def main(args):
weather = ERA5Land(args.ERA5)
nasa_firms = NASAFIRMS(args.nasa_firms, args.nasa_firms_type)
print(weather.shape)
print(nasa_firms.shape)
# Time span selection
date_range = get_intersection_range(weather.time, nasa_firms.acq_date)
weather = weather[weather.time.isin(date_range)]
nasa_firms = nasa_firms[nasa_firms.acq_date.isin(date_range)]
print(weather.shape)
print(nasa_firms.shape)
# Merge
if args.type_of_merged == "departements":
# drop redundant columns with weather datasets
nasa_firms = nasa_firms.drop(["nom"], axis=1)
merged_data = merge_datasets_by_departements(weather, "time", "code",
nasa_firms, "acq_date",
"code", "left")
to_drop = [
"acq_date",
"latitude_y",
"longitude_y",
"bright_t31",
"frp",
"acq_date_time",
"confidence",
]
else:
# drop redundant columns with weather datasets
nasa_firms = nasa_firms.drop(["code", "nom"], axis=1)
merged_data = merge_datasets_by_closest_weather_point(
weather, "time", nasa_firms, "acq_date")
to_drop = [
"closest_weather_point",
"acq_date",
"latitude_y",
"longitude_y",
"bright_t31",
"frp",
"acq_date_time",
"confidence",
"weather_lat",
"weather_lon",
]
final_data = merged_data.copy()
where = merged_data["confidence"] >= 60
final_data.loc[where, "Statut"] = 1
final_data.loc[~where, "Statut"] = 0
final_data["Statut"] = final_data["Statut"].astype(int)
# drop unnecessary columns
final_data = final_data.drop(to_drop, axis=1)
print(final_data)
def main(args):
weather = NOAAWeather(args.weather)
history = BDIFFHistory(args.wildfire)
# Time span selection
date_range = get_intersection_range(weather.DATE, history.date)
weather = weather[weather.DATE.isin(date_range)]
history = history[history.date.isin(date_range)]
# Merge
df = merge_datasets_by_departements(
weather, "DATE", "code", history, "date", "Département", "left"
)
# Label data
df.Statut = 1 - df.Statut.isna().astype(int)
df = df.filter(
items=[
"DATE",
"code",
"nom",
"LATITUDE",
"LONGITUDE",
"ELEVATION",
"DEWP",
"DEWP_ATTRIBUTES",
"FRSHTT",
"GUST",
"MAX",
"MIN",
"MXSPD",
"PRCP",
"SLP",
"SLP_ATTRIBUTES",
"SNDP",
"STP",
"STP_ATTRIBUTES",
"TEMP",
"TEMP_ATTRIBUTES",
"VISIB",
"VISIB_ATTRIBUTES",
"WDSP",
"WDSP_ATTRIBUTES",
"Statut",
]
)
print(df)
def __init__(
self, era_source_path=None, viirs_source_path=None, fwi_source_path=None
):
"""Define the merged era-fwi-viirs dataframe.
Args:
era_source_path (str, optional): Era5 data source path. Defaults to None.
viirs_source_path (str, optional): Viirs data source path. Defaults to None.
fwi_source_path (str, optional): Fwi data source path. Defaults to None.
