Comparative analysis and selection criteria for residence time distribution models in free water surface constructed wetlands

Abstract

The assessment of wetland performance requires consideration of flow characteristics. Although various systemic models (also known as network of ideal reactors) have been proposed to simulate residence time distribution (RTD) in wetlands, these models have only been tested with very few observations. In addition, the applicability of RTD models is affected by the shape characteristics of RTD curves. However, suitable indexes and quantitative relationships are absent to represent this effect. It remains unclear how to choose a suitable model when using the RTD modeling method to analyze tracer data. This study is designed to (i) test the accuracy of typical systemic models using sufficient observations and (ii) establish model selection criteria based on the shape characteristics of RTD. Firstly, the tanks in series (TIS) model, tanks in series plus delay (TISD) model, tanks in series with exchange (TISE) model, and tanks in series with delay and exchange (TISDE) model are compared using 36 sets of observations from tracer tests in two types of free water surface constructed wetlands (vegetated ditches and vegetated ponds). The results show that the widely used TIS model is unsatisfactory for more than half of the data, the TISD model is effective for rapidly rising RTD curves, the TISE model is preferred for long-tailed RTD curves, and the TISDE model can effectively fit all RTD curves except those with multiple peaks. Then, model selection criteria are established based on ∫trtpf(t)dt (the area for the rising stage of RTD), trtp (the degree of early peak), and tdtp (the degree of long tail), where tr represents the time when tracer concentration begins to rise, tp represents the time of peak concentration, and td represents the time when concentration decreases to 1/10 of peak concentration. The optimal model for an RTD curve can be identified based on the above three indexes and the corresponding critical conditions. Finally, the relationship between wetland design parameters and RTD shape indexes is examined. The results indicate that low hydraulic loading rates, small aspect ratios, and dense vegetation favor the creation of long tails. Large water depths tend to cause multiple peaks. Long-tailed and multi-peaked RTD curves are more likely to appear in ponds than ditches. Models that account for multiple nonideal flow phenomena are required to simulate water movement in ponds.