Add station_code field to ModelRunPrediction table

api/alembic/versions/5b745fe0bd7a_add_station_code_to_modelrunprediction.py

@@ -0,0 +1,34 @@
+"""Add station_code to ModelRunPrediction
+
+Revision ID: 5b745fe0bd7a
+Revises: 2442f07d975c
+Create Date: 2023-10-25 11:34:45.429269
+
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision = '5b745fe0bd7a'
+down_revision = '2442f07d975c'
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # Delete unusable data from the model_run_predictions table
+    op.execute(sa.text('DELETE FROM model_run_predictions'))
+    # ### commands auto generated by Alembic ###
+    op.add_column('model_run_predictions', sa.Column('station_code', sa.Integer(), nullable=False))
+    op.drop_constraint('model_run_predictions_prediction_model_run_timestamp_id_pre_key', 'model_run_predictions', type_='unique')
+    op.create_unique_constraint('model_run_predictions_unique_constraint', 'model_run_predictions', ['prediction_model_run_timestamp_id', 'prediction_timestamp', 'station_code'])
+    # ### end Alembic commands ###
+
+
+def downgrade():
+    # ### commands auto generated by Alembic ###
+    op.drop_constraint('model_run_predictions_unique_constraint', 'model_run_predictions', type_='unique')
+    op.create_unique_constraint('model_run_predictions_prediction_model_run_timestamp_id_pre_key', 'model_run_predictions', ['prediction_model_run_timestamp_id', 'prediction_timestamp'])
+    op.drop_column('model_run_predictions', 'station_code')
+    # ### end Alembic commands ###

api/app/db/crud/observations.py

@@ -39,7 +39,8 @@ def get_actuals_left_outer_join_with_predictions(
     """
     return session.query(HourlyActual, ModelRunPrediction)\
         .outerjoin(ModelRunPrediction,
-                   and_(ModelRunPrediction.prediction_timestamp == HourlyActual.weather_date))\
+                   and_(ModelRunPrediction.prediction_timestamp == HourlyActual.weather_date,
+                        ModelRunPrediction.station_code == station_code))\
         .outerjoin(PredictionModelRunTimestamp,
                    and_(PredictionModelRunTimestamp.id ==
                         ModelRunPrediction.prediction_model_run_timestamp_id,

api/app/db/crud/weather_models.py

@@ -62,14 +62,15 @@ def get_or_create_prediction_run(session, prediction_model: PredictionModel,
     return prediction_run
 
 
-def get_model_run_predictions(session: Session,
+def get_model_run_predictions_for_station(session: Session, station_code: int,
                               prediction_run: PredictionModelRunTimestamp) -> List:
     """ Get all the predictions for a provided model run """
     logger.info("Getting model predictions for grid %s", prediction_run)
-    return session.query(ModelRunPrediction).\
-        filter(ModelRunPrediction.prediction_model_run_timestamp_id ==
-               prediction_run.id).\
-        order_by(ModelRunPrediction.prediction_timestamp)
+    return session.query(ModelRunPrediction)\
+        .filter(ModelRunPrediction.prediction_model_run_timestamp_id ==
+               prediction_run.id)\
+        .filter(ModelRunPrediction.station_code == station_code)\
+        .order_by(ModelRunPrediction.prediction_timestamp)
 
 
 def delete_weather_station_model_predictions(session: Session, older_than: datetime):
@@ -79,6 +80,15 @@ def delete_weather_station_model_predictions(session: Session, older_than: datet
     session.query(WeatherStationModelPrediction)\
         .filter(WeatherStationModelPrediction.prediction_timestamp < older_than)\
         .delete()
+
+
+def delete_model_run_predictions(session: Session, older_than: datetime):
+    """ Delete any model run prediction older than a certain date.
+    """
+    logger.info('Deleting model_run_prediction data older than %s...', older_than)
+    session.query(ModelRunPrediction)\
+        .filter(ModelRunPrediction.prediction_timestamp < older_than)\
+        .delete()
 
 
 def get_station_model_predictions_order_by_prediction_timestamp(

api/app/db/models/weather_models.py

@@ -170,7 +170,7 @@ class ModelRunPrediction(Base):
     Each value is a numeric value that corresponds to the lat lon from the model raster """
     __tablename__ = 'model_run_predictions'
     __table_args__ = (
-        UniqueConstraint('prediction_model_run_timestamp_id', 'prediction_timestamp'),
+        UniqueConstraint('prediction_model_run_timestamp_id', 'prediction_timestamp', 'station_code'),
         {'comment': 'The prediction values of a particular model run.'}
     )
 
