Pseudo global warming experiments (PGW) were conducted with 1000m and 500m horizontal resolution for the 2012 Kameoka extreme rainfall event, which is back-building convection system. We analyzed the difference of back-building representation due to the difference of spatial resolution, the future changes of back-building, and its mechanisms from multiple spatial scales. As a result, the back-building was very well represented in both the present and PGW experiments with 500m resolution, and the rainfall intensity and total rainfall increased in PGW experiments. This is due to the fact that in the future, the amount of condensation in cumulonimbus will increase, a convective unstable field will be realized (meso γ to β), and the back-building structure will be strengthened (meso β) and that the convective unstable atmosphere was continuously supplied from the south (meso α). We also showed that the 1000m resolution can capture future changes in meso-α, but cannot capture changes smaller than meso-β sufficiently.

 In this study, we statistically analyzed the future change of Baiu heavy rainfall duration and accumulated rainfall considering its spatio-temporal characteristics by classifying them into “quite localized back-building type Baiu heavy rainfall” and “relatively localized Baiu heavy rainfall caused by large convergence of Baiu front”. We analyzed the Baiu heavy rainfall events obtained from a 5km-mesh regional climate model and past real events. As a result, we showed the parameter of present and future Baiu heavy rainfall is significantly different and that the peak of occurence frequency will shift to the longer duration in future. In addition, it is clarified that the relationship between heavy rain duration and accumulated rainfall was different by classifying the spatio-temporal characteristics. The results bring a new viewpoint of future change, and deepened the understanding of the phenomenon itself of Baiu heavy rainfall.

 In this study, future changes in formation of heavy rainfalls (1 in 100 year probability rainfalls) in Gifu Prefecture, Japan, are compared using sea-level pressure distributions derived from the database called d4PDF including the 4-K warming future climate experiments (5,400 years) and historical climate experiments (3,000 years). The sea-level pressure patterns in the heavy rainfall events are classified into three categories: (A) frontal system I, (B) frontal system II, (C) low pressure system I and (D) low pressure system II. The averaged daily rainfalls in the future climate are greater in any category than those in the historical climate. It is considered that the heavy rainfalls in the future climate in Type A and Type D are intensified by the reinforcement of southwestern typhoons and the slowdown of landfalling typhoons, respectively.

 The factor analysis method for spatial deviation of heavy rainfall was proposed in this study. This method used frequency of vertical wind and rainfall intensity as the function of vertical wind speed for analizing the factor of spatial deviation of rainfall into dynamic and thermodynamic effect. The method was applied for the annual maximum heavy rainfall events over the Tokachi River basin, which was obtained from 5 km resolution large ensemble dataset. The results show the thermodynamic effect is dominant factor of spatial deviation of rainfall, while the dynamic effect makes rainfall heavier around the Hidaka mountains. The model was also applied to the dataset under warmer climate condition. The results show general character of the factor of spatial deviation of rainfall does not change under the warmer climate, while both the dynamic effect and the thermodynamic effect are stronger than the past climate and make rainfall heavier around the Hidaka mountains.

 This study estimates the hydrological quantities such as evapotranspiration and snow depth in Hokkaido, using statistically downscaled (SDS) data, to investigate measures for adapting to climate change.The accuracy of the basin water balance estimation result was improved by changing the data referenced by SDS and correcting the snowfall water amount.Using the created mesh data, the hydrologic cycle of Hokkaido under the future climate was projected, the annual maximum snow depth is projected to decrease.It is suggested the need for adaptation measures that take into account changes in the hydrologic cycle in snowy cold regions.The results promise to be of use in evaluating the effects of climate change on local water issues and in investigating measures for adapting to climate change.

 Rain gages have problems that they don’t observe precipitation at mountain area because they are sparse there, and they can’t capture snowfall enough. Therefore, the basin average precipitation from them may not be suitable to evaluate it in climate models and so on for basin scale or more. Aiming to evaluate precipitation without influenced by these problems, this study estimates basin average precipitation from water budget using observed runoff, evapotranspiration, soil moisture and snow water equivalent ; the last three values are calculated by the land surface model SiBUC. And then, the bias in d4PDF was evaluated using both precipitation estimated from water budget and the precipitation grid data APHRODITE. It reveals that the former is better to evaluate basin average d4PDF precipitation than the latter.

