Water Conservation Routes
Inspection-route planning, archived grouping comparisons, and a later Python workflow for controlled routing and workload analysis.
Why it matters
Grouping locations and reviewing their road connections makes an inspection plan easier to examine. Keeping the planning groups separate from stop sequencing also makes clear which decisions the route solver handles.
My contribution
I developed the original ArcGIS Pro workflow for preparing inspection locations, analyzing routes, and presenting planning maps. The later Python reconstruction adds controlled route comparisons and inspection-workload estimates. The current interactive viewer makes differences between the archived grouping scenarios visible, including mixed cells and gaps in route coverage.
Result
Seven saved routes contain 846 distinct stops across six of twelve planning groups. Archived 8-, 10-, and 12-group scenarios use the same 2,100 locations.
Tools
- ArcGIS Pro
- Network Analyst
- Python
- ArcPy
- JavaScript
- Leaflet
The archive is partial. Saved solver values do not establish observed travel times, measured savings, or a globally optimal plan. The figures and viewer retain real historical geography; only the Python example is synthetic. Licensed ArcGIS integration remains unverified.
Historical route results
Figures redraw historical GIS outputs and retain real archived geography. They do not document a new route solve. 3 images. Select an image to view it full size.
How the study evolved
My original ArcGIS Pro work organized inspection locations and produced route-planning maps. The later Python reconstruction extends that work with comparable route experiments and workload estimates, while the current interactive viewer brings the saved routes and geographic groupings together. These additions analyze the archive and demonstrate methods on a separate synthetic network; they do not represent a new solve of the historical routes.
What the comparisons show
Across all 2,100 planning locations, the 8-, 10-, and 12-group scenarios have stop-count variation (CV) of 39.94%, 63.96%, and 45.18%. Eight groups is the least uneven by this measure. Counts alone cannot identify the best operational plan: travel, inspection time, and other constraints also matter. The saved route archive covers only six of the twelve groups.
A separate synthetic example
Four invented network nodes: one depot and three inspection stops. Both routes use the same endpoints and edge costs.
| Visit order | Travel |
|---|---|
| Fixed visit orderDepot → A → C → B → Depot | 15 min |
| Shortest tourDepot → A → B → C → Depot | 8 min |
The shorter tour demonstrates route sequencing on this invented network, not observed travel savings.
| Per inspection | Total workload |
|---|---|
| 2 min assumed | 19 min |
| 5 min assumed | 28 min |
| 8 min assumed | 37 min |
Each total = 8 min modeled travel + 3 × the service time above + an assumed 5 min break. Waiting and additional travel are explicitly assumed to be zero.
Missing required inputs leave the workload unknown. These examples demonstrate the method; they do not establish staffing needs or a feasible work schedule.
Technical details
Original ArcGIS work · Later Python reconstruction
Water conservation inspections needed location preparation, geographic grouping, and routes that could be reviewed on a map.
- The original work used ArcGIS Pro to prepare locations, analyze routes, and communicate inspection plans. The archive did not contain reusable Python source.
- The later Python reconstruction checks a minimal stop table for identifiers, coordinates, route groups, and optional sequence values. Coincident locations are reported without merging distinct stops.
- Optional routing requires a separately supplied local ArcGIS network dataset and licensed ArcGIS Pro environment. Groups are assigned before solving; the reconstruction does not rerun historical geocoding or clustering.
- Route comparisons reject incompatible stops, endpoints, declared network settings, units, and evidence types. Synthetic examples and ArcGIS-modeled results cannot be compared as equivalent evidence.
- Workload estimates combine modeled travel with explicit inspection, waiting, and break assumptions. Missing required inputs remain unknown rather than becoming zero.
- The viewer preserves saved geometry and exposes group balance, mixed-cell counts, tied groups, suppression totals, and partial archive coverage. Cell summaries support keyboard selection. Drawn measurements do not follow the road network.