Flood forecasting for Melaka using arima and nar modelling methods
Flooding is an annual occurring incident in Malaysia. Several states in Malaysia are strongly affected by the flooding including Melaka where the flash floods are a common occurence. A flash flood is challenging to forecast and requires a sophisticated algorithm and system compared to the seasonal f...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English English |
Published: |
2022
|
Subjects: | |
Online Access: | http://eprints.utem.edu.my/id/eprint/26975/1/Flood%20forecasting%20for%20Melaka%20using%20arima%20and%20nar%20modelling%20methods.pdf http://eprints.utem.edu.my/id/eprint/26975/2/Flood%20forecasting%20for%20Melaka%20using%20arima%20and%20nar%20modelling%20methods.pdf http://eprints.utem.edu.my/id/eprint/26975/ https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=122222 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Flooding is an annual occurring incident in Malaysia. Several states in Malaysia are strongly affected by the flooding including Melaka where the flash floods are a common occurence. A flash flood is challenging to forecast and requires a sophisticated algorithm and system compared to the seasonal flood. It is difficult to forecast the flash flood compared to the seasonal flood. In Melaka, flash flood occurs regularly and it can happen to rise and fall in pace. This is the reason that flash floods can cause more damage than the seasonal floods. This study aims to develop a flood monitoring system to provide real-time data for the flood forecast. The objective is to develop a flood forecast model by analysing the flood parameters on a specific geographical layout which is Pengkalan Rama Jetty, Melaka. Following this, the efficiency of the flood forecast model is evaluated to forecast the water level where two flood forecast models were studied in this research which are the Autoregressive Integrated Moving Average (ARIMA) and Nonlinear Autoregressive Neural Network (NAR). The water level data considered for both methods were taken from 1st July 2020 at 12:00 am until 30th July 2020 at 7.00pm . There was a total of 2782 data in this time-series. The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) were used to find the best ARIMA model. The second method using NAR as a flood forecast model. This research used the time series data in NAR training, validation and testing to forecast the flash flood. In this research, the model was set to forecast the water level in several hourly time period of 1, 3, 5, and 7 hours. The forecast accuracy were measured using the Pearson R and R-squared to find the most accurate model for this multiple time-step ahead. The model’s accuracy was determined by comparing the original and forecasted time series using Pearson R, R-Squared, Root Mean Squared Error (RMSE), Mean Squared Error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). The result of the flood forecast system were compared with 7 hours forecast ahead and it was found that the ARIMA (2, 1, 3) was the best model for the Pengkalan Rama, Jetty, with an AIC of 5653.7004 and a BIC of 5695.209. The model also produced a lead forecast of up to 7 hours for the time series. Meanwhile, the result showed that the NAR model outperforms ARIMA with the lowest value in terms of RMSE, MSE, MAE and MAPE which are 1.915715, 3.669963, 1.576785 and 1.785951 respectively. In terms of Pearson R and R-Squared, the NAR model achieved Pearson R value of 0.931505 and R-Squared was 86.77024% compared to ARIMA which achieved R's value of -0.73993 and R-Squared of 54.74961%. It can be concluded that the flood forecast model for 7 hours ahead of using NAR outperformed ARIMA and is suitable for use in the flood forecast system at Pengkalan Rama Jetty. |
---|