[SCM][SCORE TREND] add altitude visualization for the studies

experiment/meteo_france_SCM_study/visualization/studies_visualization/main_studies_visualizer.py

+from experiment.meteo_france_SCM_study.abstract_score import MannKendall, WeigthedScore, MeanScore, MedianScore
 from experiment.meteo_france_SCM_study.abstract_study import AbstractStudy
 from experiment.meteo_france_SCM_study.crocus.crocus import CrocusDepth, CrocusSwe, ExtendedCrocusDepth, \
     ExtendedCrocusSwe
 from experiment.meteo_france_SCM_study.safran.safran import SafranSnowfall, ExtendedSafranSnowfall, \
     ExtendedSafranTotalPrecip
 from experiment.meteo_france_SCM_study.visualization.studies_visualization.studies import Studies
-from experiment.meteo_france_SCM_study.visualization.studies_visualization.studies_visualizer import StudiesVisualizer
+from experiment.meteo_france_SCM_study.visualization.studies_visualization.studies_visualizer import StudiesVisualizer, \
+    AltitudeVisualizer
+from experiment.meteo_france_SCM_study.visualization.study_visualization.main_study_visualizer import ALL_ALTITUDES, \
+    study_iterator_global
 
 from experiment.meteo_france_SCM_study.visualization.study_visualization.study_visualizer import StudyVisualizer
 from collections import OrderedDict
 
-SCM_STUDIES = [SafranSnowfall, CrocusSwe, CrocusDepth]
-SCM_EXTENDED_STUDIES = [ExtendedSafranTotalPrecip, ExtendedSafranSnowfall, ExtendedCrocusSwe, ExtendedCrocusDepth]
-SCM_STUDY_TO_EXTENDED_STUDY = OrderedDict(zip(SCM_STUDIES, SCM_EXTENDED_STUDIES))
-
 
 def normal_visualization():
-    for study_type in SCM_EXTENDED_STUDIES[:1]:
+    for study_type in [ExtendedSafranTotalPrecip]:
         extended_studies = Studies(study_type)
         studies_visualizer = StudiesVisualizer(extended_studies)
         studies_visualizer.mean_as_a_function_of_altitude(region_only=True)
 
 
+def altitude_trends():
+    save_to_file = False
+    only_first_one = False
+    # altitudes that have 20 massifs at least
+    altitudes = ALL_ALTITUDES[3:-6]
+    # altitudes = ALL_ALTITUDES[:2]
+    visualizers = [StudyVisualizer(study, save_to_file=save_to_file, temporal_non_stationarity=True, verbose=True,
+                                   score_class=MedianScore)
+                   for study in study_iterator_global(study_classes=[SafranSnowfall], only_first_one=only_first_one,
+                                                      altitudes=altitudes)]
+    altitude_to_visualizer = OrderedDict(zip(altitudes, visualizers))
+    visualizer = AltitudeVisualizer(altitude_to_visualizer)
+    visualizer.negative_trend_percentages_evolution()
+
+
 if __name__ == '__main__':
-    normal_visualization()
+    altitude_trends()

experiment/meteo_france_SCM_study/visualization/studies_visualization/studies_visualizer.py

+from collections import OrderedDict, Counter
+from typing import Dict
+
 import numpy as np
 import pandas as pd
 import matplotlib.pyplot as plt
+from matplotlib.lines import Line2D
 
 from experiment.meteo_france_SCM_study.abstract_extended_study import AbstractExtendedStudy
 from experiment.meteo_france_SCM_study.visualization.studies_visualization.studies import \
     Studies
+from experiment.meteo_france_SCM_study.visualization.study_visualization.study_visualizer import StudyVisualizer
 from experiment.meteo_france_SCM_study.visualization.utils import plot_df
+from utils import cached_property, get_display_name_from_object_type
 
 
 class StudiesVisualizer(object):
@@ -29,11 +35,86 @@ class StudiesVisualizer(object):
         for study in self.studies.altitude_to_study.values():
             mean_serie = study.df_annual_total.loc[:, massif_names].mean(axis=0)
             mean_series.append(mean_serie)
-        df_mean = pd.concat(mean_series, axis=1) # type: pd.DataFrame
+        df_mean = pd.concat(mean_series, axis=1)  # type: pd.DataFrame
         df_mean.columns = self.studies.altitude_list
         plot_df(df_mean)
 
