Sto. Cristo junction · DRT Hwy × Pulilan Regional Rd × Diversion Rd
More traffic arrives at this junction than it can possibly let through. In the evening peak the average driver loses about five minutes here — and it is not a discipline problem or a staffing problem. A traffic signal would not fix it either: five roads meet here, so a signal has to give each one its own turn, and the waiting between turns eats the time it was meant to save. A roundabout never takes turns. That is the whole argument, and it is about geometry, not politics.
What today's average evening-peak wait becomes with a roundabout in place, on the same modelled traffic.
About a third of the vehicles reaching this junction in the peak hour have nowhere to go. That is not something better driving or more enforcers can solve — the road simply runs out of room to pass them. Every one of the five approaches is failing.
We modelled a signal as generously as the rules allow, and it still cannot keep up — about three minutes of average waiting. Five roads meet here, so each has to wait its turn, and the few seconds lost every time the lights change adds up to nearly a tenth of the whole cycle.
It could carry more than twice today's traffic, because nobody has to wait their turn — you join when there is a gap and keep moving. The busiest approach runs at about half its capacity, and it still works in the mid-2030s once traffic has grown. Roundabouts do not care how many roads meet.
Read off the OpenStreetMap centrelines, the five approaches sit at 17.9°, 92.5°, 135.8°, 198.1° and 273.5°. Two of those are ordinary opposing pairs: the DRT Highway runs N–S at 179.8° apart, and Pulilan Regional Road runs E–W at 179.0° apart. So four of the legs form a conventional crossroads. The fifth — the Plaridel–Pulilan Diversion Road — arrives diagonally at 135.8°, and that is where the trouble is.
Run the turn geometry and the governing movement falls out immediately. The two trunk roads exchange most of their traffic across this junction, and one direction of that exchange is easy while the other is not. Diversion Road → DRT Highway north is a 62° right turn, and it flows. The return movement — DRT Highway north → Diversion Road, about 320 vehicles an hour — is a 62° left turn that must cross the whole DRT Highway through movement, two directions of a four-lane national road, with no signal and no protected phase. That one movement is why the junction locks up, and it is exactly the movement a roundabout deletes.
Change of heading relative to straight-through. Bearings increase clockwise, so positive is a right turn. Left turns are the ones that cross opposing flow.
True bearings from OSM. Class and lane count as tagged.
Both panels are fed an identical seeded arrival stream at the modelled peak-hour volumes, so it is a fair race. This animation is illustrative. The delay figures the case actually rests on are the HCM calculations below it, not the counters here.
What to watch: the queue on the left never clears — vehicles arrive faster than the junction can release them. On the right nothing ever stops for long.
This is the real site. Background is OpenStreetMap raster imagery at zoom 19, embedded in the page and pinned to the same junction node the geometry came from — 0.29 m per pixel, so the vehicles are drawn at true size and a car really is about 4.6 m long. The approaches are straight lines on the surveyed bearings; checked against the real centrelines, that approximation is within 0.8 m out to 84 m from the junction, which is why the overlay sits on the actual carriageway. The roundabout drawn on top is the proposal, not something that exists. Map data © OpenStreetMap contributors, ODbL 1.0.
Average delay counts vehicles still queuing, not only those that got through — otherwise the layout that serves fewest vehicles would flatter itself. Arrivals are dropped once a queue backs past the edge of the frame, so the left panel understates how bad it gets. Vehicle colours follow the class mix used in the PCU conversion — motorcycles and tricycles are the small marks, jeepneys and buses the long ones. The roundabout panel the roundabout as it could actually be built: no bypass lane, because the only quadrant one could occupy is the fire station. Switch the comparison to Roundabout + bypass to see what that quadrant would have bought.
The objection to roundabouts is almost never about capacity. It is “I don’t know how to use it” and “it will be chaos.” Both are fair, and both have short answers. There is one rule that matters and four that follow from it.
Everything else follows. You are not waiting for a signal and you are not waiting your turn — you are waiting for a gap, and at this junction a gap arrives every few seconds. If the circle is clear, you do not stop at all.
The whole design is built to make you arrive slowly — the curve into the entry holds speeds to about 25–30 km/h. You are looking for a gap, not for a light.
On the two-lane entries, keep right if you are taking one of the first exits and left if you are going further round. Changing lanes inside the circle is where roundabouts go wrong.
Traffic already circulating has priority over everyone waiting to enter, which is exactly why the circle never locks. Do not stop inside it to let someone in.
It tells the driver waiting at the next entry that the gap is theirs. This single habit is most of what makes a roundabout flow.
