Merge branch 'feature/static-risk-traj' into feature/interpolated-trajectories

Author: spjuhel

climada/data/system/entity_template.xlsx

@@ -403,6 +403,9 @@ def __init__(
 
         self.description = self._consolidate(meta, "description", description)
         self.ref_year = self._consolidate(meta, "ref_year", ref_year, DEF_REF_YEAR)
+
+        if geodata.shape[0] > 0:
+            value_unit = self._consolidate(geodata.iloc[0], "value_unit", value_unit)
         self.value_unit = self._consolidate(
             meta, "value_unit", value_unit, DEF_VALUE_UNIT
         )

climada/entity/exposures/base.py

@@ -182,7 +182,7 @@ def test__init__mda_in_kwargs(self):
     def test_read_raster_pass(self):
         """from_raster"""
         exp = Exposures.from_raster(
-            HAZ_DEMO_FL, window=Window(10, 20, 50, 60), attrs={"value_unit": "USD"}
+            HAZ_DEMO_FL, window=Window(10, 20, 50, 60), attrs={"value_unit": "PKR"}
         )
         exp.check()
         self.assertTrue(u_coord.equal_crs(exp.crs, DEF_CRS))
@@ -204,7 +204,7 @@ def test_read_raster_pass(self):
         self.assertAlmostEqual(
             exp.gdf["value"].values.reshape((60, 50))[25, 12], 0.056825936
         )
-        self.assertEqual(exp.value_unit, "USD")
+        self.assertEqual(exp.value_unit, "PKR")
 
     def test_assign_raster_pass(self):
         """Test assign_centroids with raster hazard"""
@@ -429,16 +429,27 @@ def test_read_template_pass(self):
         exp_df = Exposures(df)
         # set metadata
         exp_df.ref_year = 2020
-        exp_df.value_unit = "XSD"
+        exp_df.value_unit = "PAK"
         exp_df.check()
 
+    def test_handling_unit_conflicts_pass(self):
+        """Check that the value_unit is correctly set when there are conflicting value_unit definitions in the data frame and the meta attribute."""
+        df = pd.read_excel(ENT_TEMPLATE_XLS)
+        exp_df = Exposures(df, meta={"value_unit": "XSD"}, value_unit="XSD")
+        exp_df.check()
+        self.assertEqual(exp_df.value_unit, "XSD")
+        with self.assertRaises(ValueError) as cm:
+            exp_df = Exposures(df, meta={"value_unit": "XSD"}, value_unit="PAK")
+        with self.assertRaises(ValueError) as cm:
+            exp_df = Exposures(df, meta={"value_unit": "PAK"}, value_unit="XSD")
+        with self.assertRaises(ValueError) as cm:
+            exp_df = Exposures(df, meta={"value_unit": "PAK"}, value_unit="PAK")
+
     def test_io_hdf5_pass(self):
         """write and read hdf5"""
-        exp = Exposures(pd.read_excel(ENT_TEMPLATE_XLS), crs="epsg:32632")
-
-        # set metadata
-        exp.ref_year = 2020
-        exp.value_unit = "XSD"
+        exp = Exposures(
+            pd.read_excel(ENT_TEMPLATE_XLS), crs="epsg:32632", ref_year=2020
+        )
 
         # add another geometry column
         exp.data["geocol2"] = exp.data.geometry.copy(deep=True)

climada/entity/exposures/test/test_base.py

@@ -160,7 +160,7 @@ def __init__(
 
         """
 
-        self._reset_impact_data()
+        self._init_impact_data()
         self.snapshots = snapshots
         self.impact_computation_strategy = impact_computation_strategy
         self._date_idx = pd.DatetimeIndex(
@@ -182,8 +182,10 @@ def __init__(
             error_message = str(exc).lower()
             if "need at least one array to concatenate" in error_message:
                 self._group_id = np.array([])
+            else:
+                raise
 
-    def _reset_impact_data(self):
+    def _init_impact_data(self):
         """Util method that resets computed data, for instance when
         changing the computation strategy.
 
