Modelling effects of stormwater best management practices on urban stormwater runoff phosphorus
Loading...
Date
2024-01-15
Authors
Zhou, Bowen
Journal Title
Journal ISSN
Volume Title
Publisher
University of Waterloo
Abstract
Phosphorus (P) is a key limiting nutrient for algal growth in freshwater whose excess loading to freshwater bodies contributes to cultural eutrophication and the associated symptoms of water quality deterioration. Urban stormwater is a significant contributor of P to downstream ecosystems from various point and non-point sources and via a variety of transport and emission pathways. Stormwater best management practices (BMPs) such as stormwater ponds (SWPs, a type of traditional stormwater BMP) and bioretention cells (BRCs, a type of low-impact development (LID) BMP) have the potential to attenuate P loads from urban areas and hence mitigate eutrophication risks to aquatic ecosystems. Despite their rapidly growing implementation worldwide, the effects of these stormwater BMPs on urban stormwater P concentrations and loads remain poorly understood.
In this thesis, I assess the effects of urban stormwater BMPs on P export, with the goal of determining (1) what are the knowns and unknowns regarding the sources, pathways, and influence of stormwater BMPs on urban P export, (2) what are the dominant internal processes that control P reduction in BRC, based on process-based modelling, (3) what are the general effects of BRCs on urban stormwater runoff P and how are they different from the effects on nitrogen (N), and (4) how to predict the effects of BMPs on urban stormwater runoff P through the use of data-driven models, and what are the potential influencing factors. I address these research questions by reviewing urban P sources discussed in the literature, quantifying P mass balance in a BRC facility in Mississauga, ON, assessing effects of urban stormwater BMPs on P export based on data from the International Stormwater BMP Database, and through the development of process-based and data-driven BMP P models.
In Chapter 2, I review the existing literature and analyze data from the International Stormwater BMP Database (ISBD) to summarize the sources, pathways and speciation of urban stormwater P, and the effects of urban stormwater BMPs on P export. This study acts as an introduction to the issues of P in urban stormwater runoff and identifies the research gaps associated with understanding effects of stormwater BMPs on urban stormwater P export. I show, based on both previous literature and the data in the ISBD, that the effects of stormwater BMPs on urban P export remain highly uncertain and unknown. There is a lack of predictive tools for estimating effects of stormwater BMPs on urban P export, and I go on to fill this research gap in Chapters 3, 4, and 5.
Following Chapter 2, I address my research questions by developing a process-based P model for a BRC facility in Mississauga, ON. This model is calibrated using field monitored data for flow, water quality and filter media soil chemistry (from core samples). In Chapter 3, the model simulates the multi-year P partitioning, accumulation and export in this stormwater BMP. I show, via the analysis of model simulation results, that exfiltration to underlying native soil was principally responsible for decreasing the surface water discharge from the BRC (63% runoff reduction), while accumulation in the filter media layer was the predominant mechanism responsible for the reduction in P outflow loading (57% retention of total P (TP) inflow load). Of the P retained within the filter media layer, only 11% was stored in easily mobilizable forms. There were no signs that the P retention capacity of the BRC was approaching saturation after 7 years of operation. Thus, my results demonstrate sustained efficient P load reduction by this BRC.
In Chapter 4 I evaluate the general effects of BRCs on urban stormwater runoff P concentration and loading by analyzing data from a large number of BRCs in the ISBD from across the United States. I further compare the influence of BRCs on P and N export. I also introduce the data-driven approach in Chapter 4 by training a random forest model to predict the reduction and enrichment effects of BRCs and compare the importance of different explanatory variables. I show that while BRCs typically enrich concentrations of TP and soluble reactive P (SRP), the corresponding outflow loads of TP and SRP, were generally lower, mainly because of reductions to surface runoff volumes via exfiltration to the subsurface. This finding raises questions regarding the relative importance of this infiltrating P to the subsurface environment and potential impacts to groundwater quality. Because they are generally more efficient in reducing N loads than P loads, BRCs tended to decrease the N:P ratio of stormwater runoff, potentially altering nutrient limitation patterns in receiving aquatic ecosystems. My findings also imply that the impacts of BRCs on P and N concentrations, speciation, and loads in urban runoff are highly variable. This variability can be partly accounted for by some explanatory variables related to the climate, watershed and BRC characteristics, and predicted by machine learning (ML) methods such as the random forest model. Random forest modeling identified inflow concentrations and BRC age as key variables modulating the changes in TP, SRP, and total N concentrations between inflow and outflow. For dissolved inorganic N, the BRC’s storage volume and drainage area also emerged as important explanatory variables.
Chapter 5 also focuses on the ISBD, similar to Chapter 4, but the analysis of P control performance is expanded to six categories of BMPs. I compare the accuracy of different data-driven models for the prediction of BMP P reduction/enrichment factors, through the use of different ML methods. I show that although LID BMPs are generally more efficient at reducing runoff quantity, they are more likely to enrich TP and SRP concentrations compared to traditional BMPs leading to poorer P load reduction performance amongst LID BMPs. Both traditional and LID BMPs are more likely to enrich SRP concentration when influent SRP concentration is low, in watersheds with higher imperviousness and in drier climates. The influence of LID BMPs on SRP concentration is also more sensitive to climate, watershed and BMP characteristics compared to traditional BMPs. I show that the random forest model provides the most accurate estimation of BMPs effects on urban stormwater P concentrations when compared to models produced using other ML methods. This study suggests that switching to LID BMPs has the potential to increase eutrophication risks and requires further examination. It also proposes that ML methods, especially use of the random forest model can represent a more robust approach to estimate the effects of stormwater BMPs on urban runoff P by accounting for both P reduction and enrichment effects.
My results show that that BRCs and other stormwater BMPs have highly variable effects on urban P export. I show that although the BRC I investigated in Mississauga, ON, exhibits efficient reduction of P export, it appears to be atypical and that BRCs and other LID BMPs are generally more likely to enrich P concentration compared to traditional BMPs based on data from a large number of BMP systems in the ISBD. This concentration enrichment may further impact the quality of groundwater and surface waterbodies. Considering the global environmental policy trend to promote replacement of traditional stormwater BMPs with LID BMPs, the findings of this thesis should serve as a caution to policy makers, as understanding of the effects of stormwater BMPs on urban P export remain incomplete.
Description
Keywords
Urban stormwater management, Phosphorus export control, Eutrophication, Reactive transport modelling, Machine learning