"""
weather = ERA5T(era_source_path) # ERA5Land(era_source_path)
nasa_firms = NASAFIRMS_VIIRS(viirs_source_path)
# Time span selection
date_range = get_intersection_range(weather.time, nasa_firms.acq_date)
weather = weather[weather.time.isin(date_range)]
nasa_firms = nasa_firms[nasa_firms.acq_date.isin(date_range)]
# Keep only vegetation wildfires and remove thermal anomalies with low confidence
where = (nasa_firms["confidence"] != "l") & (nasa_firms["type"] == 0)
nasa_firms = nasa_firms[where]
# Get FWI dataset for year 2019 (1st september missing)
if fwi_source_path is None:
days = [
x.strftime("%Y%m%d")
for x in pd.date_range(start="2019-01-01", end="2019-12-31")
]
days.remove("20190901")
fwi_df = GwisFwi(days_list=days)
else:
fwi_df = pd.read_csv(fwi_source_path)
# Load FWI dataset
fwi_df["day"] = pd.to_datetime(fwi_df["day"], format="%Y%m%d", errors="coerce")
# Group fwi dataframe by day and department and compute min, max, mean, std
agg_fwi_df = (
fwi_df.groupby(["day", "departement"])[cfg.FWI_VARS]
.agg(["min", "max", "mean", "std"])
.reset_index()
)
agg_fwi_df.columns = ["day", "departement"] + [
x[0] + "_" + x[1] for x in agg_fwi_df.columns if x[1] != ""
]
# Group weather dataframe by day and department and compute min, max, mean, std
agg_wth_df = (
weather.groupby(["time", "nom"])[cfg.WEATHER_ERA5T_VARS]
.agg(["min", "max", "mean", "std"])
.reset_index()
)
agg_wth_df.columns = ["day", "departement"] + [
x[0] + "_" + x[1] for x in agg_wth_df.columns if x[1] != ""
]
# Merge fwi and weather together
mid_df = pd.merge(
agg_fwi_df, agg_wth_df, on=["day", "departement"], how="inner"
)
# Count fires by day and department
fires_count = (
nasa_firms.groupby(["acq_date", "nom"])["confidence"]
.count()
.to_frame()
.reset_index()
)
fires_count = fires_count.rename({"confidence": "fires"}, axis=1)
# Merge fires
final_df = pd.merge(
mid_df,
fires_count,
left_on=["day", "departement"],
right_on=["acq_date", "nom"],
how="left",
).drop(["acq_date", "nom"], axis=1)
# Fill lines with no fires with 0
final_df["fires"] = final_df["fires"].fillna(0)
super().__init__(final_df)
def main(args):
weather = ERA5Land(args.ERA5)
nasa_firms = NASAFIRMS_VIIRS(args.nasa_firms, args.nasa_firms_type)
print(weather.shape)
print(nasa_firms.shape)
# Time span selection
date_range = get_intersection_range(weather.time, nasa_firms.acq_date)
weather = weather[weather.time.isin(date_range)]
nasa_firms = nasa_firms[nasa_firms.acq_date.isin(date_range)]
print(weather.shape)
print(nasa_firms.shape)
# Keep only vegetation wildfires and remove thermal anomalies with low confidence
where = (nasa_firms["confidence"] != "l") & (nasa_firms["type"] == 0)
nasa_firms = nasa_firms[where]
# Merge
if args.type_of_merged == "departements":
# drop redundant columns with weather datasets
nasa_firms = nasa_firms.drop(["nom"], axis=1)
merged_data = merge_datasets_by_departements(
weather, "time", "code", nasa_firms, "acq_date", "code", "left"
)
to_drop = [
"acq_date",
"latitude_y",
"longitude_y",
"bright_ti4",
"confidence",
"bright_ti5",
"frp",
"type",
"acq_date_time",
]
else:
# drop redundant columns with weather datasets
nasa_firms = nasa_firms.drop(["code", "nom"], axis=1)
merged_data = merge_by_proximity(
nasa_firms, "acq_date", weather, "time", "right"
)
to_drop = [
"latitude_x",
"longitude_x",
"closest_lat",
"closest_lon",
"acq_date",
"bright_ti4",
"confidence",
"bright_ti5",
"frp",
"type",
"acq_date_time",
]
final_data = merged_data.copy()
where = merged_data["confidence"].isna()
final_data.loc[~where, "Statut"] = 1
final_data.loc[where, "Statut"] = 0
final_data["Statut"] = final_data["Statut"].astype(int)
# drop unnecessary columns
final_data = final_data.drop(to_drop, axis=1)
final_data = final_data.rename(
columns={"latitude_y": "latitude", "longitude_y": "longitude"}
)
print(final_data)
def _test_get_intersection_range(self, s1, s2, expected_len):
date_range = utils.get_intersection_range(s1, s2)
self.assertIsInstance(date_range, pd.DatetimeIndex)
self.assertEqual(len(date_range), expected_len)