@@ -193,6 +195,8 @@ def _reset_impact_data(self):
         self._per_date_eai = None
         self._per_date_aai = None
 
+    _reset_impact_data = _init_impact_data
+
     @property
     def impact_computation_strategy(self) -> ImpactComputationStrategy:
         """The method used to calculate the impact from the (Haz,Exp,Vul)
@@ -204,7 +208,9 @@ def impact_computation_strategy(self) -> ImpactComputationStrategy:
     @impact_computation_strategy.setter
     def impact_computation_strategy(self, value, /):
         if not isinstance(value, ImpactComputationStrategy):
-            raise ValueError("Not an impact computation strategy")
+            raise ValueError(
+                "The provided value is not an ImpactComputationStrategy object. See the trajectory module documentation for more information on how to define your own impact computation strategies."
+            )
 
         self._impact_computation_strategy = value
         self._reset_impact_data()
@@ -352,9 +358,12 @@ def calc_return_periods_metric(self, return_periods: list[int]) -> pd.DataFrame:
         return rp_df
 
     def apply_measure(self, measure: Measure) -> "CalcRiskMetricsPoints":
-        """Creates a new `CalcRiskMetricsPoints` object with a measure.
+        """Creates a new `CalcRiskMetricsPoints` object by applying the effects
+        of the given measure.
 
-        The given measure is applied to both snapshot of the risk period.
+        The effects of the measure are applied to all the snapshots contained
+        in the initial `CalcRiskMetricsPoints` and a new `CalcRiskMetricsPoints`
+        containing the modified snapshots is returned.
 
         Parameters
         ----------
@@ -364,7 +373,7 @@ def apply_measure(self, measure: Measure) -> "CalcRiskMetricsPoints":
         Returns
         -------
 
-        CalcRiskPeriod
+        CalcRiskMetricsPoints
             The risk period with given measure applied.
 
         """

climada/trajectories/calc_risk_metrics.py

@@ -59,12 +59,16 @@
 
 
 class StaticRiskTrajectory(RiskTrajectory):
-    """This class implements static risk trajectories, objects that
+    """This class implements static risk trajectories: objects that
     regroup impacts computations for multiple dates.
 
-    This class computes risk metrics over a series of snapshots,
-    optionally applying risk discounting. It does not interpolate risk
-    between the snapshot and only provides results for each snapshot.
+    This class computes risk metrics over a series of `Snapshot` objects,
+    optionally applying risk discounting, and offers access to the results
+    in tidy formatted pandas DataFrames.
+
+    Contrary to InterpolatedRiskTrajectories, it does not interpolate risk
+    between the snapshot and only provides results at each snapshot specific
+    date.
 
     """
 
@@ -115,7 +119,7 @@ def __init__(
             The discount rate to apply to future risk. Defaults to None.
         impact_computation_strategy: ImpactComputationStrategy, optional
             The method used to calculate the impact from the (Haz,Exp,Vul)
-            of the two snapshots. Defaults to :class:`ImpactCalcComputation`.
+            for each snapshot. Defaults to :class:`ImpactCalcComputation`.
 
         """
         super().__init__(
@@ -137,7 +141,11 @@ def impact_computation_strategy(self) -> ImpactComputationStrategy:
     @impact_computation_strategy.setter
     def impact_computation_strategy(self, value, /):
         if not isinstance(value, ImpactComputationStrategy):
-            raise ValueError("Not an interpolation strategy")
+            raise ValueError(
+                "The provided impact computation strategy is not an ImpactComputationStrategy, "
+                "please refer to the documentation to define your own strategies or stick to the "
+                "default one"
+            )
 
         self._reset_metrics()
         self._risk_metrics_calculators.impact_computation_strategy = value
@@ -174,6 +182,15 @@ def _generic_metrics(
             A tidy formatted dataframe of the risk metric computed for the
             different snapshots.
 
+        Notes
+        -----
+
+        The computation checks that there are no duplicated rows of results
+        for the same tuples (Date, Group, Measure, Metric,
+        [Coordinates for metrics on that level]) and takes the first row in
+        this case.
+
+
         Raises
         ------
         NotImplementedError
@@ -211,7 +228,12 @@ def _generic_metrics(
         # When more than 2 snapshots, there might be duplicated rows, we need to remove them.
         # Should not be the case in static trajectory, but in any case we really don't want
         # duplicated rows, which would mess up some dataframe manipulation down the road.
-        tmp = tmp[~tmp.index.duplicated(keep="first")]
+        if tmp.index.duplicated().any():
+            LOGGER.warning(
+                "Duplicated rows were found in the results. Will keep the first one."
+            )
+            tmp = tmp[~tmp.index.duplicated(keep="first")]
+
         tmp = tmp.reset_index()
         if self._all_groups_name not in tmp[GROUP_COL_NAME].cat.categories:
             tmp[GROUP_COL_NAME] = tmp[GROUP_COL_NAME].cat.add_categories(
@@ -236,7 +258,8 @@ def eai_metrics(self, **kwargs) -> pd.DataFrame:
         Notes
         -----
 
-        This computation may become quite expensive for big areas with high resolution.
+        This computation may become quite expensive for exposures with many points
+        (e.g., big areas with high resolution).
 
         """
         metric_df = self._compute_metrics(