Typhoon Champi Shuts Ningbo and Shanghai Windows: Gates, Pilotage, Then the Backlog

The Analysis

Spread the map out and put the version from ten years ago beside it. The East China coast has not moved: two gateways, one estuary, one pilot station, a feeder web running up the Yangtze that nobody bothered to draw in 2016 because it simply worked. Ten years ago this was another matter, mostly because the load on those bones was lighter and the ships calling there were smaller and more forgiving of a lost day.

What has actually changed is the net. Typhoon Champi is now moving across Taiwan and into East China, and the forwarder advisory issued on 23 October puts Ningbo down from the evening of the 23rd until about midday on the 25th, and Shanghai down from the evening of the 24th until about midday on the 25th. Two of the busiest gateways on earth, closing inside the same thirty-six hours, on a coast that had already been running behind.

The advisory also says something else, in a subordinate clause that most people will read past: pilotage on the Yangtze estuary has already been suspended, and it is already holding up feeder movements into Shanghai. That clause, not the two closure windows, is where this story lives. Everything below is an attempt to show why.

Start with the hard lines and nothing else. Ningbo is expected to suspend operations from the evening of 23 October until around midday on 25 October, which is roughly forty hours. Shanghai is expected to suspend from the evening of 24 October until around midday on 25 October, which is roughly eighteen hours, and it begins a full day after Ningbo. Both schedules remain subject to change, and that hedge is part of the problem rather than a footnote: you are planning against a window the port has not promised you. On top of that, both ports were already congested before the storm arrived, and airfreight via Taiwan, Shanghai and the surrounding region is running twenty-four to forty-eight hours late with a backlog behind it, airports still open but cancellations reported.

Why now, and why the two ports do not close on the same clock, is a question of geometry rather than policy. Champi is tracking up the coast, and Ningbo sits south of Shanghai, so the same system reaches Ningbo's approaches first and Shanghai's later. Ports do not shut when the wind at the berth crosses a line; they shut when the sea state makes pilot transfer unsafe and crane work impossible, and that decision is taken by each harbour master against his own forecast, port by port. The result is two windows that overlap but do not coincide, with closing times published firmly and opening times wrapped in the word about. That asymmetry between a firm stop and a soft start is the first thing worth carrying into your planning.

Now the anatomy, because a port closure has two layers and they do not move together. The gate is a decision, taken by a person, reversible in an hour. The water is a physical constraint, and it has a lag on both ends. Pilotage is suspended before the worst of the weather arrives, because pilots are transferred by boat and boats stop running when the swell does; and pilotage comes back after the weather has gone, because it takes time to re-establish boarding grounds, re-crew the station, and work down the queue of ships that have been waiting in the roads. The Yangtze estuary is a single throat: one pilot station, one set of boarding grounds, one tidal regime. When that throat closes, Shanghai's feeder intake stops on days when the port itself would otherwise still be working.

Trunk lines feel this differently from feeders. A deep-sea call lost at Ningbo is not a lost voyage but a shifted rotation; the vessel either omits the port and leaves a pile of boxes behind, or it waits, and the slide then propagates down the string for the next two or three calls. Vessels on an Asia to North Europe service run round trips measured in weeks, so a berth day lost on 24 October is not recovered on 26 October; it is recovered on the next rotation, five to seven weeks out, if it is recovered at all. This is the part that surprises shippers who think in closure length. The forty hours on the notice is the time the cranes are still; it is not the time the schedule is disturbed, and the two numbers are not the same number.

Feeders take it on both ends, and worse. A Yangtze feeder that cannot transit the estuary cannot deliver boxes to the mother vessel and cannot receive them either, so the cargo sits at the origin terminal while the clock runs. For a shipper in Nanjing or Wuhan whose box was meant to reach Shanghai for a sailing on the 24th, the effective stop began on the morning of the 23rd when pilotage went, not at Shanghai's gate closure that evening, and it will not end at midday on the 25th, because the pilots return after the sea state does and then work through what accumulated while they were off station. The published Shanghai window is eighteen hours. The window this shipper actually experiences is something else entirely, and no advisory will print it.

Inside the terminal, the variable that governs the first week back is yard density rather than crane capacity. Export laden boxes keep arriving by road until the gate stops accepting them, and the yard fills past the point where it can be worked efficiently; heavy boxes sit where they were put, on top of stacks that will have to be unpicked. Import boxes that were due for collection before the storm are now sitting through it, burning free time on a terminal that is not open and cannot release them even if a truck arrives. On the first day back, the constraint is not how fast you can lift but whether you can find the box, whether there is a chassis under it, and whether the gate appointment system will let the truck in at all. Anyone who has worked a reopening knows which day is worst: not the last day of the closure, but the second day after it.

