grdwindinversion 0.2.3.post13__tar.gz → 0.2.3.post15__tar.gz

{grdwindinversion-0.2.3.post13 → grdwindinversion-0.2.3.post15}/.github/workflows/publish.yml

@@ -11,9 +11,9 @@ jobs:
     if: github.repository == 'umr-lops/grdwindinversion'
     steps:
       - name: Checkout
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
       - name: Set up Python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@v5
         with:
           python-version: "3.x"
       - name: Install dependencies
@@ -22,12 +22,12 @@ jobs:
           python -m pip install build twine
       - name: Build
         run: |
-          python -m build --sdist --outdir dist/ .
+          python -m build --sdist --wheel --outdir dist/ .
       - name: Check the built archives
         run: |
           twine check dist/*
       - name: Upload build artifacts
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
           name: packages
           path: dist/*
@@ -45,10 +45,10 @@ jobs:
 
     steps:
       - name: Download build artifacts
-        uses: actions/download-artifact@v3
+        uses: actions/download-artifact@v4
         with:
           name: packages
           path: dist/
 
       - name: Publish to PyPI
-        uses: pypa/gh-action-pypi-publish@b7f401de30cb6434a1e19f805ff006643653240e
+        uses: pypa/gh-action-pypi-publish@81e9d935c883d0b210363ab89cf05f3894778450

{grdwindinversion-0.2.3.post13/grdwindinversion.egg-info → grdwindinversion-0.2.3.post15}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: grdwindinversion
-Version: 0.2.3.post13
+Version: 0.2.3.post15
 Summary: Package to perform Wind inversion from GRD Level-1 SAR images
 License: MIT
 Classifier: Development Status :: 2 - Pre-Alpha

{grdwindinversion-0.2.3.post13 → grdwindinversion-0.2.3.post15}/grdwindinversion/config_prod.yaml

@@ -1,3 +1,4 @@
+no_subdir: True
 S1A:
     GMF_VV_NAME: "gmf_cmod5n"
     GMF_VH_NAME: "gmf_s1_v2"
@@ -5,6 +6,10 @@ S1A:
     apply_flattening: True
     recalibration: False
     ancillary: "ecmwf"
+    inc_step: 0.1
+    wspd_step: 0.1
+    phi_step: 1.0
+    resolution: "high"
 S1B:
     GMF_VV_NAME: "gmf_cmod5n"
     GMF_VH_NAME: "gmf_s1_v2"
@@ -12,6 +17,10 @@ S1B:
     apply_flattening: True
     recalibration: False
     ancillary: "ecmwf"
+    inc_step: 0.1
+    wspd_step: 0.1
+    phi_step: 1.0
+    resolution: "high"
 RS2:
     GMF_VV_NAME: "gmf_cmod5n"
     GMF_VH_NAME: "gmf_rs2_v2"
@@ -19,6 +28,10 @@ RS2:
     apply_flattening: False
     recalibration: False
     ancillary: "ecmwf"
+    inc_step: 0.1
+    wspd_step: 0.1
+    phi_step: 1.0
+    resolution: "high"
 RCM:
     GMF_VV_NAME: "gmf_cmod5n"
     GMF_VH_NAME: "gmf_rcm_noaa"
@@ -26,8 +39,7 @@ RCM:
     apply_flattening: True
     recalibration: False
     ancillary: "ecmwf"
-
-inc_step: 0.1
-wspd_step: 0.1
-phi_step: 1.0
-resolution: "high"
+    inc_step: 0.1
+    wspd_step: 0.1
+    phi_step: 1.0
+    resolution: "high"

{grdwindinversion-0.2.3.post13 → grdwindinversion-0.2.3.post15}/grdwindinversion/config_prod_recal.yaml

@@ -1,3 +1,4 @@
+no_subdir: True
 S1A:
     GMF_VV_NAME: "gmf_cmod5n"
     GMF_VH_NAME: "gmf_s1_v2"

{grdwindinversion-0.2.3.post13 → grdwindinversion-0.2.3.post15}/grdwindinversion/inversion.py

@@ -15,6 +15,7 @@ from scipy.ndimage import binary_dilation
 import re
 import string
 import os
+from grdwindinversion.streaks import get_streaks
 from grdwindinversion.load_config import getConf
 # optional debug messages
 import logging
@@ -49,7 +50,7 @@ def getSensorMetaDataset(filename):
         raise ValueError("must be S1A|S1B|RS2|RCM, got filename %s" % filename)
 
