Merge branch 'max_version_0.6.3' into 'master'

Max version 0.6.3

See merge request !15

CHANGELOG.md

 # Changelog
 All notable changes to this project will be documented in this file.
 
+## v0.6.3 - 2024-07-30 - fixed custom-sampling
+
+### general:
+* reference series now also uses the sampling type and adapts the data range accordingly
+* fixed calc_data_capture
+
+### technical:
+* custom sample statistics now returns with an index that matches the start-date instead of the start of the year
+
 ## v0.6.2 - 2024-07-22 - bugfixes
 
 ### general:

setup.cfg

 [metadata]
 name = toarstats
-version = 0.6.2
+version = 0.6.3
 author = Niklas Selke, Martin Schultz, Max Lensing
 author_email = n.selke@fz-juelich.de, m.schultz@fz-juelich.de, m.lensing@fz-juelich.de
 description = Collection of statistics for the TOAR community

toarstats/metrics/interface.py

@@ -127,7 +127,7 @@ def calculate_statistics(
         else DEFAULT_CROPS, input_parameters.required.seasons,
         input_parameters.required.crops
     )
-    ref = create_reference_series(input_parameters.data.index, daterange)
+    ref = create_reference_series(input_parameters.data.index, input_parameters.sampling, daterange)
     resample_rule = ("seasonal" if input_parameters.sampling == "vegseason"
                      else input_parameters.sampling)
     min_data_capture_value = (input_parameters.min_data_capture
@@ -145,8 +145,14 @@ def calculate_statistics(
                 min_data_capture_value
             )
         )
-    results = harmonize_time(
-        results, resample_rule,
-        "H" if "diurnal_cycle" in input_parameters.statistics else None
-    )
+
+    if resample_rule == "custom":
+        custom_index = pd.DatetimeIndex([ref.index.min().strftime("%Y-%m-%d")])
+        for res in results:
+            res["ser"].index = custom_index
+    else:
+        results = harmonize_time(
+            results, resample_rule,
+            "H" if "diurnal_cycle" in input_parameters.statistics else None
+        )
     return pd.DataFrame({res["name"]: res["ser"] for res in results})

toarstats/metrics/stats_utils.py

@@ -113,40 +113,48 @@ def calc_data_capture(ser, ref, sampling, how, mincount=0, minfrac=None,
 
     :return: A series with the data capture fraction
     """
-    if sampling == "100YS":
-        ser_tmp = ser.resample(sampling).count().values
-        fcov = ser_tmp / ref.resample(sampling).count().values
-        data_capture = pd.Series([fcov[0]], index=[ref.index[0].round('H')], name="data_capture")
-    else:
-        ser_tmp = ser.resample(sampling).count()
-        fcov = ser_tmp / ref.resample(sampling).count()
-        data_capture = fcov.reindex(ser_tmp.index)
-    return data_capture
+    # brings the series to the form of the reference time series, no values in series are represented by NaN
+    ser_com = ser.combine(ref, max, fill_value=None)
+
+    ser_tmp = ser_com.resample(sampling).count()
+    fcov = ser_tmp / ref.resample(sampling).count()
+    return fcov.reindex(ser_tmp.index)
 
 
-def create_reference_series(index, daterange=None):
-    """Create a reference series.
+def create_reference_series(index, sampling, daterange):
+    """Create a reference series by the daterange. Additional extends to the intervals specified by the sampling.
 
     :param index: the given index
 
+    :param sampling: sampling rate defines the daterange of the ref-series
+            e.g. monthly changes the daterange to fill in all months touched by daterange
+
     :param daterange: start- and end-date separated by comma
            if set, the reference series creates a series from start to end-date
 
 
-    :return: A series with a date range spanning from the beginning of
-             the earliest given year to the ending of the latest given
-             year and filled with zeros
+    :return: A series with a date range based on the specified date range. The date range is extended based on the
+            specified sampling.
     """
     min_date = index.min()
     max_date = index.max()
-    if daterange:
-        start_date = pd.to_datetime(daterange.split(",")[0])
-        start_date = start_date.replace(minute=min_date.minute)
-        end_date = pd.to_datetime(daterange.split(",")[1])
-        end_date = end_date.replace(minute=max_date.minute)
-    else:
-        start_date = f"{min_date.year}-01-01 00:{min_date.minute}"
-        end_date = f"{max_date.year}-12-31 23:{max_date.minute}"
+    # adjust start and end-date by the minute-offset of the timeseries
+    start_date = pd.to_datetime(daterange.split(",")[0])
+    start_date = start_date.replace(minute=min_date.minute)
+    end_date = pd.to_datetime(daterange.split(",")[1])
+    end_date = end_date.replace(minute=max_date.minute)
+
+    if sampling in ["annual", "seasonal", "summer", "xsummer"]:
+        start_date = start_date.replace(month=1, day=1, hour=0)
+        end_date = end_date.replace(month=12, day=31, hour=23)
+    elif sampling == "monthly":
+        start_date = start_date.replace(day=1, hour=0)
+        end_date = pd.offsets.MonthEnd().rollforward(end_date)
+        end_date = end_date.replace(hour=23)
+    elif sampling == "daily":
+        start_date = start_date.replace(hour=0)
+        end_date = end_date.replace(hour=23)
+
     reference_index = pd.date_range(start=start_date, end=end_date, freq="h")
     return pd.Series(0, reference_index)