Identifying Areas with Isolated Pockets of Different Sanitation System Types


Presenter: Kenny Hernandez

Faculty Sponsor: Jay Taneja

School: UMass Amherst

Research Area: Computer Science

Session: Poster Session 6, 4:15 PM - 5:00 PM, Auditorium, A2

ABSTRACT

Access and maintenance of sanitation infrastructure are critical for public health. However, in most of the U.S, these systems have had no comprehensive mapping/documentation since the 1990 US Census.  With the use of predictive models, we have the opportunity to uncover these unknown systems; while power, these predictive approaches have difficulties identifying exceptional circumstances where one or more properties does not have the same sanitation system (e.g., sewer system or septic system) as its neighbors. These exceptions may reflect historical practices such as redlining, underbounding, and non-connection, which may have created disrupting patterns in the placement of sanitation services. This work focuses on identifying these “donuts” in sanitation system access, which are areas where one or more septic systems are surrounded by sewer systems or vice versa. By using QGIS and Python, we are able to analyze publicly available spatial datasets to spot any areas that may follow the donut criteria. These exceptions may also correlate with confounding factors like discriminatory practices or other real-world events, which the model may have limited capabilities of taking into account for. In all, this study aims to improve the accuracy of the models, establishing criterion for identifying “donuts”, presenting the statistical prevalence of these “donut” scenarios in the publicly available datasets, also unveiling the social and historical factors that influenced the construction and establishment of sanitation infrastructure in the U.S.

RELATED ABSTRACTS