 
-def getOutputName2(input_file, outdir, sensor, meta):
+def getOutputName2(input_file, outdir, sensor, meta, subdir=True):
     """
     Create output filename for L2-GRD product
 
@@ -84,21 +85,14 @@ def getOutputName2(input_file, outdir, sensor, meta):
         match = regex.match(basename_match)
         MISSIONID, BEAM, PRODUCT, RESOLUTION, LEVEL, CLASS, POL, STARTDATE, STOPDATE, ORBIT, TAKEID, PRODID = match.groups()
         new_format = f"{MISSIONID.lower()}-{BEAM.lower()}-owi-xx-{STARTDATE.lower()}-{STOPDATE.lower()}-{ORBIT}-{TAKEID}.nc"
-        out_file = os.path.join(outdir, basename, new_format)
-        return out_file
-
     elif sensor == 'RS2':
         regex = re.compile(
             "(RS2)_OK([0-9]+)_PK([0-9]+)_DK([0-9]+)_(....)_(........)_(......)_(.._?.?.?)_(S.F)")
         template = string.Template(
             "${MISSIONID}_OK${DATA1}_PK${DATA2}_DK${DATA3}_${DATA4}_${DATE}_${TIME}_${POLARIZATION}_${LAST}")
         match = regex.match(basename_match)
-
         MISSIONID, DATA1, DATA2, DATA3, DATA4, DATE, TIME, POLARIZATION, LAST = match.groups()
         new_format = f"{MISSIONID.lower()}--owi-xx-{meta_start_date.lower()}-{meta_stop_date.lower()}-_____-_____.nc"
-        out_file = os.path.join(outdir, basename, new_format)
-        return out_file
-
     elif sensor == 'RCM':
         regex = re.compile(
             "([A-Z0-9]+)_OK([0-9]+)_PK([0-9]+)_(.*?)_(.*?)_(.*?)_(.*?)_(.*?)_(.*?)_(.*?)")
@@ -107,13 +101,16 @@ def getOutputName2(input_file, outdir, sensor, meta):
         match = regex.match(basename_match)
         MISSIONID, DATA1, DATA2, DATA3, DATA4, DATE, TIME, POLARIZATION1, POLARIZATION2, LAST = match.groups()
         new_format = f"{MISSIONID.lower()}--owi-xx-{meta_start_date.lower()}-{meta_stop_date.lower()}-_____-_____.nc"
-        out_file = os.path.join(outdir, basename, new_format)
-        return out_file
-
     else:
         raise ValueError(
             "sensor must be S1A|S1B|RS2|RCM, got sensor %s" % sensor)
 
+    if subdir:
+        out_file = os.path.join(outdir, basename, new_format)
+    else:
+        out_file = os.path.join(outdir, new_format)
+    return out_file
+
 
 def getAncillary(meta, ancillary_name='ecmwf'):
     """
@@ -227,7 +224,7 @@ def inverse(dual_pol, inc, sigma0, sigma0_dual, ancillary_wind, dsig_cr, model_v
         sigma0 to be inverted for dualpol
     ancillary_wind=: xarray.DataArray (numpy.complex28)
         ancillary wind
-            | (for example ecmwf winds), in **GMF convention** (-np.conj included), 
+            | (for example ecmwf winds), in **GMF convention** (-np.conj included),
     dsig_cr=: float or xarray.DataArray
         parameters used for
 
@@ -284,7 +281,7 @@ def inverse(dual_pol, inc, sigma0, sigma0_dual, ancillary_wind, dsig_cr, model_v
     return wind_co, None, None
 
 
-def makeL2asOwi(xr_dataset, dual_pol, copol, crosspol):
+def makeL2asOwi(xr_dataset, dual_pol, copol, crosspol, add_streaks):
     """
     Rename xr_dataset variables and attributes to match naming convention.
 
