Tomorrow.io's radar satellites use machine learning to punch well above their weight

Those of We lucky enough has be session by A window can predict THE weather report just by look out, but For THE less privileged, weather report forecast And analysis East get better And better. Tomorrow.io just released THE results Since It is First of all two radar satellites, which, THANKS has machine learning, turn out has be competitive with bigger, more old school forecast technology on Earth And In orbit.

THE business has has been planning This assignment Since he was called ClimaCell, back In 2021, And THE results be released Today (And officially present has A meteorology conference Soon) to show that their advanced technology approach it works.

Weather report prediction East complex For A plot of the reasons, but THE reciprocal between high power but legacy material (as radar networks And older satellites) And modern software East A big A. That Infrastructure East powerful And precious, but has improve their to go out requires A plot of work on THE calculation side — And has a few indicate You to start get decrease returns.

This is not it just "East he while going has rain This afternoon" but more complex And important predictions as which direction A tropical storm will move, Or Exactly how a lot rain fell on A given region on A storm Or drought. Such knowledge are more and more important as THE climate changes.

Space East, of course, THE obvious place has invest, but weather report Infrastructure East prohibitively big And heavy. from NASA Global Precipitation The measure Satellite, THE gold standard For This field spear In 2014, uses both Ka (26-40 GHz) And Ku (12-18 GHz) band radar, And weighs a few 3,850 kilograms.

Tomorrow.io plan East has create A new space based radar Infrastructure with A modern twist. It is satellites are little (only 85 kilograms) And to use THE Ka band exclusively. THE two satellites, Tomorrow R1 And R2, spear In April And June of last year, are just NOW, After A long period of shaken And essay, beginning has to show their quality.

In A series of experiences that THE business East planning has publish In A newspaper later This year, Tomorrow complaints that with only A radar band And A fraction of THE mass, their satellites can produce results on by with from NASA GPM And on the ground systems. Through A variety of Tasks, THE R1 And R2 satellites were able has TO DO in the same way accurate Or even better And more accurate predictions And observations as GPM, And their results Also account closely with THE ground radar data.

Examples of data Since THE R1 And R2 satellites. Picture Credits: Tomorrow.io

They accomplish This However THE to use of A machine learning model that, as Chief Weather report Officer Arun Chowla describe he, actions as two instruments In A. He was qualified on data Since both of THE GPM radars, but by learning THE relationship between THE observation And THE difference between THE two radar signals, he can TO DO A similar prediction using just A band. As their Blog job puts this:

THE algorithm East qualified with these derivative of the double frequency precipitation profiles but only uses THE Ka band observations as to input. Nevertheless, THE complex relationship between THE reflectivity profile shape And precipitation East "learned" by THE algorithm, And THE complete precipitation profile East recovered even In case Or THE Ka band reflectivity East completely attenuated by heavy precipitation.

It is A big success For Tomorrow.io if these results stove ...