We developed a machine learning-based system for detecting anomalies and quantifying the reliability of data recorded by an automatic in-situ observation system. Dissolved oxygen measurements in Osaka Bay, a hypoxic estuary in Japan, were decomposed by principal component analysis to reduce the dimensionality, thereby allowing 90% of the variance of the dissolved oxygen to be reconstructed by 10% of the total principal components. Then a gradient decision tree algorithm was applied to predict the principal components of dissolved oxygen, based on explanatory variables including meteorological factors, tide, and river discharge, thereby allowing efficient representation of dissolved oxygen fluctuation cycles with periods ranging from diurnal to annual. To evaluate the performance of the prediction model, cross-validation and comparison with a baseline linear model were conducted. Cross-validation yielded an RMSE of 1.14 mg L^–1, as compared to 1.61 mg L^–1 for the baseline. Spatially, the data observed at monitoring stations near river mouth were found to have larger errors than those of other coastal sites. This reliability evaluation system enables not only early detection of anomalies in the monitoring observation system, but also objective evaluation of the quality of data used for time-series analysis or data assimilation.

Annual patterns in mountain runoff in Japan and its long-term trends were analyzed from dam inflow observational data. In this study, inflow into dams, which is often found in mountainous areas of Japan, was regarded as runoff in mountainous areas. Cluster analysis based on dam inflow revealed that mountain runoff in Japan has distinct regional characteristics. For dams on the Japan Sea side, inflow was predominately during the snowmelt season. For dams on the Pacific Ocean side, inflow was predominately during the summer rainy season. Greater inter-annual variability in annual inflow was identified for dams on the Pacific Ocean side due to the substantial inflow variability in summer. An analysis of dam inflow trends found annual dam inflows showed no significant trends for the majority of dams. In contrast, inflow in March and May at many dams in snowy regions showed notable changes. The findings suggest that the total available water resources in mountainous regions in Japan have not seen significant changing trends, but that climate warming has already led to changes in the timing of dam inflows (i.e. mountain runoff) due to earlier snowmelt in snowy regions in Japan.

This study investigated the spatial and temporal patterns of surface water quality in the Greater Houston Area, Texas. It addressed two key questions: (1) How does population density at the county level influence water quality over time and across space? (2) Have ongoing pollution control efforts effectively maintained water quality over the past 30 years? The Water Quality Index (WQI) method was utilized to transform large quantities of water quality data collected at 20 stations during 1990–2020 into sets of indicative values to investigate spatiotemporal variabilities. Mann-Kendall and Sen’s slope were employed to detect long-term trends. Spatial analysis reveals that water quality is inversely correlated with population density, suggesting areas with higher population density are more prone to water quality degradation. Mann-Kendall trend analysis shows that water quality remained stable for 12 stations, with 6 stations showing an improving trend, possibly reflecting positive effects of ongoing pollution control measures. However, 2 stations in Harris County exhibit a declining trend, highlighting that pollution control has not been equally effective across all areas. These findings highlight the challenges in densely populated counties and underscore the necessity of sustained water pollution control and long-term monitoring efforts.

Dissolved iron (dFe) transported by rivers supports the growth of phytoplankton in coastal regions, yet the terrestrial sources of dFe remain poorly understood. To better understand the factors controlling the concentration of riverine dFe, we investigated the chemical composition (water temperature, pH, electrical conductivity, dissolved oxygen, dFe, dissolved manganese, and dissolved organic carbon) in 24 rivers in northern Hokkaido, Japan, during the autumn low-water season, and also analyzed the watershed characteristics (land cover, soil type, and lowland area percentage). The dFe concentrations ranged from 0.020 to 0.506 mg/L and were significantly correlated with dissolved organic carbon (r = 0.83, p < 0.01) and dissolved manganese (r = 0.60, p < 0.01), suggesting that dFe is exported as an organically complexed form from reducing environments. In watersheds dominated by pastureland, dFe concentrations were significantly correlated with a greater presence of reducing soils, such as peat soils and gray lowland soils (r = 0.81, p < 0.01). Our results indicate that reducing soils serve as a legacy source of iron to rivers even after conversion to pastureland, and their distribution primarily determines the potential dFe supply from watersheds in this region.