@@ -322,6 +319,7 @@ def makeL2asOwi(xr_dataset, dual_pol, copol, crosspol):
         'winddir_co': 'owiWindDirection_co',
         'ancillary_wind_speed': 'owiAncillaryWindSpeed',
         'ancillary_wind_direction': 'owiAncillaryWindDirection',
+        'sigma0_detrend': 'owiNrcs_detrend'
     })
 
     if "offboresight" in xr_dataset:
@@ -350,6 +348,9 @@ def makeL2asOwi(xr_dataset, dual_pol, copol, crosspol):
     xr_dataset.owiNrcs.attrs['long_name'] = 'Normalized Radar Cross Section'
     xr_dataset.owiNrcs.attrs['definition'] = 'owiNrcs_no_noise_correction - owiNesz'
 
+    xr_dataset['owiMask_Nrcs'] = xr_dataset['sigma0_mask'].sel(pol=copol)
+    xr_dataset.owiMask_Nrcs.attrs = xr_dataset.sigma0_mask.attrs
+
     # NESZ & DSIG
     xr_dataset = xr_dataset.assign(
         owiNesz=(['line', 'sample'], xr_dataset.nesz.sel(pol=copol).values))
@@ -395,14 +396,20 @@ def makeL2asOwi(xr_dataset, dual_pol, copol, crosspol):
             'winddir_dual': 'owiWindDirection',
             'windspeed_cross': 'owiWindSpeed_cross',
             'windspeed_dual': 'owiWindSpeed',
+            'sigma0_detrend_cross': 'owiNrcs_detrend_cross'
         })
         # nrcs cross
         xr_dataset['owiNrcs_cross'] = xr_dataset['sigma0_ocean'].sel(
             pol=crosspol)
+
         xr_dataset.owiNrcs_cross.attrs['units'] = 'm^2 / m^2'
         xr_dataset.owiNrcs_cross.attrs['long_name'] = 'Normalized Radar Cross Section'
         xr_dataset.owiNrcs_cross.attrs['definition'] = 'owiNrcs_cross_no_noise_correction - owiNesz_cross'
 
+        xr_dataset['owiMask_Nrcs_cross'] = xr_dataset['sigma0_mask'].sel(
+            pol=crosspol)
+        xr_dataset.owiMask_Nrcs_cross.attrs = xr_dataset.sigma0_mask.attrs
+
         # nesz cross
         xr_dataset = xr_dataset.assign(owiNesz_cross=(
             ['line', 'sample'], xr_dataset.nesz.sel(pol=crosspol).values))  # no flattening
@@ -429,6 +436,11 @@ def makeL2asOwi(xr_dataset, dual_pol, copol, crosspol):
 
             xr_dataset.owiNrcs_cross.attrs['definition'] = 'owiNrcs_cross_no_noise_correction_recalibrated - owiNesz_cross'
 
+    if add_streaks:
+        xr_dataset = xr_dataset.rename({
+            'streaks_direction': 'owiStreaksDirection',
+        })
+
     #  other variables
 
     xr_dataset['owiWindQuality'] = xr.full_like(xr_dataset.owiNrcs, 0)
@@ -487,7 +499,7 @@ def makeL2asOwi(xr_dataset, dual_pol, copol, crosspol):
     return xr_dataset, encoding
 
 
-def preprocess(filename, outdir, config_path, overwrite=False, resolution='1000m'):
+def preprocess(filename, outdir, config_path, overwrite=False, add_streaks=False, resolution='1000m'):
     """
     Main function to generate L2 product.
 
@@ -507,7 +519,7 @@ def preprocess(filename, outdir, config_path, overwrite=False, resolution='1000m
     Returns
     -------
     xarray.Dataset
-        final dataset   
+        final dataset
     """
 
     sensor, sensor_longname, fct_meta, fct_dataset = getSensorMetaDataset(
@@ -529,16 +541,19 @@ def preprocess(filename, outdir, config_path, overwrite=False, resolution='1000m
 
     recalibration = config["recalibration"]
     meta = fct_meta(filename)
-    out_file = getOutputName2(filename, outdir, sensor, meta)
+
+    no_subdir_cfg = config_base.get("no_subdir", False)
+    out_file = getOutputName2(filename, outdir, sensor,
+                              meta, subdir=not no_subdir_cfg)
 
     if os.path.exists(out_file) and overwrite is False:
-        raise FileExistsError("out_file %s exists already")
+        raise FileExistsError("outfile %s already exists" % out_file)
 
     ancillary_name = config["ancillary"]
     map_model = getAncillary(meta, ancillary_name)
     if map_model is None:
         raise Exception(
-            f'the weather model is not set `map_model` is None -> you probably don"t have access to f{ancillary_name} archive')
+            f"the weather model is not set `map_model` is None -> you probably don't have access to {ancillary_name} archive")
 
     try:
         if ((recalibration) & ("SENTINEL" in sensor_longname)):
@@ -592,6 +607,25 @@ def preprocess(filename, outdir, config_path, overwrite=False, resolution='1000m
         copol_gmf = 'HH'
         crosspol_gmf = 'VH'
 
