Performance evaluation of Auto-Regressive Integrated Moving Average models for forecasting saltwater intrusion into Mekong river estuaries of Vietnam

Abstract

The Mekong Delta is the most severely affected area by saltwater intrusion in Vietnam. Recent studies have focused on predicting this disaster with weekly and decade lead times without many seasonal forecasts, which is important for planning crop selection, crop structure, and sowing time. This study aims to forecast the spatial distribution of saltwater intrusion into the Mekong river estuaries of Vietnam during the dry season of 2021 by integrating Auto-Regressive Integrated Moving Average with Geographic Information System. ARIMA models were trained with a single input of water salinity measurements from 2012 to 2020. Compared to the weekly salinity observations in 2021, these models predicted very well in the My Tho and Ham Luong rivers but unsatisfactory performance in the Co Chien river. The deepest saltwater intrusion will occur between March 19th and April 16th of 2021, when the 4‰ saline front will move the farthest distance of 41,41 and 44 kilometers inland from the sea through My Tho, Ham Luong Co Chien rivers, respectively. The entire river system will be exposed to moderate risk of saltwater intrusion. Freshwater zones decreased significantly to 0.73% of the whole area of Ben Tre province. These findings could provide a valuable scientific foundation for the appropriate management of coastal aquifers to control or reduce saltwater intrusion.