ExtremaPointFeature#

class rojak.atmosphere.feature_composite.ExtremaPointFeature(field, kind, *, threshold=None, footprint_radius=24, time_dim_name='time', lat_dim_name='latitude', lon_dim_name='longitude')[source]#

Lows or highs: one point feature per extremum of a field, found by identify_circular_extrema().

Each connected component (8-connectivity) of local_extrema in a time step is one centre, so separate extrema with equal values stay separate and a flat extremum spanning several grid cells is one centre (D7). Each centre owns the cells of extrema_regions within the footprint of any pixel of its component for which it is the deepest centre in range, which is the rule that defines extrema_regions, written as a label array (D11). A component is represented by its first pixel in row-major order, and where equally deep centres are in range of a cell, the one with the smallest centre_id owns it (D19).

centre_id is (time_index * n_lats + lat_index) * n_lons + lon_index + 1 for the representative pixel, with time_index along the whole time axis. It is unique across chunks without a global count, independent of the chunking, increasing in row-major (time, lat, lon) order, and at least 1, since 0 means “no owner” in the labels.

Distances are in grid cells and do not wrap around in longitude, as in the extrema filter’s default boundary mode. Deferred: orientation from a steering flow (phase 5).

field#

Field to find the extrema in, e.g. mean sea level pressure, with exactly the dimensions time_dim_name, lat_dim_name and lon_dim_name. It may be chunked along time; it is rechunked to one chunk in latitude and longitude.

Type:: xarray.DataArray

kind#

Minima for lows, maxima for highs.

Type:: ExtremaKind

threshold#

Only extrema below (minima) or above (maxima) it are kept. None keeps all.

Type:: float | None

footprint_radius#

Radius in grid cells of the circular footprint of the extrema filter, and of the region each centre may own. At least 1.

Type:: int

time_dim_name#

Name of the time dimension of field.

Type:: str

lat_dim_name#

Name of the latitude dimension of field.

Type:: str

lon_dim_name#

Name of the longitude dimension of field.

Type:: str

Raises:

ValueError – If field does not have exactly the three named dimensions, or footprint_radius is below 1.

Parameters:

field (xr.DataArray)
kind (ExtremaKind)
threshold (float | None)
footprint_radius (int)
time_dim_name (str)
lat_dim_name (str)
lon_dim_name (str)

__init__(field, kind, *, threshold=None, footprint_radius=24, time_dim_name='time', lat_dim_name='latitude', lon_dim_name='longitude')[source]#

Parameters:

field (xr.DataArray)
kind (ExtremaKind)
threshold (float | None)
footprint_radius (int)
time_dim_name (str)
lat_dim_name (str)
lon_dim_name (str)

Return type:

None

Methods

`__init__`(field, kind, *[, threshold, ...])
`chunk_inputs`()	The field values, `local_extrema` and `extrema_regions`, built lazily without computing anything.
`tabulator`()	The parameters-only tabulator sent to each chunk task.

Attributes

`footprint_radius`
`geometry`	Each low or high is a point.
`has_labels`	The tabulator labels the region that each centre owns.
`lat_dim_name`
`lon_dim_name`
`threshold`
`time_dim_name`
`field`
`kind`