sync embedded/arc_tz_weather from arc_tz_weather

Author: actionresearchprojects

embedded/arc_tz_weather/index.html

@@ -295,6 +295,38 @@ def spike_filter(series, max_val):
     return series.where(series <= max_val)
 
 
+def peak_wind_qc(df):
+    """Apply quality control to peak_wind_kph in-place and return the DataFrame.
+
+    Two filters are applied:
+      1. Reed switch bounce: peak_wind_kph / avg_wind_kph > 8 AND peak_wind_kph > 25.
+         A ratio this extreme is physically implausible and indicates sensor artefact.
+         Rows where avg == 0 and peak > 25 are treated as infinite ratio and flagged.
+      2. Ceiling filter: peak_wind_kph > 100, regardless of average speed.
+
+    Flagged rows have peak_wind_kph replaced with NaN. avg_wind_kph is unaffected.
+    A boolean column 'peak_wind_flagged' is added (True = flagged).
+
+    Args:
+        df: DataFrame with columns avg_wind_kph and peak_wind_kph.
+    Returns:
+        The same DataFrame with peak_wind_kph cleaned and peak_wind_flagged added.
+    """
+    import numpy as np
+
+    # Ratio-based bounce filter: peak > 8x average AND peak > 25 km/h.
+    # Rows with avg == 0 and peak > 25 are treated as ratio-infinite, so flagged directly.
+    zero_avg = df["avg_wind_kph"] == 0
+    ratio = np.where(zero_avg, np.inf, df["peak_wind_kph"] / df["avg_wind_kph"].replace(0, np.nan))
+    bounce_mask = (np.array(ratio) > 8) & (df["peak_wind_kph"] > 25)
+    ceiling_mask = df["peak_wind_kph"] > 100
+    flagged = bounce_mask | ceiling_mask
+
+    df["peak_wind_flagged"] = flagged
+    df.loc[flagged, "peak_wind_kph"] = np.nan
+    return df
+
+
 def compass_bin(degrees, n_points=16):
     """Assign a compass direction label to a bearing in degrees."""
     dirs = COMPASS_DIRS_16 if n_points == 16 else COMPASS_DIRS_8

embedded/arc_tz_weather/modules/common.py

@@ -13,7 +13,7 @@
     COMPASS_DIRS_16, COMPASS_DIRS_8, WIND_SPEED_BINS, WIND_SPEED_LABELS,
     WIND_SPEED_COLORS, BEAUFORT_SCALE, VENTILATION_COLORS, WIND_CLASSIFICATIONS,
     KN_TO_KPH, TIMEZONE, CALM_THRESHOLD_KPH,
-    spike_filter, compass_bin, beaufort_number, weibull_fit, to_eat_ms,
+    peak_wind_qc, compass_bin, beaufort_number, weibull_fit, to_eat_ms,
     get_season_boundaries,
 )
 
@@ -28,8 +28,11 @@ def process(df):
     """
     wdf = df.copy()
 
-    # Apply spike filter to peak wind
-    wdf["peak_wind_kph"] = spike_filter(wdf["peak_wind_kph"], 150)
+    # Quality-control peak wind before any analysis:
+    # - flag reed switch bounce (avg < 1.0 km/h and peak > 25 km/h)
+    # - flag implausible ceiling (peak > 100 km/h)
+    # Flagged values are set to NaN; avg_wind_kph is unchanged.
+    wdf = peak_wind_qc(wdf)
 
     # Compass direction labels
     wdf["compass_16"] = wdf["wind_dir"].apply(lambda d: compass_bin(d, 16))