Project area
The study site was the central business district (CBD) of Wellington City, New Zealand, demarcated by street perimeters (Figure 1). The CBD is an area of high volumes of pedestrian traffic, shopping, nightlife, dining and recreation due to its compact urban design. The relatively flat topography, narrow streets and limited parking mean that foot traffic is the most common source of transportation and this increases the potential for visual exposure to smoking in the downtown area. In Wellington smoking is currently permitted in outdoor public spaces (except for parks and playgrounds) and in the outdoor dining areas at bars/cafés.
Data included in the viewshed database
Visibility or viewshed analysis is a method for quantifying visible areas across a geographical region. By providing a computer program with a representation of terrain and building footprints and heights, one is able to determine which areas are visible from defined observer locations. Typically, a Digital Elevation Model (DEM) is included in this type of visibility analysis to represent terrain, however, nearly all the Wellington CBD is located on flat terrain, nestled between the coastline of Wellington Harbour and low lying hills. As such, incorporating a DEM was not necessary for the viewshed analysis in this study.
Rather, using a computer program, we generated a digital grid of 1 metre by 1 metre cells or pixels (called a raster file) across the study area to represent ground level elevation. This gridded raster file provides the base layer for visibility analysis in the program. Each cell/pixel is treated as an independent entity which may or may not be visible from defined observer locations. Then, various visibility obstructions were added to this digital layer. Extruded building footprints obtained from the Wellington City Council (2012) were merged with the raster file to represent the CBD in a 3-dimensional space.
Next, we added a geographic data layer of the locations of outdoor smoking. We compiled all hospitality business addresses with pavement licences registered with the City Council. These addresses within the CBD were geocoded (n = 70) using ArcGIS v10 (Redlands, CA, USA). These pavement license points were located on the ground level raster file and then were assigned an elevation of 1.5 metres (i.e., average height of a seated patron) for the 3-dimensional viewshed analysis. The locations of streets with high numbers of retail shops and hospitality venues were digitally added to the base map, based on local knowledge of the area.
We then defined observer locations (viewpoints) as all locations within public areas (e.g., sidewalks along streets, parking lots, streets, parks, etc.). These data allowed us to calculate the number of visible smokers (occurring at outdoor bar/dining areas) for each observer location, accounting for obstructions in vision caused by the presence of buildings.
Observational data collection
Observational data were collected in order to estimate the prevalence of smoking in licensed pavement areas of bars/cafés/restaurants [13]. Different observational times were used to capture temporal fluctuations, both by time of day and day of week. Methods for observations followed methods trialled in a previous study [9], but were slightly modified to increase data accuracy. Data were collected over 15-minute periods twice per day (during 12-1 pm and 7-8 pm) on Wednesdays and Fridays over two consecutive weeks in April 2013, for a total of eight observation periods for each location. Sites were selected for observation from the list of licenced pavement leases within the CBD. Selected sites operated between the hours of 11 am and 11 pm. Nineteen sites met the inclusion criteria, of which 14 were randomly selected (using a random number generator using Excel 2010 software). Fourteen was the maximum number of sites that could be observed with the resources available, as two sites each were assigned to seven pairs of observers. At each observation period, the number of patrons and number of lit cigarettes at five minute intervals were recorded, using 30-second scans. To minimise conspicuousness, counts of smokers were tallied using smartphones with note-taking applications such as ‘Note’ and ‘S Memo’. From these observations, averages, and ranges were calculated by time of day and day of the week for the 15-minute observation periods [13].
Extrapolation from observed sites to other sites
In order to assign smoking levels to venues for which we did not collect observational data, we divided the pavement licenses into three categories: Bars, Cafés and Restaurants. Then, using the observational data, we calculated averages for each category by daytime, night time, mid-week and end of week. We then assigned these averages to the remaining venues according to their category.
Viewshed analyses
Visibility analyses were conducted to provide a smoking visibility index for the Wellington CBD using the software ArcGIS v10 (Redlands, CA, USA). The purpose of the viewshed tool is to classify a region into visible or non-visible areas from a single or multiple defined observer points. This is achieved by generating lines of sight (LoS) between observer point(s) to pavement lease locations, accounting for obstructions using the gridded raster surface. Typically, viewshed analysis produces a binary raster file (1 for visible cells, 0 for non-visible cells). However, in order to reflect the number of smokers visible, positive cell values were multiplied by the number of smokers for a specific pavement lease based on the observational data.
A novel iterative script (see supplementary material) was developed in Python programming language to conduct the viewshed analysis outlined here. The output raster maps highlighting which areas were visibly exposed to smoking from each pavement license were consecutively overlaid, resulting in maps of the cumulative exposure to visible smoking across the CBD. In order to identify weekly and diurnal variation in smoking visibility, the process was independently repeated six times, to estimate visible smoking based on observed smoking counts during four different time periods. This method produced an estimated number of visible smokers (located at pavement bars/dining areas) from the viewpoint of each defined observer location – in public areas (roads, sidewalks, and other public spaces) throughout the CBD in a 15-minute period. Specifically, the following maps and estimated visual exposure to smoking at pavement licenses from a given viewpoint were generated:
-
a)
Visibility of smoking at pavement licenses at midday (12-1 pm - Average of Wednesday and Friday)
-
b)
Visibility of smoking at pavement licenses in the evening (7-8 pm - Average of Wednesday and Friday)
-
c)
Visibility of smoking at pavement licenses Wednesday or Friday at midday (12-1 pm)
-
d)
Visibility of smoking at pavement licenses Wednesday or Friday evening (7-8 pm)
Ethics approval
Ethical approval for this research was obtained through the ethics approval process of the University of Otago, in April 2013.