 Climate change is a serious issue resulting in global variation in the precipitation pattern. One of the major problems hindering research in examining the change and variation in precipitation trend in Middle East and North Africa (MENA) is a lack of high-quality, long-term data. Therefore, in the current study, the spatial and temporal annual, seasonal, and extreme precipitation trend were analyzed over 118 years (1901-2018) using high resolution Climate Research Unit data (CRU TS 4.03) over the entire region. Durbin Watson was used to test the autocorrelation in the annual and seasonal series, the modified MannKendall (MMK) and Sen's slope estimator tests were used to detect trend for the autocorrelated time series data. The monthly CRU data was then converted to daily using MODAWEC model in order to test the trend in extreme precipitation and for evaluating the change in four indices (R5D, R20, RR1, and SDII). The results showed decreasing trend affecting most of the countries in the region especially in annual and winter series with Yamen, Palestine and Lebanon got the highest significant negative trend. There were no signal toward negative or positive trend in extreme precipitation series. Arabian Peninsula (Kuwait, Qatar, Yamen, and Bahrain) showed positive trend in extreme indices (R5D, R20, SDII) while the western side exhibited negative trend. North Africa had stationary condition except for Egypt where negative trend detected.

 In previous studies, reanalysis data has been used instead of observational data for bias correction of GCMs outputs in observational data scarce basin. However, reanalysis data does not match well the real weather characteristics (e.g. precipitation) in a local scale since it is targeted at a global scale. Therefore, this study aims to apply a statistical correction method for re-analysis data based on observational data. And also, we compared accuracy of statistical correction method which is developed in 4 calibration periods (observational datasets for 5 years, 10 years, 15 years and 20 years), to clarify the number of observation period required for development of statistical correction method. As a result, it is revealed that observational datasets at least 15 years should be used as calibration period, to reproduce the statistical characteristics (average, standard deviation, frequency of precipitation, and so on) in observational data.

 Extremes of rainfall at several sites are important elements to examine the total amount of precipitation for the drainage basin, and the correlation should be taken into consideration to evaluate the flood risk. Dependence function has been investigated in the context of theoretical statistics for multivariate extremes, because such item can be only defined for the excess over either threshold of the components. However it gives the indirect properties of the component-wise maxima which have occurred in the same year and which were not caused by the same rainfall. Here a newly defined correlation function is proposed in this paper by employing the occurrence numbers which includes the joint exceedance of all components and the partial exceedance of at least one component. Numerical simulations are also implemented and demonstrated to show the statistical variabilities.

 Recent studies report that ongoing climate change has enhanced frequency and amount of heavy precipitation, however, current design storms rely on statistics derived from long-term observed precipitation without considering the effect of climate change. In this study, we investigated how the design storms and associated maximum inundation depth and extent may change under 4-degree warmer climate condition using a large ensemble climate experiment d4PDF over the Zenpukuji river in Tokyo, which is a typical urban river basin. Daily precipitation of 20-year return period increased 1.2 times higher under 4-degree climate. Simulations of a flood analysis model using the obtained design storms showed that that maximum inundation depth and inundation extent increased up to 1.2 times and 1.5 times, respectively. We also showed that simulations using a centralized shape of design storm produced inundation extent about 10% larger than that using design storm based on scaling method.

 In recent years unusually heavy rainfall has been occurring with greater frequency and intensity throughout Japan, phenomena considered to be a result of climate change caused by the emission of greenhouse gases. It is thus necessary to evaluate the degree of such impact on stability of rail transpotation that may also be affected by such phenomena, and study possible countermeasures to incorporate into train operating procedures.Therefore, the purpose of this study is to understand the effects of climate change in terms of stability in railway transportation in times of heavy rainfall. The results of an experiment (NHRCM02) conducted by the Meteorological Agency’s research institute to obtain projections of Japan’s climate around the end of the 21st century based on the RCP8.5 scenario for greenhouse gas emission was used in the analysis. According to a quantitative study of railway lines in Japan’s Tōhoku region, the stability of railway transportation will suffer, due to higher frequency of heavy rainfall at greater intensity by the end of the 21st century and the amount of time required for stopping operation is expected to increase by 3.5 to 6.5 times; that is, an average of 470% over current time lengths.

 We present a study on the future changes of landslide risk in Japan by applying the critical line method to the 5-km regional climate model of NHRCM05. From the six datasets of NHRCM05, we extracted effective rainfall events from the parameter of surface precipitation without any interpolation. In all extracted rainfall events, the critical line method was applied to obtain the frequency of landslide occurrence at each 1 by 1-km grid in the whole Japan. We exhibit the future changes of nationwide landslide risk distribution, monthly occurrence frequency in each geographical region, and occurrence trend in each prefecture. Additionally, from a different perspective, we reveal the relation between landslide risk trend and the geological feature of plate tectonics. As a result, the landslide risk is higher in early summer of July in west Japan and in late summer of September in east Japan. Particularly, the analysis shows a significant increasing trend of landslide risk in Hokkaido area.

