1 of 30

HOUSEKEEPING

This presentation will be recorded and published on the website https://sites.google.com/view/encouraging-recycling/home.

Questions/Answers will be summarized and included in the meeting minutes.

Questions can be asked through the chat box and raise hands function available.

Please type your name and affiliation in the chat so that we can get a record of attendance. Thanks!

2 of 30

Encouraging Better Curbside Recycling Behaviors and Improvement of the Cart-tagging Programs

Jiannan (Nick) Chen, Assistant Professor, University of Central Florida

Debra Reinhart, Pegasus Professor Emerita, University of Central Florida

Mert Gokgoz, Ph.D. Student, University of Central Florida

Technical Advisory Group (TAG) Meetings

The William W. “Bill” Hinkley Center for Solid and Hazardous Waste Management

Nov. 28, 2023

3 of 30

Presentation Overview

Project Rationale and Background
Research Objectives
Tasks and Methodology, Timeline, and Deliverables
Benefit to End Users
Q & A

3

4 of 30

Contamination of recyclables has become an emerging issue, reducing the efficiency and benefits of recycling (US EPA, 2021).

State and local governments have implemented programs that provide simpler educational instructions to residents on what products can be included in their curbside recyclable collection programs (Becker 2014).

Using curbside inspection of residential carts, i.e., cart-tagging program, is one of the most effective ways to reduce contamination (Staub 2021).

For example, Recycling Partnership funded several cities on the cart-tagging programs in the states of Florida, Ohio, Michigan, North Carolina, etc.

4

Project Rationale and Background

5 of 30

5

Trained recycling monitors inspected recycling bins.
Leaving “oops” tags on contaminated recycling carts and providing positive feedback to those with clean carts.

What is Cart-tagging Program?

Cart Inspection and Photo Taking

(Orange County’s Recycling Pilot Provides Pivotal Data, 2018)

Clean Cart

Rejection Cart

Change the recycling behavior, and reduce the contamination, e.g., from 45 to 32% for single-family recycling (OCU 2021).
Currently, Orange County, City of Sarasota, City of Jacksonville, etc. had implemented the program.

6 of 30

6

Clean carts before and after tagging (OCU 2021)

Contamination in the cart, setout rate,

Participation – behavior change by educating the residents

Major contaminants (OCU 2021)

Major Contaminants in the Cart from the photo

What we have learned

7 of 30

7

What else can we learn?

What are the habits and attitudes of households toward cart-tagging?

External factors: Demographic and socio-economic characteristics
Internal Factors: Attitudes, beliefs, and norms
Policy and programmatic characteristics: cost and convenience

(Iyer & Kashyap 2007, Vicente & Reis 2008, etc.)

Is one type of tagging effective for all demographics?

What are the eco-environmental impacts of the tagging program in terms of recycling improvement and cost? How can we improve the program?

Economical-environmental assessment
Expected gains from tagging require an increase in recycled material tonnage to offset these costs.

8 of 30

8

Research Objectives

Understand the behaviors of residents using the tagging program data and identify solutions for recycling quality improvement,

Assess the economic and environmental impacts of the tagging program,

Make recommendations to optimize Florida’s residential curbside collection programs based on the tagging program results and eco-environmental assessment.

9 of 30

9

Overview of the Research Tasks

External variables

Internal variables

Demographic and socio-economic characteristics

Attitudes, beliefs, and norms

U.S. Census

Task 2 Survey the Residents after the Tagging Program and Recycling Outreach

Task 1 Analyze Data from the Current Tagging Programs in Florida

Task 3 Eco-environmental Assessment of the Tagging Programs and Recommendations for the Design of RCC Programs.

Task 4 Using Machine Learning to Predict the Recycling Score of Communities (for designing future tagging-program)

10 of 30

10

Address	Accepted Bins	Rejected Bins	Coordinates	Tract	Block
4000 Central Florida Blvd, Orlando, FL 32816	3	1	28.603721, -81.199652	17121	2502

Cart-tagging Program

Census Geocoder

Doctorate (%)	Masters (%)	Bachelors (%)	Household Ownership (%)	SNAP Enrollments (%)	Extracted Features from Images
10%	15%	60%	55%	10%	Quantifiable results from processing

U.S. Census

Task 1 Results

Images

Task 1 Analyze Data from the Current Tagging Programs in Florida

Identify the demographic and socio-economic factors that induce the differences in recycling behavior among the communities

11 of 30

11

Task 1 Analyze Data from the Current Tagging Programs in Florida

Images taken during the Orange County tagging program will be included in the analysis step.