+    model_vv = config["GMF_"+copol_gmf+"_NAME"]
+    model_vh = config["GMF_"+crosspol_gmf+"_NAME"]
+
+    # need to load gmfs before inversion
+    gmfs_impl = [x for x in [model_vv, model_vh] if "gmf_" in x]
+    windspeed.gmfs.GmfModel.activate_gmfs_impl(gmfs_impl)
+    sarwings_luts = [x for x in [model_vv, model_vh]
+                     if x.startswith("sarwing_lut_")]
+
+    if len(sarwings_luts) > 0:
+        windspeed.register_sarwing_luts(getConf()["sarwing_luts_path"])
+
+    nc_luts = [x for x in [model_vv, model_vh] if x.startswith("nc_lut")]
+
+    if len(nc_luts) > 0:
+        windspeed.register_nc_luts(getConf()["nc_luts_path"])
+
+    if (model_vv == "gmf_cmod7"):
+        windspeed.register_cmod7(getConf()["lut_cmod7_path"])
     #  Step 2 - clean and prepare dataset
 
     # variables to not keep in the L2
@@ -681,21 +715,32 @@ def preprocess(filename, outdir, config_path, overwrite=False, resolution='1000m
     # nrcs processing
     xr_dataset['sigma0_ocean'] = xr.where(xr_dataset['mask'], np.nan,
                                           xr_dataset['sigma0'].compute()).transpose(*xr_dataset['sigma0'].dims)
-    xr_dataset['sigma0_ocean'] = xr.where(
-        xr_dataset['sigma0_ocean'] <= 0, np.nan, xr_dataset['sigma0_ocean'])
-
     xr_dataset['sigma0_ocean'].attrs = xr_dataset['sigma0'].attrs
-    #  we forced it to nan
-    xr_dataset['sigma0_ocean'].attrs['comment'] = "clipped, no values <=0"
+    #  we forced it to 1e-15
+    xr_dataset['sigma0_ocean'].attrs['comment'] = "clipped, no values <=0 ; 1e-15 instread"
+
+    # rajout d'un mask pour les valeurs <=0:
+    xr_dataset['sigma0_mask'] = xr.where(
+        xr_dataset['sigma0_ocean'] <= 0, 1, 0).transpose(*xr_dataset['sigma0'].dims)
+    xr_dataset.sigma0_mask.attrs['valid_range'] = np.array([0, 1])
+    xr_dataset.sigma0_mask.attrs['flag_values'] = np.array([0, 1])
+    xr_dataset.sigma0_mask.attrs['flag_meanings'] = 'valid no_valid'
+    xr_dataset['sigma0_ocean'] = xr.where(
+        xr_dataset['sigma0_ocean'] <= 0, 1e-15, xr_dataset['sigma0_ocean'])
 
     xr_dataset['sigma0_ocean_raw'] = xr.where(xr_dataset['mask'], np.nan,
                                               xr_dataset['sigma0_raw'].compute()).transpose(*xr_dataset['sigma0_raw'].dims)
-    xr_dataset['sigma0_ocean_raw'] = xr.where(
-        xr_dataset['sigma0_ocean_raw'] <= 0, np.nan, xr_dataset['sigma0_ocean_raw'])
+
     xr_dataset['sigma0_ocean_raw'].attrs = xr_dataset['sigma0_raw'].attrs
 