@@ -183,6 +183,8 @@ class ModelRunPrediction(Base):
         "PredictionModelRunTimestamp", foreign_keys=[prediction_model_run_timestamp_id])
     # The date and time to which the prediction applies.
     prediction_timestamp = Column(TZTimeStamp, nullable=False, index=True)
+    # The station code representing the location (aka weather station).
+    station_code = Column(Integer, nullable=True)
     # Temperature 2m above model layer.
     tmp_tgl_2 = Column(Float, nullable=True)
     # Relative humidity 2m above model layer.

api/app/jobs/common_model_fetchers.py

@@ -9,10 +9,11 @@
 from app.db.crud.weather_models import (get_processed_file_record,
                                         get_processed_file_count,
                                         get_prediction_model_run_timestamp_records,
-                                        get_model_run_predictions,
+                                        get_model_run_predictions_for_station,
                                         get_weather_station_model_prediction,
                                         delete_weather_station_model_predictions,
-                                        refresh_morecast2_materialized_view)
+                                        refresh_morecast2_materialized_view,
+                                        delete_model_run_predictions)
 from app.weather_models.machine_learning import StationMachineLearning
 from app.weather_models import SCALAR_MODEL_VALUE_KEYS, ModelEnum, construct_interpolated_noon_prediction
 from app.schemas.stations import WeatherStation
@@ -162,6 +163,7 @@
         # keeping 21 days (3 weeks) of historic data is sufficient.
         oldest_to_keep = time_utils.get_utc_now() - time_utils.data_retention_threshold
         delete_weather_station_model_predictions(session, oldest_to_keep)
+        delete_model_run_predictions(session, oldest_to_keep)
 
 
 def accumulate_nam_precipitation(nam_cumulative_precip: float, prediction: ModelRunPrediction, model_run_hour: int):
@@ -349,7 +351,7 @@
         machine.learn()
 
         # Get all the predictions associated to this particular model run.
-        query = get_model_run_predictions(self.session, model_run)
+        query = get_model_run_predictions_for_station(self.session, station.code, model_run)
 
         nam_cumulative_precip = 0.0
         # Iterate through all the predictions.

api/app/tests/weather_models/test_process_grib.py

@@ -40,8 +40,9 @@ def test_read_single_raster_value():
                                                geo_to_raster_transformer)
 
     raster_band = dataset.GetRasterBand(1)
-    value = next(processor.yield_value_for_stations(raster_band))
+    station, value = next(processor.yield_value_for_stations(raster_band))
 
     assert math.isclose(value, 55.976, abs_tol=0.001)
+    assert station.code == 322
 
     del dataset

api/app/weather_models/process_grib.py

@@ -171,7 +171,7 @@
                 logger.warning('coordinate not in raster - %s', station)
                 continue
 
-            yield value
+            yield (station, value)
 
     def yield_uv_wind_data_for_stations(self, u_raster_band: gdal.Dataset, v_raster_band: gdal.Dataset, variable: str):
         """ Given a list of stations and 2 gdal datasets (one for u-component of wind, one for v-component
@@ -191,11 +191,11 @@
                 v_value = v_raster_band.ReadAsArray(x_coordinate, y_coordinate, 1, 1)[0, 0]
 
                 if variable == 'wdir_tgl_10':
-                    yield calculate_wind_dir_from_u_v(u_value, v_value)
+                    yield (station, calculate_wind_dir_from_u_v(u_value, v_value))
                 elif variable == 'wind_tgl_10':
                     metres_per_second_speed = calculate_wind_speed_from_u_v(u_value, v_value)
                     kilometres_per_hour_speed = convert_mps_to_kph(metres_per_second_speed)
-                    yield kilometres_per_hour_speed
+                    yield (station, kilometres_per_hour_speed)
             else:
                 logger.warning('coordinate not in u/v wind rasters - %s', station)
 
