def _find_overlapping_points_in_network(points: gpd.GeoDataFrame) -> Dict[int, List[int]]:
"""
seeks unique points in network
Parameters
----------
points:
geodataframe with all points of the network
Returns
-------
"""
unique_points = []
same_points = {}
for index, point in points.iterrows():
point_list = []
for index2, point2 in points.iterrows():
if index2 not in unique_points:
if point.geometry.distance(point2.geometry) < 1e-1 and index != index2:
point_list.append(index2)
if len(point_list) == 0:
unique_points.append(index)
else:
same_points[index] = point_list
return same_points
def from_data_frame(self, trace_data: GeoDataFrame):
"""
:param trace_data:
:return:
"""
if geom_check(trace_data, "Point"):
assert all(
element in list(trace_data.columns.values)
for element in [
"trace_id",
]
), (
"Expected ['trace_id',] key to be in Data",
"got %s",
list(trace_data.columns.values),
)
for idx, feature in trace_data.iterrows():
feature_geometry, feature_property = decompose_data_frame_row(feature)
if "trace_point_id" in feature_property:
trace_point_id = feature_property["trace_point_id"]
else:
trace_point_id = uuid.uuid1().int
self.add(
x=feature_geometry["coordinates"][0],
y=feature_geometry["coordinates"][1],
trace_point_id=trace_point_id,
**feature_property,
)
elif geom_check(trace_data, "LineString"):
for idx, feature in trace_data.iterrows():
trace_id = uuid.uuid1().int
feature_geometry, feature_property = decompose_data_frame_row(feature)
line_string_coordinate = feature_geometry["coordinates"]
for nodes in line_string_coordinate:
trace_point_id = uuid.uuid1().int
self.add(
x=nodes[0],
y=nodes[-1],
trace_point_id=trace_point_id,
trace_id=trace_id,
**feature_property,
)
else:
raise ValueError("Expected Geometry Type to be in ['Point', 'LineString']")
return self
def test_cluster_points2(self):
pts = [
Point(0, 0),
Point(6, 0),
Point(6, 6),
Point(0.2, 0.2),
Point(6.2, 0.2),
Point(6.2, 6.2)
]
expected = GeoDataFrame(
pd.DataFrame([{
'geometry': Point(0.1, 0.1),
'n': 2
}, {
'geometry': Point(6.1, 0.1),
'n': 2
}, {
'geometry': Point(6.1, 6.1),
'n': 2
}]))
actual = PointClusterer(pts, max_distance=5,
is_latlon=False).get_clusters()
actual = [(c.centroid, len(c.points)) for c in actual]
expected = [(c.geometry, c.n) for key, c in expected.iterrows()]
assert len(actual) == len(expected)
for pt in expected:
assert pt in actual
def __points2grids(gdf: geopandas.GeoDataFrame,
shift: float = 1 / 32,
crs: str = '4326') -> geopandas.GeoDataFrame:
'''
Converts livneh centroid 'lat' 'lon' coordinates to polygon grids.
Inputs:
(1) a geopandas GeoDataFrame containing 'lat' and 'lon' series
(2) a float resenting the shortest distance between the centroid and the grid edge
(3) the coordinate system identified to be used, '4326' by default
Output:
(1) a geopandas GeoDataFrame containing the grid cells polygon geometry
Notes:
The 'shift' parameter is used to identify the north, east, south and west most extents of the polygon, as so...
ne---nw
| |
se---sw
'''
grids_geometry: list = []
for i, row in gdf.iterrows():
w: float = float(row['lon']) - shift
e: float = float(row['lon']) + shift
n: float = float(row['lat']) + shift
s: float = float(row['lat']) - shift
lons, lats = [w, e, e, w], [n, n, s, s]
grids_geometry.append(Polygon(zip(lons, lats)))
gdf['geometry'] = grids_geometry
return gdf.to_crs(epsg=crs)
def s_test_gdf_series(bin_gdf: GeoDataFrame,
test_config: dict,
N_norm: float = 1.0):
return [
s_test_bin(row.SpacemagBin, test_config, N_norm)
for i, row in bin_gdf.iterrows()
]
def calc_onborder(geoms_gdf: gpd.GeoDataFrame,
border_bounds: Tuple[float, float, float, float],
onborder_column_name: str = "onborder") -> gpd.GeoDataFrame:
"""
Add/update a column to the GeoDataFrame with:
* 0 if the polygon isn't on the border and
* 1 if it is.