+    xr_dataset['sigma0_detrend'] = xsarsea.sigma0_detrend(
+        xr_dataset.sigma0.sel(pol=copol), xr_dataset.incidence, model=model_vv)
+
     # processing
     if dual_pol:
+
+        xr_dataset['sigma0_detrend_cross'] = xsarsea.sigma0_detrend(
+            xr_dataset.sigma0.sel(pol=crosspol), xr_dataset.incidence, model=model_vh)
         if config["apply_flattening"]:
             xr_dataset = xr_dataset.assign(nesz_cross_final=(
                 ['line', 'sample'], windspeed.nesz_flattening(xr_dataset.nesz.sel(pol=crosspol), xr_dataset.incidence)))
@@ -721,24 +766,50 @@ def preprocess(filename, outdir, config_path, overwrite=False, resolution='1000m
         sigma0_ocean_cross = None
         dsig_cross = 0.1  # default value set in xsarsea
 
-    model_vv = config["GMF_"+copol_gmf+"_NAME"]
-    model_vh = config["GMF_"+crosspol_gmf+"_NAME"]
-
     if ((recalibration) & ("SENTINEL" in sensor_longname)):
-        xr_dataset["path_aux_pp1_new"] = os.path.basename(os.path.dirname(
+        xr_dataset.attrs["path_aux_pp1_new"] = os.path.basename(os.path.dirname(
             os.path.dirname(xsar_dataset.datatree['recalibration'].attrs['path_aux_pp1_new'])))
-        xr_dataset["path_aux_cal_new"] = os.path.basename(os.path.dirname(
+        xr_dataset.attrs["path_aux_cal_new"] = os.path.basename(os.path.dirname(
             os.path.dirname(xsar_dataset.datatree['recalibration'].attrs['path_aux_cal_new'])))
 
-        xr_dataset["path_aux_pp1_old"] = os.path.basename(os.path.dirname(
+        xr_dataset.attrs["path_aux_pp1_old"] = os.path.basename(os.path.dirname(
             os.path.dirname(xsar_dataset.datatree['recalibration'].attrs['path_aux_pp1_old'])))
-        xr_dataset["path_aux_cal_old"] = os.path.basename(os.path.dirname(
+        xr_dataset.attrs["path_aux_cal_old"] = os.path.basename(os.path.dirname(
             os.path.dirname(xsar_dataset.datatree['recalibration'].attrs['path_aux_cal_old'])))
 
+    if add_streaks:
+        xsar_dataset_100 = fct_dataset(
+            meta, resolution='100m')
+        xr_dataset_100 = xsar_dataset_100.datatree['measurement'].to_dataset()
+        xr_dataset_100 = xr_dataset_100.rename(map_model)
+
+        # adding sigma0 detrend
+        xr_dataset_100['sigma0_detrend'] = xsarsea.sigma0_detrend(
+            xr_dataset_100.sigma0.sel(pol=copol), xr_dataset_100.incidence, model=model_vv)
+
+        xr_dataset_100['sigma0_detrend_cross'] = xsarsea.sigma0_detrend(
+            xr_dataset_100.sigma0.sel(pol=crosspol), xr_dataset_100.incidence, model=model_vh)
+
+        sigma0_detrend_combined = xr.concat(
+            [xr_dataset_100['sigma0_detrend'],
+                xr_dataset_100['sigma0_detrend_cross']],
+            dim='pol'
+        )
+        sigma0_detrend_combined['pol'] = [copol, crosspol]
+
+        xr_dataset_100['sigma0_detrend'] = sigma0_detrend_combined
+        xr_dataset_100.land_mask.values = binary_dilation(xr_dataset_100['land_mask'].values.astype('uint8'),
+                                                          structure=np.ones((3, 3), np.uint8), iterations=3)
+        xr_dataset_100['sigma0_detrend'] = xr.where(
+            xr_dataset_100['land_mask'], np.nan, xr_dataset_100['sigma0'].compute()).transpose(*xr_dataset_100['sigma0'].dims)
+
+        xr_dataset['streaks_direction'] = get_streaks(
+            xr_dataset, xr_dataset_100)
+
     return xr_dataset, dual_pol, copol, crosspol, copol_gmf, crosspol_gmf, model_vv, model_vh, sigma0_ocean_cross, dsig_cross, sensor_longname, out_file, config
 
 
-def makeL2(filename, outdir, config_path, overwrite=False, generateCSV=True, resolution='1000m'):
+def makeL2(filename, outdir, config_path, overwrite=False, generateCSV=True, add_streaks=False, resolution='1000m'):
     """
     Main function to generate L2 product.
 
