From f632ea35aa16e424b0cfef0fc240d3df31e1719a Mon Sep 17 00:00:00 2001
From: Felix Kleinert <f.kleinert@fz-juelich.de>
Date: Wed, 31 Mar 2021 17:18:52 +0200
Subject: [PATCH] prepare sector assingmnet
---
mlair/data_handler/data_handler_wrf_chem.py | 72 ++++++++++++++++++---
1 file changed, 62 insertions(+), 10 deletions(-)
diff --git a/mlair/data_handler/data_handler_wrf_chem.py b/mlair/data_handler/data_handler_wrf_chem.py
index c7313ba6..176fdd8f 100644
--- a/mlair/data_handler/data_handler_wrf_chem.py
+++ b/mlair/data_handler/data_handler_wrf_chem.py
@@ -548,16 +548,17 @@ class DataHandlerSingleGridColumn(DataHandlerSingleStation):
self.make_history_window(self.target_dim, self.window_history_size, self.time_dim)
if self.var_logical_z_coord_selector is not None:
self.history = self.history.sel({self._logical_z_coord_name: self.var_logical_z_coord_selector})
- self.history = self.modify_history()
self.make_labels(self.target_dim, self.target_var, self.time_dim, self.window_lead_time)
self.make_observation(self.target_dim, self.target_var, self.time_dim)
if self.targetvar_logical_z_coord_selector is not None:
self.label = self.label.sel({self._logical_z_coord_name: self.targetvar_logical_z_coord_selector})
self.observation = self.observation.sel(
{self._logical_z_coord_name: self.targetvar_logical_z_coord_selector})
+
+ self.remove_nan(self.time_dim)
+ self.history = self.modify_history()
self.label = self.modify_label()
self.observation = self.modify_observation()
-
self.remove_nan(self.time_dim)
# self.input_data = self.input_data.compute()
# self.label = self.label.compute()
@@ -686,11 +687,11 @@ class DataHandlerSectorGrid(DataHandlerSingleGridColumn):
trafo_opts = self._transformation[pos_of_trafo]
return trafo_opts
- def apply_transformation_on_wind_sector_edges(self, ws_edges):
- ws_trafo_opts = self.get_transformation_opts()
- ws_trafo, _ = self.transform(ws_edges, dim=self.aggregation_dim, inverse=False,
- opts=ws_trafo_opts, transformation_dim=self.target_dim)
- return ws_trafo
+ def apply_transformation_on_data(self, data):
+ trafo_opts = self.get_transformation_opts()
+ trafo, _ = self.transform(data, dim=self.aggregation_dim, inverse=False,
+ opts=trafo_opts, transformation_dim=self.target_dim)
+ return trafo
def modify_history(self):
if self.transformation_is_applied:
@@ -699,12 +700,63 @@ class DataHandlerSectorGrid(DataHandlerSingleGridColumn):
sector_allocation = self.windsector.get_sect_of_value(value=wind_dir_of_interest, external_edges=ws_edges)
existing_sectors = np.unique(sector_allocation.data)
with self.loader as loader:
- pass
- #circular_data = loader.data[self.variables].where(loader.geo_infos.dist.squeeze() <= self.radius)
+ # setup sector history
+ sector_history = xr.ones_like(self.history)
+ sector_history_var_names = [f"{var}_sect" for var in sector_history.coords[self.target_dim].values]
+ # sector_history = sector_history.assign_coords({self.target_dim: sector_history_var_names})
+
+ grid_data = self.preselect_and_transform_neighbouring_data_based_on_radius(loader)
+
+ for sect in existing_sectors:
+ # select data in wind sector
+ sec_data = self.get_section_data_from_circle(grid_data, loader, sect)
+ sec_data = self.apply_aggregation_method_on_sector_data(sec_data, loader)
+
+ # loader.data.T2.where(loader.geo_infos.dist.sel({'points': 0}).drop('points') <= self.radius).where(
+ # self.windsector.is_in_sector(sect, loader.geo_infos.bearing))
+ #
+ # loader.data[self.variables].sel(
+ # {self.time_dim: self.history.coords[self.time_dim].values}).where(
+ # loader.geo_infos.dist.sel({'points': 0}).drop('points') <= self.radius).where(
+ # self.windsector.is_in_sector(sect,
+ # loader.geo_infos.bearing.drop('points').squeeze()))
return self.history
+ def get_section_data_from_circle(self, grid_data, loader, sect):
+ sec_data = grid_data.where(
+ self.windsector.is_in_sector(sect, loader.geo_infos.bearing.drop('points').squeeze()))
+ return sec_data
+
+ def preselect_and_transform_neighbouring_data_based_on_radius(self, loader):
+ """
+ Select neighbouring grid boxes which have a maximal distance of pre-selected radius from full model field
+ :param loader:
+ :type loader:
+ :return:
+ :rtype:
+ """
+ # get data from loader
+ grid_data = loader.data[self.variables]
+ # select correct time steps and vertical layers
+ grid_data = grid_data.sel({
+ # self.time_dim: self.history.coords[self.time_dim].values, # get data which is used in history
+ self.time_dim: slice(self.start, self.end), # get data which is used in history
+ self._logical_z_coord_name: self.var_logical_z_coord_selector, # get only used vertical layers
+ })
+ # select grid boxes which are closer than given radius
+ grid_data = grid_data.where(
+ loader.geo_infos.dist.sel({"points": 0}).drop("points") <= self.radius)
+ # apply transformation on variables
+ grid_data = self.apply_transformation_on_data(grid_data.to_array(self.target_dim)).to_dataset(self.target_dim)
+
+ return grid_data
+
+ def apply_aggregation_method_on_sector_data(self, data, loader):
+ data = data.mean(dim=(loader.logical_x_coord_name, loader.logical_y_coord_name))
+ return data
+
def compute_wind_dir_of_interest(self):
wind_dir_of_intrest = self.history.sel({self.target_dim: self.wind_dir_name, self.window_dim: 0})
return wind_dir_of_intrest
@@ -712,7 +764,7 @@ class DataHandlerSectorGrid(DataHandlerSingleGridColumn):
@TimeTrackingWrapper
def get_applied_transdormation_on_wind_sector_edges(self):
ws_edges = self._get_left_and_right_wind_sector_edges(return_as='xr.da', dim=self.wind_sector_edge_dim_name)
- ws_edges = self.apply_transformation_on_wind_sector_edges(ws_edges)
+ ws_edges = self.apply_transformation_on_data(ws_edges)
return ws_edges
# def set_inputs_and_targets(self):
--
GitLab