Args
geoms_gdf: input GeoDataFrame
border_bounds: the bounds (tupple with (xmin, ymin, xmax, ymax)
to check against to determine onborder
onborder_column_name: the column name of the onborder column
"""
# Split geoms that need unioning versus geoms that don't
# -> They are on the edge of a tile
if geoms_gdf is not None and len(geoms_gdf.index) > 0:
# Check
for i, geom_row in geoms_gdf.iterrows():
# Check if the geom is on the border of the tile
geom_bounds = geom_row['geometry'].bounds # type: ignore
onborder = 0
if (geom_bounds[0] <= border_bounds[0]
or geom_bounds[1] <= border_bounds[1]
or geom_bounds[2] >= border_bounds[2]
or geom_bounds[3] >= border_bounds[3]):
onborder = 1
geoms_gdf.loc[i, onborder_column_name] = onborder # type: ignore
geoms_gdf.reset_index(drop=True, inplace=True)
return geoms_gdf
def add_nodes_from_gdf(self, gdf: gpd.GeoDataFrame, node_index_attr=None):
"""Add nodes with the given `GeoDataFrame` and fill nodes attributes with the geodataframe columns.
Parameters
----------
gdf :
GeoDataFrame representing nodes to add (one row for one node).
node_index_attr :
Node index attribute for labeling nodes. If ``None``, the dataframe index is used, else
the given column is used. (Default value = None)
See Also
--------
add_edges_from_gdf
"""
if not (gnx.is_null_crs(self.crs) or gnx.is_null_crs(gdf.crs)
or gnx.crs_equals(gdf.crs, self.crs)):
gdf = gdf.to_crs(self.crs, inplace=False)
if node_index_attr is not None:
gdf = gdf.set_index(node_index_attr, drop=True, inplace=False)
if gdf._geometry_column_name != self.nodes_geometry_key:
gdf = gdf.rename(
columns={gdf._geometry_column_name: self.nodes_geometry_key},
inplace=False)
gdf.set_geometry(self.nodes_geometry_key, inplace=True)
self.add_nodes_from(gdf.iterrows())
self.check_nodes_validity()
def add_edges_from_gdf(self,
gdf: gpd.GeoDataFrame,
edge_first_node_attr=None,
edge_second_node_attr=None):
"""Add edges with the given `GeoDataFrame`. If no dataframe columns are specified for first and second node,
the dataframe index must be a multi-index `(u, v)`.
Parameters
----------
gdf :
GeoDataFrame representing edges to add (one row for one edge).
edge_first_node_attr :
Edge first node attribute. If ``None``, the dataframe index is used, else the given
column is used. Must be used with ``edge_second_node_attr``. (Default value = None)
edge_second_node_attr :
Edge second node attribute. If ``None``, the dataframe index is used, else the
given column is used. Must be used with ``edge_first_node_attr``. (Default value = None)
See Also
--------
add_nodes_from_gdf
"""
if not (gnx.is_null_crs(self.crs) or gnx.is_null_crs(gdf.crs)
or gnx.crs_equals(gdf.crs, self.crs)):
gdf = gdf.to_crs(self.crs, inplace=False)
if edge_first_node_attr is not None and edge_second_node_attr is not None:
gdf = gdf.set_index([edge_first_node_attr, edge_second_node_attr],
drop=True,
inplace=False)
if gdf._geometry_column_name != self.edges_geometry_key:
gdf = gdf.rename(
columns={gdf._geometry_column_name: self.edges_geometry_key},
inplace=False)
gdf.set_geometry(self.edges_geometry_key, inplace=True)
self.add_edges_from((*r[0], r[1]) for r in gdf.iterrows())
def disagg(vec: gpd.GeoDataFrame):
"""Dissagregate collections and multi geomtries"""
# Split GeometryCollections
no_coll = []
for i, row in vec.iterrows():
geom = row.geometry
if geom.type == 'GeometryCollection':
for part in geom:
row2 = row.copy()
row2.geometry = part
no_coll.append(row2)
else:
no_coll.append(row)
