Hurricane Simon 2026 (Mexico)¶

Hurricane Simon was a Category 4 tropical cyclone of the 2026 eastern North Pacific hurricane season. It formed as Tropical Depression Twenty-E on 7 October 2026 about 285 mi south-southeast of Lázaro Cárdenas, became a tropical storm the same day, and then tracked slowly north-northwest parallel to Mexico's southwestern coast. Simon peaked on 11 October with maximum sustained winds of 155 mph (250 km/h) and a central pressure of 928 mb. Its eye was then roughly 50 km offshore between Manzanillo and Puerto Vallarta, and its inner core brushed the coast. The storm never made landfall. It weakened over the following days and dissipated about 100 mi southwest of Cabo San Lucas on 16 October. Its outer bands brought heavy rain, flash flooding and landslides from Guerrero to Jalisco, flooded Manzanillo, closed the Manzanillo and Puerto Vallarta airports, and cut power to about 65,000 customers in Guerrero. At least one death has been directly attributed to the storm. Vantor's (formerly Maxar) Open Data Program released very high resolution post-event imagery of Puerto Vallarta, Manzanillo and Lázaro Cárdenas, catalogued here as an Earth Engine collection.

Event Summary¶

NHC has not yet published its Tropical Cyclone Report for Simon, so the intensity figures are provisional. Death tolls reported in the days after the storm often combine Simon with Tropical Storm Rachel, which struck Baja California at the same time. No International Charter or Copernicus EMS activation was found for this event.

Vantor (Maxar) Open Data¶

Vantor published the event as Mexico-Hurricane-Simon-Oct-2026 under CC BY-NC 4.0, with the first scenes released on 10 October 2026. All imagery is post-event, at panchromatic resolutions of about 0.32 to 0.82 m, from WorldView Legion, WorldView-2, WorldView-3 and GeoEye-1. Coverage is concentrated on three areas:

Vantor labels every scene in this event as post-event. Three Puerto Vallarta scenes from 10 October were taken before Simon's closest approach on 11 and 12 October, so they are better read as near-event than as post-event imagery. The collection contains 28 scenes acquired between 10 and 21 October. A further WorldView-2 scene of Puerto Vallarta from 22 October will be added.

// Vantor (Maxar) open data, post-event imagery of Hurricane Simon.
var vantorSimon2026 = ee.ImageCollection("projects/sat-io/open-datasets/VANTOR-DISASTER-DATA/HURRICANE_SIMON_2026");

Preprocessing¶

Vantor's open data imagery ships without a cloud mask of any kind. Every scene in this collection was therefore passed through OmniCloudMask during ingestion, and two derived bands were added after the imagery bands:

OmniCloudMask is a sensor-agnostic deep learning model for cloud and cloud shadow segmentation. Vantor imagery has no near-infrared band, so blue is passed in its place. Comparisons on PlanetScope imagery found this costs nothing measurable. The model is trained for 10 to 50 m imagery, so scenes are resampled to 10 m for inference and the mask is resampled back onto the native grid. Only solid cloud counts as cloud in cloud_mask, because thin cloud and haze are the model's least reliable classes. Pixels deep inside large, bright, saturated clouds are sometimes classed as clear, so check the imagery before relying on the mask over heavy cloud.

Wright, N., Duncan, J.M.A., Callow, J.N., Thompson, S.E., & George, R.J. (2025). Training sensor-agnostic deep learning models for remote sensing: Achieving state-of-the-art cloud and cloud shadow identification with OmniCloudMask. Remote Sensing of Environment, 322, 114694. https://doi.org/10.1016/j.rse.2025.114694

Masking no data, and cloud with the threshold passed in as a percentage:

// No data is black fill: every visual band is 0.
function maskNoData(image) {
  var rgb = image.select(['red', 'green', 'blue']);
  // copyProperties returns an Element, so cast back to an Image.
  return ee.Image(rgb
    .updateMask(rgb.reduce(ee.Reducer.max()).gt(0))
    .copyProperties(image, image.propertyNames()));
}

// Cloud mask, threshold passed in as a percentage.
function maskClouds(maxCloudPct) {
  return function (image) {
    var clear = image.select('cloud_probability').lte(maxCloudPct);
    return maskNoData(image).updateMask(clear);
  };
}

var noData = vantorSimon2026.map(maskNoData);
var masked = vantorSimon2026.map(maskClouds(30));

Map.setCenter(-104.32, 19.08, 13); // Manzanillo
Map.addLayer(noData.mosaic(), {min: 0, max: 255}, 'No-data masked only');
Map.addLayer(masked.mosaic(), {min: 0, max: 255}, 'No-data + cloud masked');

Scene-level cloud cover as reported by Vantor ranges from 0 to 74 percent and is stored as cloud_percent. Other useful properties are acquired, view_angle, sun_elevation and sun_azimuth. To pick the clearest scenes over one area, filter on bounds and sort by Vantor's cloud figure:

var puertoVallarta = ee.Geometry.Point([-105.23, 20.62]);
var clearest = vantorSimon2026
  .filterBounds(puertoVallarta)
  .sort('cloud_percent', false) // mosaic() draws the last image on top
  .map(maskNoData);
Map.addLayer(clearest.mosaic(), {min: 0, max: 255}, 'Puerto Vallarta, clearest first', false);

Sample code: https://code.earthengine.google.com/?scriptPath=users/sat-io/awesome-gee-catalog-examples:disaster-response/VANTOR-OPEN-DATA-HURRICANE-SIMON-2026

License¶

The collection on this page is released under Creative Commons Attribution Non Commercial 4.0.

Data provided by: Vantor (Maxar) Open Data Program

Curated in GEE by: Samapriya Roy

Keywords: disaster, hurricane, tropical cyclone, eastern Pacific, Mexico, Jalisco, Colima, Michoacán, Guerrero, Puerto Vallarta, Manzanillo, Lázaro Cárdenas, Vantor, Maxar, WorldView Legion, WorldView, GeoEye

Last updated on GEE: 2026-10-22

Ask AI