Real-time flood forecasting by employing artificial neural network based model with zoning matching approach

Abstract

Flood forecasting models are a necessity, as they help in planning for flood events, and thus help prevent loss of lives and minimize damage. At present, artificial neural networks (ANN) have been successfully applied in river flow and water level forecasting studies. ANN requires historical data to develop a forecasting model. However, long-term historical water level data, such as hourly data, poses two crucial problems in data training. First is that the high volume of data slows the computation process. Second is that data training reaches its optimal performance within a few cycles of data training, due to there being a high volume of normal water level data in the data training, while the forecasting performance for high water level events is still poor. In this study, the zoning matching approach (ZMA) is used in ANN to accurately monitor flood events in real time by focusing the development of the forecasting model on high water level zones. ZMA is a trial and error approach, where several training datasets using high water level data are tested to find the best training dataset for forecasting high water level events. The advantage of ZMA is that relevant knowledge of water level patterns in historical records is used. Importantly, the forecasting model developed based on ZMA successfully achieves high accuracy forecasting results at 1 to 3 h ahead and satisfactory performance results at 6 h. Seven performance measures are adopted in this study to describe the accuracy and reliability of the forecasting model developed

Statistical downscaling and disaggregation for supporting regional climate change impact studies

“…As based on our own test (see Section 8.6.1), daily runoff prediction for long-term period for this study case is hardly reliable by using data-driven approaches. Similar findings are also reported by Sulaiman et al (2011) and Hassan et al (2012). The detailed descriptions for each individual component are provided in the followed sections.…”

“…As mentioned by the previous studies, the daily runoff prediction for a long-term period would notably underestimate the extreme data (Sulaiman et al, 2011;Hassan et al, 2012). To verify this for our own study, the result of simulating daily runoff directly using the BNN method is presented.…”

“…The major reason is that the forecasting performance of data-driven approaches (like ANN) for daily extreme data in a long-term time frame (e.g. more than thirty years) is relatively poor (Sulaiman et al, 2011;Hassan et al, 2012). Secondly, flood risk assessment is an important task for watersheds in central-south part of China as this region has suffered from serious flooding problems.…”

Effect of individual factors on the extraction of phenolic compounds from turmeric rhizomes (Curcuma longa) and moringa leaves (Moringa oleifera

“…The decrease in total polyphenol content in extracts of Curcuma longa rhizomes and Moringa oleifera leaves at extraction times of less than 30 min can be explained by the dissolution of these compounds in the extraction medium [33]. However, the drop in total polyphenol content at extraction times greater than 30 min could be explained by oxidation of the polyphenols due to their prolonged exposure to temperature and oxygen [34].…”