This study aims to examine the spatiotemporal variations of extreme rainfall events in Seoul, South Korea, utilizing subhourly rainfall data collected from an extensive observation network. Employing the Mann–Kendall trend detection test and Moran’s I index, this study analyzed temporal trends and spatial distributions in the extreme rainfall time series at 37 weather stations. The findings revealed that the annual maximum precipitation (AMP) series for various durations in Seoul generally exhibited a decreasing trend, with a marked decline observed for durations of 300 and 360 min. Furthermore, the study identified notable changes in the spatial patterns of AMP events, particularly for subhourly AMP events. The results indicate an intensification in the spatial concentration of AMP distribution. Consequently, despite the overall reduction in AMP, the likelihood of urban flooding may increase. This suggests that existing hydraulic design criteria, based on past concepts that temporal and spatial variability are constant, may no longer adequately represent extreme rainfall events. Therefore, revising these criteria to account for the evolving nature of extreme rainfall patterns is imperative.

This research assesses four Shared Socioeconomic Pathway (SSP) scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5) concerning precipitation quantiles across Korea, utilizing the CMIP6 multi-model ensemble comprising 23 General Circulation Models alongside observational data to project future changes. Precipitation quantiles, derived from regional frequency analysis conducted at 615 sites, are calculated as annual averages for the period from 2015 to 2024. Each SSP scenario is evaluated for its spatial distribution through the application of observational data and chi-square tests, with the results indicating that the SSP3-7.0 ensemble most accurately reflects the current quantile estimates derived from observational data. Furthermore, interannual precipitation quantiles are projected to extend to the year 2100 to discern long-term trends within each reproducible period. It is anticipated that precipitation associated with the 100-year reproducible period will increase by over 20% in most regions across the nation by the century’s end, with this increase becoming more pronounced in accordance with the severity of the pathway. These findings underscore significant increases in extreme rainfall events under high-emission scenarios and highlight the critical need for the integration of enhanced flood mitigation, water resource management, and climate adaptation strategies within Korea’s planning framework.

Reliable inflow forecasting represents a challenging and representative problem in long-horizon time series forecasting. Although long-term time series forecasting (LTSF) algorithms have shown strong performance in other domains, their applicability to hydrological inflow prediction has not yet been systematically assessed. Therefore, this study examined two LTSF linear models for inflow forecasting: NLinear and DLinear. LTSF models were trained with a 24 h input window and evaluated for 24 h lead times at eight major dams in South Korea. Long Short-Term Memory (LSTM) network and eXtreme Gradient Boosting (XGBoost) were employed as a conventional AI model. LSTM consistently achieved the highest coefficient of determination (R²) and the lowest normalized root mean square error, DLinear minimized normalized mean square error, and NLinear delivered superior hydrological consistency as measured by Kling–Gupta efficiency. XGBoost showed comparatively larger variability across sites. Spatial heterogeneity was evident; sites were grouped into high-performing, transition, and vulnerable groups. Peak-flow analysis revealed amplitude attenuation and phase lag at longer horizons.

Recent climate change has been linked to more frequent and more intense short-timescale rainfall extremes, increasing exposure to urban pluvial flooding. Because many urban catchments respond within minutes, rainfall information at sub-hourly resolution is often needed for hydrologic analyses. An AI-driven temporal downscaling approach is introduced here to derive 10-minute rainfall series from hourly observations using a conditional diffusion generative model. Rain-gauge observations at Seoul Gwanaksan (#1917), operated by the Korea Forest Service, were used. The record covers the years 2015 through 2024. Paired hourly totals and observed 10-minute series were prepared to examine whether sub-hourly rainfall sequences can be reconstructed from hourly totals while preserving realistic within-hour variability. The feasibility of loss function variation was investigated. The experiments indicate that incorporating distributional and temporal statistics into the objective function can enhance the realism of sub-hourly rainfall structure under hourly constraints. The proposed framework is expected to provide more reliable 10-minute rainfall inputs for urban hydrologic analyses and pluvial-flood–relevant applications in rapid-response catchments.

For long-term time series forecasting of hydro-meteorological variables, physical models and artificial intelligence (AI)-based models have been used. Long-term forecast using physical models may have a limited predictive performance due to assumptions and conditions used in the physical model. Although AI-based models for long-term forecast of hydro-meteorological variables have a restricted capacity to explain phenomena, they have practical strengths such as high precision and short computation time. Zeng et al. (2022) proposed Long-Term Time Series Forecasting (LTSF) models and showed that they outperformed transformer-based AI models for long-term forecast In this study, a long-term forecasting model was developed using the LTSF model applied to dam inflow data and assess the feasibility of LTSF in the long-term forecast of inflow data. For comparison, The Long Short Term Memory (LSTM) algorithm was employed. The results show that the LTSF showed a comparable performance of long-term forecast to the LSTM although the structures of LTSF models are much simpler than LSTM. The LSTF models can be considered as a good alternative of LSTM when the forecast task need prompt computation.Zeng, A., Chen, M., Zhang, L., & Xu, Q. (2022). Are Transformers Effective for Time Series Forecasting? The Chinese University of Hong Kong and International Digital Economy Academy (IDEA).