Major Contaminants in the Cart from the photo

12 of 30

12

Task 1 Analyze Data from the Current Tagging Programs in Florida

Out of the 4 inspections in 4 weeks:

Dark blue: 4/4 Accepted bins

Blue: 3/4 Accepted bins

White: 2/4 Accepted bins

Pink: 1/4 Accepted bin

Dark red: 0/4 Accepted bins

Recycling performance varies in communities across Orange County as can be observed by analyzing each Census tract (4 inspections were conducted)

13 of 30

13

Task 1 Analyze Data from the Current Tagging Programs in Florida

Data consist of house address, and number of accepted and rejected bins over a month (4 weeks)
US Census Geocoder is used, and each household now has their US census tract and block found, for the 78% of the households
Address formatting did not match exactly for the rest, these households are currently being matched by other means
Data spans over 188 tracts and 365 blocks, presenting a diverse set of backgrounds

Annual income of $200,000 or more, (CV: 0.53)

Education attainment level bachelor’s Degree, (CV: 0.24)

14 of 30

14

Task 1 Analyze Data from the Current Tagging Programs in Florida

The aggregated data for communities is being analyzed for correlations between demographics and accepted tags
The strongest correlations are shown below:

Education levels
Age groups

15 of 30

15

Waste Composition from the Cart and Image Analysis

Examples of accepted bins

1,000 images

Examples of rejected bins

52,000 images

Image dataset resulting from the Cart Tagging program can be used to develop a model that can classify between the rejected and accepted bins
This can allow users to check their bins prior to set out, reducing the possibility of rejection at the source
Multispectral image dataset created within the scope of the previous project will be expanded, and can be implemented in training of this model too

16 of 30

16

Waste Composition from the Cart and Image Analysis

https://www.ibm.com

Convolutional Neural Networks (CNNs) are deep learning models for grid data like images.

CNNs have layers for detecting features (convolutional), reducing data size (pooling), and classifying (fully connected).

CNNs learn features from images, fine-tuning through backpropagation to improve predictions or classifications

Recycling Quality Improvement Program Report – Orange County

Wide Applications: CNNs are extensively used in image and video recognition, image classification, medical image analysis, natural language processing, and other auto-encoded learning tasks. Their ability to learn complex patterns and structures makes them exceptionally powerful for visual perception tasks.�What are Convolutional Neural Networks? | IBM��Convolutional Neural Networks (CNNs) are a type of deep learning algorithm primarily used for processing data with a grid-like topology, such as images.
CNNs consist of multiple layers, each designed to perform specific functions. The primary types of layers are convolutional layers, pooling layers, and fully connected layers. The convolutional layers apply a set of filters to detect patterns, edges, or features in the input data. Pooling layers reduce the spatial dimensions (width and height) of the input volume for the next convolutional layer. Finally, fully connected layers combine these features to classify the input into various categories.
Unlike traditional algorithms, CNNs automatically and adaptively learn spatial hierarchies of features from input images. This is achieved through the backpropagation algorithm, which adjusts the weights of the network to minimize the difference between the predicted and actual outputs.

17 of 30

Waste Composition from the Cart and Image Analysis

17

Attention heatmap during classification

Visualizing Attention: Heatmaps highlight the areas of the input image that the model focuses on to make a prediction, with brighter areas indicating higher importance.

Superimposed Representation: The resulting heatmap is superimposed on the original image, providing a visual representation of the model's "attention" during the prediction process.

Special attention can be diverted to False Negatives and False Positives, presenting a picture of the areas of failure for the model. Frequently miss classification causing objects will then be imaged with the multispectral camera, with the methodology developed in the previous project, and model will be re-trained with additional data.

Attention heatmap

Input Image

Superimposed representation

Your feedback on the application of image processing is welcomed

18 of 30

18

Survey is planned to understand the internal factors (attitudes, beliefs, and norms) and policy and programmatic factors (cost and convenience) that affect the recycling behavior.

Survey is planned to initially target the underperforming communities identified by the Task 1. Approximately 28% of households received warning or rejection tags and either continued setting out unclean recyclables or else stopped setting out (OUC 2021).

Task 2 Survey the Residents after the Tagging Program and Recycling Outreach

19 of 30

19

10 undergraduate students for the curbside survey during the normal recycling collection schedule, and questionnaires will be placed in the recycling bin after the collection.
The survey questions: environment awareness, recycling awareness, satisfaction with current recycling management programs and services (drop-off locations), preferred service changes to result in cleaner recyclables, what products are considered recyclables, and feedback on education and engagement.
Additional flyers on how to recycle.
UCF has submitted the Institutional Review Board training and protocol approval prior to surveying.

Task 2 Survey the Residents after the Tagging Program and Recycling Outreach

20 of 30

20

Example Survey

How often do you recycle?

a. Always-4 times a month

b. Often-3 times a month

c. Rarely -2 times a month

e. Never – Once a month

Do you think that your recycling efforts make a difference?

Yes

No

Would you participate if you knew recycling does make a difference?

If the Q2 is “often” or above and Q1 is yes,

Conclude the participant is providing all their generated recyclables

Q2

Q1

If the Q2 is “rarely” or below and Q1 is no, conclude the participant might require more engagement

Yes

Questions aimed to gain insights on the current program’s performance

Extract the maximum value of the tagging program data

We would like your feedback when we finish the survey

21 of 30

21

Cart tagging is effective in reducing curbside recycling contamination but is also labor-intensive.
Some case studies found that carts did not necessarily need to be individually inspected and tagged to reduce contamination (Heffernan 2022).
An increase in collection times due to tagging might result in increased fuel usage would increase fuel usage and environmental impacts.
Preliminary LCA found results in favor of tagging but greater precision and accuracy for cart tagging results will improve the conclusions.

