Therefore, this study aimed to analyze the spatio-temporal dynamics of suspended sediment in the 4.2 km long confluence zone of Tisza and Maros Rivers using Sentinel-2 images to support the sediment sampling in the future. Altogether 143 Sentinel-2 images were analyzed covering the period between June 2015 to May 2021. Throughout the classification of the images the K-means clustering, which is considered as one of the most popular unsupervised classification techniques, was applied to demonstrate the spatial distribution of surficial suspended sediment dynamism in the confluence zone. The pixels were classified into three classes, representing the water of the Tisza (TW), the water of the Maros (MW) and their mixture (MIX). The areal distribution of these classes (the number of pixels) was compared to the hydrological parameters of the two rivers (e.g., water levels, slopes and discharges).

The results showed that the highest Pearson correlation (r) existed between the number of TW and MIX pixels and water slope, and slope difference between the two rivers (r= -0.65 and -0.64 respectively). On the other hand, the area of the Maros water (MW) pixels showed less sensitivity to slope and higher sensitivity to discharge (r=0.5). The derived correlation graphs and the cluster analysis suggest that during low and medium discharges (QTisza ≤ 747 m3/s and QMaros ≤ 330 m3/s) the imaginary regression line has a lower slope compared to higher or flood discharges. Therefore, separating data into two clusters (Cluster I: Low and medium dischargers and Cluster II: High discharges) would produce better predictive models.

Predictive models for the areal distribution of TW, MW and the longitudinal extend of Maros into the Tisza were produced considering the two data clusters and the estimated coefficient of determination (R2) were 0.61, 0.61 and 0.82 for Cluster I., and 0.6, 0.7 and 0.64 for Cluster II. Based on the data, the shape of mixing in the confluence zone could be categorized into four groups: (1) MW covers the whole width of the Tisza in the confluence; (2) MW could be identified till the end of the study area; (3) MW covers a limited area in the Tisza; and (4) unusual mixing shapes. The histograms show that Group 2 and Group 3 have the highest probabilities (52.3%) and (52.6%) in clusters I and II respectively. Based on these results it is recommended to sample sediments from both shores of the Tisza in the confluence zone paying attention to the actual mixing status, for representative sampling mission.