 This study assessed the latent risk of large-scale flooding and the effectiveness of flood control facilities in low-lying river areas of the Chitose River, a tributary of the Ishikari River. As this river flows through low-lying areas, it is affected by backwater from the main river. To estimate the inundation risk of the Chitose River basin area, a large volume of ensemble rainfall information, acquired by downscaling d4PDF data, was used to estimate climate uncertainty. A case similar to floods observed in the past was confirmed among the 3,000 cases in the d4PDF current climate prediction data. Additionally, a case was found that predicted outflow patterns for both the Chitose River and Ishikari River which exceeded past flood records. These results indicated the latent possibility of inundation damages exceeding those from the past floods for the Chitose River, while acknowledging the effectiveness of flood control facilities.

 It is essential to predict river discharge under the climate change for adaptation to water related hazard and water resources. In this study, bias correction method to runoff data has been developed to estimate river discharge more accurately. Quantile-Quantile mapping method was applied to correct bias of runoff data from GCM output. Reference runoff data for bias correction was created by land surface model utilizing atmospheric reanalysis and observation data. Bias corrected runoff data was converted to river discharge by river routing model, then the accuracy of bias correction method was evaluated from the viewpoint of river discharge. Moreover, the bias correction function which was determined by certain GCM ensemble was applied to the other ensembles for evaluation of applicability of developed bias correction method.

 This study estimated the damage cost and exposed population caused by inland flood in Japan modeled by extreme precipitation data. We used two inundation scenarios (natural drainage scenario and poor drainage scenario) based on the presence or absence of drainage by gravity to rivers. In the current climate, in the natural drainage scenario, the expected damage cost and exposed population were estimated to be 618 billion JPY/year and 52,695 people/year, respectively. On the other hand, in the poor drainage scenario, the expected damage cost and exposed population were estimated to be 9,422.4 billion JPY/year and 870,388 people/year, respectively. We then estimated the damage cost and exposed population in the nearfuture climate (2031-2050) using the poor drainage scenario. As a result, the damage cost increased about 2-fold and exposed population (considering population change) increased about 1.6-fold from current climate to near-future climate. The increase rate of the damage cost tended to be higher on the Japan Sea side. The increase rate of exposed population was higher in Hokuriku region, Tokyo and Saitama prefecture.

 This paper shows a bias correction method for runoff estimation by a global climate model that considers land cover characteristics. In GCM simulation processes, runoff data is generated by a land surface model that incorporates the direct influence of land cover. This makes the bias correction possible to adapt to the different types of land cover setting in GCMs. In this study, a bias correction method considering land cover characteristics with a combination of a linear scaling factor and an empirical quantile mapping was implemented to MRI-AGCM3.2S runoff data over the Chao Phraya River basin in Thailand. The daily runoff spatial pattern and daily river discharge were used to evaluate the performance of the bias correction. The result shows that the proposed bias correction has better performance than a direct application of empirical quantile mapping bias correction in a sub-catchment and grid to grid-scale grouping for the rainy season. These results underpin the incorporation of land cover information into a bias correction method to improve the performance of runoff and river discharge predictions.

 d4PDF runoff data was applied to river discharge simulation in the Chao Phraya River basin. The simulated discharge underestimated and overestimated observed one in the upstream Bhumibol and Sirikit dams and downstream Nakhon Sawan station, respectively. These different biases were corrected separately through 1) bias correction of wet and dry season volumes of dam inflow, 2) dam operation and river routing simulation with the bias-corrected dam inflow, 3) and bias correction of flood volume (over 2000 m³/s) at the Nakhon Sawan station. As the result, the return period of 2011 flood was evaluated in terms of flood volume as around 100 years. Inundation by the top 60 flood volume in d4PDF was simulated with a flood-inundation model accelerated with the ADU algorithm. Maximum flood volume varied among the events while flood area was similar to each other, implying that largest-class floods cause inundation in similar area to the 2011 flood with deeper flood depth.

 Urbanization is an essential, yet underrepresented, parameter when investigating futuristic climate change of cities. The change in 2 m air temperature in August between the 2006–2015 period and the 2046–2055 period for 33 megacities and 10 emerging megacities under RCP8.5 emission forcing and SSP3 was projected with the consideration of both global climate change (using pseudo-global warming method) and local urbanization (using global urban sprawling map, distributed urban morphological parameters, and hourly anthropogenic heat emission dataset).

