The first operational version was implemented to forecast waves at Guaecá Beach, in São Sebastião. At that stage, the main objective was to test whether a high-resolution coastal grid could provide more useful local information than global wave models alone.

GUAECAST V1.0 used the SWAN model nested inside NOAA's WaveWatch III global forecast. The global model supplied offshore wave conditions and wind forcing, while the local SWAN grid represented the coastal transformation of waves near Guaecá and the surrounding São Sebastião/Ilhabela region.

This version also included astronomical tide information from the São Sebastião tide gauge, allowing water-level variations to be considered in shallow-water wave propagation. The model was operated mainly during the autumn and winter seasons, when larger swell events are more frequent and more relevant for surf forecasting.

The main lesson from V1.0 was that the nested approach worked. Even with a relatively limited domain, the higher-resolution model improved the ability to represent local effects caused by bathymetry, headlands, islands and shoreline geometry.

GUAECAST V2.0 expanded the system beyond its original single-beach focus. The forecast domain was extended to cover a larger portion of the São Paulo coast, including Santos, Guarujá, São Sebastião and Ubatuba. The name GUAECAST was retained because it had already become associated with the project.

To cover this larger region without losing local detail, the model was reorganized into a hierarchy of nested grids. A broad outer grid receives the offshore wave conditions, an intermediate grid resolves the central and northern São Paulo coast, and local grids provide higher-resolution results for the main forecast regions.

This multi-grid structure made it possible to keep computational costs reasonable while still resolving the nearshore areas where wave transformation is most important. The local grids were used to produce maps, time series and analysis products for specific beaches and coastal sectors.

During the V2.0 period, the system also became a year-round operational forecast. Updates to bathymetry, forcing datasets and model configuration were incorporated as new data and global model products became available.

GUAECAST V3.0 consolidated the current regional configuration and extended the forecast horizon to 14 days. This was possible because global wave forecast products evolved to provide longer outlooks, allowing GUAECAST to generate a two-week forecast while still preserving the local downscaling strategy.

The V3.0 configuration keeps the broad spatial coverage established in V2.0, with regional and local products for Baixada Santista, São Sebastião and Ubatuba. The main improvement is not a new geographic domain, but a more complete forecast window and a clearer set of products for interpretation.

Version 3.1 is an operational and visual update. On the data side, it adapts the processing workflow to changes in how NCEP makes the data available. Earlier versions accessed data through OPeNDAP, while the current operational workflow uses GRIB files. This change required adjustments in the data ingestion and preprocessing steps, but preserves the overall modeling approach.

On the website side, V3.1 reorganizes the forecast interface, improves mobile responsiveness, and modernizes the visual presentation of maps, charts and navigation. The goal is to make the forecast products easier to browse without reducing the technical content behind them.

Version 3.2 introduced layout adjustments to the website only. The forecast methodology, numerical configuration, grid structure and generated products remained unchanged.

Version 3.3 introduces another layout update and adds a Directional Spectra tab to the forecast interface. The new tab presents directional wind and swell spectra alongside the existing maps and local analyses, while the numerical model configuration remains unchanged.

GUAECAST V4.0 extends the forecast to 16 days and updates the nested system to eight SWAN grids. The existing regional grids are joined by dedicated refinements for Bonete, Laranjeiras/Praia do Sono and Guaecá/Barequeçaba, where the finest grid has a spacing of approximately 100 m. The grids and bathymetry have been revised to improve the representation of the coastal domains.

The 13 municipalities are Paraty, Ubatuba, Caraguatatuba, São Sebastião, Ilhabela, Bertioga, Guarujá, Santos, São Vicente, Praia Grande, Mongaguá, Itanhaém and Peruíbe. The coastal forecasts use eight nested SWAN grids, with dedicated refinements for Bonete, Laranjeiras/Praia do Sono and Guaecá/Barequeçaba. The G0 layer provides broader Atlantic context, from 69.5°W to 20°W and from 55°S to the Equator, with wave height, period and direction, wind and pressure.

The forecast now covers 88 locations, with hourly outputs through the first five days and three-hourly outputs thereafter: 209 forecast times per run. Each beach combines wave and swell information, regional offshore directional spectra, wind and atmospheric conditions. Astronomical tides from FES2022b and sea-level variations from HYCOM are presented as individual time series for each location. Wave calculations use a representative water level for each coastal grid; Ressacast retains its regional water-level references at this stage.

The new interface brings the forecast into an interactive map with wave height and period, wind and pressure, directional overlays and time playback. City and beach selectors simplify navigation between locations. Beach markers show the requested geographic positions, while the forecast series are extracted from the selected wet model nodes.

The interactive map below shows the current eight-grid GUAECAST configuration. G1 receives the global wave boundary conditions and supplies G3. G3 supplies G4A, G4B, G4C, Bonete and Laranjeiras; G4B supplies the finer G5 Guaecá grid. Toggle individual grids, show or hide the full set, and select a grid to inspect its coverage and spacing.

Colours identify the grids. Spacings are approximate and calculated between grid nodes. Select a grid to see its parent and zoom to the region.