The splitter islands break each crossing into two short halves, and you only face one direction of traffic at a time — instead of five streams at once, as today.
You gain the most and you also carry the most risk. Entry speeds drop, which helps — but do not filter up the inside of a vehicle that is about to turn out of the circle. Ride the lane, not the gap beside it.
No more waiting for an enforcer to wave you across the highway. You enter on a gap like everyone else, and the gaps come often because nothing is being held at a red light.
The centre island has a mountable apron — a low, paved ring your rear wheels can track over on the tight turns. It is built for articulated trucks, so a bus or a ten-wheeler has room.
Filipino drivers already use roundabouts every day. Fuente Osmeña Circle in Cebu City is routinely cited as one that works; Metro Manila has the Welcome Rotonda and, at Quezon Memorial Circle, the largest in the country. In fairness: none of them publishes capacity or crash figures, so they are offered here as evidence that the manoeuvre is familiar — not as evidence of how well they perform. The performance case on this page rests on the measurements above, not on those examples.
Seconds per vehicle, peak hour 2026. Sorted by today's delay. Bars share a zero baseline; above roughly 120 s the HCM formula only means "failing", not a precise forecast.
Volume-to-capacity ratio by approach. The 1.0 line is the point of breakdown; 0.85 is the design target DPWH would normally hold to.
Capacity in pcu/h, delay in s/veh, level of service per HCM thresholds (signalised A–F at 10/20/35/55/80 s; roundabout at 10/15/25/35/50 s).
Four phases at 4 s lost time each removes 16 s from every cycle. At the 180 s maximum practical cycle that is 9% of capacity gone before anything moves — and the diagonal leg needs a phase of its own that serves only 27% of the traffic.
No phases, so no lost time. Entry capacity depends only on the gap in circulating flow, and four of the five entries are sized to two lanes, circulating flow peaks at about 980 pcu/h — comfortably inside the range where entry capacity is still high.
A 4% capacity penalty at every entry for pedestrian crossings, PCU factors that count a tricycle as a full car and a bus as 2.5, and a saturation flow for the signal case that is higher than the enforcer case. The comparison is deliberately unkind to the proposal.
Conflict points are countable, not arguable: draw every permitted movement as a path and count where paths cross, merge and diverge. For this five-leg layout that is 78 conflict points, 48 of them crossing conflicts — the type that produces right-angle and head-on collisions, the ones that injure and kill. A roundabout converts every crossing conflict into a merge. Five entries, five exits, ten conflict points, none of them crossing.
All 20 movements drawn at true bearings. Red marks are crossing conflicts.
Same movements, same bearings. Every path now merges instead of crossing.
FHWA crash modification factors for converting a stop-controlled or uncontrolled junction to a roundabout. These are transferred factors, not Philippine ones — no local CMF database exists. Enter a real baseline once Pulilan MPS releases junction records.
The 12 is a placeholder. If you hold real crash counts for this junction — from Pulilan MPS, the municipal engineering office, or a barangay blotter — type the annual figure in and every bar below recalculates from your number.
Four marked but uncontrolled crossings and two bus stops sit at this junction today.
Roundabouts are worse than signals for blind and low-vision pedestrians, who rely on the sound of stopped traffic to judge a gap. Any design here must include audible or tactile treatment at the crossings. That is a real cost of the proposal and it belongs in the record, not in a footnote.
Pulilan Fire Station stands 25 m from the centre of this junction. Every appliance that leaves it in an emergency has to cross the thing this page is about, so the question deserves its own section rather than a line in the methodology. It cuts both ways — and on balance it cuts in favour.
Today an appliance turning out of the station in the evening peak meets stopped traffic with nowhere to go — vehicles boxed in on all five approaches, and no green phase anyone can give it. Drivers cannot pull over because there is no space to pull into. A roundabout has no red phase and therefore no standing queue: entry delay falls to about eight seconds, traffic keeps moving, and vehicles can mount the truck apron to open a path. The same property that makes it efficient makes it clearable.
A fire pumper needs a wider swept path than a car, and an aerial ladder more again. At 46 m inscribed diameter this is comfortably inside the range routinely designed for articulated trucks, which are longer than any appliance the station is likely to run — so it fits. What matters is that the design keeps a genuinely mountable apron rather than a kerbed island, and that entry deflection is not tightened past what a long wheelbase can take.
This is a desk analysis. It has not been discussed with the station, and their answer outranks it. Before this goes any further somebody should ask them three things: which legs their appliances actually use most, whether the apron and entry radii suit the vehicles they run, and whether they would want a dedicated exit onto the highway. If the station objects on operational grounds, that objection should be recorded here in full — not argued with.
