A framework for designing and optimizing green infrastructure network under uncertainty

Heidari Haratmeh, Bardia

Permalink

https://hdl.handle.net/2142/106151 Copy

Description

Title

A framework for designing and optimizing green infrastructure network under uncertainty

Author(s)

Heidari Haratmeh, Bardia

Issue Date

2019-09-12

Director of Research (if dissertation) or Advisor (if thesis)

• Minsker, Barbara
• Schmidt, Arthur

Doctoral Committee Chair(s)

• Minsker, Barbara
• Schmidt, Arthur

Committee Member(s)

• Brunner, Robert
• Uster, Halit

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Green infrastructure
• urban hydrology
• SWMM
• Stromwater management

Abstract

Green infrastructure (GI) is becoming a common solution to mitigate stormwater-related problems. Despite wide acknowledgement of GI benefits, there is a lack of decision support tools that allow practitioners to interactively identify and evaluate the performance of small GI practices using hydrologic models under uncertainty. Also, the benefits and costs of GI practices are not fully understood when the analysis scale changes from a household to a subwatershed to an entire watershed. Moreover, recognition of optimal locations in a watershed, given the uncertainty in modelling parameters, is also another challenge for GI planning and design. To address these needs, an online Cloud-based interactive tool — called Interactive DEsign and Assessment System for Green Infrastructure (IDEAS_GI)— has been developed. This study demonstrates the application of the tool, using hydrologic and empirical models, to estimate life cycle cost, stormwater volume reduction and treatment, and air pollutant deposition. The tool was applied in two small watersheds in the Baltimore metropolitan area. The results show that GI properties do not significantly affect performance of individual GI practices during design storm events due to the intensity of the storms exceeding the capacity of GI practices to treat and capture stormwater. Using the tool to identify potential locations for GI placement, the study then provides a quantitative and comparative analysis of environmental benefits and economic costs of GI using two metrics [Benefit-Cost Ratios (BCRs) and nutrient removal costs] at household, subwatershed, and watershed scales. The results for a case study in Baltimore show that the unit cost of nutrient removal in some of the subwatersheds is lower than the unit costs at either the watershed or household scales, calling for optimization frameworks to determine the features that dictate optimality at the subwatershed level. Moreover, rain gardens provide far more efficient stormwater treatment at the household scale in comparison to watershed scale, for which large-scale dry or wet basins are more efficient. The results show that for BCR, smaller subwatersheds are more cost effective for GI implementation, while for nutrient removal cost, upstream subwatersheds are more suitable. Furthermore, self-installation of rain gardens greatly reduces nutrient removal costs. Finally, to identify preferable locations for GI implementation, the numerical hydrologic model used in IDEAS_GI, SWMM, has been merged with a probabilistic noisy genetic algorithm (GA). The GA uses a probabilistic selection method that requires numerous sampling realizations to estimate the uncertainties associated with the fitness (objective function) values, which are cumulative stormwater volume reduction and GI life cycle cost. To overcome the computational challenge and to identify significant features for preferable locations, the GA is merged with artificial neural networks, which act as surrogates for the numerical models. The surrogate models use GA-generated archives as training datasets to predict the mean and standard deviation of cumulative stormwater volume reduction. The results show that the addition of meta-models decreases average computational time required to reach Pareto frontiers similar to the ones generated by the noisy GA by more than 95%.

Graduation Semester

2019-12

Type of Resource

text

Permalink

http://hdl.handle.net/2142/106151

Copyright and License Information

Copyright 2019 Bardia Heidari Haratmeh

Owning Collections