The air option deserves its own look because it is the reflex. The advisory puts air delays at twenty-four to forty-eight hours across Taiwan, Shanghai and the surrounding area, with the airports still operating but cancellations reported. Nadia Pryce, who tracks what a twenty-four to forty-eight hour air slippage does to the parcel side, would read that number as a shed problem rather than a flight problem: the aircraft still move, the ground side does not, and the backlog sits in the warehouse. So the modal switch people reach for when the sea gate shuts is itself running two days behind, and it is being reached for simultaneously by every shipper in the same catchment.

You are not buying the difference between ocean and air transit; you are buying that difference minus two days, minus the space that was gone before you called, in a market that prices scarcity faster than you can get a quote back. For genuinely urgent, high-value, low-volume cargo it is still the right call. For the mid-value cargo that everyone will try to push upstairs on the 25th, it is a very expensive way to arrive two days earlier than the ship.

Timing is where this becomes concrete, so work backwards from your own documents. Export boxes with a cut-off on the 23rd or 24th are the ones caught inside the closure window; anything you planned to gate in on the 25th or 26th arrives into a rationed appointment system, so assume the gate-in window you had before the storm is not the one you will get after it. On imports, boxes that were released and ready before the storm should have been pulled before the gate shut, and those still on the terminal will join a collection rush on the first day back that competes with every other truck in the catchment.

For second legs, a missed mother vessel means a rebooking onto the next string, and on a lane that was already congested the next sailing is itself tight, so the rebooking cost is measured in sailing intervals rather than in days. On inventory, the honest test is cover: if your destination DC is running on less than about twelve days of stock, an event of this shape puts you into a stockout before the second vessel arrives, because a five to seven day slip plus a weekend plus the rebooking interval eats twelve days faster than most planners expect.

Here is the part the coverage does not say, and it is the one that costs money. The binding constraint is pilotage, not the gates. Feeders into Shanghai stop before the port does and start after it reopens, so the effective outage is longer than the published window, and it lands on a system that was already queued before the first crane stopped. Read the advisory as a gate document and you will plan for forty hours at Ningbo and eighteen at Shanghai. Read it as a pilotage document and you are planning for a three-day outage on the estuary that began before either port closed and will finish after both have reopened. The second reading is the one your cargo experiences.

The second half of the same point is that a storm closure is not symmetric. Vessels do not stop arriving when a port shuts; they anchor. So the queue that exists at the moment of reopening is the queue that was already there before the storm plus everything that would have worked through the terminal during the closure, and that combined pile does not drain at the rate it formed, because the terminal comes back below full speed for the first day or two and the pilots are clearing the roads backlog at the same time. Recovery is therefore a multiple of closure length, never equal to it, and the multiple is worse when the port was congested going in. Both ports were congested going in. That is the whole of the schedule slip, and it is invisible in the headline.

Let me put numbers on it, with the assumptions stated so you can argue with them. Assume your box is an export laden forty-footer out of Ningbo, gated in on 22 October, booked on a sailing with a cut-off on the evening of the 23rd and a planned load on the 24th. Assume the published Ningbo closure holds at forty hours, which is 1.7 days. Assume estuary pilotage, already down on the morning of the 23rd, comes back twelve hours after Shanghai's midday reopening on the 25th, so roughly midnight into the 26th; that is a three-day pilotage outage for Shanghai feeders against a published Shanghai window of eighteen hours, a ratio of about 1.8 times.

Assume that before the storm both ports carried about one day of berth queue, which is my number, put on a congestion the advisory describes but does not quantify. Assume arrivals continue through the closure, so the queue grows by about one day of work per day of closure. Assume finally that on reopening the terminal runs at about 0.9 times normal throughput for the first two days while yard density and gate appointments and pilot slots are sorted out.

Run it. At Ningbo, the queue at reopening is one day pre-existing plus 1.7 days formed during the closure, so 2.7 days of work. Draining that at 0.9 times takes about 3.0 days. Add the 1.7 days of gate outage and the total schedule disturbance is about 4.7 days, call it five. For a Shanghai feeder box, the outage is the three-day pilotage window, the queue at reopening is one plus three, so four days of work, draining at 0.9 times takes about 4.4 days, and the total is around seven days. Now put that against a delivery promise. Assume a thirty-day transit to North Europe and a promise to have the goods at the DC three days after arrival: a 24 October load would have put arrival at 23 November and delivery on the 26th.

Five days of slip puts arrival at 28 November and delivery on 1 December. That is not five days late, it is the wrong month, and for any customer working to a December shelf plan or a month-end cut it is a different conversation. On the feeder box, seven days of slip puts delivery around 4 December and puts a DC holding ten days of cover into a stockout before the replacement arrives.

