4 Case Study Methodology
4.4 Traffic Analysis Methods
4.4.1 Corridor Moving Speed
The aim of the corridor speed analysis is to understand the typical speed at which freight tricycles and trucks move on local streets, and to compare the results for each mode. The estimated values represent a moving speed; they do not account for vehicle stopped-time, which will be evaluated and discussed in Section 4.4.2.
This study seeks adapt methodologies from the previous studies discussed above for application in Manhattan’s congested urban corridors. Given the relatively small size of the collected data set, speed measures were aggregated across blocks within each corridor. To account for traffic flow disruptions imposed by intersections, and to
distinguish directional movements where “used” roadways intersect, spot speeds within 60 ft of any point of intersection were removed from the analysis dataset. A buffer distance of 60 ft was chosen to account for corridor widths (including vehicle and bicycle travel lanes, medians, parking lanes, and sidewalks) up to 120 ft. The procedure for removal of these data points is shown in Figure 4.
Figure 4. Process for Removing Intersection Spot Speeds
In traditional traffic studies, a harmonic mean speed is estimated to characterize traffic speeds; this mean is
calculated by dividing the number of observations by the sum of the inverses of each observed spot speed. However, when using GPS data, extremely small speeds resulting from congested conditions result in a very large inverse, and ultimately skew the mean speed toward these low measures. In Manhattan conditions, where congestion is common, the impact of these points on the mean will be extreme. Quiroga and Bullock (1998) concluded that the median, or the 50th percentile speed, is a more robust estimator of central tendency (typical flow conditions) than harmonic mean speed. To account for the potential distortion due to the heavily congested conditions in the study area, median observed speeds have been estimated to characterize speeds on each road.
While the median provides an estimate of the typical speed on a roadway, it does not provide any information about the distribution of speeds along the roadway. In uncongested travel conditions, vehicle speeds are generally expected to be distributed approximately normally. However, when driver’s speeds are limited by traffic congestion and other obstructions, the distribution of speeds will likely include a high frequency of low-speed observations and few high- speed observations. To visualize the distribution of speeds within each corridor, sample probability density functions have been plotted for each mode and for various categorizations within each mode. Corridor characteristics
examined include directionality, classification as a truck route, presence of dedicated bicycle facility, and neighborhood.
The Manhattan road network is primarily a grid. Generally north-south “Avenues” are relatively wide corridors with much lower intersection densities than east-west “Streets.” The majority of both Avenues and Streets carry one-directional traffic. In New York City, trucks are required to travel on designated local truck routes; they may only deviate from these routes to take the shortest path to their final destination. Figure 5 shows the designated local truck routes in the study area. In the last decade, the City of New York has also installed considerable mileage of designated on-street bicycle facilities. Figure 6 shows the Class 1 (buffer protected or grade separated) and Class 2 (standard on-street bicycle lanes) bicycle facilities in the study area.
Figure 5. Study Area Truck Routes
Figure 6. Study Area Dedicated Bicycle Infrastructure
To plot the sample PDF for a filtered speed data set, the data had to first be sorted into speed range bins. For freight tricycles, speeds generally ranged from 0 to between 11 and 19 miles per hour, so a bin size of 2 mph was used to develop initial PDFs. For trucks, which mostly range from 0 to 30 mph, a bin size of 3 mph was used to maintain the same quantity of bins (10). For plotting of comparative distributions, a bin size of 3 mph was also used for freight tricycles. For each bin, the frequency of observations within the given speed range was determined. To allow for comparison of distributions across roadways with varying numbers of speed observations, the observed frequencies in each bin were divided by the total number of speed observations in the total dataset to obtain the percentage of observations belonging to that bin (Equation 1). Finally, the estimated percent of observations was plotted vs. the centroid of the speed range (e.g., for bin 0 to 2 mph, the centroid is 1). Figure 7 provides an example of a sample PDF.
𝑝𝑏= ∑ 𝐹𝐹𝑏 𝑥
𝐵 (1) where:
𝑝
𝑏= share of observation belonging to bin b𝐹
𝑏 = frequency of observations belonging to bin b𝐹
𝑋 = frequency of observations belonging to bin x B = set of all observed binsFigure 7. Sample PDF Example
0 0.05 0.1 0.15 0.2 0.25 0 5 10 15 20 SH AR E OF OB SER VA TI ON S B EL ON GI N G TO BIN
BIN CENTROID (SPEED IN MPH)
Speed observations were also aggregated across corridors to evaluate time-of-day differences. Four time periods were defined for analysis: the morning peak from 6:30 to 9:30 a.m.; the midday peak from 12:00 noon to 2:00 p.m., the evening peak from 4:00 p.m. to 7:30 p.m.; and the off-peak, which includes all other time periods between 9:30 a.m. and 10:00 p.m. These periods were identified based both on general operating hours for City Bakery and City Harvest and on typical Manhattan traffic conditions. While City Harvest trucks do operate in later hours, no truck data were collected during these periods due to limited battery life of the GPS device.
Once speed behavior was evaluated for each variable, the cumulative distribution of truck speeds was evaluated to examine the share of truck observations that feasibly could be reached by a freight tricycle. This evaluation also required the development of data bins; however, rather than identifying bin categories as regular intervals, bins were established with boundaries relating to the observed speeds for City Bakery and City Harvest freight tricycles. Operational characteristics impacting these observed speeds are discussed in detail in Section 5.