Considering the high random and non-static property of the rainfall-runoff process, lots of models are being developed in order to learn about such a complex phenomenon. Recently, Machine learning techniques such as the Artificial Neural Network (ANN) and other networks have been extensively used by hydrologists for rainfall-runoff modelling as well as for other fields of hydrology. However, deep learning methods such as the state-of-the-art for LSTM networks are little studied in hydrological sequence time-series predictions. We deployed ANN and LSTM network models for simulating the rainfall-runoff process based on flood events from 1971 to 2013 in Fen River basin monitored through 14 rainfall stations and one hydrologic station in the catchment. The experimental data were from 98 rainfall-runoff events in this period. In between 86 rainfall-runoff events were used as training set, and the rest were used as test set. The results show that the two networks are all suitable for rainfall-runoff models and better than conceptual and physical based models. LSTM models outperform the ANN models with the values of R 2 and N S E beyond 0.9, respectively. Considering different lead time modelling the LSTM model is also more stable than ANN model holding better simulation performance. The special units of forget gate makes LSTM model better simulation and more intelligent than ANN model. In this study, we want to propose new data-driven methods for flood forecasting.
Abstract:In China, 9Ð5% of the landmass is karst terrain and of that 47,000 km 2 is located in semiarid regions. In these regions the karst aquifers feed many large karst springs within basins of thousands of square kilometres. Spring discharges reflect the fluctuation of ground water level and variability of ground water storage in the basins. However, karst aquifers are highly heterogeneous and monitoring data are sparse in these regions. Therefore, for sustainable utilization and conservation of karst ground water it is necessary to simulate the spring flows to acquire better understanding of karst hydrological processes. The purpose of this study is to develop a parsimonious model that accurately simulates spring discharges using an artificial neural network (ANN) model. The karst spring aquifer was treated as a non-linear input/output system to simulate the response of karst spring flow to precipitation and applied the model to the Niangziguan Springs, located in the east of Shanxi Province, China and a representative of karst springs in a semiarid area. Moreover, the ANN model was compared with a previous time-lag linear model and it was found that the ANN model performed better.
The Xinanjiang model has been widely used in the humid regions in southern China as a basic tool for rainfall–runoff simulation, flood forecasting and water resources planning and management. However, its performance in the arid and semi-arid regions of northern China is usually not so good as in the humid regions. A modified Xinanjiang model, in which runoff generation in the watershed is based on both infiltration excess and saturation excess runoff mechanisms, is presented and discussed. Three different watersheds are selected for assessing and comparing the performance of the Xinanjiang model, the modified Xinanjiang model, the VIC model and the TOPMODEL in rainfall–runoff simulation. It is found that the modified Xinanjiang model performs better than the Xinanjiang model, and the models considering the Horton and Dunne runoff generation mechanisms are slightly better than those models considering the single runoff generation mechanism in semi-arid areas. It is suggested that the infiltration excess runoff mechanism should be included in rainfall–runoff models in arid and semi-arid regions.
Runoff reduction in most river basins in China has become a hotpot in recent years. The Gushanchuan river, a primary tributary of the middle Yellow river, Northern China, showed a significant downward trend in the last century. Little is known regarding the relative contributions of changing environment to the observed hydrological trends and response on the runoff generation process in its watershed. On the basis of observed hydrological and meteorological data from 1965–2010, the Mann-Kendall trend test and climate elasticity method were used to distinguish the effects of climate change and human activities on runoff in the Gushanchuan basin. The results indicate that the runoff in the Gushanchuan Basin has experienced significant declines as large as 77% from 1965 to 2010, and a mutation point occurred around 1997; the contribution rate of climate change to runoff change is 12.9–15.1%, and the contribution rate of human activities to runoff change is 84.9–87.1%. Then we divided long-term data sequence into two stages around the mutation point, and analyzed runoff generation mechanisms based on land use and cover changes (LUCC). We found that the floods in the Gushanchuan Basin were still dominated by Excess-infiltration runoff, but the proportion in 1965–1997 and 1998–2010 decreased gradually (68.46% and 45.83% in turn). The proportion of Excess-storage runoff and Mixed runoff has increased, which means that the runoff is made up of more runoff components. The variation law of the LUCC indicates that the forest area increased by 49.61%, the confluence time increased by 50.42%, and the water storage capacity of the watershed increased by 30.35%.
Continual learning has gained increasing attention in recent years, thanks to its biological interpretation and efficiency in many realworld applications. As a typical task of continual learning, continual relation extraction (CRE) aims to extract relations between entities from texts, where the samples of different relations are delivered into the model continuously. Some previous works have proved that storing typical samples of old relations in memory can help the model keep a stable understanding of old relations and avoid forgetting them. However, most methods heavily depend on the memory size in that they simply replay these memorized samples in subsequent tasks. To fully utilize memorized samples, in this paper, we employ relation prototype to extract useful information of each relation. Specifically, the prototype embedding for a specific relation is computed based on memorized samples of this relation, which is collected by K-means algorithm. The prototypes of all observed relations at current learning stage are used to re-initialize a memory network to refine subsequent sample embeddings, which ensures the model's stable understanding on all observed relations when learning a new task. Compared with previous CRE models, our model utilizes the memory information sufficiently and efficiently, resulting in enhanced CRE performance. Our experiments show that the proposed model outperforms the state-of-the-art CRE models and has great advantage in avoiding catastrophic forgetting. The code and datasets have been released on https://github.com/fd2014cl/RP-CRE.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.