 In newly urbanized area, the urbanization effect will be significant, accounting for (13.5 ± 5.9) % of the total temperature change. In existing urban areas, the effect will vary depending on the current degree of urbanization. When viewed over a regional scale, the effect will be rather insignificant. It was observed in some cities that urbanization effect originating from urban area was advected by the wind to non-urban area located kilometers downwind.

 For a more accurate understanding of the impact of climate change and human activities on the hydrological cycle, a hyper-resolution global water resource assessment is necessary. Functions such as intergrid water transfer and medium-sized reservoir operation, which are essential for hyper-resolution simulations, are also available in the latest version of global hydrological model H08. In this study, we conducted a series of global hydrological simulations at spatial resolution of 5 minutes and 0.5 degrees in H08. By comparing the model responses with different resolutions, we obtained insight into the importance of water allocation functions and changes in water withdrawal sources in a hyper-resolution global hydrological model. In addition, our results showed a certain level of feasibility of water supply and demand assessment in large irrigation areas and large cities by using the hyper-resolution global hydological model.

 World population would generally increase until 2100. To cope with the population growth, it is necessary to build a sustainable agricultural production system. Salinization is one of the factors that threaten sustainable land use in farmlands. In this study, we focused on potential risk of salinization to both irrigated and rainfed farmlands. We selected five factors which were soil type, dryness, groundwater level, groundwater condition and irrigation method on the basis of salinaization mechanism. As a result, we could make distribution map of salinization of potential risk on a global scale, and clarified that a high potential risk areas of salinization are 20% of the irrigated farmland which have some potential risk. In rainfed farmland, we showed potential risk areas of salinization are 65%.

 Recently, data assimilation has been introduced into climate reconstruction. In this study, we conducted several experiments with different prior estimates in order to reconstruct climate change over the last millennium by data assimilation using oxygen isotope ratios of proxies. The results show global surface air temperature decline after the large-scale volcanic eruptions in the past and the global distribution of surface air temperature difference between the Medieval Warm Period (MWP) and the Little Ice Age (LIA). The temperature decrease after the volcanic eruptions was confirmed by results of other studies using data assimilation. On the other hand, the global distribution of surface air temperature difference (MWP-LIA) was different from the other studies, mainly in the tropical Pacific. In this study, proxy models corresponding to process of isotopic variation are used, while statistical proxy models are used in the other studies. As well as the climate models and proxies used for data assimilation, it is suggested that the difference of the methods is reflected in the results.

 Two forest operation techniques with an flood peak reduction effect were discussed using a tank-model and two small artificial forested basins where forest physiognomy was different. One operation was the way to strengthen the insulation evaporative rate by changing its physiognomy, other was the way to make forest floor vegetation and ground roughness increased. The effect when artificial cedar forest was replaced with a mixed conifer-hardwood forest was discussed. As a result, it was estimated that the flood peak flow decreased around 20% before the replacement.

 Risk evaluation of bridge collapse requires a regression model for river discharge that not only covers the situation during precipitation but also an ordinary situation without it. Although data-driven approaches using machine learning have been proposed to construct regression models for forecasting river discharge following precipitation, the distribution of discharge data is imbalanced. Thus, these regression models usually cannot track rare behaviors, such as a sudden increase in discharge because of heavy precipitation. This study aims to improve the performance of regression models by resampling learning data. The discharge data of Doki river targeted in this study has an imbalanced distribution because of rare precipitation. This data imbalance is alleviated using the resampling technique by oversampling synthetically minor data and undersampling major data. Regression models were constructed by learning the original data and resampled data, respectively. The performances of these models were compared to show that the alleviation of imbalanced learning data significantly improves the regression accuracy of the high-discharge region while maintaining the regression accuracy of the low-discharge region.

 To manage intensification of flood risk accompanied with climate change, it is necessary to predict flood discharge with high accuracy. In this research, we aimed to apply rainfall-runoff model considering vertical infiltration into the large-scale flood which we have not experienced in the past. Therefore, parameter estimations of this rainfall-runoff model are conducted for dam basins in upper Tone river basin, and parameter characteristics are analyzed in the perspective of flood scale. Considering results of this analysis, reproduction calculations of targeted flood events are conducted using estimated parameters. As a result, high reproductivity of runoff analysis which gives Nash coefficient exceeding 0.7 is shown by comparison of runoff calculations using average value of parameters with small scale and medium scale flood events respectively. Also, the tendency of underestimation of calculated discharge near peaks is improved in the case of using parameters derived from medium scale and large scale flood events.

