Remote Sensing Application for Assessing Salinity Intrusion in the Mekong Delta, Vietnam

Abstract

Salinity intrusion is a complex issue in coastal areas. Currently, remote sensing techniques have been widely used to monitor water quality changes, ranging from inland river networks to deep oceans. The Vietnamese Mekong Delta (VMD) is an important rice-growing area and intrusion of saline water into irrigated freshwater- based agriculture areas is one of the most crucial constraints for agriculture development. This study aimed at building a numerical model to realize the salinity intrusion through the relationship between reflectance from the Landsat-8 Operational Land Imager (OLI) images and salinity levels measured in-situ. 103 observed samples were divided into 50% training and 50% test. The Multiple Linear Regression (MLR), Decision Trees (DTs) and Random Forest (RF) approaches were applied in the study. The result showed that the RF approach was the best model to estimate salinity along the coastal river network in the study area. However, the large samples size needed was a significant challenge to circumscribe the predicting ability of the RF models. The reflectance was found good to have a correlation with salinity when locations (latitude - longitude) of salinity measured station were added as a parameter of the Step-wise model. The R-square values were 77.48% in training and 74.16% in test while RMSE was smaller than 3. The reflectance - Location model was employed for mapping salinity intrusion on 24th Jan 2015 and 09th Feb 2015 recognized changes of salinity concentration in the whole study area. However, locality issue was a limitation for mapping salinity by using latitude and longitude as parameters. On the other hand, the real data was used for a re-sampling routine where data was performed re-sampling two times and four times by using bootstrap method. Four statistical models including the DTS, the MLR the RF and the Neural Network (ANN and ANT) were applied. Larger sample sizes that are regularly updated are needed to more fully develop the model. The best model was performed by the RF in re-sampling data four times which was employed for mapping salinity in early dry season 2015. The salinity map on 24th Jan and 09th Feb distinguished the tendency of salinity level as well as salinity dynamics and recognized changes of salinity concentration from upstream to downstream. This study proved the possibility of using the Landsat-8 images for mapping salinity as a useful tool to support the early warning system in the future in the VMD.