The counterfactuals matter, because the forecast will move. If Champi tracks further east and the closures shorten to a single day each, the gate outage drops to 1.0 day and the queue at reopening to 2.0 days of work, which drains in about 2.2 days, so the total disturbance falls from five days to roughly three. Better, but still three times the published closure, because the multiplier does not go away when the storm does. If the track shifts west and Ningbo takes a closer pass, add a day to the Ningbo window and the arithmetic above gives about six days of slip rather than five, and the yard-density effect gets worse than linear, since an extra day of trucks arriving into a closed gate compounds what can be unpicked on reopening.

If recovery runs fast, say 1.3 times normal because the port pushes a surge shift, the drain on 2.7 days of work takes about 2.1 days and the total falls to around 3.8 days. If recovery runs slow, say 0.7 times because yard density is worse than expected or pilots are short, the drain takes about 3.9 days and the total climbs past 5.6 days. Same storm, same closure notice, a two-day spread in outcome depending entirely on a parameter nobody publishes.

There is a disagreement worth naming, because you will hear both versions this week. One reading, common among forwarders who work from the port notice, is that this is a forty-eight hour event and the backlog clears by the weekend. The other, held by anyone who has watched an estuary reopen, is that this is a week. Neither side is disputing the weather. They are reading different clocks: one reads the gate, the other reads the queue. The disagreement also has a structural edge that regional diversification does not solve. Ningbo and Shanghai are close enough to sit under the same system, and the same system is crossing Taiwan, which is why the air alternative out of the region is late at the same time the sea gates are shut.

Moving a load port south to Xiamen or Shenzhen buys distance from the track and costs days of feeder or road movement into a different congestion. Transshipping via Busan keeps you outside the weather and adds a leg and a hand. None of those are wrong; all of them are more expensive than the alternative they are replacing, and the cheapest is still expensive when the whole region is moving at once.

Over a ten-year scale, is this a pulse or a turn? Spread the map out again and compare 2016 with today: the bones are the same, and what has changed is that the ships calling are larger, the berth windows are tighter, the feeder web up the estuary is denser, and each call moves more boxes than it used to. A forty-hour closure in 2016 was absorbed by slack that no longer exists, which is why the same weather now buys a longer tail. That makes Champi a pulse, not a turning point. What would turn it into one is frequency rather than intensity: if multi-port East China closures of this shape start arriving several times a season, carriers will be forced to build buffer back into schedules they have spent a decade compressing, and the baseline transit you quote your customers will have to be restated. That is the test, and it is a test you can watch from your own booking data rather than from a forecast track.

So the practical posture is not panic and not shrugging. Treat the published windows as the optimistic case and plan the queue, not the gate. For export, anything with a cut-off inside the 23rd to 25th range should be assumed to miss the original sailing, and rebooking should start now rather than on the 26th, because the queue at the booking desk forms faster than the queue at the berth. Do not rebook onto the next available sailing without checking whether its cut-off has itself moved and whether your box will physically be out of the yard in time; the common failure after a reopening is not missing a vessel, it is booking one you then miss twice.

For import, pull what you can before the gate shuts and expect the first collection day back to be rationed. Watch free time on boxes already on the terminal, since it burns through a closure the same way it burns through a weekend, and check whether your contract treats a port closure as an event that suspends it; the answer varies and it is worth knowing before you argue about the bill rather than after. For inventory, the twelve-day cover line is the one I would work to on lanes touching this coast over the next three weeks, and if you are under it, the honest options are to move the next order forward by one sailing interval or to accept a stockout and tell the customer now, while it is still a forecast rather than a failure.

The most common mistake this week will be believing the reopening. A port that opens at midday on the 25th is not a port that works normally on the 25th; it is a port that begins a two or three day unwind of a pile that was already there before the storm and grew while the cranes were stopped. The second most common mistake will be treating the air number as clean. Twenty-four to forty-eight hours of air delay is the delay before you add the backlog, and it applies to a mode that everyone in the catchment is trying to use at once. Watch the pilotage notice rather than the port notice, because the pilotage notice is the one that tells you when feeders actually move, and it is the one that has already moved against you.

Spread the map out one more time in a week, when the cranes are turning again and the yards look calm from the outside. The two gateways will be in the same place, the estuary will be in the same place, and the ships will be moving. What will have changed is the queue, invisible on any map, working its way through the net for another few days while the published timetable insists that everything is back to normal. Ten years ago this coast had slack enough to absorb that. It does not now, and the difference between the two maps is not the storm. The storm is weather. What changed is the net.

↑ Back to the key points

— By Tomás Ferreira