Improve dataset for L3

Author: ghiggi

gpm/dataset/attrs.py

@@ -56,6 +56,8 @@
     "Sensor",
     # DPRKuInfo, DPRKaInfo,
     "eqvWavelength",
+    # IMERG
+    # "title",
 )
 
 
@@ -183,7 +185,11 @@ def get_granule_attrs(dt):
     # Flatten attributes (without group)
     if _has_nested_dictionary(nested_attrs):
         attrs = {}
-        _ = [attrs.update(group_attrs) for group, group_attrs in nested_attrs.items()]
+        for k, v in nested_attrs.items():
+            if isinstance(v, dict):
+                attrs.update(v)
+            else:
+                attrs[k] = v
     else:
         attrs = nested_attrs
     # Subset only required attributes

gpm/dataset/coords.py

@@ -110,11 +110,19 @@ def get_grid_coords(dt, scan_mode):
     """
     attrs = decode_string(dt.attrs["FileHeader"])
     start_time = attrs["StartGranuleDateTime"][:-1]  # 2016-03-09T10:30:00.000Z
-    # end_time = attrs["StopGranuleDateTime"][:-1]    # 2003-05-01T23:59:59.999Z
-    time_interval = attrs["TimeInterval"]
-    time_delta = get_time_delta_from_time_interval(time_interval)
-    start_time = np.array([start_time]).astype("M8[ns]")
-    end_time = start_time + time_delta
+    end_time = attrs.get("StopGranuleDateTime", "")[:-1]  # 2003-05-01T23:59:59.999Z
+    time_interval = attrs.get("TimeInterval", "")
+    if time_interval in {"ORBIT", ""}:
+        start_time = np.array([start_time]).astype("M8[ns]")
+        end_time = np.array([end_time]).astype("M8[ns]")
+    elif time_interval == "MONTH":
+        end_time = pd.to_datetime(start_time) + pd.DateOffset(months=1)
+        start_time = np.array([start_time]).astype("M8[ns]")
+        end_time = np.array([end_time]).astype("M8[ns]")
+    else:
+        time_delta = get_time_delta_from_time_interval(time_interval)
+        start_time = np.array([start_time]).astype("M8[ns]")
+        end_time = start_time + time_delta
 
     # Define time coordinate
     time = xr.DataArray(end_time, dims="time")
@@ -128,18 +136,24 @@ def get_grid_coords(dt, scan_mode):
     time_bnds = np.concatenate((start_time, end_time)).reshape(1, 2)
     time_bnds = xr.DataArray(time_bnds, dims=("time", "nv"))
 
+    # Retrieve coordinates
+    lon = dt[scan_mode].get("lon")
+    lat = dt[scan_mode].get("lat")
+
     # Define dictionary with coordinates (DataArray)
-    return {
+    coords_dict = {
         "time": time,
-        "lon": dt[scan_mode]["lon"],
-        "lat": dt[scan_mode]["lat"],
+        "lon": lon,
+        "lat": lat,
         "time_bnds": time_bnds,
     }
+    coords_dict = {k: v for k, v in coords_dict.items() if v is not None}
+    return coords_dict
 
 
 def get_coords(dt, scan_mode):
     """Get coordinates from GPM objects."""
-    return get_grid_coords(dt, scan_mode) if scan_mode in ["Grid"] else get_orbit_coords(dt, scan_mode)
+    return get_grid_coords(dt, scan_mode) if scan_mode in ["Grid", "GRID", ""] else get_orbit_coords(dt, scan_mode)
 
 
 def _subset_dict_by_dataset(ds, dictionary):

gpm/dataset/dataset.py

@@ -565,6 +565,8 @@ def open_files(
     # Infer product from file
     product = _infer_product_name(dict_scan_modes[scan_modes[0]])
 
+    # TODO: Specialized sanitizer for L3 SLH, CSH, ...
+
     # Warn if product is unknown
     if product is None:
         msg = "GPM-API didn't apply specialized variables decoding because product is unknown !"

gpm/dataset/datatree.py

@@ -88,7 +88,7 @@ def check_non_empty_granule(dt, filepath):
     """Check that the datatree (or dataset) is not empty."""
     attrs = dt.attrs
     attrs = decode_string(attrs["FileHeader"])
-    is_empty_granule = attrs["EmptyGranule"] != "NOT_EMPTY"
+    is_empty_granule = attrs.get("EmptyGranule", "NOT_EMPTY") != "NOT_EMPTY"
     if is_empty_granule:
         raise ValueError(f"{filepath} is an EMPTY granule !")
 

gpm/dataset/dimensions.py

@@ -93,6 +93,14 @@
     "npixelht": "cross_track",
     "npixelcs": "cross_track",
     "npixelfr": "cross_track",  # S4 mode
+    # L3 CSH, SLH
+    "nlon": "lon",
+    "nlat": "lat",
+    # L3 DPR
+    "ltL": "lat",
+    "lnL": "lon",
+    "ltH": "lat",
+    "lnH": "lon",
 }
 
 SPATIAL_DIMS = [

gpm/dataset/granule.py

@@ -225,6 +225,8 @@ def _get_scan_mode_dataset(
     )
     closer = ds._close
 
+    # L3 DPR: GridTime
+
     # If GRID, time in coords and not a dimension of ds, add time dimension (L3 products)
     if "along_track" not in ds.dims and "time" in coords and "time" not in set(ds.dims):
         vars_to_expand = [
@@ -283,6 +285,9 @@ def get_scan_modes_datasets(filepath, groups, variables, decode_cf, chunks, pref
         ]
         scan_modes = set(nodes) - set(invalid_nodes)
 
+    if dt.depth == 0:  # IMERG L3 DAILY
+        scan_modes = [""]
+
     # Retrieve scan mode dataset (without cf decoding)
     dict_scan_modes = {}
     for scan_mode in scan_modes: