FORECASTING OF CHANGES IN SALINITY INTRUSION IN THE VIETNAMESE MEKONG DELTA BY THE COMBINED MODEL OF LSTM (Long Short-Term Memory) AND SRM (Sinusoidal Regression Model)

  • Uyen T. Huynh
Keywords: Mekong Delta, Salinity intrusion, LSTM and SRM models

Abstract

The salinity intrusion in the Vietnamese Mekong Delta (VMD) has become

more complex and temporally heterogeneous. This could seriously threaten the

livelihoods of local residents and agricultural activities. Therefore, the research

was conducted by using a combined model of LSTM (Long Short-Term Mem-

ory) and SRM (Sinusoidal Regression Model) to assess the trends and anomalies

of salinity intrusion, with a series of data collected from main stations in the VMD

in the year of 2021. The findings showed that the combined model exhibited high

predictive (R2 = 0.9299, M SE = 2.0861, and M AP E = 0.1276) in fore-

casting the increasing and decreasing trends of salinity intrusion and effectively

detecting anomalous variations. Consequently, these results could be helpful to

policymakers in predicting and responding to future salinity intrusion and to likely

widespread implications for other regions impacted by saline intrusion.

Published
2025-07-28