Hydrological Drought Forecasting Using Modified Surface Water Supply Index (SWSI) and Streamflow Drought Index (SDI) in Conjunction with Artificial Neural Networks (ANNs)

2019 ◽  
Vol 10 (4) ◽  
pp. 39-57
Author(s):  
Raphael M. Wambua
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Surface Water ◽  
Water Supply ◽  
Lead Time ◽  
Drought Index ◽  
Hydrological Drought ◽  
Lead Times ◽  
Artificial Neural ◽  
Surface Water Supply

Hydrological drought in upper Tana River basin adversely affects water resources. In this study, a hydrological drought was forecasted using a Surface Water Supply Index (SWSI), a Streamflow Drought Index (SDI) and an Artificial Neural Networks (ANNs). The best SWSI involved combinations of rainfall and the index values integrated into ANNs. The best forecasts with SDI entailed composite functions of rainfall, stream flow and SDI. Different ANN models for both SWSI and SDI with lead times of 1 to 24 months were tested at hydrometric stations. Results show that the forecasting ability of all the networks decreased with the increase in lead-time. The best ANNs with specific architecture performed differently based on forecasting lead-time. SWSI drought forecasts were better than those of the SDI for all lead-times. The SWSI and SDI depicted R values of 0.752 and 0.732 for station 4AB05 for one-month lead-time. The findings are useful as an effective hydrological-drought early warning for viable mitigation and preparedness approaches to minimize the negative effects of drought.

scholarly journals The Probabilistic Drought Forecast Based on the Ensemble Technique Using the Korean Surface Water Supply Index

2017 ◽  
Author(s):  
Suk Hwan Jang ◽  
Jae-Kyoung Lee ◽  
Ji Hwan Oh ◽  
Jun Won Jo ◽  
Younghyun Cho
Keyword(s):  
Surface Water ◽  
Water Supply ◽  
Drought Index ◽  
Probability Distributions ◽  
Hydrological Drought ◽  
Normal Distributions ◽  
Ensemble Technique ◽  
Flood Season ◽  
Selection Of ◽  
Surface Water Supply

Abstract. This study proposes the new hydrological drought index, Korean Surface Water Supply Index (KSWSI), which overcomes some of the limitations in the calculation of previous SWSI applied in Korea and conducts the probabilistic drought forecasts using KSWSI. In this study, all hydrometeorological variables in the Geum River basin were investigated and appropriate variables were selected as KSWSI components for each sub-basin. And whereby only the normal distributions are applied to all drought components, probability distributions suitable for each KSWSI component were estimated. As a result of verifying KSWSIs, the accuracy of KSWSIs showed better drought phenomenon in drought events. The monthly probabilistic drought forecasts were also calculated based on ensemble technique using KSWSI. In 2006 and 2014 drought events, the accuracy of the drought forecasts using KSWSIs were higher than those using previous SWSI, demonstrating that KSWSI is able to enhance the accuracy of drought forecasts. The influence of expanding hydrometeorological components and selecting appropriate probability distributions for each KSWSI component were analyzed. It is confirmed that the accuracy of KSWSIs may be affected by the choice of hydrometerological components, the station data obtained, the length of used data for each station, and the probability distributions selected. Furthermore, the uncertainty quantification of the KSWSI calculation procedure was also carried out using the Maximum Entropy (ME) theory. The large MEs and standard deviations of KSWSIs in the flood season cause uncertainties, implying that the selection of the appropriate probability distributions for selected drought components in the flood season is very important.

Drought projection based on a hybrid drought index using Artificial Neural Networks

2014 ◽  
Vol 29 (11) ◽  
pp. 2635-2648 ◽  
Author(s):  
Tao Yang ◽  
Xudong Zhou ◽  
Zhongbo Yu ◽  
Valentina Krysanova ◽  
Bo Wang
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Drought Index ◽  
Artificial Neural

scholarly journals Wavelets-based non-linear model for real-time daily flow forecasting in Krishna River

2013 ◽  
Vol 15 (3) ◽  
pp. 1022-1041 ◽  
Author(s):  
R. Maheswaran ◽  
Rakesh Khosa
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Linear Model ◽  
Volterra Model ◽  
Lead Times ◽  
Discrete Wavelet ◽  
Promising Alternative ◽  
Daily Streamflow ◽  
Non Linear ◽  
Artificial Neural

In this study, a multi-scale non-linear model based on coupling a discrete wavelet transform (DWT) and the second-order Volterra model, i.e. the wavelet Volterra coupled (WVC) model, is applied for daily inflow forecasting at Krishna Agraharam, Krishna River, India. The relative performance of the WVC model was compared with regular artificial neural networks (ANN), wavelet-artificial neural networks (WA-ANN) models and other baseline models such as auto-regressive moving average with exogenous variables (ARMAX) for lead times of 1–5 days. The models were applied for the forecasting of daily streamflow at Krishna Agraharam Station at Krishna River. The WVC performed very well, especially when compared with the WA-ANN model for lead times of 4 and 5 days. The results indicate that the WVC model is a promising alternative to the other traditional models for short-term flow forecasting.

Sign in / Sign up

Export Citation Format

Share Document