Search Results (4)

This paper advances machine learning (ML)-based streamflow prediction by strategically selecting rainfall events, introducing a new loss function, and addressing rainfall forecast uncertainties. Focusing on the Iowa River Basin, we applied the stochastic storm transposition (SST) method to create realistic rainfall events, which were input into a hydrological model to generate corresponding streamflow data for training and testing deterministic and probabilistic ML models. Long short-term memory (LSTM) networks were employed to predict streamflow up to 12 h ahead. An active learning approach was used to identify the most informative rainfall events, reducing data generation effort. Additionally, we introduced a novel asymmetric peak loss function to improve peak streamflow prediction accuracy. Incorporating rainfall forecast uncertainties, our probabilistic LSTM model provided uncertainty quantification for streamflow predictions. Performance evaluation using different metrics improved the accuracy and reliability of our models. These contributions enhance flood forecasting and decision-making while significantly reducing computational time and costs. Full article

Given the fact that the high frequency of extreme weather events globally, in particular typhoons, has more of an influence on flood forecasting, there is a great need to further understand the impact of typhoon events on design storms. The main objectives of this paper are to examine the magnitude, occurrence, and mechanism of typhoon events in southeast coastal China and their contribution to the design storm study. We take Shanghai, which is a typical metropolitan region in the Yangtze River Delta, China, as an example. The impact of typhoons on the rainfall frequency analysis is quantitatively evaluated using stochastic storm transposition (SST)-based intensity–duration–frequency (IDF) estimates with various temporal and spatial structures under different return periods. The results show that there is significant variability in the storm magnitude within the transposition domain across different durations, highlighting the spatiotemporal heterogeneity over the coastal area. Moreover, the probability of random storm transposition exhibits an uneven distribution. The frequency of typhoon rainfall events within the transposition domain is notably high, and there is considerable variability in the structure of rainfall. Typhoon rainfall amplifies the intensity of design storms, and its contribution increases with return periods. The variability in design storms increases accordingly. Based on the advantages of SST, which retains the spatiotemporal structure of the rainfall in the generated scenarios, the overall framework provides an effective way to examine the impact of diverse characteristics of typhoon rainfall on frequency analysis and facilitate a deeper exploration of the direct impact of various types of extreme storms on the intensity, spatial, and temporal distributions of design storms amidst evolving environmental conditions over this metropolitan region. Full article

Nonstationary is one of the prominent phenomena in the current hydrological time series due to climate change and urban expansion. In this study, using the long time series rainfall data from rain gauges and satellite rainfall data, the trend and abrupt change of rainfall in the Taihu Lake basin, China, are examined by the Mann–Kendall (MK) test and the Pettitt test, using rain gague data. For seven water conservancy zones in this basin, the intensity–duration–frequency curves (IDFs) are obtained using satellite rainfall and the stochastic storm transposition (SST) method, providing a method for rainfall frequency analysis based on nonstationary assumption. The IDFs results between the conventional frequency analysis method with the stationary assumption and the SST-based method are compared. The results show an overall increasing trend of annual total rainfall in the Taihu Lake Basin, with significant changes at most stations. The SST-based results show a significant difference of IDFs in seven conservancy zones, which are linked to nonstationary changes in rainfall series. Our results provide an important reference for understanding the nonstationary changes and nonstationary frequency analysis of extreme rainfall in the Taihu Lake basin. Full article

Frequency analysis of rainfall data is essential in the design and modelling of hydrological systems but is often statistically limited by the total observation period. With advances in weather radar technology, frequency analysis of areal rainfall data is possible at a higher spatial resolution. Still, the observation periods are short relative to established rain gauge networks. A stochastic framework, “stochastic storm transposition” shows great promise in recreating rainfall statistics from radar rainfall products, similar to rain gauge-derived statistics. This study estimates intensity–duration–frequency (IDF) relationships at both point and urban catchment scales. We use the stochastic storm transposition framework and a single high-resolution, 17-year long (however, discontinuous), radar rainfall dataset. The IDF relations are directly compared to rain gauge statistics with more than 40 years of observation, and rainfall extremes derived from the original, and untransposed, radar dataset. An overall agreement is discovered, however, with some discrepancies in short-duration storms due to scaling errors between gauge and radar. Full article