This study performed in Sapanca Lake catchment area, used as a drinking water resource. Two highways located at northern and southern shores, and a railway at its south are significant sources of pollution. As a possible solution for protecting water quality a pocket wetland constructed and operated. Performances statistically interpreted by Spearman's Correlation test and univariate analysis of variance on collected data. The mean removal efficiencies obtaited were 52% (TSS), 4% (Nitrate), 26% (TN), -5% (TOC), 63% (TP), 4.5% (Chloride), 3% (Sulfate), 33% (Cr), 39% (Co), -19.5% (Ni), 7% (Cu), 55% (Zn), 36% (As), 38% (Cd) and 18% (Pb). TSS removal was in positive significant medium correlation with Co, Cu, Zn, and Pb removal respectively (p < 0.05). Other statistically significant positive high correlations calculated between removal efficiency of Nitrate-TN, Chloride-Sulfate, Cr-Co-Cu-As-Cd. According to ANOVA and Kruskal-Wallis test results, removal efficiencies of TSS and TOC partially affected by different temperature (p < 0.1 for TSS and p < 0.05 for TOC) and pH ranges (p < 0.1 for both removal efficiencies), TP removal efficiency significantly affected by different pH ranges (p < 0.001), and Chloride and Sulfate removal efficiencies were significantly (p < 0.001) affected by different temperature ranges. Regardless of geographical location and climatic factors, pocket wetland systems can be relied upon for minimizing heavy metals such as Cr, Co, Zn, As, Cd and Pb and critical pollutants such as TP and TSS caused by highway runoff. (C) 2018 Elsevier Ltd. All rights reserved.