Snow Processors

Estimate seasonal snow cover so the texture generator can paint snow onto the selection texture. The model works in two steps: it estimates a surface temperature for each pixel — either from a simple built-in climatology (warmer near the equator, colder at altitude, with a summer/winter swing) or by interpolating a CAMS atmospheric profile onto the terrain — and then turns that temperature into a snow probability that rises smoothly as it drops below freezing. Two seasons are supported, June and December. Configured via SnowConfig (see Scene config).

s2gos_generator.processors.terrain_texture.snow.get_day_of_year

get_day_of_year(month: Month) -> int

Map Month enum to approximate day of year.

s2gos_generator.processors.terrain_texture.snow.apply_spatial_smoothing

apply_spatial_smoothing(
    data: ndarray, sigma: float = 10.0
) -> np.ndarray

Apply Gaussian spatial smoothing.

s2gos_generator.processors.terrain_texture.snow.calculate_seasonal_amplitude

calculate_seasonal_amplitude(
    abs_lat: ndarray,
) -> np.ndarray

Seasonal temperature swing as a function of latitude.

Small near the equator, large near the poles (linear interpolation between the two reference latitudes, clamped to that range).

s2gos_generator.processors.terrain_texture.snow.interpolate_cams_temperature

interpolate_cams_temperature(
    elevations: ndarray, thermoprops: Dataset
) -> np.ndarray

Interpolate temperature from CAMS atmospheric profile.

Temperature is extracted from CAMS vertical profile and interpolated to terrain elevations.

Parameters:

Name	Type	Description	Default
`elevations`	`ndarray`	Terrain elevation in meters, shape (height, width)	required
`thermoprops`	`Dataset`	CAMS dataset after squeeze(drop=True), containing: - t: air temperature [z] in Kelvin - z: height [z] in kilometers (0-120km typical)	required

Returns:

Type	Description
`ndarray`	Temperature in Celsius, same shape as elevations

s2gos_generator.processors.terrain_texture.snow.calculate_temperature_field

calculate_temperature_field(
    latitudes: ndarray,
    elevations: ndarray,
    day_of_year: int,
) -> np.ndarray

Estimate surface air temperature from latitude, elevation, and day of year.

Synthetic climatology: start from a warm equatorial reference, cool toward the poles and with altitude, and add a summer/winter swing whose size grows with latitude and whose timing depends on the hemisphere.

s2gos_generator.processors.terrain_texture.snow.calculate_snow_probability_map

calculate_snow_probability_map(
    latitudes: ndarray,
    elevations: ndarray,
    day_of_year: int,
    smooth_sigma: float = 0.0,
    thermoprops: Optional[Dataset] = None,
) -> Tuple[np.ndarray, np.ndarray]

Estimate the per-pixel probability of snow from temperature.

Temperature comes from a CAMS profile when thermoprops is given, otherwise from the synthetic model. Probability follows a smooth transition around freezing — close to 1 well below 0 °C, falling to 0 just above it — and any pixel warmer than the hard freeze limit (0.3 °C) is left snow-free.

Parameters:

Name	Type	Description	Default
`latitudes`	`ndarray`	Latitude in degrees, shape (height, width)	required
`elevations`	`ndarray`	Elevation in meters, shape (height, width)	required
`day_of_year`	`int`	Day of year (1-365)	required
`smooth_sigma`	`float`	Gaussian smoothing sigma (0 = no smoothing)	`0.0`
`thermoprops`	`Optional[Dataset]`	Optional CAMS dataset (after squeeze). If provided, uses CAMS temperature; otherwise uses synthetic model.	`None`

Returns:

Type	Description
`Tuple[ndarray, ndarray]`	(probabilities, temperatures) - both same shape as inputs