What this shows

Oslo Bysykkel riders end each day lower than they started. This chart tracks that in bike-metres — one bike moved one vertical metre — across a single day.

The top panel is hourly flux: metres descended below the line, climbed above. The bottom panel accumulates it. The gap between where that curve ends and zero is the lift obligation: metres of elevation someone has to put back.

Across July 2026 (175,048 trips) riders net-descend an average of 33,566 bike-metres per day. The worked example here is 8 July — a dry, mid-week, peak-season day with 5,893 trips and a closing deficit of −43,526 bike-m. Note it runs about 24% hot per trip against the monthly average, so treat the monthly figure as the headline and this day as illustrative.

The striking part is the shape. Every hour from 04:00 to 20:00 is net downhill. Not just the morning peak — the evening commute never reverses it. Climbing does peak in the late afternoon (17:00 is the biggest climbing hour of the day at 5,105 bike-m) but descent that same hour is 6,388, so the deficit still deepens. The curve bottoms at −44,476 around 20:00 and claws back barely 950 metres overnight, from a handful of late trips heading home uphill.

So this isn't an imbalance that self-corrects on a daily cycle. It accumulates and stays.

Why I'm looking at this

I'm testing whether Oslo could run something like Citi Bike's Bike Angels — the scheme where riders earn points or credit for moving bikes from full stations to empty ones. In New York it handles roughly 30–40% of all rebalancing, so it's a serious alternative to van fleets rather than a gimmick.

This chart is the first check on whether that transplants. And it points somewhere awkward.

In Manhattan, angels work dense flat clusters — surplus and deficit stations sit a few hundred metres apart on level ground, so a rebalancing move is a three-minute hop. In Oslo the deficit is uphill, everywhere, all day. Every rebalancing move is a climb. That's precisely the move an unassisted cyclist is worst at, and it's why the naïve "pay people to ride bikes back" model may not survive contact with Oslo's terrain even though it works in New York.

The corollary is more interesting: an electric bike does that same lift for a few øre of electricity. Which reframes the white bikes as rebalancing machines that customers pay to operate, rather than as a comfort upgrade.

How it's computed

Per-trip Δh is elevation[end station] − elevation[start station], summed by hour. Over a closed cycle where the fleet begins and ends distributed identically, Σ(rider Δh) + Σ(operator Δh) = 0 — so the rider side, sign-flipped, is the operator's obligation. Trip data records only riders; rebalancing moves generate no trip record.

Measured: station elevations, trip endpoints, timestamps. Derived: the debt figure, by arithmetic on those — not a fitted or assumed model. That makes this firmer than any cost-based estimate, which moves as soon as you change an assumed wage or vehicle capacity.

Endpoint caveat: Δh is start-to-finish, not metres physically climbed — a route between two low stations can still cross a ridge. The conservation identity depends only on endpoints, so the result holds, but nothing here should be read as a claim about rider effort.

Known issues and next steps

Occupancy correction not applied to the daily figure. Bikes don't end the day exactly where they started, so part of −43,526 reflects a shifted fleet rather than debt requiring physical lift. This washes out over a month, which is the second reason to lead with 33,566.

A separate rebalancing-move calculation for the same day produces a similar total lift. That near-agreement is a useful bug check — it confirms neither calculation has a sign error or dropped records. It is not independent corroboration: any valid rebalancing solution must undo exactly the displacement riders created, so the two figures could not have disagreed.

One numeric discrepancy is open. A companion station-level breakdown reports mean pickup and drop-off elevations implying roughly 29,800 bike-m for that same solution, against 42,800 cited here. Most likely weighted versus unweighted means, but unresolved — don't multiply those figures against these until it's settled.

Concentration is unresolved. Ranking station pairs by contribution put the top 10 at 2.3% of monthly debt. That's roughly 97× a uniform share, so it is concentrated — but pairs is the wrong unit, since one problem sink spreads its debt across every pair feeding it, and you can't relocate a pair. The decision-relevant number is how many stations account for 50% of the debt. Until that exists, no conclusion about fleet-wide fixes versus station relocation.

Fleet changes unquantified. Bikes entering or leaving maintenance move elevation without a trip record. Reported as a residual, not folded in.

Season. July 2026, full fleet. The winter studded-tyre fleet is effectively a different network and isn't pooled.

Sources

Oslo Bysykkel GBFS and historical trips, NLOD 2.0 (data.urbansharing.com), Client-Identifier: novem-bike-angel-analysis. Station elevation: Open-Meteo (CC BY 4.0).