Tutorials

Tutorials#

Four runnable notebooks — each end-to-end on the public archives, no credentials. Take them in order; each one builds on what the last one taught.

Tip

Run Verify your install once before the first tutorial to confirm import xarray, icechunk and the connection to s3://nexrad-arco work in your environment.

Open weather radar archives in 5 lines of code

Open the KLOT (Chicago) archive on the AWS Open Data Registry as one xarray.DataTree, narrow it to a single VCP and sweep, and read back one scan of a July 2016 hailstorm as reflectivity.

radar datatree: Cloud-native, time-aware weather radar datasets

Get only the sweep you need

Only need the cuts closest to the ground? Open the KLOT-lowsweeps virtual reference archive and get sweep_0 and sweep_1 as continuous time axes over the whole record — no VCP bookkeeping, and no bytes copied. See why both sit at the same 0.5° elevation (the split cut), then plot Z, ZDR, RHOHV and PHIDP.

Get only the sweep you need

Reproduce Ryzhkov et al. (2016): traditional vs ARCO

Compute the QVP for the May 20, 2011 KVNX MCS via the traditional file workflow vs ARCO streaming, assert both produce the identical profile, and measure the gap — hours of preparation become seconds of analysis. The notebook re-measures on your machine rather than quoting a fixed number.

QVP Analysis: Traditional vs ARCO Workflows

QPE scaling: from hours to months

Marshall–Palmer Z–R rainfall accumulation, live for one day. Cluster-recommended copy-paste templates for 7d / 30d / 6mo runs included.

QPE Scaling Benchmark: From Hours to Months