# Split Multi geomries
res = []
for row in no_coll:
geom = row.geometry
if geom.type.startswith('Multi'):
for part in geom:
row2 = row.copy()
row2.geometry = part
res.append(row2)
else:
res.append(row)
return gpd.GeoDataFrame(res, crs=vec.crs).reset_index(drop=True)
def road_network_from_data_frame(road_data: GeoDataFrame) -> RoadNetwork:
assert supported_crs(road_data), (
"Supported CRS ['epsg:26910', 'epsg:32649']" "got %s",
(road_data.crs,),
)
road_table = RoadTable()
assert geom_check(
road_data, "LineString"
), "Expected all geometries in to be LineString"
for idx, feature in road_data.iterrows():
feature_geometry, feature_property = decompose_data_frame_row(feature)
assert "u" in feature_property and "v" in feature_property, (
"Expected 'u' and 'v' to be present in the property"
"indicating the start node and the end node of the provided"
"geometry"
)
if "fid" in feature_property:
idx = feature_property["fid"]
road_table.add(
idx, feature_property, feature_geometry, shape(feature_geometry).length
)
return RoadNetwork(
road_table=road_table,
maximum_distance=compute_diagonal_distance_of_extent(road_data),
)
def test_cluster_vertical_points(self):
pts = [
Point(0, 2),
Point(6, 2),
Point(6, 2),
Point(0.2, 2),
Point(6.2, 2),
Point(6.2, 2),
]
expected = GeoDataFrame(
pd.DataFrame([
{
"geometry": Point(0.1, 2),
"n": 2
},
{
"geometry": Point(6.1, 2),
"n": 4
},
]))
actual = PointClusterer(pts, max_distance=5,
is_latlon=False).get_clusters()
actual = [(c.centroid, len(c.points)) for c in actual]
expected = [(c.geometry, c.n) for key, c in expected.iterrows()]
assert len(actual) == len(expected)
print([str(c) for p in actual for c in p])
for pt in expected:
assert pt in actual
def fetch_ecodivision_name(latitude: str, longitude: str,
ecodivisions: geopandas.GeoDataFrame):
""" Returns the ecodivision name for a given lat/long coordinate """
station_coord = Point(float(longitude), float(latitude))
for _, ecodivision_row in ecodivisions.iterrows():
geom = ecodivision_row['geometry']
if station_coord.within(geom):
return ecodivision_row['CDVSNNM']
return None
def get_model_annual_eq_rate(bin_gdf: gpd.GeoDataFrame) -> float:
annual_rup_rate = 0.0
for i, row in bin_gdf.iterrows():
sb = row.SpacemagBin
min_bin_center = np.min(sb.mag_bin_centers)
bin_mfd = sb.get_rupture_mfd(cumulative=True)
annual_rup_rate += bin_mfd[min_bin_center]
return annual_rup_rate
def fetch_ecodivision_name(lat: str, long: str,
ecodivisions: geopandas.GeoDataFrame):
""" Returns the ecodivision name for a given lat/long coordinate """
station_coord = Point(float(long), float(lat))
for index, row in ecodivisions.iterrows(): # pylint: disable=redefined-outer-name, unused-variable
geom = row['geometry']
if station_coord.within(geom):
return row['CDVSNNM']
return None
def _check_has_overlaps_old(polygons: gpd.GeoDataFrame):
for i, (inda, pola) in enumerate(polygons.iterrows()):
for (indb, polb) in polygons.iloc[i + 1:].iterrows():
if pola.geometry.intersects(polb.geometry):
warnings.warn(
f"List of shapes contains overlap between {inda} and {indb}. Points will be assigned to {inda}.",
UserWarning,
stacklevel=5,
)
def assign_images_to_tiles(
df_tiles: gpd.GeoDataFrame, gdf: gpd.GeoDataFrame
) -> gpd.GeoDataFrame:
"""
For each tile, specify which images it can be obtained from.
Adds a "pre-event" and "post-event" column to `df_tiles`.
Arguments:
df_tiles: The output of `generate_tiles`.
gdf: The output of `get_extents`.
Returns the `df_tiles` input argument.