The north-west quadrant — the one place a free-flow bypass lane could physically go — is Pulilan Fire Station, whose building edge sits 25 m from the centre of the junction. (OpenStreetMap records the name as "Pulinan Fire Station". That spelling is wrong; the label was baked into the map tiles, so it has been repainted on this page — the only edit made to the imagery. Worth fixing at the source on openstreetmap.org.) That rules the bypass out, and it is the right call: a fire station needs its apron more than this junction needs a slip lane. The good news is that it costs surprisingly little. Without the bypass the roundabout still runs at level of service A; what it loses is about two years at the far end of its service life.
Measuring every mapped building footprint against the junction centre turns up a harder problem. The south-west quadrant closes to 16.4 m, where a row of retail frontage stands. The largest roundabout that clears every existing building is about 33 m across — and at that size it fails, badly. So the honest position is this: a roundabout works here, and it needs land acquired on the south-west corner. Saying otherwise would not survive the first site visit.
A roundabout that works here needs frontage on the south-west corner — the row of small retail buildings on the Pulilan Regional Road side, the nearest of which stands 16 m from the centre. Some of those businesses would lose part of their lot, and possibly their building. That is a real cost borne by identifiable people, and pretending otherwise would get this proposal thrown out the first time somebody walked the site.
Land taken for a national road project is acquired under the right-of-way law, which provides for payment at market value plus compensation for structures and for disturbance to a business. Whether that is adequate is a fair argument — but it is an argument to have with the affected owners in the room, early, not one to discover at a public hearing.
The junction is a living: TODA queues, the market frontage, the stalls and the two bus stops. A roundabout changes where vehicles can legally stop, and a splitter island can end a loading spot that has existed for years. None of that appears anywhere else on this page, and it should — the operators are organised, they will be at the barangay session, and they are right to ask.
The design answer is usually laybys beyond the splitter islands rather than a ban, which can end up better than the present arrangement of stopping in the conflict area. But that is a promise to be made concretely, not in principle.
A central island sits where water currently runs across the carriageway, and a roundabout regrades the whole junction. Done carelessly it becomes the low point of the intersection; done properly the island carries the drainage and the approaches shed to it. This page has done no hydraulic work at all — it is flagged so that the question is asked at design stage rather than discovered in the first heavy rain.
Distance from the junction centre to the nearest mapped building edge, by direction. The shaded ring is the 46 m roundabout; anything inside it has to be acquired.
Peak-hour 2026. The compact option is the only one that fits inside the existing right of way, and it is over capacity on two approaches on the day it opens.
Traffic is grown on a tapering schedule — 4% a year for five years, 3% for the next five, 2% thereafter — which is how DPWH and JICA handle long-range forecasts. The result is not that a roundabout is permanent. It is that a roundabout is the only option that is still working when today's children are driving through it.
Average wait per vehicle in the evening peak. The same three options, at today's traffic and after it has grown. No special axis — these are just minutes.
Weighted across all approaches. Level-of-service bands shaded behind the lines.
Delay is not free — it is paid in fuel, in wages, and in hours nobody gets back. Every assumption below is adjustable, and every one has been set low on purpose: three equivalent peak hours a day rather than the full day, 250 operating days rather than 365, and a value of time well under what NEDA uses for private vehicles. Change them and see how little it matters — the conclusion survives assumptions far more conservative than these.
Set against an indicative construction cost of ₱48–85 million for a five-leg roundabout with truck apron, splitter islands, drainage, lighting and signage — plus land acquisition on the south-west corner, which this study cannot price and which may well exceed the civil works. Signalisation would cost ₱9–16 million to install plus roughly ₱350,000 a year to run — and by this analysis it would still leave the junction over capacity. The enforcer posts that direct it today cost on the order of ₱0.9–1.4 million a year, indefinitely, and stop working the moment nobody is standing there.
This matters more than any chart above it. A proposal that overstates its evidence gets dismissed once; a proposal that labels its gaps gets taken seriously. The geometry is measured. The volumes are not yet. Everything below is colour-coded on exactly that basis — green is measured, red still needs finding.
Read the red rows as a to-do list rather than a warning. None of them is unknowable; they are simply things nobody has gone and got yet. Every one of them can be closed by someone willing to send a letter or stand at a corner for fifteen minutes.
Turning-movement counts. Every capacity figure on this page scales with them, and nobody in the Philippines publishes them for provincial junctions. Two fifteen-minute counts — one AM peak, one PM peak — move this entire analysis from modelled to measured, and it takes one person half an hour.
