Skip to the timeline
Let's go to Mars
c. 3.8 billion years ago

Rivers run, and an ocean nobody can prove

Thousands of branching valleys and a delta in Jezero crater show water flowed for a long time. Whether it ever pooled into a northern ocean is still argued.

An overhead view of rust-brown highland terrain cut by a fine, many-branched web of dry channels draining out of high ground, with a large flat-floored crater at the left edge.
Dry valleys branching across the southern highlands, imaged from orbit by Mars Express at about 14 metres per pixel — the landform that needs rain or groundwater, and a long time. · Image: European Space Agency

Valley networks cut the southern highlands of Mars: dendritic drainage systems, some thousands of kilometres long. The landform needs sustained precipitation or groundwater sapping, not a single flood, and a long time to carve. Most date to the late Noachian and early Hesperian, 3.8 to 3.6 billion years ago.

Perseverance resolved the sediment fan in Jezero crater into a true delta: inclined strata laid down where a river lost speed entering standing water, capped by boulders no lazy river moves. One of two confirmed lakes among hundreds of candidates, it confirms the landform, not the climate.

Climate models do not reproduce it. The Sun then was 25 percent fainter, and models struggle to lift the surface above freezing even with a thick carbon dioxide atmosphere: thicker CO₂ scatters sunlight away while trapping heat, and condenses into cloud that flattens the lapse rate. Proposed rescues are partial: hydrogen with CO₂ giving collision-induced absorption, bursts of methane, impacts delivering transient warmth. A growing camp argues Mars was mostly frozen, only episodically wet.

Since the 1980s two candidate shorelines have been traced around the northern lowlands: the Arabia and Deuteronilus contacts. A sea surface is an equipotential; these wander by kilometres in elevation. Robert Citron and colleagues argued in 2018 that Tharsis did it: the shorelines formed before and during the rise's emplacement, and its growing load flexed the crust. The older Arabia contact should then be deformed more than the younger Deuteronilus, and it is. If that holds, any martian ocean is early, contemporaneous with the valley networks.

In April 2026 Abdallah Zaki and Michael Lamb changed the target. On Earth an ocean's clearest topographic fingerprint is not the shoreline but the broad flat band of coastal plain and continental shelf behind it, roughly −410 to −15 metres. They found a martian equivalent: a circumglobal flat zone between about −1,800 and −3,800 metres, 10.2 million km², around 7 percent of the surface. The strongest geomorphic case so far, and not proof.