cloudnetpy 1.95.2__tar.gz → 1.95.4__tar.gz

{cloudnetpy-1.95.2/cloudnetpy.egg-info → cloudnetpy-1.95.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cloudnetpy
-Version: 1.95.2
+Version: 1.95.4
 Summary: Python package for Cloudnet processing
 Author: Simo Tukiainen
 License: MIT License

{cloudnetpy-1.95.2 → cloudnetpy-1.95.4}/cloudnetpy/categorize/model.py

@@ -71,10 +71,12 @@ class Model(DataSource):
             for ind, (alt, prof) in enumerate(
                 zip(self.model_heights, data_in, strict=True),
             ):
-                if prof.mask.all():
+                # Use only valid levels so that fill values don't leak in
+                valid = ~ma.getmaskarray(prof)
+                if np.count_nonzero(valid) < 2:
                     datai[ind, :] = ma.masked
                 else:
-                    fun = interp1d(alt, prof, fill_value="extrapolate")
+                    fun = interp1d(alt[valid], prof[valid], fill_value="extrapolate")
                     datai[ind, :] = fun(self.mean_height)
             return CloudnetArray(datai, key, units)
 

{cloudnetpy-1.95.2 → cloudnetpy-1.95.4}/cloudnetpy/instruments/kazr.py

@@ -39,6 +39,13 @@ KEYMAP_CFR = {
     "signal_to_noise_ratio_crosspolar_v": "SNRx",
 }
 
+# Same format with short SNR names (e.g. olikazrgeM1.a1)
+KEYMAP_CFR_SHORT = {
+    **{k: v for k, v in KEYMAP_CFR.items() if v not in ("SNR", "SNRx")},
+    "snr_copol": "SNR",
+    "snr_xpol": "SNRx",
+}
+
 # KAZR moments (e.g. sgpkazrgeC1.a1) and corrected moments (sgpkazrcorgeC1.c1)
 KEYMAP_COR = {
     "reflectivity_copol": "Zh",
@@ -66,7 +73,8 @@ def kazr2nc(
     screened using the ARM significant detection mask and LDR is calculated
     from the cross- and co-polar reflectivities. LDR is screened using the
     cross-polar SNR and removed completely if the cross-polar channel is found
-    unreliable. Velocities of the non-corrected formats are dealiased using
+    unreliable. Stationary clutter layers (zero velocity and spectral width)
+    are removed. Velocities of the non-corrected formats are dealiased using
     velocity continuity.
 
     Args:
@@ -105,6 +113,7 @@ def kazr2nc(
     kazr.screen_ldr()
     kazr.correct_ldr_leakage()
     kazr.screen_noise(snr_limit)
+    kazr.screen_stationary_clutter()
     kazr.dealias_velocity()
     kazr.add_correction_bits()
     kazr.mask_invalid_data()
@@ -187,6 +196,46 @@ class Kazr(CloudnetInstrument):
             if cloudnet_array.data.ndim == 2:
                 cloudnet_array.mask_indices(is_noise)
 
+    def screen_stationary_clutter(
+        self,
+        v_lim: float = 0.03,
+        width_lim: float = 0.05,
+        min_fraction: float = 0.05,
+    ) -> None:
+        """Masks stationary clutter layers.
+
+        Some KAZRs (e.g. Oliktok 2015-2019) show persistent echoes with zero
+        Doppler velocity and minimal spectral width in fixed range gates, also
+        well above the ground (1.5-2 km at Oliktok). Pixels with near-zero
+        velocity and width are masked in range gates where such pixels are
+        found in at least `min_fraction` of the profiles. Real hydrometeors
+        with zero velocity have larger spectral width due to turbulence.
+        """
+        if "width" not in self.data:
+            return
+        v = self.data["v"][:]
+        width = self.data["width"][:]
+        is_stationary = ma.filled(
+            (ma.abs(v) < v_lim) & (width < width_lim), fill_value=False
+        )
+        fraction = np.mean(is_stationary, axis=0)
+        is_clutter = is_stationary & (fraction >= min_fraction)
+        if not np.any(is_clutter):
+            return
+        gates = np.nonzero(fraction >= min_fraction)[0]
+        height = self.data["range"].data
+        logging.info(
+            "Masking stationary clutter in %s range gates (%.0f-%.0f m)",
+            len(gates),
+            height[gates[0]],
+            height[gates[-1]],
+        )
+        is_valid = ~ma.getmaskarray(self.data["Zh"][:]) & ~is_clutter
+        is_valid = utils.remove_small_objects(is_valid, max_size=20, connectivity=2)
+        for cloudnet_array in self.data.values():
+            if cloudnet_array.data.ndim == 2:
+                cloudnet_array.mask_indices(~is_valid)
+
     def dealias_velocity(self) -> None:
         """Unfolds aliased Doppler velocities using continuity."""
         if self.corrected:
@@ -282,6 +331,8 @@ class Kazr(CloudnetInstrument):
     def _init_metadata(self, nc: netCDF4.Dataset) -> None:
         if "reflectivity" in nc.variables:
             self.keymap = KEYMAP_CFR
+            if "snr_copol" in nc.variables:
+                self.keymap = KEYMAP_CFR_SHORT
         elif "reflectivity_copol" in nc.variables:
             self.keymap = KEYMAP_COR
             self.corrected = "significant_detection_mask" in nc.variables

{cloudnetpy-1.95.2 → cloudnetpy-1.95.4}/cloudnetpy/instruments/nc_radar.py

@@ -213,16 +213,26 @@ class ChilboltonRadar(NcRadar):
 
 
 def estimate_snr_limit(
-    snr: ma.MaskedArray, n_gates: int = 50, percentile: float = 99.99
+    snr: ma.MaskedArray,
+    n_gates: int = 50,
+    percentile: float = 99.99,
+    max_sigma: float = 8,
 ) -> np.ndarray:
     """Estimates per-profile SNR threshold from noise in the top range gates.
 
     The noise level of each profile is the median SNR of its top gates. The
     margin above it is the given percentile of the noise scatter around the
-    level, pooled over all profiles.
+    level, pooled over all profiles, but at most `max_sigma` times the robust
+    (median absolute deviation) scatter so that signal in the top gates does
+    not inflate it.
     """
     top = snr[:, -n_gates:]
     noise_level = ma.median(top, axis=1)
     residual = (top - noise_level[:, np.newaxis]).compressed()
-    margin = np.percentile(residual, percentile) if residual.size else 0
+    if residual.size:
+        margin = np.percentile(residual, percentile)
+        sigma = 1.4826 * np.median(np.abs(residual))
+        margin = min(margin, max_sigma * sigma)
+    else:
+        margin = 0
     return ma.filled(noise_level + margin, fill_value=np.inf)

{cloudnetpy-1.95.2 → cloudnetpy-1.95.4}/cloudnetpy/version.py

@@ -1,4 +1,4 @@
 MAJOR = 1
 MINOR = 95
-PATCH = 2
+PATCH = 4
 __version__ = f"{MAJOR}.{MINOR}.{PATCH}"

{cloudnetpy-1.95.2 → cloudnetpy-1.95.4/cloudnetpy.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cloudnetpy
-Version: 1.95.2
+Version: 1.95.4
 Summary: Python package for Cloudnet processing
 Author: Simo Tukiainen
 License: MIT License