[Pycon] [new paper] "Michael Salib" - Let's predict the weather! Building a real time global rain forecaster.

[info a pycon.it]

Title: Let's predict the weather! Building a real time global rain forecaster.
Duration: 60 (includes Q&A)
Q&A Session: 15
Language: en
Type: Talk

Abstract: We're going to write a simple short term precipitation forecaster together. Along the way, we'll learn about how to handle weather and geospatial data.

We're going to use free data from NASA's Global Precipitation Mission (GPM). That will give us precipitation rates for every 0.1 degree tile on Earth, updated every 30 minutes with a 3 hour delay. That data tells us exactly where and when it is raining. Starting from that data, we're going to perform a simple optical flow to determine where rain is moving to.

Along the way, we'll learn how to deal with rasterized gridded weather data products. That means learning:

 - how to handle weird data formats
 - how to perform simple geospatial transforms since data variables are stored in an array where each point corresponds to a point on a grid projection
 - because GPM data covers the entire world and has many data fields, it is a massive data set, so we'll use a special protocol to pull down only the data fields we need in the region of time and space that we care about

I'm focusing on intermediate Python programmers who have had some experience with Numpy or Pandas but know little about weather or geospatial data processing. That definitely includes data scientists who are not application developers. There is a huge amount of weather data and satellite imagery available for free that could drive all sorts of cool applications and analyses but there are too many barriers to entry in this field. I want to dismantle some of those barriers.

I want attendees to walk away knowing how to find and fetch free weather data, the basics of how geophysical datasets are stored (i.e., gridded data rasters with metadata), and how to extract a geographical and temporal subset from a large rain dataset. I also want them to develop a sense for how we measure meteorological data and how reliable those measurements are (and when they're not).



Tags: [u'pandas', u'pydata']