Skip to content

Hurricane tracking — Hurricane Ian, September 2022

Reconstruct a major hurricane's track and intensity from buoy, satellite, and tropical-cyclone-heat-potential data.

What this teaches

  • How to combine multiple complementary collections in one analysis
  • Using aggregate with geometry_geohash_grid_frequency to find observation hot-spots
  • Using column_stats_summary to filter buoys by extreme wave heights
  • Following the same storm across in-situ and satellite sources

Sources used: NDBC, GOES_R, NOAA_AOML_HDB, NOAA_GDP

Prompt

Hurricane Ian made landfall in southwest Florida on September 28, 2022.
Find me NDBC buoys along its track that recorded wave heights over 6 meters
during the storm, GOES-R satellite imagery for that day, and any tropical
cyclone heat potential data from AOML for the days leading up to landfall.
Also tell me which Global Drifter Program drifters were in the path.

Transcript

1. Aggregate first — find observation density along the track

aggregate(
  aggregations = "collection_frequency",
  bbox = "-86,24,-80,30",
  datetime = "2022-09-26T00:00:00Z/2022-10-01T00:00:00Z"
)

2,448 items across 34 collections. Top contributors: - IOOS_SENSORS — 1,420 items - SECOORA — 207 - UAF — 200 - GOES_R — 153 (product-family items) - POLARWATCH — 65 - COASTWATCH — 56 - NOAA_AOML_HDB — 42 - NDBC — 6, NOAA_GDP — 2

2. NDBC buoys with wave heights over 6 m

search_datasets(
  collections = "NDBC",
  bbox = "-86,24,-80,30",
  datetime = "2022-09-26T00:00:00Z/2022-10-01T00:00:00Z",
  filter = {
    "op": ">=",
    "args": [
      {"property": "aquaview:column_stats_summary.variables.Wave Height.max"},
      6.0
    ]
  },
  limit = 10
)

Returns one buoy:

Buoy Location Lifetime max wave height (column_stats_summary)
ndbc_42097 Pulley Ridge, FL (25.7°N, 83.7°W) 6.5 m across 2,137 observations since 2016

⚠ Caveat: 6.5 m is the buoy's lifetime max — not confirmed for Sep 28, 2022 specifically. The catalog filter narrows you to the right candidate buoy; you confirm the storm-day value by pulling the timeseries from https://www.ndbc.noaa.gov/data/realtime2/42097.txt or the NDBC THREDDS archive.

3. GOES-R imagery for landfall day

search_datasets(
  collections = "GOES_R",
  bbox = "-86,24,-80,30",
  datetime = "2022-09-28T00:00:00Z/2022-09-28T23:59:59Z",
  fields = "id,properties.datetime,assets",
  include_assets = true,
  limit = 20
)

Returns 153 GOES-R product items (e.g., goes18_abi_l2_dmwm Derived Motion Winds Mesoscale, goes18_abi_l2_ctpf Cloud Top Pressure Full Disk). Each item represents a product family across the bbox/time, not an individual scene. Each product's gcs_data asset points to its bucket — actual NetCDF scenes for Sep 28, 2022 (day-of-year 271) are at paths like gs://gcp-public-data-goes-16/ABI-L2-CMIPF/2022/271/.

4. Tropical cyclone heat potential leading up to landfall

search_datasets(
  collections = "NOAA_AOML_HDB",
  bbox = "-90,15,-80,30",
  datetime = "2022-09-20T00:00:00Z/2022-09-28T23:59:59Z",
  limit = 10
)

42 matches. Top items:

  • xbt_ax7_2022 — Expandable Bathythermograph data along the AX7 transect line, 2022
  • ships_atlantic_5days, ships_atlantic_7days — SHIPS (Statistical Tropical Cyclone Intensity) forecast products with SST, OHC, isotherm depths (CD20, CD26), wind shear, intensity history
  • ships_eastpacific_5days, ships_eastpacific_7days — same for East Pacific basins

The SHIPS series carries the storm-context oceanographic fields (Tropical Cyclone Heat Potential proxies via COHC, CD26).

5. Drifters in the path

search_datasets(
  collections = "NOAA_GDP",
  bbox = "-86,24,-80,30",
  datetime = "2022-09-26T00:00:00Z/2022-10-01T00:00:00Z",
  limit = 12
)

Returns 2 items, both product-level (drifter_hourly_qc and drifter_6hour_qc — the entire interpolated QC drifter products). AQUAVIEW indexes GDP at the product level, not per-drifter. To list which specific drifters were in Ian's path on Sep 28, download one of these products from https://erddap.aoml.noaa.gov/gdp/erddap and filter trajectories by lat/lon/time.

Result the agent gave

"Hurricane Ian made landfall as a Category 4 storm near Cayo Costa, FL on Sep 28, 2022. AQUAVIEW indexes 2,448 items in the eastern Gulf bbox during the storm window, distributed across 34 collections.

Buoys* — Filtering NDBC items in the bbox/time-window for lifetime max wave height ≥ 6 m returns one buoy: ndbc_42097 (Pulley Ridge, FL, 25.7°N, 83.7°W)*, U.S. Army Corps of Engineers-owned. Its column_stats_summary shows max wave height 6.5 m across 2,137 observations since 2016. I can't confirm from the catalog alone that the 6.5 m was hit during Ian — that requires pulling the timeseries from https://www.ndbc.noaa.gov/data/realtime2/42097.txt and checking Sep 27–28.

Satellite — 153 GOES-R items match. Each is a product family (Cloud Top Pressure full disk and CONUS, Derived Motion Winds at full-disk / CONUS / mesoscale, etc.) covering both GOES-18 (West) and GOES-19. Per-scene NetCDFs for Sep 28, 2022 are at gs://gcp-public-data-goes-16/ABI-L2-CMIPF/2022/271/ and the equivalent paths for GOES-18.

AOML — 42 items in the basin/window. The most useful for Ian's intensification context are the SHIPS forecasting products (ships_atlantic_5days, ships_atlantic_7days), which include OHC, isotherm depths (CD20, CD26), and SST fields used for tropical-cyclone intensity prediction. Plus 2022 XBT line data along AX7.

GDP — AQUAVIEW indexes the Global Drifter Program at the product level (2 items: hourly and 6-hourly QC products), not per-drifter. To enumerate drifters in Ian's path, download drifter_hourly_qc from https://erddap.aoml.noaa.gov/gdp/erddap and filter by trajectory.

AQUAVIEW narrows the search to the right collections and product families very quickly, and column_stats_summary is enough to identify the candidate buoy. Confirming time-specific extreme values still requires the source data."

Variations

  • Hurricane Helene 2024: bbox = "-86,28,-80,36", datetime = "2024-09-25T00:00:00Z/2024-09-28T23:59:59Z"
  • Just the eyewall: tighten bbox and use geometry_geohash_grid_frequency at precision=5 to map observation density
  • Pre-storm baseline: same bbox, shift datetime 14 days earlier
  • Compare two storms: run the same query with two different datetime ranges