Urban Stormwater: First Flush Analysis and Treatment by an Undersized Constructed Wetland
TUCKER, ROBERT S. Urban Stormwater: First Flush Analysis and Treatment by an Undersized Constructed Wetland. (Under the direction of Jean Spooner and William F. Hunt.)
Nonpoint stormwater runoff remains a leading threat to surface water quality in the U.S. Increased impervious surfaces, climate change, and increasing water demands put even more pressure on stormwater managers to improve stormwater management practices with regards to cost effectiveness, removal performance, and ecological sustainability. More effective BMPs can be designed by understanding the nature of pollutant runoff loads with respect to the hydrograph. Many studies have been performed on the first flush (FF) phenomenon (the assumption that the initial portion of a rainfall-runoff event is more polluted than the later portions). However, controversy remains on whether or not the first flush truly exists, which environmental factors influence a first flush, and how best to define the first flush phenomenon. The objective of this study as to evaluate the first flush occurrence in two small urban watersheds (differing in extent of impervious area) using multiple analytical methods and definitions previously published in the literature. The first watershed is 4.8 acres with 67% impervious roadway and the second watershed is 5.6 acres with 87% wooded land cover. Statistical tests were performed to analyze for site-specific correlations between first flush strength and rainfall characteristics (e.g. rainfall depth, peak flow rate, runoff volume, peak rainfall intensity, and antecedent dry period) and determine differences in FF strength between different land uses and pollutants. Furthermore, the FF study was utilized to perform an annual treatable load analysis in order to evaluate the effectiveness of two hypothetical BMPs sized to treat stormwater from the highly impervious watershed based on the 1.3 cm, 1.9 cm, 2.6 cm and 3.2 cm of rainfall water quality volumes.
A year-long study captured stormwater samples from over 33 storm events and analyzed for TSS, turbidity, nutrients and heavy metals. Samples were collected using flow-based sampling frequencies that yield a more accurate quantification of pollutant mass transport throughout the storm event than time-paced sampling used in most of the previous first flush studies. For each collected storm, normalized cumulative pollutant load (L') and runoff volume curves (V') were generated for each pollutant with a minimum of seven discrete curve points to quantify the first flush effect and evaluate for first flush occurrence based on several published methods and definitions (e.g., max L'>V', max L'-V' > 0.2, and 80% of total load runoff in first 30% of total runoff volume). Linear regression analyses were performed to determine the first flush coefficient (b) for the power function L' = V'b to quantify the FF strength. Relationships between the first flush strength and rainfall characteristics were examined. Analysis of covariance (ANCOVA) was utilized to determine if FF strength (i.e. b-value) significantly differed between wooded and impervious watersheds, and a nonparametric ANOVA (Kruskal-Wallis test) was used to evaluate differences in FF strength among the pollutants within each watershed.
As indicated by all the methods utilized in this study, most pollutants exhibit a slight FF effect on average but substantial pollutant loading still occurred in the latter portion of the storm?s total runoff volume. Thus, to treat the majority of a storm?s total pollutant load requires capturing almost the same fraction of runoff volume. Although the FF phenomenon was not dominant, this study did define a ?most efficient? design volume (first 40% of runoff or approximately 1.3 cm of rainfall), which was where the fraction of total pollutant load was greatest compared to the fraction of total runoff volume. Of the rainfall characteristics analyzed, rainfall and runoff volume both inversely affected the FF strength of TSS and heavy metals on the impervious watershed. Although the runoff nature of orthophosphate (O-PO4) at the first portion of the storms did not have a first flush, the relative FF strength for O-PO4 actually increased with increasing rainfall or runoff. Land use did not influence the first flush strength of the pollutants except for Pb, which had significantly stronger FF effect on the more impervious watershed compared to the heavily wooded one. Disregarding the estimated pollutant load from a large 18.2 cm tropical storm, the 1.3 cm and 2.6 cm of rainfall design volumes could potentially capture and treat on average 64% and 82% of the annual runoff volume and 67% and 85% of the annual pollutant load for all the pollutants on the impervious watershed. Although stormwater BMPs designed to capture the first 2.5 cm of rainfall can potentially treat a substantial fraction of yearly pollutant load, this study suggests that in watersheds with limited and expensive land area it is more efficient to use multiple smaller BMPs near the source that capture the smaller, more frequent storms (1.3 cm of rainfall or less).
Among all the stormwater BMPs currently utilized to reduce peak runoff volumes and remove contaminants from urban runoff, constructed wetlands have emerged as an optimal choice because of their high performance of water quality improvement and ecological benefits. Due to a stormwater wetland?s high land requirement, however, they are often difficult to size properly in urban environments. This study also evaluated the pollutant removal efficiencies of a newly constructed flow-through stormwater wetland that is sized to only capture 20% of runoff generated by the recommended design storm in N.C. (first 2.5 cm of rainfall). The wetland was constructed in December 2006 as a retrofit to a failing level-spreader in order to repair a rapidly eroding head-cut between two stormwater outfall channels and the receiving stream. Following addition of the extended detention function, the undersized wetland has initially removed, on average, 71% of the TSS load, between 39% and 60% of the nutrient load, and 60% of the heavy metal load. For all the pollutants except TP and TKN, lower removal efficiencies were observed for the storms with flow bypass (5 out of 9 events since the extended detention function) on average as compared to the smaller events without bypass. When evaluated for all the sampled storms, event mean concentrations proved to be statistically lower at the outlet for all pollutants except Ti and V. For most pollutants, higher influent EMCs were observed for the storms with higher removal efficiencies. It is important to note, however, that these initial results are skewed by a small number of storms with low rainfall depths (median = 0.46 cm). With that considered, the average removal efficiencies are actually higher than predicted for the 1.3 cm design storm BMP. Undersized wetlands with ?flow through? design might provide effective and efficient pollutant removals if designed to safely pass the larger storms. Longer term research will provide a more definitive evaluation.
Advisor:Jean Spooner; Jean Spooner; William F. Hunt; William F. Hunt; Aziz Amoozegar; Aziz Amoozegar
School:North Carolina State University
School Location:USA - North Carolina
Source Type:Master's Thesis
Keywords:biological and agricultural engineering
Date of Publication:10/17/2007