Task 3 Eco-environmental Assessment of the Tagging Programs

22 of 30

22

Solid waste LCA models set up within the scope of the ongoing Hinkley Center Project will be expanded to include the effects of the tagging program
Evaluate the value from better recycling and reduced greenhouse gas emissions vs. the costs of the tagging program.

Task 3 Eco-environmental Assessment of the Tagging Programs

Results from the current Hinkley Project – calculate benefit from low contamination

23 of 30

23

A machine learning model will be constructed to predict the recycling performance of communities (i.e., the potential contamination and recycling behavior) and their potential response to outreach based on results from Tasks 1 (social-economical) and 2 (survey).
Additional data from other municipalities across Florida will be requested to test and improve the model.

Task 4 Using Machine Learning to Predict the Recycling Score of Communities

24 of 30

24

Task 2 Survey the Residents after the Tagging Program and Recycling Outreach

Task 1 Analyze Data from the Current Tagging Programs in Florida

Task 4 Using Machine Learning to Predict the Recycling Behavior of Communities

Socio-demographic

Factors

Socio-phycological

Factors

Major Contaminant

from the Photo

Cart-tagging Database

Recycling Behavior

Potential Response to

Cart-tagging

Example: Decision Tree Model

25 of 30

Benefit to End Users

25

An in-depth examination of current tagging programs to improve the recycling quality of RCC programs

Economical and environmental effects of incorporating tagging and inspection projects on curbside collection

Aid decision-making for solid waste division managers to select the best educational strategies

Understanding the tradeoffs between the cost of tagging (e.g., labor, material, fuel expenditure) and gaining from a better quality of recyclables

This approach can be used to evaluate other interventions

26 of 30

Project Timeline

26

Month	1	2	3	4	5	6	7	8	9	10	11	12
Task 1: Analyze Data from the Current Tagging Programs in Florida
Task 2: Survey the Residents after the Tagging Program and Recycling Outreach
Task 3: Eco-environmental Assessment of the Tagging Programs
Task 4: Using Machine Learning to Predict the Recycling Score of Communities
Annual and Quarterly Reports
TAG Meetings

27 of 30

27

Deliverables

Journal articles, conference presentations

A minimum of two TAG meetings will be held, and the Draft and Final Reports will be reviewed by the TAG members.

Metrics, Abstract, Quarterly Progress Reports, and Final Technical Reports.

Final technical report in Hinkley Center approved formats

Machine learning model will be made publicly available

28 of 30

Teaching and Education

Prepare the rising stars for solid waste industry.

The research outcomes will be shared through courses including:

ENV3001 Introduction to Environmental Engineering,

ENV 4341: Solid Waste Management,

ENV4300 Solid Waste Design.

ENV 5700 Geoenvironmental Engineering

28

Mert Gokgoz, �PhD Student,

Experience: LCA SWOFLpy, landfill sensing, leachate treatment survey.

Méry Mbengue,

Undergraduate Researcher,

A&WMA Scholar,

Experience: leachate treatment survey, LCA SWOFLpy

Nimna Manage,

Master’s Students

Experience: : LCA SWOLFpy, data analysis

Students will work on this research

29 of 30

Questions?

29

30 of 30

References

Florida 75% Recycling Goal | Florida Department of Environmental Protection
FDEP Florida Municipal Solid Waste Management (2021b). Retrieved 12/3/2021, from Florida Department of E https://floridadep.gov/sites/default/files/2020_MSW_Managed_Chart_2.pdf.
Recycling contamination: Understanding it and preventing it | LARAC
REMX_Product_Sheet_A4.pdf (micasense.com)
Miafodzyeva, S., Brandt, N., 2013. Recycling behaviour among householders: synthesizing determinants via a meta-analysis. Waste and Biomass Valorization 4, 221e235. https://doi.org/10.1007/s12649-012-9144-4
Saphores, J. D. M., & Nixon, H. (2014). How effective are current household recycling policies? Results from a national survey of U.S. households. Resources, Conservation and Recycling, 92, 1–10. https://doi.org/10.1016/j.resconrec.2014.08.010
Green Clean. (n.d.). Home. https://www.orangecountyfl.net/WaterGarbageRecycling/GreenClean.aspx
Orange County’s Recycling Pilot Provides Pivotal Data. (2018, October 9). OCFL Newsroom. https://newsroom.ocfl.net/2018/10/orange-countys-recycling-pilot-provides-pivotal-data/
Orange County Utilities Launches Recycling Improvement Pilot Program. (2023, June 27). OCFL Newsroom. https://newsroom.ocfl.net/media-advisories/press-releases/2018/06/orange-county-utilities-launches-recycling-improvement-pilot-program/
Recycling Education Initiative | Apopka, FL. (n.d.). Apopka, FL | Official Website. https://www.apopka.gov/650/Recycling-Education-Initiative

30