 The purpose of this study is to apply two kinds of inundation analysis models to the Shiranego area of Niigata prefecture, which is a low-lying paddy field area, and to compare their model performance. The following two inundation analysis models are adopted: (1) the cell model which fine topography is represented by arbitrary polygons called topographically adjusted cells, and the propagation of flood water can be represented by water movement between adjacent cells, (2) the grid model that traces two-dimensional flow with a uniform mesh of 50m, which has been used for creating inundation area maps. The models were verified by the flood in July 2011 and obtained results are as follows. The cell model was superior to the grid model in the estimation accuracy of the river/drainage water level. In the grid model using the uniform mesh, the water level recession in the latter half of the flood was faster than the actual results. The inundation area was underestimated by the cell model and overestimated by the grid model.

 Predicting snowmelt runoff with a high accuracy is important in water utilization and prevention of floods. Temperature lapse rate is one of uncertainty in snowmelt runoff analysis. In this study, we estimated temperature lapse rate in each month of each year in two regions and assessed seasonality and locality of temperature lapse rate. The purpose of this study is that we showed the importance of considering of temperature lapse rate in snowmelt runoff analysis to compared snowmelt runoff analysis which using estimated temperature lapse rate and snowmelt runoff analysis which using constant temperature lapse rate in four large and small regions in upper and middle area in Shinano river basin. As a result, temperature lapse rate has seasonality and locality. Using estimated temperature lapse rate led to high accuracy in snowmelt runoff analysis. Additionally, regions having large amount of snowfall were affected by temperature lapse rate than regions having small amount of snowfall in a heavy snowfall region.

 As a damage reduction plan with the ice jam, we aim to develop practical ice jam occurrence prediction model. It is important to collect examples of recent ice jams and to solve the problems by applying the model on actual rivers for practical use. At first we investigated the mechanism of ice jam which occurred in several rivers in Hokkaido in March 2020. Time of outbreak and disappearance of ice jam was obtained successfully with video image by river surveillance camera and telemeting water level. Next we tried to evaluate the occurance risk evaluation of this ice jam. We compared with factor which influenced ice jam occurance in 2018 and in 2020. We also examined problems and improvement about outbreak prediction with the river ice thickness prediction model.

 This study aimed to clarify the situation of a river ice jam occurred on the Tokoro river in Hokkaido. We carried out water level observation, video and aerial photography, and also visually monitored the river ice area. The phenomenon of ice jam became clear by this observation. The water level rapidly increased to 3.1 m in 2 hours due to the ice jam phenomenon. River ice size is large near ice jam occurrence point. And this size is small at an upstream point more distant than an ice jam occurrence point. River ice velocity before and after the occurrence of the ice jam was calculated by PIV analysis. The speed of river ice flow before and after the ice jam was determined by CCTV cameras. The maximum velocity of the river ice is 3.5 m/s. As the flow rate increases, more water flows downstream with the ice, which clogs the river channel where it tends to accumulate, blocking the flow and causing the water level to rise.

 This study aims to clarify the ice jam phenomenon around piers. We conducted an ice jam experiment which used mixed real ice with ice sheet and frazil slush. In order to clarify the effect of frazil slush, the experimental results show that ice velocity, number of ice, floods area variations, water level variations. In the case of containing frazil slush, they accumulate underneath the flowing ice sheet, sticking to each other and flowing down as large blocks. And, Ice sheet velocity is slowed down because the shape resistance increases. In the case of containing frazil slush, the ice flows downstream as large ice blocks. When ice jam is generated, the upstream floods area velocity increases and floods rapidly.

 Recently, due to the climate change, Japan has suffered from devastating hydro-meteorological hazards caused by localized heavy rainfall known as Guerrilla heavy rainfall (GHR). GHR causes a disadvantage of very short lead time for early warning and evacuation. For accurately predicting disasters triggered by GHR, it is necessary to develop a methodology integrating early detection and quantitative risk prediction methods. By using the early detection method, the existence of the first echo of hydrometeors in a convective cell was identified. Then, to predict the risk more precisely, the output of risk prediction is categorized into four risk levels based on the maximum rainfall intensity of rain-cells. To accomplish accurate prediction, a variable selection method is applied to identify the relevant variables obtained by radar observation data. With selected relevant variables, multiple linear regression equations for risk levels are fitted by 10 GHR events. The most appropriate multiple linear regression equation is verified by a Receiver Operating Characteristic (ROC) analysis. As the results, it is possible to predict the risk quantitatively with a high accuracy of 81%.

 Radar observation is commonly used for rainfall observation and estimate the structure in clouds through wind speed. In this study, we observed rainfall and sea spray in the atmospheric boundary layer in the summer of 2018. A normalized radar cross section, which is an important parameter for radar observation, can be derived by assuming an arbitrary drop size distribution in the Mie scattering theory. Therefore, in this study, we propose an algorithm to estimate the parameters of the drop size distribution of rainfall and sea spray by comparing the normalized radar cross section obtained from observation with that obtained from the Mie scattering theory.