"""
df_tiles['pre-event'] = [[] for x in range(len(df_tiles))]
df_tiles['post-event'] = [[] for x in range(len(df_tiles))]
for ixt, rowt in df_tiles.iterrows():
pre_event_images, post_event_images = [], []
for ix, row in gdf.iterrows():
bounds_image = rasterio.coords.BoundingBox(*row['geometry'].bounds)
bounds_tile = rasterio.coords.BoundingBox(*rowt['geometry'].bounds)
if not rasterio.coords.disjoint_bounds(bounds_image, bounds_tile):
if row['pre-post'] == 'pre-event':
pre_event_images.append(row['file'])
else:
post_event_images.append(row['file'])
if len(pre_event_images) > 0:
df_tiles.at[ixt, 'pre-event'] = pre_event_images
else:
df_tiles.at[ixt, 'pre-event'] = np.nan
if len(post_event_images) > 0:
df_tiles.at[ixt, 'post-event'] = post_event_images
else:
df_tiles.at[ixt, 'post-event'] = np.nan
# drop tiles that do not contain both pre- and post-event images
df_tiles = df_tiles[(~pd.isna(df_tiles['pre-event'])) & (~pd.isna(df_tiles['post-event']))]
# The pre-event and post-event columns contain lists of filenames, but geopandas
# doesn't allow lists as GeoJSON properties to maintain compatibility with
# shapefiles. We can work around this by converting them to dictionaries.
df_tiles.loc[:, "pre-event"] = df_tiles["pre-event"].map(lambda l: dict(enumerate(l)))
df_tiles.loc[:, "post-event"] = df_tiles["post-event"].map(lambda l: dict(enumerate(l)))
return df_tiles
def _calculate_row_uncertainties(self, grid: GeoDataFrame,
period: DateRange) -> list[float]:
"""Combine daily grid uncertainties to value for the full period for all cells."""
# TODO It would be great to make this faster using clever iteration/numpy/CPython
uncertainties = Series([TEMIS_CELL_UNCERTAINTY] * len(period))
return [
self._combine_uncertainties(row[-(len(period) + 3):-3],
uncertainties)
for _, row in grid.iterrows()
]
def subset_cophub_from_esa(
esa_df: gpd.GeoDataFrame,
cop_df: gpd.GeoDataFrame,
outdir: Optional[Union[Path, str]] = Path(os.getcwd()),
save_file: Optional[bool] = True) -> gpd.GeoDataFrame:
"""Subsets Cophub list based on the ESA list if their Geometry(Footprint) intersects
for a given day.
:param esa_df: The GeoDataFrame with attributes for ESA S3 listing.
:param cop_df: The GeoDataFrame with attributes for Cophub(NCI) file listing.
:param save_file: Flag to save the subset download list of not.
:return:
GeoDataFrame with Cophub FRP product footprint that intersects with ESA FRP
footprint for a given day.
"""
# convert datetime string to datetime stamp
esa_df['start_date'] = pd.to_datetime(esa_df["start_date"],
format="%Y%m%dT%H%M%S")
cop_df['start_date'] = pd.to_datetime(cop_df["start_date"],
format="%Y%m%dT%H%M%S")
# assign crs to GeoDataFrame
esa_df.crs = 'EPSG:4326'
cop_df.crs = 'EPSG:4326'
esa_df['date'] = pd.to_datetime(esa_df['start_date']).dt.date
cop_df['date'] = pd.to_datetime(cop_df['start_date']).dt.date
column_names = [
"title", "start_date", "sensor", 'relative_orbit', 'geometry'
]
cop_download_df = gpd.GeoDataFrame(columns=column_names)
for idx_esa, esa_row in esa_df.iterrows():
cophub_df_subset = cop_df[cop_df['date'][:] == esa_row['date']].copy()
cophub_df_subset['intersects'] = False
esa_geom = esa_row['geometry'].buffer(0)
for idx_cop, cop_row in cophub_df_subset.iterrows():
cop_geom = cop_row['geometry'].buffer(0)
if esa_geom.intersects(cop_geom):
cop_download_df = cop_download_df.append(
{
'title': cop_row["title"],
'start_date': cop_row['start_date'],
'sensor': cop_row['sensor'],
'relative_orbit': cop_row['relative_orbit'],
'geometry': cop_row['geometry'],
'esa_geometry': esa_row['geometry']
},
ignore_index=True)
subset_cop_download_df = cop_download_df.drop_duplicates(['title'])
if save_file:
subset_cop_download_df.to_csv(
outdir.joinpath('cophub_download_list.csv'))
return subset_cop_download_df
def yield_top_point_slices(
data: geopandas.GeoDataFrame
