For attributes slope or ramp width, standard error of up to 5% is considered tolerable (in light of the large number of curbs to be surveyed and limited budget). For attributes such as whether a curb cut leads into a hazardous part of the intersection, inputs are qualitative and require human input. Location accuracy is important since each curb must be locatable for remedial works.
3.2 Data collection methods
With these requirements under consideration, data collection was conducted through three methods: (i) smartphone-based survey; (ii) Structure Sensor; and (iii) LiDAR scanner.
(i) Smartphone-based survey
A survey tool was built using ESRI Survey123 software. The tool comprised 12 questions in a web-based survey form. Publicly available on the internet, the tool may be utilized by anyone with a smartphone, allowing multiple team members to gather data independently. Data is entered for each curb inspected by a team member, making use of smartphone GPS and measurements using a tape measure. A photo is taken of each curb cut. The survey tool may be viewed at
http://bit.ly/2I85t5Z.
Construction of the survey questions required several iterations of field testing. Challenges included:
- Strict measurement definitions required. Physical measurements for lip / bump height and ramp width proved, in initial field testing, to be ambiguous. The team agreed on standardized definitions for the measurements. For lip / bump height, it was agreed to measure height at maximum point of the drop from sidewalk to street - since the maximum height determines the difficulty faced by wheelchair users.
- Condition classification scheme. For ease of interpretation by decision-makers the team opted to include an 'overall condition' attributes based on an ordinal scale from 0 to 5. At first, subjective judgment about what constitutes categories such as 'good' or 'very poor' raised concerns over data consistency, assuming multiple team members conducting the survey. To resolve this, a classification scheme was introduced with decision rules for inclusion of a curb in a given category.
- Challenge of slope angle measurement. A best effort was made to generate measurements of slope angle. This comprised measuring height from street to maximum elevation of sidewalk using a tape measure. Combined with a measurement of ramp length, simple Pythagorean geometry allowed the extraction of the angle of ramp elevation. However, measuring elevation by tape measure proved difficult. Team members were not confident in the reliability of this method and hence of the manual slope angle measurements generated.
- Location accuracy. Field testing revealed major problems with using iPhone GPS. The survey instrument was constructed for the user to hit 'locate', at which point the phone GPS would record longitude and latitude coordinates, attaching this to the curb record. However, in post-processing, curbs ended up 50 meters from their true. Given that typical 4-year intersections have 8 curb cuts, and complicated intersections with mid-street elevated sidewalk may have 12 or more, this degree of inaccuracy is not tolerable. New fields were introduced to allow for manual snapping of curbs to their true location. The fields were: (i) cross street; and (ii) crossing orientation [north-south or east-west].
Having refined the survey tool, curb cuts were surveyed between 26th St and Union Square.