Using Machine Learning to “Nowcast” Precipitation in High Resolution

I chime in on this because I&#x27;m working on deep learning for precipitation nowcasting using radar for my Ph.D. I was very excited when google released the press statement at NeurIPS about their work in this area. Unfortunately, after reading the paper, I have to say that their approach is fairly basic. Basically they threshold the precipitation in 4 thresholds (no rain, light, medium, heavy) and then use a U-Net like architecture, treating it as a classification problem. I think that the works of Shi et al are much more interesting in this regard:<p><a href="https:&#x2F;&#x2F;papers.nips.cc&#x2F;paper&#x2F;5955-convolutional-lstm-network-a-machine-learning-approach-for-precipitation-nowcasting.pdf" rel="nofollow">https:&#x2F;&#x2F;papers.nips.cc&#x2F;paper&#x2F;5955-convolutional-lstm-network...</a><p><a href="http:&#x2F;&#x2F;papers.nips.cc&#x2F;paper&#x2F;7145-deep-learning-for-precipitation-nowcasting-a-benchmark-and-a-new-model.pdf" rel="nofollow">http:&#x2F;&#x2F;papers.nips.cc&#x2F;paper&#x2F;7145-deep-learning-for-precipita...</a><p>What I think is that Google wanted to use a lighter model that can be applied to the whole continental US. I expect them to integrate this in google assistant, like: &quot;hey google, tell me when it&#x27;s going to rain&quot;

I don&#x27;t get why the put nowcast in quotes. Nowcasting is pretty common term in meteorology.<p>It&#x27;s also cool that they could get this far without physics. Of course, &quot;HRRR model begins to outperform our current results when the prediction horizon reaches roughly 5 to 6 hours&quot;. Simple associations that could be discovered by neural networks work in the short term, but the atmospheric physics is needed to understand the long term evolution of storms.

In the UK, netweather.tv shows weather radar with a 5 minute temporal resolution, 50 meter spacal resolution, and a 10 minute latency (less if you pay I think).<p>I regularly use it &#x27;by eye&#x27; to predict when a big band of rain is coming. I can very effectively figure out if there will be more or less rain 5 minutes from now. &quot;Shall I walk to the car now, or should I wait 5 mins?&quot;.<p>In comparison, the results of this work seem disappointing.

A physics-free approach to precipitation nowcasting using a CNN.

Sounds a lot like darksky.net they have an interesting story starting from kickstarter to getting acquired by longnow.

dup: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=22051889" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=22051889</a>

Now cast, let me look outside