pvdeg.geospatial.elevation_stochastic_downselect

pvdeg.geospatial.elevation_stochastic_downselect(meta_df: DataFrame, kdtree, downselect_prop: float, k_neighbors: int = 3, method: str = 'mean', normalization: str = 'linear')

Downsample function.

Downsample, assigning each point a weight associated with its neighbors changes in height. Randomly choose points based on weights to preferentially select points next to or in mountains while drastically lowering the density of points in flat areas to find a non-uniformly dense sub-sample of original data points.

Parameters:

meta_dfpd.DataFrame

Dataframe of metadata as generated by pvdeg.weather.get for geospatial

kdtreesklearn.neighbors.KDTree or str

kdtree containing latitude-longitude pairs for quick lookups Generate using pvdeg.geospatial.meta_KDTree. Can take a pickled kdtree as a path to the .pkl file.

downselect_propfloat

proportion of original datapoints to keep in output gids list

k_neighborsint, (default = 3)

number of neighbors to check for elevation data in nearest neighbors

methodstr, (default = ‘mean’)

method to calculate elevation weights for each point. Options : ‘mean’, ‘sum’, ‘median’

normalizationstr, (default = ‘linear’)

function to apply when normalizing weights. Logarithmic uses log_e/ln options : ‘linear’, ‘log’, ‘exp’, ‘invert-linear’

Returns:

gidsnp.array

numpy array of downselected gids.