@@ -766,7 +837,7 @@ def makeL2(filename, outdir, config_path, overwrite=False, generateCSV=True, res
     """
 
     xr_dataset, dual_pol, copol, crosspol, copol_gmf, crosspol_gmf, model_vv, model_vh, sigma0_ocean_cross, dsig_cross, sensor_longname, out_file, config = preprocess(
-        filename, outdir, config_path, overwrite, resolution)
+        filename, outdir, config_path, overwrite, add_streaks, resolution)
 
     kwargs = {
         "inc_step_lr": config.pop("inc_step_lr", None),
@@ -778,23 +849,6 @@ def makeL2(filename, outdir, config_path, overwrite=False, generateCSV=True, res
         "resolution": config.pop("resolution", None),
     }
 
-    # need to load gmfs before
-
-    gmfs_impl = [x for x in [model_vv, model_vh] if "gmf_" in x]
-    windspeed.gmfs.GmfModel.activate_gmfs_impl(gmfs_impl)
-    sarwings_luts = [x for x in [model_vv, model_vh]
-                     if x.startswith("sarwing_lut_")]
-    if len(sarwings_luts) > 0:
-        windspeed.register_sarwing_luts(getConf()["sarwing_luts_path"])
-
-    nc_luts = [x for x in [model_vv, model_vh] if x.startswith("nc_lut")]
-
-    if len(nc_luts) > 0:
-        windspeed.register_nc_luts(getConf()["nc_luts_path"])
-
-    if (model_vv == "gmf_cmod7"):
-        windspeed.register_cmod7(getConf()["lut_cmod7_path"])
-
     wind_co, wind_dual, windspeed_cr = inverse(dual_pol,
                                                inc=xr_dataset['incidence'],
                                                sigma0=xr_dataset['sigma0_ocean'].sel(
@@ -857,7 +911,7 @@ def makeL2(filename, outdir, config_path, overwrite=False, generateCSV=True, res
         xr_dataset["winddir_cross"].attrs["model"] = "No model used ; content is a copy of dualpol wind direction"
 
     xr_dataset, encoding = makeL2asOwi(
-        xr_dataset, dual_pol, copol, crosspol)
+        xr_dataset, dual_pol, copol, crosspol, add_streaks=add_streaks)
 
     #  add attributes
     firstMeasurementTime = None
@@ -910,7 +964,7 @@ def makeL2(filename, outdir, config_path, overwrite=False, generateCSV=True, res
     }
 
     for recalib_attrs in ["path_aux_pp1_new", 'path_aux_pp1_old', "path_aux_cal_new", "path_aux_cal_old"]:
-        if recalib_attrs in xr_dataset:
+        if recalib_attrs in xr_dataset.attrs:
             attrs[recalib_attrs] = xr_dataset.attrs[recalib_attrs]
 