) -> Generator[geopandas.GeoDataFrame, None, None]:
"""Yields slices of the GeoDataFrame containing the top FRP point and the corresponding nearby points"""
start = 0
for index, row in data.iterrows():
if index == 0: # avoid yielding first row by itself
continue
if row.is_top_point is True:
stop = index
yield data.iloc[start:stop, :]
start = stop
def __init__(self,
fault_geodataframe: gpd.GeoDataFrame,
exclude_regions: List[Polygon] = None):
for field in required_fields:
if field not in fault_geodataframe.columns:
raise ValueError("Missing required field: {}".format(field))
for field in expected_fields:
if field not in fault_geodataframe.columns:
print("Warning: missing expected field: {}".format(field))
self._faults = []
# If appropriate, clip out data that fall within exclude_regions
if exclude_regions is not None:
assert isinstance(exclude_regions, list)
assert all([isinstance(a, Polygon) for a in exclude_regions])
exclude_regions_nztm = []
# Check that polygons are in NZTM, otherwise convert them
for poly in exclude_regions:
x, y = poly.exterior.xy
if all(np.array(y) < 0):
# Assume in WGS (Lon Lat), convert to NZTM
new_x, new_y = transform_wgs2nztm.transform(
np.array(x), np.array(y))
new_poly = Polygon([(xi, yi)
for xi, yi in zip(new_x, new_y)])
exclude_regions_nztm.append(new_poly)
else:
# Assume NZTM, do nothing
exclude_regions_nztm.append(poly)
# Make list of faults outside region
trimmed_fault_ls = []
for i, row in fault_geodataframe.iterrows():
if not any([
row.geometry.within(poly)
for poly in exclude_regions_nztm
]):
trimmed_fault_ls.append(row)
trimmed_fault_gdf = gpd.GeoDataFrame(trimmed_fault_ls)
else:
trimmed_fault_gdf = fault_geodataframe
# Sort alphabetically by name
sorted_df = trimmed_fault_gdf.sort_values("FZ_Name")
# Reset index to line up with alphabetical sorting
sorted_df = sorted_df.reset_index(drop=True)
for i, row in sorted_df.iterrows():
self.add_fault(row)
self.df = sorted_df
def build(self, geo_df: GeoDataFrame):
gdp_list = []
df_columns = list(geo_df.columns)
for i, row in geo_df.iterrows():
polygons: List[Polygon] = row["geometry"].geoms
# TODO: Check for non-polygon entries
for poly in polygons:
meta = {
column: row[column]
for column in list(
filter(lambda x: x != "geometry", df_columns))
}
gdp_list.append(GeoDataPoint(meta, poly))
self.strtree.build(gdp_list)
def bboxes_per_year(
df: geopandas.GeoDataFrame, box_date: date,
surrounding_metres: int) -> List[Tuple[EEBoundingBox, date]]:
boxes: List[Tuple[EEBoundingBox, date]] = []
for _, row in df.iterrows():
boxes.append((
bounding_box_from_centre(
mid_lat=row["lat"],
mid_lon=row["lon"],
surrounding_metres=surrounding_metres,
),
box_date,
))
return boxes
def create_get_urls(county_gdf: gpd.GeoDataFrame) -> List[str]:
"""
Create a list of urls to issue get requests
Args:
county_gdf: GeoDataFrame representing Indiana counties
Returns:
List of formatted urls to issue GET requests
"""
urls = []
for index, row in county_gdf.iterrows():
urls.append(config.ARDA_ESRI_REST_ENDPOINT.format(xmin=row.minx, ymin=row.miny,
xmax=row.maxx, ymax=row.maxy))
return urls
def feature_collection(self, ft_name, **kwargs):
""" Call bdq wfs feature_collection and format the result to a pandas DataFrame
"""
global cache
cv_list = None
if 'ts' in kwargs:
cv_list = kwargs['ts']
del kwargs['ts']
fc = self.wfs.feature_collection(ft_name, **kwargs)
geo_data = pd.DataFrame(fc['features'])
geo_data = GeoDataFrame(geo_data, geometry='geometry', crs=fc['crs']['properties']['name'])
metadata = {'total': fc['total'], 'total_features': fc['total_features']}
# retrieve coverage attributes
if cv_list is not None:
# TODO: paralelizar essas chamadas
for cv in cv_list:
if type(cv['attributes']) is str:
cv['attributes'] = tuple([cv['attributes']])
# create columns as object to support lists
for c in cv['attributes']:
name = '{}.{}'.format(cv['coverage'], c)
