How do we know that it will remain cloudy but won’t rain? Or if the clouds will surmount to any precipitation?
In: 121
I’ll preface this by saying that I’m a meteorologist and am still amazed almost every day that we can forecast with the accuracy that we do.
The atmosphere is a fluid that follows well known physical processes like fluids do (for an Explain Like I’m in College, see the Navier-Stokes equations). For example, if air rises, it will expand and cool. If it cools enough, eventually water vapor will nucleate or deposit into tiny things, and we get cloud droplets. With time, continued cooling will let those droplets grow large enough to start to fall relative to the nearby tiny droplets, and that means it’ll collide with and “absorb” other tiny droplets, growing larger and falling faster. Etc. Air will tend to move from high to low pressure. Things like that. At the short end (hours to days), weather forecasting is essentially an initial conditions problem, in that the future state (e.g., the temperature, moisture, and pressure at any location in 12 hours) depends mostly on the current state. This means that we need to know the state of the atmosphere everywhere *now* to know the state in the future. To do this, we needs lots and lots of observations. We use in-situ measurements (that is, measurements made where the instrument is) with things like thermometers, hygrometers, pressure sensors, and anemometers on the ground (well, 2-10 m above the ground), instruments on weather balloons, instruments on airplanes, etc. Unfortunately, we can’t physically put instruments to make measurements *everywhere*, so we also rely on remote sensing observations. These include things like radar and satellite that can “see” over large areas and volumes.
As a first order guess of tomorrow’s weather, we can just look at today’s weather and assume it’ll be the same tomorrow. In some places and/or at some times when the atmosphere is relatively steady or controlled by very slowly varying things (think of a small island surrounded by 80 deg ocean water in the tropics), this is a pretty good guess. We can also use climatology, or use the average of the past 30 yrs of data for Jun 24th to guess what tomorrow will be. For many areas in the mid latitudes or in the shoulder seasons, these may not be good forecasts at all.
Most of the time, climatology or persistence can get you in the ballpark, but more accurate forecasts require that we use the physics equations that describe the movement and characteristics of fluids. This is done on supercomputers. It takes a *lot* of data and processing time, particularly when we try to narrow down how precise of an area we want to start making forecasts for. Maybe we can model the atmosphere on a grid that has points every 50 km, but that won’t be great if we want to capture things that occur on scales smaller than that. We then need to increase the resolution by, for example, changing the grid spacing to every 5 km, which also requires that we update the model more frequently in time as well (which means more calculations and processing time). To model things like individual thunderstorms, we need to get down closer to 100 m grid spacing, and even smaller than that where we start having to deal with turbulent flows.
As the forecast time window increases to months and years, the nature of the forecast changes. Of course the initial state is important, but the forecast progressively becomes a boundary condition problem, wherein the energy balance on the “edges” of the atmosphere (e.g., the ground and ocean surface down low and the sun at the top of the atmosphere) starts to drive the future state. Of course, given the turbulent and “chaotic” nature of the atmosphere, we lose our ability to model very small things (like individual storms) for very specific areas when we go to get far off forecasts, but we can get a handle on the general, large-scale structure of the atmosphere.
Latest Answers