mesmer.core.weighted.equal_scenario_weights_from_datatree

mesmer.core.weighted.equal_scenario_weights_from_datatree(dt, ens_dim='member', time_dim='time')

Create a DataTree isomorphic to dt, holding the weights for each scenario to weight the ensemble members of each scenario such that each scenario contributes equally to some fitting procedure. The weight of each member = 1 / number of members in the scenario, so weights = 1 / ds[ens_dim].size.

Thus, if all scenarios have the same number of members, all weights will be equal. If one scenario has more members than the others, its weights will be smaller. Weights are always along the time and ens dim, if there are more dimensions in a dataset, they will be dropped.

Parameters:

dtDataTree

DataTree holding the xr.Datasets for which the weights should be created. Each dataset must have at least ens_dim and time_dim as dimensions, but can have more dimensions.

ens_dimstr

Name of the dimension along which the weights should be created. Default is “member”.

time_dimstr

Name of the time dimension, will be filled with equal values for each ensemble member. Default is “time”.

Returns:

DataTree

DataTree holding the weights for each scenario isomorphic to dt, where each dataset has dimensions (time_dim, ens_dim).

Example:

>>> dt = DataTree()
>>> dt["ssp119"] = DataTree(xr.Dataset({"tas": xr.DataArray(np.ones((20, 3)), dims=("time", "member"))}))
>>> dt["ssp585"] = DataTree(xr.Dataset({"tas": xr.DataArray(np.ones((20, 2)), dims=("time", "member"))}))
>>> weights = equal_scenario_weights_from_datatree(dt)
>>> weights
DataTree('None', parent=None)
   ├── DataTree('ssp119')
   │       Dimensions:  (time: 20, member: 3)
   │       Coordinates:
   │         * time     (time) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
   │         * member   (member) int64 0 1 2
   │       Data variables:
   │           weights  (time, member) float64 0.3333 0.3333 0.3333 ... 0.3333 0.3333
   └── DataTree('ssp585')
           Dimensions:  (time: 20, member: 2)
           Coordinates:
           * time     (time) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
           * member   (member) int64 0 1
           Data variables:
               weights  (time, member) float64 0.5 0.5 0.5 0.5 0.5 ... 0.5 0.5 0.5 0.5 0.5