Projections of soil conditions using general circulation models (GCMs) and soil heat and water transport simulations are ongoing. The accuracy of GCM-based meteorological projections can be improved through site-specific bias correction using local meteorological observations. However, this approach is not feasible in areas lacking weather stations or with limited data. This study examines the impact of site-specific bias correction on future soil condition projections. Bias correction was applied to two GCMs (MIROC5 and MRI-CGCM3) in the plain and mountainous regions of Gifu Prefecture, Japan. Current (2012–2021) and future (2091–2100) soil temperature, volumetric water content, and matric potential were simulated with HYDRUS-1D using both corrected and uncorrected meteorological inputs. Results indicate that site-specific bias correction influenced future soil condition projections, with the magnitude of changes varying by location, GCM type, and soil variable. Maximum monthly variations at 10 cm depth reached 3.8°C, 0.01 m³ m^–3 and 257 cm for soil temperature, water content, and matric potential, respectively. However, when focusing on relative changes from current to future conditions, the impact of bias correction diminished. This suggests that, in data-scare regions, future soil conditions can be estimated without site-specific bias correction by focusing on relative changes.

Due to the increasing frequency and intensity of heavy rainfall events observed globally, the impact of climate change on critical heavy rainfall information, such as probable maximum precipitation (PMP), requires urgent update to ensure its utility in protecting lives and informing infrastructure design.

We applied the depth-area-duration (DAD) analysis on a 5-km d4PDF large ensemble dataset under historical and warming scenarios (+2K, +4K) with 720-year climate periods to investigate climate change impacts on PMP using the change factor approach. Annual Maximum Rainfall (AMR) in different durations over Northwest (NW) and Southeast (SE) Kyushu were analyzed. Depth-area relationships were derived and enveloped to estimate DAD-based PMP. Change factors were calculated as the ratio of warming scenario PMP to historical PMP.

Results indicate that AMR magnitude increases in NW Kyushu by approximately 10% at +2K and 20% at +4K across all durations. In SE Kyushu, it increases approximately 10% overall, with a larger increase (10–20%) above the 75th percentile. PMP change factors were higher in NW Kyushu (1.17 at +2K, 1.28 at +4K) compared to SE Kyushu (1.12 at +2K, 1.22 at +4K). These estimates highlight the need to address climate change impact on PMP used for flood risk management.

A U-shaped convolutional network (U-NET) based deep learning model is developed to predict weather fronts over Japan and the surrounding seas in summer (June, July, and August) using upscaled gridded frontal datasets from observations between 1980 and 2020 and reanalysis data corresponding to the same period. We justify the applicability of the deep learning model for predicting weather fronts in summer based on outputs from coarse-scale General Circulation Models (i.e. GCMs) from two perspectives. First, the coarse resolution of GCMs (e.g. 1.25 degrees) can capture the general morphological features of weather fronts. Second, models trained in a colder climate are applied to predict fronts in a warmer climate with some decrease in predicted peak frequency of fronts, but the general features of the spatial distribution of fronts can be represented by the deep-learning model predictions. By applying the deep learning models to predict weather fronts of multiple ensemble members from past and future climate experiments of GCMs, we can see that the locations of peak frequency tend to move slightly more southwesterly in a slant zone within the belt region between 25°N to 40°N as climate warms in the future.

Flooding is a persistent challenge in the Upper Citarum River Basin (UCRB) in Indonesia. It is exacerbated by rapid urbanization, deforestation, and climate change. We evaluated the effectiveness of two flood mitigation measures, river dredging and paddy field dams (PFDs), under varying land-use and climate change scenarios. A coupled 1D-2D model and a rainfall-runoff model simulated flood conditions for various scenarios, integrating land-use projections for 2050 and future return period rainfall. The results show that river dredging reduces inundation areas and volumes by up to 54% for lower return period floods, whereas the combination of dredging and PFDs achieves a 24% reduction in far-future scenarios. However, both measures exhibited reduced effectiveness under extreme hydrological conditions.

River dredging involves high operational costs and long-term environmental risks. Although PFDs are a low-cost alternative, their effectiveness depends on land availability, outlet design, and community participation. These findings emphasize the potential role of river dredging and PFDs in future changes. This study also demonstrates the critical role of incorporating future changes into infrastructure planning to enhance resilience and operational efficiency.

The trade-off problem in numerical simulations for flood forecasting is defined by an increased computational load with increasing resolution of the calculation, which improves accuracy. The Average Component Acceleration (ACA) method addresses this issue by proposing a new axis for acceleration calculation. In the ACA method, to compensate for the shortening of the time scale of the discharge hydrographs, the temporal change in water depth is divided into average and local components, amplifying only the average component. However, the ACA method has not yet been applied to several downstream conditions. This study proposes a novel evaluation method for the average component term and applies it to several downstream conditions with various calculation domain length to flood wavelength ratios. The accuracy of the previous ACA method decreased with the ratio of the channel domain length to the wavelength. The enhanced ACA method, employing a quadratic curve for the average component, efficiently reproduced the temporal variations in water surface distributions with the original analysis under several downstream conditions.