Flood Magnitude Prediction Model using Autoregressive and Integrated Moving Average
Domaine:
climateType de record:
papermodel
Créateur:
AdaEmmUkeAde
Éditeur:
Hôte:
Disaster is imminent and humans have to learn to manage it. Flood is one of the most common
natural disasters in Nigeria with adequate negative impact on people, properties and farmlands.
Flooding tends to occur during heavy down pour (rainfall) when water absorption is low. The
extent of flood is proportional to the negative impact that is experienced by the victims. The
adverse effects of flood cannot be overemphasized. It causes loss of life, properties and
economic degradation to mention but a few. These effects can be mitigated via Flood Prediction
Models (FPMs). Several successful Flood Prediction Models have been developed, but there are
still limitations in the prediction of the magnitude of the flood when it occurs. Flood magnitude
is the degree (extent) of the flood. The aim of this study is to build an efficient time series model
in forecasting flood magnitude as a natural phenomenon that occurs frequently in nature. In this
study, secondary data was collected and used for model training and testing. An autoregressive
integrated moving average (ARIMA) Time Series was used to develop a flood magnitude
prediction model. The ARIMA model is a combination of autoregressive (AR) and Moving
average (MA) models. The implementation was done using Python Spider IDE in ANACONDA
for the model simulation. The model was trained, successfully tested and evaluated via
coefficient of determination and standard error diagnostic tools. The ARIMA model prediction
results indicate that in the year 2024 the flood magnitude will be low level, while that of 2025
will be normal. The result also shows that the year 2026 will have high levels of flood. The
results show that the coefficient of AR technique produced a score of 0.0554 indicating a higher
and better performance than the MA model with a score value of -0.9077. The higher the
coefficient of determination value, the better and more successful the models’ performance.
Keywords: Disaster, Flood, Flood Magnitude, Machine Learning, Autoregressive, Moving Average, Time Series.
natural disasters in Nigeria with adequate negative impact on people, properties and farmlands.
Flooding tends to occur during heavy down pour (rainfall) when water absorption is low. The
extent of flood is proportional to the negative impact that is experienced by the victims. The
adverse effects of flood cannot be overemphasized. It causes loss of life, properties and
economic degradation to mention but a few. These effects can be mitigated via Flood Prediction
Models (FPMs). Several successful Flood Prediction Models have been developed, but there are
still limitations in the prediction of the magnitude of the flood when it occurs. Flood magnitude
is the degree (extent) of the flood. The aim of this study is to build an efficient time series model
in forecasting flood magnitude as a natural phenomenon that occurs frequently in nature. In this
study, secondary data was collected and used for model training and testing. An autoregressive
integrated moving average (ARIMA) Time Series was used to develop a flood magnitude
prediction model. The ARIMA model is a combination of autoregressive (AR) and Moving
average (MA) models. The implementation was done using Python Spider IDE in ANACONDA
for the model simulation. The model was trained, successfully tested and evaluated via
coefficient of determination and standard error diagnostic tools. The ARIMA model prediction
results indicate that in the year 2024 the flood magnitude will be low level, while that of 2025
will be normal. The result also shows that the year 2026 will have high levels of flood. The
results show that the coefficient of AR technique produced a score of 0.0554 indicating a higher
and better performance than the MA model with a score value of -0.9077. The higher the
coefficient of determination value, the better and more successful the models’ performance.
Keywords: Disaster, Flood, Flood Magnitude, Machine Learning, Autoregressive, Moving Average, Time Series.