 The investigation of single cell to multicell in the cluster thunderstorms using vertical vorticity analysis will be proposed. The techniques of vortex tube structure and identification of initial stage values of the first radar echo were applied to rainfall events with the unstable atmospheric conditions in Kinki region, Japan. Cluster thunderstorms cases during the two events were selected and the analysis of single cells to multicells were performed by investigating the initial stage of vorticity values and their structures. The vertical vortex tube of vorticity analysis was shown to examine the signature of merging of single cells to mutlicells. In the case of two events, the criteria index of persistence development for cells and dissipating cells were established and it was found that the positive vorticity greater than 0.015s^-1 possibly prolonged the duration of lifetime more than one hour. In the meantime, the positive vorticity less than 0.0025s^-1 might shorten the lifetime of single cells in the multicell clusters. The signature for merging of single cell to multicell was difficult to discover due to the change of core vorticity at upper level height before single cells merge into multicells.

 In September 2014, during Ishikari and Iburi heavy rainfall, Line-Shaped Rain Bands was occured. The relationship between the three dimensional wind field calculated using the three Doppler radars and precipitation intensity was shown. Heavy rains show typical back-building characteristics, with new precipitation cells being generated one after another and propagating northeast.

 In the lower levels, the warm and humid air masses from the southeast converged and caused an upward motion at the upper layer. When the upward airflow grew to the vertical vortex structure, the precipitation intensity was stronger. The relationship may provide a physical explanation for the back-building mechanism.

 This study aims to develop a measurement strategy for drop size distribution (DSD) in the cloud. It is challenging to estimate quantitative DSD using a video sonde because it has a small window to capture precipitation particles. The present study proposes a sampling number to ensure the quantitative DSD with numerical simulation using the bin type simulation model, SCALE-RM. Then, it compares the DSD measured by a video sonde observation and radar data through the cloud water content.

 The type of precipitation on the ground in winter season is the most important information to estimate the risk of snow damages. As a commonly used method to determine type of precipitation, there is a method using the temperature as a threshold value. But the accuracy becomes poor near 0°C. Therefore, we consider to use the snowflakes melting model, and use polarimetric radar for making initial value, freezing level and particles diameter. In this study, we develop the method to estimate freezing level using polarimetric radar. In winter precipitation cases, it is confirmed that our developed method can estimate freezing level in more cases and improve the accuracy than previous studies method. And we applied developed method to snowflakes melting model as initial value and compered with ground observation. As a result, it was found that by using liquid water content calculated by the melting model, it is possible to make a more detailed determination of the precipitation type than using only the temperature.

 A compact low-cost Doppler rainfall sensor towards vehicle precipitation measurements is presented in this study. The applicability of the Doppler sensor for mobile measurements was examined in the vehicle observation conducted in Chiba on 25^th June 2020. Two Doppler rainfall sensors was prepared and mounted on the roof carrier of a car, but one was fixed to face upwards and another did horizontally, to see the sensitivity of sensors installed in the different direction. To compare the signals from the Doppler sensors with rainfall intensity, a reliable laser drop-sizing gauge was also placed on the roof carrier, together with a cup anemometer, a wind vane and GPS to get the traveling speed of the car. The signals seemingly responded to rainfall intensity but were apparently influenced by the relative wind speed measured by the wind anemometer, which can represent the relative motion of raindrops against the sensors on the moving car. Thus, a wind correction was applied to the signals in a simple way that the values were divided by the relative wind speed. As results, the corrected signals showed a clear linear relation to the rainfall intensity measured by the laser drop-sizing gauge, particularly true for the vertically-facing Doppler sensor. These results show that, by applying a linear regression model (R² = 0.84) to the corrected signals from the Doppler rainfall sensor, rainfall intensity can be estimated.

 This research focused on investigation of impacts of model structures and soil parameters of Land Surface Models (LSMs) from the view point of runoff characteristics and simulated river discharge. In this study, MRI-SiB output was used as reference data. While, SiBUC model was utilized to investigate the impacts of changing model structures and soil parameters. The findings of this study indicated that: (1) introducing a structure for direct infiltration into root zone resulted in decreased of surface runoff by 60% and 7% using SiBUC and MRI-SiB soil parameters, respectively, and (2) consideration of gravitational drainage for soil water flux calculation contributed to an increase of surface runoff by 56% and decrease of evapotranspiration by 5%. Impact of changing soil parameters and model structures was further investigated by analyzing the change of soil moisture movement and evapotranspiration components.

