Salt Lake City Public Utility (SLCPU) shared with researchers at the University of Utah’s Metropolitan Research Center four years of anonymized monthly meter data for the more than 70K customers they serve city water to. I led a team of 3 student interns in cleaning and normalizing these data, and aggregating accounts to tax-lot parcels and then to 2010 census blocks, so we could analyze the data.

Preparing these data for visualization and analysis was tricky, for a few reasons. Along northern Utah’s urbanized Wasatch Front, outdoor water use accounts for roughly two-thirds of annual household water use. Because outdoor use is so significant, it was important to disaggregate it from indoor use. We estimated outdoor use for each parcel by subtracting the minimum-use month over the year from all months (this assumes that indoor use is constant).

Much trickier was to get estimates of outdoor watered area over which to normalize outdoor water use. This GIS effort was complicated by the representation of condominium properties in the county parcel feature class, where legally distinct condominiums show up as nominal circles or rectangles perforating the land parcel on which buildings stand but without regard to actual building geometry. It was further complicated by apartments and other buildings that span parcel boundaries, and by large parks where we frequently noted that the parcel associated with meter was not the park itself. A further challenge was the limited availability of building footprints, which we had for Salt Lake City but not the other municipalities served by SLCPU (see below).

With data cleaned and geocoded, I produced a video illustrating key patterns of city-wide water use for audiences including SLCPU internal staff, research team members of the statewide iUTAH research project, and the public.

In one analysis of these data with co-authors Philip Stoker, Robin Rothfeder, Kenneth Dudley, and Philip Dennison, we published findings about the utility of multi-spectral imagery alone v. multi-spectral imagery + lidar-derived canopy measures as predictors of water-use:

In this paper we examine whether land-cover measures derived from multi-spectral (MS) imagery in combination with light detection and ranging (LiDAR) data sources better predict parcel scale urban water consumption than measures derived solely from MS imagery. Land-cover measures such as the percentage of impervious surface and vegetative cover are important predictors of household level water use. This study found that the additional effort required to obtain LiDAR data does not appear to add predictive power for water demand modeling. We suggest that MS imagery is just as useful estimating household level water demand.

From Stoker, Rothfeder, Dudley, Dennison, and Buchert. “Comparing the utility of LiDAR data vs. multi-spectral imagery for parcel scale water demand modeling“. Urban Water Journal, 2017. https://doi.org/10.1080/1573062X.2015.1111915

In another analysis of these data with co-authors Philip Stoker, Sarah Hinners, Douglas Jackson-Smith, and Zachariah Levine, we published research finding that, even after accounting for parcel-scale factors (i.e. building and occupant variables), about a quarter of the parcel-to-parcel variation in water use is attributable to neighborhood effects:

Planning for urban water conservation requires an understanding of how and where water is used in cities. There is significant evidence that urban water use is related to the characteristics of the residents, housing types, and landscaping patterns. At the same time, a growing body of research has shown geographic clustering of high or low water use at the neighborhood scale. This paper explores how the characteristics of neighborhoods influence water use at the parcel scale. We hypothesized that neighborhood characteristics influence water use through social dynamics as well as the physical structure of the neighborhood. Using a dataset for almost 75,000 parcels across 248 neighborhoods in Salt Lake City, Utah, we used multilevel modeling to determine how nine characteristics of neighborhoods influenced parcel-level water use. Almost a quarter (24%) of the variation in parcel-scale water use was explained by neighborhood characteristics. Controlling for key parcel-level drivers of water use, we determined that several neighborhood factors were significantly associated with parcel-level water use. For residential properties, parcels in more homogeneous suburban neighborhoods dominated by detached single family-owned homes and family households used more water than comparable parcels in neighborhoods with mixed housing and household types. Neighborhood effects were more pronounced for residential parcels than commercial, and more for outdoor than indoor water use. We suggest that planning and design strategies at the neighborhood level can contribute to urban water conservation.

From Stoker, Hinners, Jackson-Smith, Buchert, and Levine “Neighborhood Effects on Parcel-level Water Use“. Sustainable Water Resources Management 2019. https://doi.org/10.1007/s40899-019-00306-5