Land use/land cover change (LULCC) in general and green space change in particular play an important role in flood control. Therefore, understanding the effects of LULCC and green space change is considered essential for sustainable land use planning, especially in the context of climate change and urban growth. The objective of this study is to evaluate the effects of green space dynamics on flood susceptibility using machine learning and remote sensing, namely the multilayer perceptron (MLP) and the recurrent neural network (RNN) in Vietnam’s Thanh Hoa Province. The novelty of this study lies in the integration of multi-temporal remote sensing data (2017 and 2024) with deep learning to evaluate the effects of green space dynamics on flood susceptibility, a problem that has not been sufficiently addressed in previous studies, which have mainly focused on single-period analysis. A total, 1164 flood inventory points and 14 conditioning factors in 2017 and 2024 were used as input data for the machine-learning model. Various statistical indices—namely root mean square error (RMSE), mean absolute error (MAE), area under the curve (AUC), and coefficient of determination (R2)—were applied to evaluate the computational performance of the MLP and the RNN. The results showed that the MLP model performed better than the RNN model with an R2 value of 0.812. The results also highlighted that the green space area decreased from about 71% to 64% from 2017 to 2024. Although the area of very high flood susceptibility area showed a decreasing trend, decreasing from 3042.192 km2 in 2017 to 1713.748 km2 in 2024, high flood susceptibility areas increased from 1,285 km2 in 2017 to 1,971 km2 in 2024. Moderate flood susceptibility areas also increased from 1,204 km2 in 2017 to 1382 km2 in 2024. The results indicate that about 100.91 km2 transitioned from low to high flood susceptibility. At the same time, the loss of forest-mainly converted to agricultural land (7756 km2) and barren land (1592 km2)—suggests a decline in green space quality, which reduces infiltration and increases flood. The methodological framework in this study can support the identification of priority urban areas and, thus, provide information to make decision-making on green strategies more effective. The results of this study serve as a guide for supporting decision-makers in sustainable urban development.

Effects of Green Space Dynamics on Flood Susceptibility in Thanh Hoa Province, Vietnam

期刊论文
发表日期：

2026年4月30日

查看全文

Huu Duy Nguyen,
Thi Sen Tran,
Van Hong Nguyen,
Quang-Thanh Bui

摘要

Land use/land cover change (LULCC) in general and green space change in particular play an important role in flood control. Therefore, understanding the effects of LULCC and green space change is considered essential for sustainable land use planning, especially in the context of climate change and urban growth. The objective of this study is to evaluate the effects of green space dynamics on flood susceptibility using machine learning and remote sensing, namely the multilayer perceptron (MLP) and the recurrent neural network (RNN) in Vietnam’s Thanh Hoa Province. The novelty of this study lies in the integration of multi-temporal remote sensing data (2017 and 2024) with deep learning to evaluate the effects of green space dynamics on flood susceptibility, a problem that has not been sufficiently addressed in previous studies, which have mainly focused on single-period analysis. A total, 1164 flood inventory points and 14 conditioning factors in 2017 and 2024 were used as input data for the machine-learning model. Various statistical indices—namely root mean square error (RMSE), mean absolute error (MAE), area under the curve (AUC), and coefficient of determination (R²)—were applied to evaluate the computational performance of the MLP and the RNN. The results showed that the MLP model performed better than the RNN model with an R² value of 0.812. The results also highlighted that the green space area decreased from about 71% to 64% from 2017 to 2024. Although the area of very high flood susceptibility area showed a decreasing trend, decreasing from 3042.192 km² in 2017 to 1713.748 km² in 2024, high flood susceptibility areas increased from 1,285 km² in 2017 to 1,971 km² in 2024. Moderate flood susceptibility areas also increased from 1,204 km² in 2017 to 1382 km² in 2024. The results indicate that about 100.91 km² transitioned from low to high flood susceptibility. At the same time, the loss of forest-mainly converted to agricultural land (7756 km²) and barren land (1592 km²)—suggests a decline in green space quality, which reduces infiltration and increases flood. The methodological framework in this study can support the identification of priority urban areas and, thus, provide information to make decision-making on green strategies more effective. The results of this study serve as a guide for supporting decision-makers in sustainable urban development.