 Rainfall infiltration and surface run-off processes has been researched for many years under the uniform field condition. In a real surface, however, hydraulic conductivity and other physical quantity have spatial variation, and it is considered that surface run-off occurs not only the bottom of the slope but other territory. In this study, a special distribution model of the hydraulic conductivity is used to generate aquifer with pseudo dispersion and bias, and surface run-off quantity and its composition in some patterns of rainfall, in three-dimensional field is evaluated by the numerical analysis. As a result, it was shown that surface run-off quantity in uniform field tend to evaluate less than the result of analysis in non-uniform field, and it was also shown that hydraulic conductivity and surface run-off ratio in heavy rain can be closely resembled by a logarithmic function.

 Comparison of land surface temperatures from the land surface model and satellite data in a tropical monsoon climate region. It is suggested that the land surface temperature may be overestimated by the land process model in this region. In order to improve the accuracy of the surface temperature analysis, it was found that it is very important to improve the accuracy of parameters related to meteorological forcing and land cover. Especially in the case of a water body, the input data of wind speed does not represent the local wind changes and surface temperature analysis is not accurate. Also, the soil may have dried out too early in SiBUC in green areas.

 In October 2019, Typhoon Hagibis traveled over eastern Kanto area, provided record-breaking rainfall in the wide area of Kanto and Tohoku regions, resulted in severe flood disasters including Chikuma River. We applied our ensemble flood forecasting method for Chikuma River flood event to investigate the predictability of our system. Our method consists of regional ensemble numerical weather prediction, and Rainfall-Runoff-Inundation (RRI) model for hydrological part. The method could predict the arrival and water level five days before the flood arrival. The forecast simulations of 1.5 to 2.5 days lead-time were in good accuracy with water level error of 1 meter and flood arrival time error within 3 hours. The accuracy of the typhoon track forecasts by JMA global weather forecasts were particularly good in this event resulted in our good ensemble flood forecasting. This result introduced a possibility of an advanced flood forecasting.

 We conducted hindcast over the southern region of Chile using ROMS-NPZD, a regional ocean physical-biological model to investigate the effect of regional atmosphere-ocean interaction on the ocean nitrogen cycle. Firstly, the validation of the regional atmospheric model RSM and the regional atmosphere-ocean coupled model RSM-ROMS showed that RSM-ROMS captured the seasonal and daily variations of observations of terrestrial variables, and the feedback of the atmosphere-ocean interaction reduced the error compared to that of RSM, while the error of sea surface temperature increased. Next, we validated the experiments using the atmospheric field of RSM and RSM-ROMS as boundary conditions for ROMS-NPZD (UCPL and CPL) by comparing with satellite observations, CPL captured a tendency of the deviation of chlorophyll-a concentration from the annual mean, but the error of the horizontal distribution became larger then that of UCPL. It is also suggested that the feedback of atmosphere-ocean interactions may affect the ocean circulation field through the light environment.

 This study developed a method to measure spatial distribution of the surface sensible heat flux by means of the thermal image velocimetry (TIV) and bulk transfer coefficient. TIV reproduced a spatial distribution of horizontal wind speed near the ground surface. This velocity distribution is used to calculate the sebsible heat flux by multiplying the bulk transfer coefficient of heat. This method is tested in Saitama 2002 stadium, which has a natural turf ground for the football. The method can make accurate measurements of the velocity and the sensible heat flux near the ground as as validated in the ground-based observations. In addition, the method can clarify the spatial heterogeneity of the wind speed and sensible heat flux on the identical soccer ground.

 Hight-resolution pollutant model embedded into Lattice Boltzmann method (LBM) is constructed. We focuse on Particle pollutants. Flow field is calculated using D3Q27 model of LBM and particle is calculated by Lagrangian method. Using this model, we discuss the change in concentration distribution when there is a huge building (GARUDA) in Jakarta as a application.

 As a result, we can find the relation of differences in particle density and differences in flow velocity due to GARUDA. When the velocity in the case w/o GARUDA is faster than the other, particle velocity in the case w/o GARUDA is reduced. The reverse is also possible. And also, we can find the velocity near the solid boundary is underestimated and the particle density is higher than theoretical value. However, this model is valid far away from the solid boundary.

