The Cloud Layer

Every cloud on Earth, as five weather satellites saw it, ten minutes at a time.

Five satellites hang over the equator, each watching its side of the planet and looking again every ten minutes. Their views are joined into one sky, and each cloud is set back over the ground it is really above.

The land is this week's clearest look from the same satellites. Day, dusk and night, the glint on the sea and the shadows of the clouds follow the Sun of that moment. The stars and the Moon are where they were.

The poles stay open: from the equator the satellites see them too obliquely. Night cloud is lit softly so it can be followed. The globe runs 75 minutes behind now, when the last of the data is shared.

Not sure where to look? Moments is a short list of places and times worth a look: a hurricane at its peak, a typhoon off Tokyo, the overcast over Burning Man. Each one opens the globe there, at that hour.

What is drawn, and where it comes from

The cloud. GOES-19 and GOES-18 (NOAA), Himawari-9 (JMA, shared by JAXA), MTG-I1 and Meteosat-9 (EUMETSAT) each send a full picture of their side of the Earth every ten minutes (Meteosat-9 every fifteen). We read each agency's own cloud retrievals: how thick the cloud is by day (optical depth, which the Indian Ocean's Meteosat-9 does not make), where there is cloud at all (the cloud mask) and how high its top is. From these we compute how opaque each patch of cloud is: by day from its optical depth, at night from the infrared alone, since no agency measures thickness in the dark. The five views are moved back over the ground beneath them and joined into one map, about 10 km to a pixel, from 66° north to 66° south. Everything runs 75 minutes behind now, because EUMETSAT shares its data freely once it is an hour old; the other satellites could be sooner, but the globe keeps one clock so the whole Earth is the same moment.

The motion between frames. Ten minutes is a long time for a cloud. We measure how each part of the sky moved from one frame to the next (optical flow, computed at a quarter of the map's width) and slide the clouds along that path as the globe plays, so they drift instead of dissolving. It is our estimate, not a measurement: the fastest cloud, a front or a jet stream, moves too slowly and then catches up.

Cloud-top height. The agencies' cloud-top heights, or where they have none, an estimate from how cold the top is. We use the height to cast each cloud's shadow on the ground (drawn three times longer than it really is, so it can be seen), to give the cloud relief, and to keep high tops lit after sunset below them.

The zoom. Close in, sharper tiles take over: about 2.4 km over the Americas (from GOES-19 and GOES-18) for every day the globe holds, and for the newest picture only, 1.2 km over Europe and Africa (MTG-I1) and about 0.6 km by day from four of the five satellites, made from the full-resolution scan when you stop to look. Meteosat-9 has no sharper level.

The ground. The land is the week's clearest look from the same satellites: for every spot, the clearest daytime views of the last ten days, averaged and balanced to match from one satellite to the next. Where the week never cleared, and for the sea, the ice and the poles, it is NASA's Blue Marble; close in, the week's colour is laid over Blue Marble's finer texture. The city lights at night are NASA's VIIRS map of the Earth at night.

The sky. The stars are the Yale Bright Star Catalogue, every star the eye can see, each at its place and brightness. The Milky Way is ESO's photograph by Serge Brunier (CC BY 4.0). The Moon is NASA's CGI Moon Kit, from the Lunar Reconnaissance Orbiter, placed and lit where it was at that moment. The Sun is not a picture: its position is computed for each moment, and it decides day, dusk and night.

Lightning. Every flash seen by the lightning mappers on GOES-19 and GOES-18 (NOAA) and the Lightning Imager on MTG-I1 (EUMETSAT), counted in cells about 80 km across for each ten minutes. Each cell is drawn as white light, brighter the more flashes it held. No lightning imager watches the Pacific or Asia, so there is none there. By day the light is lost against the sunlit cloud.

Rain. Over the Americas, GOES-19 and GOES-18's own rain-rate estimate every ten minutes, as late as the clouds. It is provisional, read from the infrared tops alone, and is replaced by NASA's IMERG when IMERG arrives about six hours later. IMERG merges microwave and infrared satellites; elsewhere it is the only rain, so the newest hours there are empty. Drawn as a blue tint through the cloud, deeper where it rains harder.

Water vapour. The satellites' own mid-level water-vapour band (6.9 µm on GOES and Himawari, 7.3 µm on the Meteosats), every ten minutes, day and night. We turn it into how moist the air is a few kilometres up, compared with the driest air nearby over the last two days, and draw it as a soft cool wash in the air beneath the cloud.

Lightning, rain and water vapour cover the last three days. The full list of sources, with their licences, is in the layer archive. NOAA's GOES and GLM data are in the public domain; Himawari-9 is credited to JMA and JAXA; IMERG and the NASA maps carry no restrictions; the Meteosat and MTG pictures contain modified EUMETSAT data.

Derek Lomas · j.d.lomas@tudelft.nl · dereklomas.me