 
+class AltitudeVisualizer(object):
+
+    def __init__(self, altitude_to_study_visualizer: Dict[int, StudyVisualizer]):
+        assert isinstance(altitude_to_study_visualizer, OrderedDict)
+        self.altitude_to_study_visualizer = altitude_to_study_visualizer
+
+    @property
+    def altitudes(self):
+        return list(self.altitude_to_study_visualizer.keys())
+
+    @cached_property
+    def all_percentages(self):
+        return [v.percentages_of_negative_trends()[0] for v in self.altitude_to_study_visualizer.values()]
+
+    @property
+    def any_study_visualizer(self) -> StudyVisualizer:
+        return list(self.altitude_to_study_visualizer.values())[0]
+
+    def get_item_fct(self, year):
+        idx = self.any_study_visualizer.starting_years.index(year)
+        f = lambda s: s[idx]
+        return f
+
+    @cached_property
+    def starting_year(self):
+        return self.any_study_visualizer.starting_years[0]
+
+    def get_top_potential_years(self, reverse=False):
+        top_n = 5
+        top_top = 3
+        # keep the top_n for each altitude
+        all_years = [[year for year, _ in sorted(enumerate(p), key=lambda s:s[1], reverse=reverse)[-top_n:]] for p in self.all_percentages]
+        from itertools import chain
+        all_years = list(chain(*all_years))
+        years = [y for y, _ in sorted(Counter(all_years).items(), key=lambda s:s[1])[-top_top:]]
+        years = [y + self.starting_year for y in years]
+        return years
+
+    def negative_trend_percentages_evolution(self):
+        curve_name__metric_and_color = [
+            ('max', np.max, 'r'),
+            ('mean', np.mean, 'b'),
+            ('median', np.median, 'c'),
+            ('min', np.min, 'g'),
+        ]
+        # Add some years
+        # spotted_years = [1963, 1976]
+        # years_to_display = spotted_years
+        str_markers = ['o'] + [m for m in Line2D.markers if isinstance(m, str)][3:]
+        # for year, marker in zip(years_to_display, str_markers):
+        #     new = (str(year), self.get_item_fct(year), 'y', marker + ':')
+        #     curve_name__metric_and_color.append(new)
 
+        for year, marker in zip(self.get_top_potential_years(), str_markers):
+            new = (str(year), self.get_item_fct(year), 'm', marker + ':')
+            curve_name__metric_and_color.append(new)
+        for year, marker in zip(self.get_top_potential_years(reverse=True), str_markers):
+            new = (str(year), self.get_item_fct(year), 'y', marker + ':')
+            curve_name__metric_and_color.append(new)
 
+        fig, ax = plt.subplots(1, 1, figsize=self.any_study_visualizer.figsize)
+        for curve_name, metric, color, *marker in curve_name__metric_and_color[:]:
+            marker, curve_name = (marker[0], curve_name + ' starting year') if marker \
+                else ('-', curve_name + ' over the starting years')
+            values = [metric(p) for p in self.all_percentages]
+            ax.plot(self.altitudes, values, color + marker, label=curve_name)
+        ax.legend()
+        ax.set_xticks(self.altitudes)
+        ax.set_yticks(list(range(0,101, 10)))
+        ax.grid()
 
+        ax.axhline(y=50, color='k')
+        ax.set_ylabel('% of negative trends')
+        ax.set_xlabel('altitude')
+        scoer_class_name = get_display_name_from_object_type(self.any_study_visualizer.score_class)
+        title = 'Distribution of negative trend for the {}'.format(scoer_class_name)
+        ax.set_title(title)
+        plt.show()

experiment/meteo_france_SCM_study/visualization/study_visualization/study_visualizer.py

@@ -402,7 +402,7 @@ class StudyVisualizer(object):
 
     def visualize_score_wrt_starting_year(self, ax, massif_name):
         if massif_name is None:
-            percentage, title = self.percentage_of_negative_trends()
+            percentage, title = self.percentages_of_negative_trends()
             scores = percentage
             ax.set_ylabel('% of negative trends')
             # Add two lines of interest
@@ -420,24 +420,27 @@ class StudyVisualizer(object):
         ax.set_title(title)
         ax.xaxis.set_ticks(self.starting_years[2::20])
 
-    def percentage_of_negative_trends(self):
+    def percentages_of_negative_trends(self):
+        print('start computing percentages negative trends')
         # scores = np.median([np.array(v) < 0 for v in self.massif_name_to_scores.values()], axis=0)
+        # Take the mean with respect to the massifs
+        # We obtain an array whose length equal the length of starting years
         scores = np.mean([np.array(v) < 0 for v in self.massif_name_to_scores.values()], axis=0)
-        percentage = 100 * scores
+        percentages = 100 * scores
         # First argmin, first argmax
         argmin, argmax = np.argmin(scores), np.argmax(scores)
         # Last argmin, last argmax
         # argmin, argmax = len(scores) - 1 - np.argmin(scores[::-1]), len(scores) - 1 - np.argmax(scores[::-1])
         top_starting_year_for_positive_trend = self.starting_years[argmin]
         top_starting_year_for_negative_trend = self.starting_years[argmax]
-        top_percentage_positive_trend = round(100 - percentage[argmin], 0)
-        top_percentage_negative_trend = round(percentage[argmax], 0)
+        top_percentage_positive_trend = round(100 - percentages[argmin], 0)
+        top_percentage_negative_trend = round(percentages[argmax], 0)
         title = "Global trend; > 0: {}% in {}; < 0: {}% in {}".format(top_percentage_positive_trend,
                                                                       top_starting_year_for_positive_trend,
                                                                       top_percentage_negative_trend,
                                                                       top_starting_year_for_negative_trend)
 
-        return percentage, title
+        return percentages, title
 
     def visualize_all_mean_and_max_graphs(self):
         specified_massif_ids = [self.study.study_massif_names.index(massif_name)