@@ -240,22 +240,26 @@
         return variable_name
 
     def store_prediction_value(self,
+                               station_code: int,
                                value: float,
                                preduction_model_run: PredictionModelRunTimestamp,
                                grib_info: ModelRunInfo,
                                session: Session):
         """ Store the values around the area of interest.
         """
         # Load the record if it exists.
-        prediction = session.query(ModelRunPrediction).\
-            filter(
-                ModelRunPrediction.prediction_model_run_timestamp_id == preduction_model_run.id).\
-            filter(ModelRunPrediction.prediction_timestamp == grib_info.prediction_timestamp).first()
+        prediction = session.query(ModelRunPrediction)\
+            .filter(
+                ModelRunPrediction.prediction_model_run_timestamp_id == preduction_model_run.id)\
+            .filter(ModelRunPrediction.prediction_timestamp == grib_info.prediction_timestamp)\
+            .filter(ModelRunPrediction.station_code == station_code)\
+            .first()
         if not prediction:
             # Record doesn't exist, so we create it.
             prediction = ModelRunPrediction()
             prediction.prediction_model_run_timestamp_id = preduction_model_run.id
             prediction.prediction_timestamp = grib_info.prediction_timestamp
+            prediction.station_code = station_code
 
         variable_name = self.get_variable_name(grib_info)
         setattr(prediction, variable_name, value)
@@ -267,13 +271,13 @@
         # for GDPS, RDPS, HRDPS models, always only ever 1 raster band in the dataset
         raster_band = dataset.GetRasterBand(1)
         # Iterate through stations:
-        for value in self.yield_value_for_stations(raster_band):
+        for station, value in self.yield_value_for_stations(raster_band):
             # Convert wind speed from metres per second to kilometres per hour for Environment Canada
             # models (NOAA models handled elswhere)
             if grib_info.variable_name.lower().startswith("wind_agl") or grib_info.variable_name.lower().startswith('wind_tgl'):
                 value = convert_mps_to_kph(value)
 
-            self.store_prediction_value(value, prediction_run, grib_info, session)
+            self.store_prediction_value(station.code, value, prediction_run, grib_info, session)
 
     def get_raster_bands(self, dataset, grib_info: ModelRunInfo):
         """ Returns raster bands of dataset for temperature, RH, U/V wind components, and 
@@ -308,22 +312,22 @@
         tmp_raster_band, rh_raster_band, u_wind_raster_band, v_wind_raster_band, precip_raster_band = self.get_raster_bands(
             dataset, grib_info)
 
-        for tmp_value in self.yield_value_for_stations(tmp_raster_band):
+        for station, tmp_value in self.yield_value_for_stations(tmp_raster_band):
             grib_info.variable_name = 'tmp_tgl_2'
-            self.store_prediction_value(tmp_value, prediction_run, grib_info, session)
-        for rh_value in self.yield_value_for_stations(rh_raster_band):
+            self.store_prediction_value(station.code, tmp_value, prediction_run, grib_info, session)
+        for station, rh_value in self.yield_value_for_stations(rh_raster_band):
             grib_info.variable_name = 'rh_tgl_2'
-            self.store_prediction_value(rh_value, prediction_run, grib_info, session)
+            self.store_prediction_value(station.code, rh_value, prediction_run, grib_info, session)
         if precip_raster_band:
-            for apcp_value in self.yield_value_for_stations(precip_raster_band):
+            for station, apcp_value in self.yield_value_for_stations(precip_raster_band):
                 grib_info.variable_name = 'apcp_sfc_0'
-                self.store_prediction_value(apcp_value, prediction_run, grib_info, session)
-        for wdir_value in self.yield_uv_wind_data_for_stations(u_wind_raster_band, v_wind_raster_band, 'wdir_tgl_10'):
+                self.store_prediction_value(station.code, apcp_value, prediction_run, grib_info, session)
+        for station, wdir_value in self.yield_uv_wind_data_for_stations(u_wind_raster_band, v_wind_raster_band, 'wdir_tgl_10'):
             grib_info.variable_name = 'wdir_tgl_10'
-            self.store_prediction_value(wdir_value, prediction_run, grib_info, session)
-        for wind_value in self.yield_uv_wind_data_for_stations(u_wind_raster_band, v_wind_raster_band, 'wind_tgl_10'):
+            self.store_prediction_value(station.code, wdir_value, prediction_run, grib_info, session)
+        for station, wind_value in self.yield_uv_wind_data_for_stations(u_wind_raster_band, v_wind_raster_band, 'wind_tgl_10'):
             grib_info.variable_name = 'wind_tgl_10'
-            self.store_prediction_value(wind_value, prediction_run, grib_info, session)
+            self.store_prediction_value(station.code, wind_value, prediction_run, grib_info, session)
 
     def process_grib_file(self, filename, grib_info: ModelRunInfo, session: Session):
         """ Process a grib file, extracting and storing relevant information. """