In an age when automated weather stations upload data every minute, someone here still opens a wooden door four times a day and lifts paper from a drum.

Wind on paper

The weather station stands two kilometres north of the end of the road. A white Stevenson screen, a tipping-bucket rain gauge, and an anemometer mast held by three guy wires are scattered across unevenly cut grass. The station hut is small: there is room for one table, two banks of drawers, and three people turning at once.

At 6:50, keeper Xu Cheng wound the mechanical clock and opened the wind-direction recorder. The drum turned at a constant speed while the ink needle moved sideways with the vane outside. There were no numbers on the paper, only a blue line that ran straight, then suddenly doubled back.

“Automated data tells you the wind is seven metres per second. This line tells you how many times it hesitated in ten seconds.”

Manual observation does not mean rejecting electronics. Automated sensors also sit on the roof, and the two systems have operated in parallel since 2003. The value of the manual instruments is a continuous, tangible reference: when sensors are replaced, algorithms updated, or networks interrupted, the paper can still show what actually happened at that moment.

Geometric illustration of a manual weather station and grassland
The Stevenson screen south of the hut must remain ventilated, shaded from direct sun, and about 1.2 metres above the ground.

Failure is part of the record

The archive drawers hold dozens of paper rolls marked with red triangles. They do not indicate extreme weather, but days when an instrument failed: frozen ink, a clock running seven minutes slow, insects inside the rain gauge. The keeper never cuts those sections away, but writes the cause on the back.

At eight, the cloud base began to rise. Xu Cheng entered visibility, cloud cover, and ground condition in the observation book, then compared the thermometer outside with the automated station. The readings differed by 0.2 degrees, within tolerance. He closed the book. The door would next open at noon.

Four times a day

06:50Morning reading

Minimum temperature, precipitation, cloud cover, and ground condition.

12:00Noon check

Compare the manual thermo-hygrometer with the automated sensors.

18:00Change paper roll

Archive the wind and pressure records and refill the ink.

00:00Remote watch

The automated station uploads overnight; a person attends if an anomaly appears.