scholarly journals Online multistep-ahead inundation depth forecasts by recurrent NARX networks

2013 ◽  
Vol 17 (3) ◽  
pp. 935-945 ◽  
Author(s):  
H.-Y. Shen ◽  
L.-C. Chang
Keyword(s):  
Nonlinear Dynamical Systems ◽  
Flood Inundation ◽  
Forecast Period ◽  
Nonlinear Dynamical ◽  
Inundation Area ◽  
Inundation Depth ◽  
Dimensional Simulation ◽  
Online Feedback ◽  
Hydrological Fields ◽  
Nonlinear Autoregressive

Abstract. Various types of artificial neural networks (ANNs) have been successfully applied in hydrological fields, but relatively scant on multistep-ahead flood inundation forecasting, which is very difficult to achieve, especially when dealing with forecasts without regular observed data. This study proposes a recurrent configuration of nonlinear autoregressive with exogenous inputs (NARX) network, called R-NARX, to forecast multistep-ahead inundation depths in an inundation area. The proposed R-NARX is constructed based on the recurrent neural network (RNN), which is commonly used for modeling nonlinear dynamical systems. The models were trained and tested based on a large number of inundation data generated by a well validated two-dimensional simulation model at thirteen inundation-prone sites in Yilan County, Taiwan. We demonstrate that the R-NARX model can effectively inhibit error growth and accumulation when being applied to online multistep-ahead inundation forecasts over a long lasting forecast period. For comparison, a feedforward time-delay and an online feedback configuration of NARX networks (T-NARX and O-NARX) were performed. The results show that (1) T-NARX networks cannot make online forecasts due to unavailable inputs in the constructed networks even though they provide the best performances for reference only; and (2) R-NARX networks consistently outperform O-NARX networks and can be adequately applied to online multistep-ahead forecasts of inundation depths in the study area during typhoon events.

scholarly journals On-line multistep-ahead inundation depth forecasts by recurrent NARX networks

2012 ◽  
Vol 9 (10) ◽  
pp. 11999-12028 ◽  
Author(s):  
H.-Y. Shen ◽  
L.-C. Chang
Keyword(s):  
Nonlinear Dynamical Systems ◽  
Flood Inundation ◽  
Forecast Period ◽  
Nonlinear Dynamical ◽  
Inundation Area ◽  
Inundation Depth ◽  
On Line ◽  
Dimensional Simulation ◽  
Hydrological Fields ◽  
Nonlinear Autoregressive

Abstract. Various types of artificial neural networks (ANNs) have been successfully applied in hydrological fields, but relatively scant on flood inundation forecast. This study proposes a recurrent configuration of nonlinear autoregressive with exogenous inputs (NARX) network, called R-NARX, to forecast multistep-ahead inundation depths in an inundation area. The proposed R-NARX is constructed based on the recurrent neural network (RNN), which is commonly used for modeling nonlinear dynamical systems. The models were trained and tested based on a large number of inundation data generated by a well validated two-dimensional simulation model at thirteen inundation-prone sites in Yilan County, Taiwan. We demonstrate that the R-NARX model can effectively inhibit error growth and accumulation when being applied to on-line multistep-ahead inundation forecasts over a long lasting forecast period. For comparison, a feedforward time-delay and an on-line feedback configuration of NARX networks (T-NARX and O-NARX) were performed. The results show that (1) T-NARX networks cannot make on-line forecasts due to unavailable inputs in the constructed networks even though they provide the best performances for reference only; and (2) R-NARX networks consistently outperform O-NARX networks and can be adequately applied to on-line multistep-ahead forecasts of inundation depths in the study area during typhoon events.

scholarly journals Flood Inundation Assessment in the Low-Lying River Basin Considering Extreme Rainfall Impacts and Topographic Vulnerability

Water ◽  
2021 ◽  
Vol 13 (7) ◽  
pp. 896
Author(s):  
Thanh Thu Nguyen ◽  
Makoto Nakatsugawa ◽  
Tomohito J. Yamada ◽  
Tsuyoshi Hoshino
Keyword(s):  
River Basin ◽  
Extreme Rainfall ◽  
Flood Damage ◽  
Water Levels ◽  
Flood Inundation ◽  
Inundation Area ◽  
Inundation Depth ◽  
Study Results ◽  
Inundation Duration ◽  
Short Time

This study aims to evaluate the change in flood inundation in the Chitose River basin (CRB), a tributary of the Ishikari River, considering the extreme rainfall impacts and topographic vulnerability. The changing impacts were assessed using a large-ensemble rainfall dataset with a high resolution of 5 km (d4PDF) as input data for the rainfall–runoff–inundation (RRI) model. Additionally, the prediction of time differences between the peak discharge in the Chitose River and peak water levels at the confluence point intersecting the Ishikari River were improved compared to the previous study. Results indicate that due to climatic changes, extreme river floods are expected to increase by 21–24% in the Ishikari River basin (IRB), while flood inundation is expected to be severe and higher in the CRB, with increases of 24.5, 46.5, and 13.8% for the inundation area, inundation volume, and peak inundation depth, respectively. Flood inundation is likely to occur in the CRB downstream area with a frequency of 90–100%. Additionally, the inundation duration is expected to increase by 5–10 h here. Moreover, the short time difference (0–10 h) is predicted to increase significantly in the CRB. This study provides useful information for policymakers to mitigate flood damage in vulnerable areas.

Development of flood inundation area GIS database for Samsung-1 drainage sector, Seoul, Korea

2016 ◽  
Vol 49 (12) ◽  
pp. 981-993 ◽  
Author(s):  
Minkwan Oh ◽  
Dongryul Lee ◽  
Hyunhan Kwon ◽  
Dongkyun Kim
Keyword(s):  
Flood Inundation ◽  
Inundation Area ◽  
Gis Database

Linear and Nonlinear Dynamic Behaviour

2018 ◽  
Author(s):  
Ray Huffaker ◽  
Marco Bittelli ◽  
Rodolfo Rosa
Keyword(s):  
Dynamic Behavior ◽  
Dynamic Systems ◽  
Dynamic Behaviour ◽  
Nonlinear Dynamical Systems ◽  
Logistic Map ◽  
Nonlinear Dynamical ◽  
Deterministic Dynamic ◽  
Random Behavior ◽  
Single Solution ◽  
Equivalent Systems

In this chapter, we describe how highly erratic dynamic behavior can arise from a nonlinear logistic map, and how this apparently random behavior is governed by a surprising order. With this lesson in mind, we should not be overly surprised that highly erratic and random appearing observed data might also be generated by parsimonious deterministic dynamic systems. At a minimum, we contend that researchers should apply NLTS to test for this possibility. We also introduced tools to analyze dynamic behavior that form the foundation for NLTS. In particular, we have stressed the quite unexpected capability to achieve some form of predictability even with only one trajectory at hand. In subsequent chapters, we treat known nonlinear dynamical systems as unknown, and investigate how NLTS methods rely on a single solution (or multiple solutions) generated by them to reconstruct equivalent systems. This is a conventional approach in the literature for seeing how NLTS methods work since we know what needs to be reconstructed.

Sign in / Sign up

Export Citation Format

Share Document