Subtropical climatic conditions can contribute to the death of the aerial parts of constructed wetland plants in winter. This presents a barrier to the widespread application of constructed wetland and is an issue that urgently needs to be solved. Three contrasting experiments, the plant-intercropping model (A), the warm-seasonal plant model (B), and the non-plant model (C), were studied in terms of their efficiency in removing pollutants, and the change in root structure of plants in the plant-intercropping model within the vertical-flow constructed wetlands. The results indicate that model A was able to solve the aforementioned problem. Overall, average removal rates of three pollutants (COD_Cr, total nitrogen (TN) and total phosphorous (TP)) using model A were significantly higher than those obtained using models B and C (P<0.01). Moreover, no significant differences in removal rates of the three pollutants were detected between A and B during the higher temperature part of the year (P>0.05). Conversely, removal rates of the three pollutants were found to be significantly higher using model A than those observed using model B during the lower temperature part of the year (P<0.01). Furthermore, the morphologies and internal structures of plant roots further demonstrate that numerous white roots, whose distribution in soil was generally shallow, extend further under model A. The roots of these aquatic plants have an aerenchyma structure composed of parenchyma cells, therefore, roots of the cold-seasonal plants with major growth advantages used in A were capable of creating a more favorable vertical-flow constructed wetlands media-microenvironment. In conclusion, the plant-intercropping model (A) is more suitable for use in the cold environment experienced by constructed wetland during winter.