geo_data[name] = ''
geo_data[name] = geo_data[name].astype(object)
s_date = None
if 'start_date' in cv:
s_date = cv['start_date'];
e_date = None
if 'end_date' in cv:
e_date = cv['end_date'];
for idx, row in geo_data.iterrows():
ts, ts_metadata = self.time_series(cv['coverage'],
cv['attributes'],
row['geometry'].y,
row['geometry'].x,
s_date,
e_date)
for c in cv['attributes']:
name = '{}.{}'.format(cv['coverage'], c)
if s_date == e_date and s_date is not None:
if ts_metadata['total'] == 1:
geo_data.set_value(idx, name, ts[c].values.tolist()[0])
else:
geo_data.set_value(idx, name, None)
else:
geo_data.set_value(idx, name, ts[c].values.tolist())
return geo_data, metadata
def prepare_neighbor_net(gdf: GeoDataFrame, nbr: dict):
for index, row in gdf.iterrows():
name = row[1]
index = name
region = gdf[gdf.name == index]
print(index)
nbr[index] = {"coords": baricenter(region)}
disjoints = gdf.geometry.disjoint(row.geometry)
print("disjoint", row, disjoints)
neighbors = gdf[~disjoints].name.tolist()
# neighbors = gdf[disjoints].name.tolist()
if index in neighbors:
neighbors.remove(index)
print(index, neighbors)
nbr[index]["nbr"] = neighbors
nbr[index]["count"] = len(neighbors)
def build(self, geo_df: GeoDataFrame):
self.gt = GeoTrie(self.gh_len)
df_columns = list(geo_df.columns)
for i, row in geo_df.iterrows():
polygons: List[Polygon] = row["geometry"].geoms
# TODO: Check for non-polygon entries
for poly in polygons:
meta = {
column: row[column]
for column in list(
filter(lambda x: x != "geometry", df_columns))
}
gdp = GeoDataPoint(meta, poly)
geos = self.__gh_intersecting(poly)
for gh in geos:
self.gt.insert(gh, gdp)
def get_total_obs_eqs(bin_gdf: gpd.GeoDataFrame,
prospective: bool = False) -> list:
"""
Returns a list of all of the observed earthquakes within the model domain.
"""
obs_eqs = []
for i, row in bin_gdf.iterrows():
sb = row.SpacemagBin
if prospective is False:
for mb in sb.observed_earthquakes.values():
obs_eqs.extend(mb)
else:
for mb in sb.prospective_earthquakes.values():
obs_eqs.extend(mb)
return obs_eqs
def get_source_bins(bin_gdf: gpd.GeoDataFrame) -> gpd.GeoDataFrame:
"""
Returns a subset of all of the spatial bins, where each bin actually
contains earthquake sources.
:param bin_gdf:
GeoDataFrame of bins
:returns:
GeoDataFrame of bins with sources
"""
source_list = []
for i, row in bin_gdf.iterrows():
cum_mfd = row.SpacemagBin.get_rupture_mfd(cumulative=True)
if sum(cum_mfd.values()) > 0:
source_list.append(i)
source_bin_gdf = bin_gdf.loc[source_list]
return source_bin_gdf
def split_plots(
geodataframe: gpd.GeoDataFrame,
target_col: str,
separator: str = ',',
) -> gpd.GeoDataFrame:
"""
Splits plot rows in parts by based on separator
:param geodataframe:
:param target_col: column name
:param separator: separator, default ','
:return: GeoDataFrame with split rows
"""
new_geodataframe = gpd.GeoDataFrame(columns=geodataframe.columns)
for _, row in geodataframe.iterrows():
for c in str(row[target_col]).split(separator):
new_row = row
new_row[target_col] = c
new_geodataframe = new_geodataframe.append(new_row)
return new_geodataframe.reindex()
def points2grids(gdf: geopandas.GeoDataFrame, shift: int = 1 / 32):
'''
Turns Livneh gridcell point geometry into a Livneh grid polygon.
Inputs:
(1) gdf: geopandas.GeoDataFrame contining the Livneh grid cell centroids.
(2) shift: a radius shift from gridcell centroid to the grid polygon edges. 1/32nd degree by default.
Output: a geopandas.GeoDataFrame containing the Livneh grids polygon geometry.
'''
geometry: list = []
for i, row in gdf.iterrows():
w: float = float(row['lon']) - shift
e: float = float(row['lon']) + shift
n: float = float(row['lat']) + shift
s: float = float(row['lat']) - shift
lons, lats = [w, e, e, w], [n, n, s, s]
geometry.append(Polygon(zip(lons, lats)))
gdf['geometry'] = geometry
return gdf