The data model

A parsed record is a BestTracks bundle of three frames: observations (one row per fix), storms (one row per storm), and remarks (free-text notes).

The tidy observations frame

column	meaning
storm_id	stable id, "<year>-<serial>" e.g. 2020-001
year, serial	year and IMD serial number of the system
basin	BOB (Bay of Bengal), ARB (Arabian Sea), or LAND
name	cyclone name (blank for unnamed / older systems)
time	observation time (UTC, datetime64)
lat, lon	position (°N, °E)
ci_no	Dvorak CI / T-number
pressure	estimated central pressure (hPa)
wind	maximum sustained surface wind (knots)
pressure_drop	pressure drop / ΔP (hPa)
grade	ordered category: D < DD < CS < SCS < VSCS < ESCS < SuCS
oci, oci_diameter	outermost closed isobar pressure (hPa) & diameter (°)
step	0-based fix index within the storm
pos_suspect, date_suspect	QC flags (see Data quality)

The xarray Dataset

Laid out like IBTrACS: a ragged track becomes a 2-D (storm, step) grid. Per-fix variables span both dims; storm-level metadata (name, basin, year, …) are storm coordinates.

ds = bt.to_xarray()
ds.sel(storm="2020-001")["wind"].max()      # Amphan peak intensity
ds.where(ds.basin == "ARB", drop=True)      # Arabian Sea storms only

IMD intensity grades

grade	name	sustained wind
D	Depression	17–27 kt
DD	Deep Depression	28–33 kt
CS	Cyclonic Storm	34–47 kt
SCS	Severe Cyclonic Storm	48–63 kt
VSCS	Very Severe Cyclonic Storm	64–89 kt
ESCS	Extremely Severe Cyclonic Storm	90–119 kt
SuCS	Super Cyclonic Storm	≥ 120 kt