Runa had mapped the same stretch of seafloor fourteen times.
Not because anyone asked her to. The first survey was a contract — cable route assessment, 120 kilometers of continental shelf. By the third pass she could have drawn the bathymetry from memory. Sediment drape over limestone, a basalt intrusion at kilometer 47, the sand-wave field west of the ridge where the current sorted everything into parallel lines.
By the fourteenth pass, the bottom looked exactly like the thirteenth.
This was supposed to be the point. Convergence. The survey resolving into its answer. Each pass confirming the last until the map and the territory were indistinguishable. Her supervisor called it survey closure — when additional data no longer changed the product.
What she had not expected was the way the water changed the signal every time.
Same transducer. Same pulse. Same bottom. But the sound took a different path each pass — bending through thermoclines that shifted with the tide, scattering off plankton layers that hadn't been there on Tuesday, refracting through a lens of cold water the current pushed in overnight. The ray trace was never the same twice.
Her colleagues treated this as noise. The sound velocity profile was a correction — something you measured, compensated for, and discarded. The water column was the thing between you and your answer.
Runa had started saving the profiles.
Not the corrected ones. The raw bends. Fourteen passes through the same water, and no two of them agreed. She told herself she was being thorough. Mertens, who ran the processing lab, told her she was being inefficient. The bottom was the deliverable. The water was the error.
She did not argue with Mertens. He was not wrong about the deliverable.
What she could not explain — what she hadn't yet found language for — was that the bottom had stopped surprising her, and the water hadn't. The bottom gave the same answer every time, and each confirmation felt less like verification and more like the signal had been absorbed. Whatever she found in the fourteenth pass came from a different processing chain, a new frequency, a different viewing angle. It came from her. The bottom hadn't moved.
The water was different. It surprised her because it had also changed. The resistance was not in her approach to it but in its own motion. It pushed back from a position she hadn't generated.
"The bottom converges," she told Mertens once. "The water doesn't."
He looked at her the way people look at hydrographers who've been at sea too long.
What she had never done was point the transducer upward.
You could, in theory, image the sea surface from below. The air-water boundary was the strongest reflector in the ocean — the impedance mismatch between water and air was orders of magnitude greater than anything on the bottom. Every sonar operator knew the surface return. They called it surface ghost. They filtered it out.
Runa had been filtering out the strongest signal in her data for twenty years.
She didn't turn the transducer. That would have required a proposal, a justification, a paragraph explaining why a hydrographer wanted to image the sky. Instead she stopped removing the surface ghost from her downward surveys. She let it back in.
The bottom: stable, fully known. The water column: shifting, alive. And above it — reflected in the pulse that came back too early, arriving before the seafloor return — the surface. The boundary where her medium met one it could not enter.
The strongest echo came from the place where the signal could no longer continue in the same form.
She filed the unfiltered surveys alongside the corrected ones. The corrected ones went to the client. The unfiltered ones went nowhere. They contained the bottom, the water, and the surface — the answer, the path, and the boundary where the question changed medium.
No one asked about them. The bottom hadn't changed, so the product was the same.
The water was different that day. It always was.