[katieHaylor (OP)]

Jiten asks:

How is the weather predicted? With weather balloons?

Can you help?

[alancalverd]

Balloons give us data on temperature humidity and wind vector at various heights. From this data plus surface, aircraft and satellite measurements, we can create a map of the movement of air masses. The interaction of, say, a cold dry mass with a warm wet one will push the wet air upwards to form clouds and eventually rain or snow along the line of contact.

The overall movement of air masses is governed by surface temperature which initiates large scale convection, say between the poles (cold, sinking air) and the equator (warm, rising air) that is distorted by the Coriolis force  - equatorial air is generally moving from west to east relative to the surface,  polar air has very little W-E momentum - so the convection cells break into swirls that mostly travel W - E, with precipitation along the boundaries of warm and cold air. Knowing the position and speed of these swirls gives us a good idea of likely weather  along their track.

I can recommend "Meteorology for Glider Pilots"  (Wallington) as a very readable introduction that takes you gently from  "red sky at night" up to private pilot/yachtmaster.  If it grabs you, there are plenty of degree courses for professional forecasting, and advanced courses for long-haul aviators and mariners.

[evan_au]

Weather forecasting today is done in two iterative phases:
1) Collect all the data you can (from balloons, ground-based weather stations, shipboard weather stations, planes and satellites) and integrate it into a computer model which tells you what the weather is right now. This gives you a world-wide 3-dimensional grid of flowing masses of hot and cold air with a known temperature and pressure (at various heights), passing over over land and sea with a known temperature.
2) Use this computer model to continue the path of these masses of air, accounting for the day/night cycle of sunlight input, to predict the temperature, pressure and velocity of these masses of air at points of time into the future.

According to reference B (below), phase 1 takes about 60% of the computer power.

There is a fair amount of randomness in weather processes, and there is some uncertainty in the input data. The best weather prediction models run (say) 10 different simulations, using input data which is modified within the known range of input data uncertainty. This can result in a weather forecast that says "60% chance of rain next Tuesday".
 
Running these complex computer models takes huge supercomputers. State-of-the-art weather models can now predict the weather up to 2 weeks into the future with reasonable accuracy.

Weather forecasting still has problems with "twisters" (common in the US midwest), which are perhaps 1km across. They are basically invisible on a world-wide grid with data points spaced 10km apart.

References:
A) If you want an overview: https://en.wikipedia.org/wiki/Weather_forecasting
B) If you want to know a lot about it (3 hour interview): https://omegataupodcast.net/326-weather-forecasting-at-the-ecmwf/ 

[Petrochemicals]

I believe it is mostly done through radar, current conditions and computer sumulations with reference to historic happening these days, many many senarios are played out and the most likely wins.