Figure 3 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
![Tract-level commuting networks: (a) actual reported flow, and (b) optimal flow [Line width is proportional to flow volume, and bubble size represents total throughput at a tract]](publication/341851412/figure/fig2/AS:898180610801664@1591154353659/Tract-level-commuting-networks-a-actual-reported-flow-and-b-optimal-flow-Line.png)
Tract-level commuting networks: (a) actual reported flow, and (b) optimal flow [Line width is proportional to flow volume, and bubble size represents total throughput at a tract]
Source publication
Excess or wasteful commuting is measured as the proportion of actual commute that is over minimum (optimal) commute when assuming that people could freely swap their homes and jobs in a city. Studies usually rely on survey data to define actual commute, and measure the optimal commute at an aggregate zonal level by linear programming (LP). Travel t...
Contexts in source publication
Context 1
... a result, the optimal commuting pattern has only 24.4% (i.e., 44,590/182,705) within-tract commute trips when travel time is used, and 19.5% (i.e., 35,597/182,705) within-tract trips when distance is used. Figure 3 provides a visual comparison between actual and optimal commuter flows at the census tract level. Note that trips are substantially trimmed after optimization, and mostly are within tracts and between tracts in proximity. ...
Context 2
... aggregated both simulated flows and optimized flows into the census tract level, shown in Figure 4(a) and 4(b), respectively. Since the simulated flows at the individual level were intentionally designed to be proportional to the actual flows, the same pattern is observed between Figures 3(a) and 4(a), confirming that our simulation of trips worked well. In contrast to Figure 3(b) with a simpler pattern, Figure 4(b) shows the optimal commute flows at the census tract level that were aggregated back from the individual optimal pattern, which is far more complex. ...
Context 3
... the simulated flows at the individual level were intentionally designed to be proportional to the actual flows, the same pattern is observed between Figures 3(a) and 4(a), confirming that our simulation of trips worked well. In contrast to Figure 3(b) with a simpler pattern, Figure 4(b) shows the optimal commute flows at the census tract level that were aggregated back from the individual optimal pattern, which is far more complex. Individual workers are now free to swap houses for individual job locations, instead of being confined to a tract centroid for a group of workers (or a group of jobs). ...
Context 4
... the flexibility enables more choices in the optimized pairing between workers and jobs, and further brings down the total (average) minimum commuting. Figure 4b (aggregated from optimized individual trips) also shows far more interzonal trips than Figure 3b (directly optimized from zonal data) as individual workers are more likely to be paired with individual jobs in adjacent tracts instead of within the same tracts, and thus a more realistic optimization pattern. Measured at the simulated individual level, we obtained 62.45% excess commuting time and 64.48% excess commuting distance. ...
