The map that was never measured
A weather map arrives as a smooth sheet of color. Temperature, pressure, wind, and rain seem to exist at every point, all at the same instant. It is natural to read the image as a photograph of the atmosphere. Now erase every location that was not directly sampled. What remains is a scatter of surface stations, balloon profiles, aircraft tracks, radar volumes, and satellite swaths, each with its own timing, footprint, and error.
The continuous field did not come from a continuous instrument. It was assembled by asking what atmospheric state is compatible with those measurements and with a model of how nearby places, variables, and times are related. The same is true below the ocean surface, where profiles are separated by immense unsampled volumes, and below the ground, where no instrument can directly see the structures inferred from waves.
A reconstruction can be extraordinarily useful without being a photograph. It combines evidence with an observation process, a dynamical or statistical model, and assumptions about error. Dense observations can anchor much of the result. Sparse regions depend more strongly on relationships supplied by the model. If those relationships are wrong, an apparently seamless map can carry a seamless error.
The complete picture was never observed. It was reconstructed. That sentence is not a reason to distrust every map. It is a reason to ask a better set of questions: where did the information come from, which directions were invisible to the observing system, how did assumptions fill the gaps, and how uncertain is the answer?