If you do only one thing after reading this page, do that one. It is the difference between a neighbour's opinion and a document an engineer has to answer.
Everything else on this page is work for whoever is putting the case together. This part is not. If you cross Crossing Pulilan on the way to work, any one of these is worth more than agreeing with the argument.
Note the clock when you join the queue and when you clear the junction. One number, from one trip. A dozen people doing that is a dataset, and it is the kind of evidence that reads as real to a councillor because it came from residents rather than a model.
A near miss, a crash, an ambulance stuck, a child crossing between bumpers. Date it as closely as you can. Police records will not capture most of it, and the pattern matters even when the paperwork does not exist.
Your barangay captain, your councillor, the municipal engineer, the DPWH district office. One link, one sentence about your own trip. Proposals move when several people mention the same junction in the same month.
You do not need to argue the capacity figures. “More traffic arrives here than can get through, a signal would not fix it, and a roundabout would — can we ask DPWH for a study?” is the whole ask. A commissioned study is the realistic win, not a decision.
Not a roundabout. A funded traffic study. Nobody builds a five-leg junction off a community page, and this document says so in its own footer. What a page like this can realistically do is make it embarrassing for the question to keep going unasked — and give whoever eventually does the study a running start on the geometry, the constraints and the right-of-way problem.
Junction node and all five approach centrelines pulled from the OpenStreetMap Overpass API on 31 August 2026. Bearings computed by great-circle initial bearing from the junction node to the first geometry point beyond 35 m. Lane counts and road classification are the OSM tags as mapped.
Mixed traffic converted to PCU before any capacity calculation, using factors from Indonesian HCM 1997 as applied in Philippine practice: motorcycle 0.5, tricycle 1.0, car 1.0, jeepney 1.5, bus 2.5, two-axle truck 2.0, three-axle-plus 3.0. The assumed mix yields 1.02 PCU per vehicle — motorcycles offset the heavy vehicles almost exactly.
Both modelled as phased control with opposing pairs running together. Webster optimal cycle for the signal case, capped at 180 s; a fixed 150 s cycle for the enforcer case with 8 s lost time per phase against the signal's 4 s. Saturation flow 1,900 pcu/h/lane adjusted for lane width, kerbside bus stops and lane discipline — 1,345 for the enforcer case, 1,542 for the signal. Delay by HCM uniform-plus-incremental.
HCM 6th Edition Chapter 22 entry capacity against conflicting circulating flow: single-lane entry 1420·e−0.85×10⁻³·vc; two-lane entry adds 1350·e−0.92×10⁻³·vc. Entry lane counts were sized, not assumed — the smallest count holding v/c ≤ 0.85 in the 2036 design year, which returns two-lane entries on four of the five legs. Bypass lanes were tested rather than assumed, twice over: a slip lane can only reach the next leg counter-clockwise, and of those five pairs only DRT Hwy north → Pulilan Rd west deflects under 90° — and that quadrant is the fire station, so the modelled design carries none. A 4% capacity reduction is applied at every entry for pedestrian crossings.
Every building footprint within 200 m was pulled from OpenStreetMap and its nearest edge measured to the junction centre — not its centroid — giving a clear radius in every direction. The aerial view under the simulation is OpenStreetMap raster imagery at zoom 19, embedded in the page rather than fetched, pinned to the same junction node and drawn at 0.2885 m per pixel. The simulation runs at that same scale, which is why the vehicles are the right size and the approaches land on the real carriageway. One correction is painted onto that imagery: OpenStreetMap misspells the fire station as "Pulinan", so the building is repainted in the style's own colours and relabelled. It is the only alteration to the map.
Computed geometrically, not quoted. All 20 movements drawn as chords between right-hand entry and exit offsets at true bearings; pairwise intersections counted, coincident points merged. Merge and diverge counts follow n(n−2) for five legs. The roundabout figure is the standard 2n.
Volumes are estimated, not counted — this is the central weakness and the reason for the data gaps flagged above. Delay above roughly 120 s/veh should be read as "failing" rather than as a forecast, since HCM formulas lose meaning beyond v/c ≈ 1.2 and do not model queue spillback into adjacent junctions. Crash modification factors are transferred from FHWA and are not calibrated to Philippine conditions or to a traffic mix this motorcycle-heavy. The right-of-way check uses mapped building footprints only — it finds no boundary, title or setback that is not drawn in OpenStreetMap, and a real land survey will find more constraints than this one did, not fewer.