     # new one to match convention

grdwindinversion-0.2.3.post15/grdwindinversion/streaks.py

@@ -0,0 +1,79 @@
+import xarray as xr
+import xsarsea.gradients
+import xarray as xr
+from scipy.ndimage import binary_dilation
+import numpy as np
+
+
+def get_streaks(xr_dataset, xr_dataset_100):
+    """
+    Get the streaks from the wind field.
+
+    Parameters
+    ----------
+    xr_dataset : xarray.Dataset
+        dataset at user resolution.
+    xr_dataset_100 : xarray.Dataset
+        dataset at 100m resolution.
+
+    Returns
+    -------
+    xarray.Dataset
+        Extract wind direction from Koch Method using xsarsea tools.
+    """
+
+    # return empy dataArray, waiting for solution
+    return xr.DataArray(data=np.nan * np.ones([len(xr_dataset.coords[dim]) for dim in ['line','sample']]),
+                        dims=['line','sample'],
+                        coords=[xr_dataset.coords[dim] for dim in ['line','sample']])
+    #
+
+    """
+    gradients = xsarsea.gradients.Gradients(xr_dataset_100['sigma0_detrend'], windows_sizes=[
+                                            1600, 3200], downscales_factors=[1, 2], window_step=1)
+
+    # get gradients histograms as an xarray dataset
+    hist = gradients.histogram
+
+    # get orthogonals gradients
+    hist['angles'] = hist['angles'] + np.pi/2
+
+    # mean
+    hist_mean = hist.mean(['downscale_factor', 'window_size', 'pol'])
+
+    # smooth
+    hist_mean_smooth = hist_mean.copy()
+    hist_mean_smooth['weight'] = xsarsea.gradients.circ_smooth(
+        hist_mean['weight'])
+
+    # smooth only
+    # hist_smooth = hist.copy()
+    # hist_smooth['weight'] = xsarsea.gradients.circ_smooth(hist_smooth['weight'])
+
+    # select histogram peak
+    iangle = hist_mean_smooth['weight'].fillna(0).argmax(dim='angles')
+    streaks_dir = hist_mean_smooth.angles.isel(angles=iangle)
+    streaks_weight = hist_mean_smooth['weight'].isel(angles=iangle)
+    streaks = xr.merge(
+        [dict(angle=streaks_dir, weight=streaks_weight)]).drop('angles')
+
+    # streaks are [0, pi]. Remove ambiguity with anciallary wind
+    ancillary_wind = xr_dataset_100['ancillary_wind'].sel(line=streaks.line,
+                                                          sample=streaks.sample,
+                                                          method='nearest').compute()
+    streaks_c = streaks['weight'] * np.exp(1j * streaks['angle'])
+    diff_angle = xr.apply_ufunc(np.angle, ancillary_wind / streaks_c)
+    streaks_c = xr.where(np.abs(diff_angle) > np.pi/2, -streaks_c, streaks_c)
+    streaks['weight'] = np.abs(streaks_c)
+    streaks['angle'] = xr.apply_ufunc(np.angle, streaks_c)
+
+    streaks_dir = xr.apply_ufunc(
+        np.angle, streaks_c.interp(line=xr_dataset.line, sample=xr_dataset.sample))
+    streaks_dir = xr.where(
+        xr_dataset['land_mask'], np.nan, streaks_dir)
+    streaks_dir.attrs['comment'] = 'angle in radians, anticlockwise, 0=line'
+    streaks_dir.attrs['description'] = 'wind direction estimated from local gradient, and direction ambiguity removed with ancillary wind'
+    
+    return streaks_dir
+    
+    """

{grdwindinversion-0.2.3.post13 → grdwindinversion-0.2.3.post15/grdwindinversion.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: grdwindinversion
-Version: 0.2.3.post13
+Version: 0.2.3.post15
 Summary: Package to perform Wind inversion from GRD Level-1 SAR images
 License: MIT
 Classifier: Development Status :: 2 - Pre-Alpha

{grdwindinversion-0.2.3.post13 → grdwindinversion-0.2.3.post15}/grdwindinversion.egg-info/SOURCES.txt

@@ -39,6 +39,7 @@ grdwindinversion/data_config.yaml
 grdwindinversion/inversion.py
 grdwindinversion/load_config.py
 grdwindinversion/main.py
+grdwindinversion/streaks.py
 grdwindinversion/utils.py
 grdwindinversion.egg-info/PKG-INFO
 grdwindinversion.egg-info/SOURCES.txt
@@ -47,5 +48,6 @@ grdwindinversion.egg-info/entry_points.txt
 grdwindinversion.egg-info/requires.txt
 grdwindinversion.egg-info/top_level.txt
 grdwindinversion/.github/ISSUE_TEMPLATE.md
+recipe/meta.yaml
 tests/__init__.py
 tests/test_grdwindinversion.py

grdwindinversion-0.2.3.post15/recipe/meta.yaml

@@ -0,0 +1,34 @@
+package:
+  name: "grdwindinversion"
+  version: {{ environ.get('GIT_DESCRIBE_TAG', 0)}}
+
+source:
+  path: ../.
+
+build:
+  noarch: python
+  number: 0
+  script: {{ PYTHON }} -m pip install . --no-deps -vv
+
+requirements:
+  build:
+    - python >=3.9,<3.11
+    - setuptools_scm
+    - setuptools
+
+  run:
+    - python >=3.9,<3.11
+    - xsar
+    - xsarsea
+    - "xarray==2024.2.0"
+    - xarray-datatree
+    - "rioxarray<=0.15.5"
+    - "numpy<=1.26"
+    - pyyaml
+    - scipy
+    - fsspec
+    - aiohttp
+
+about:
+  home: https://github.com/umr-lops/grdwindinversion
+  summary: 'python library to compute wind speed from GRD SAR images'