analysis of flood flows, the more detailed the calculation is, the longer the computation time is required. In this study, we focused on the difference between the time interval required by the spatial resolution and the time interval required for the output of the flood flow prediction to accelerate the calculation by omitting this part. We first conducted calculations reducing the time intervals of the inflow hydrograph from the upstream end boundary. The results revealed that the shortening hydrograph delayed the arrival time of the flood. Next, we proposed the average component acceleration method that accelerates only the gradually changing mean component of the time evolution of the water depth. This method achieved comparable results to conventional computations while reducing the computation time by shortening the hydrograph. However, as the acceleration rate increased, oscillations occurred in the calculation results due to the relationship between the magnitudes of the average and fluctuating components.