 Analyzing storm genesis, which contains persistent and complex patterns of deep convection, is important for forecasting linear-shaped heavy rainfall several hours ahead. It is suggested that the baroclinic atmosphere in the localized region generated by the temperature and pressure gradients realizes deep convection. In this research, the data assimilation method of vertical vorticity was used to determine the effect of the baroclinic atmosphere on the initial values of model computations. First, applying the variational method enables us to obtain a diagnostic equation in which the equation of motion, conservation law of mass, and entropy are considered as constraints. Its formalism is applied to the observation operator for assimilation of the estimated vertical vorticity. Next, idealized experiments are used to clarify how the data assimilation method, represented above, affects the atmosphere in the model computations. As a result, the change in the potential temperature and vertical wind thorough the error covariance matrix generates coherent convection in the computations. Compared with the computation in which horizontally uniform vertical shear is applied to the initial state, the results of data assimilation are also similar to those from the viewpoint of the relationship between potential temperature distribution and convection.

 Large-eddy simulation based on lattice-Boltzmann method has been conducted, using different inflow conditions, over a realistic urban geometry including high-rise building area in Shinjuku. Three types of inflow conditions regarding the boundary-layer height and turbulence intensities were reproduced by extending fetch with roughness blocks. The effect of boundary-layer development on mean air flow fields under neutral atmospheric condition has been investigated in a target domain with identical building geometries. The ratio of mean wind speed at the same positions in the different cases corresponded well with the ratio of the friction velocities. This means that mean wind speed within the urban canyon normalized by a bulk friction velocity is solely determined by the building geometry. Therefore, we can reproduce a three dimensional mean wind distribution within the urban canyon diagnostically by referencing a regionally representative velocity scale, which is the friction velocity, and the preprocessed microscale simulation of the target urban district.

 We proposed a formula to estimate WBGT from the brightness temperature of Himawari-8. The brightness temperature of Band 7, 13, and 15, which correspond to the thermal infrared bands, were used for this purpose since they were highly correlated with the WBGT observations near the ground. The effects of thick clouds were removed based on the thresholds for the brightness temperatures. The formula of the WBGT was derived from the multiple regression analysis. The RMSE of the estimated WBGT was less than 3.0 ℃ at about 75% of 74 observation points in Kanto region.

 Municipality budgets cannot cover all the necessary public works for environmental problems. Under the limited budget, it is important to reasonably prioritize environmental public projects according to the number of beneficiaries and the extent of environmental impacts. We define ‘street potential’ as comprehensive street’s attractiveness considering the existing environmental problems and citizens’ utilization in each street. Here we aimed to development the assessment system of street potential which consists of a low cost aerosol sensor, digital camera, and objects detection algorhithm YOLOv3 for measuring the concentration of air contaminant and the number of pedestrians. The performance of the proposed system was validated with a conventional aerosol monitor for air pollutant level and visual inspections of the pedestrians. The aerosol sencer tended to overestimate the number concentration. Automatic extraction with YOLOv3 is performed accurately and as effective as visual inspection. We conducted mobile observation with this system and confirm that it can measure the concentration of the air contaminant and the number of citizens in each street at the same time.

 Heat-risk evaluations are needed by societies for adapting to and mitigating the social/health impacts attributed to climate change and urbanization. Meanwhile, big data and numerical models are advancing.

 In this study, we propose an integrated approach to utilize both big data and existing models for estimating heat-related risk. By combining products of a 3D-City irradiance model and Google big data services, a geospatial framework for mapping heat-related risks(Hazard×Exposure×Vulnerability) was developed .Hazard and exposure(individual HR factor) were estimated using thresholds of wetbulb globe temperature(WBGT) and mobility from frequented locations to nearest station exits; Vulnerability(HR probability factor) was estimated using popular times of frequented locations. Through this framework, multiple perspectives for understanding heat risks may be observed at various districts. For example, ATMs were found to have high heat risks mainly because of its high HR probability factor.

 This study estimated the potential shortage for irrigation water in Tone-gawa and Ara-kawa basin area by using the integrated water resources model H08 including the calculation module relating agricultural activities. The estimated shortage volume was compared with the hydrogeological change signals which groundwater or land subsidence level. The time series of calculated potential value was corresponded with the actual groundwater level change and transition of land subsidence in drought year. It suggests that H08 has the possibility to predict the groundwater level change associated with temporal groundwater pumping, although there are challenges in quantitatively deriving the relationship between the estimated values and groundwater level fluctuations.

 Locations of foundation leakage at the levee toe can be identified by the creation of landform classification maps. This map explains the processes of former river channels form, and classifies former river channels into old ones and new ones. Investigations on the soil compositions and permeability coefficients of former river channels and flood plains clarified that former river channels are associated with locations where water pressure is relatively high. With regard to how the locations of former river channels and levees are related to foundation leakage, it was found that foundation leakage with sand boiling tended to occur at levees situated about 200m off former river channels.