I agree. My theory is it has to do with more long-period, synoptic-scale weather in the mid latitudes. That versus what we see in the tropics, which is more ephemeral - think airmass-type thunderstorms that ebb and flow throughout the day. I notice the longer and larger a contiguous weather pattern exists, the higher the chances it will be in the sim.
I think this goes back to the general issue the sim has with producing thunderstorms - we see these discrepancies in the southern plains and southeastern US in the summer as well. We don’t really see accurate depiction and evolution of those airmass-type storms because, again, they’re so ephemeral - random atmospheric chance of initiation dictates downstream behavior, until nighttime (usually). But these micro evolutions are often happening on a timescale of an hour or less - too fast for even computer models to predict accurately, other than just spamming a general area. That is, unless they form a solid, persistent line along a synoptic-scale frontal boundary, but even that seems to take a few hours to show up in the sim.
So apply that notion to the equatorial and subequatorial regions, where storm evolution is almost always ephemeral. Here’s a good example from the GOES East satellite:
During the day the visual picture is a mess of cirrus that obscure the actual storms, but infrared at night it’s easier to see the higher cloud tops that denote the individual cells. Note there are other satellite bands that can better indicate the storm structure - IR water vapor and cloud top phase (below) being among the best:
But again, we don’t know what MB is using to ingest and compute the output. This is where models fail and clash with observations. And satellite is going to miss a lot of smaller-scale, non-storm-related cloud structure.
I still think the best way to do storms is by using radar, which is the closest to real-time granularity we’re going to get. Everything else is kind of a WAG. But unfortunately radar is not super available worldwide (and there might be data-use limitations on top of that as well).
Bottom line, this all shows there’s a lot going on under the hood with regard to live clouds. It would be really nice to know what the official line is so we can all stop wondering and trying to figure it out. What kind of data drives the output - satellite, radar, METAR, modeling? All? What kind of timescale does it follow? Why do some areas seem to have better “in-between” coverage and less obvious METAR augmentation than others?
