import pandas as pd
import matplotlib.pyplot as plt
import cartopy.crs as ccrs
### Setting path
data_base_dir = os.path.join('/data2', 'sehyun', 'Data')
path_grid_raw = os.path.join(data_base_dir, 'Raw', 'grid')
path_ea_goci = os.path.join(data_base_dir, 'Preprocessed_raw', 'EA_GOCI6km')
path_station = os.path.join(data_base_dir, 'Preprocessed_raw', 'Station')
path_stn_jp = os.path.join(path_station, 'Station_JP')
path_stn_cn = os.path.join(path_station, 'Station_CN')
path_stn_kr = os.path.join(path_station, 'Station_KR')
path_output = os.path.join(data_base_dir, 'output', 'RealTimeTraining', 'EastAsia')
matlab.check_make_dir(path_output)
tg = ['PM10','PM25']
## Load grid
mat = matlab.loadmat(os.path.join(path_grid_raw, 'grid_goci.mat'))
lon_goci, lat_goci = mat['lon_goci'], mat['lat_goci']
del mat
##
YEARS = [2016]
for yr in YEARS:
#mat = matlab.loadmat(os.path.join(path_stn_kr, 'stn_GOCI6km_location_weight_v2018.mat'))
mat = matlab.loadmat(os.path.join(path_stn_kr, 'stn_GOCI6km_location_weight_v201904.mat'))
dup_scode2_GOCI6km = mat['dup_scode2_GOCI6km']
df = pd.DataFrame(mat['stn_GOCI6km_location'], columns=mat['header_stn_GOCI6km_location'])
mask=maskarr,
transform=masksrc.transform):
shapes.append(json.loads(json.dumps(geometry)))
YEARS = [2016]
for yr in YEARS:
print(yr)
flist = glob.glob(
os.path.join(path_myd_processed, '02prj_GCS_WGS84', str(yr), "*.tif"))
flist.sort()
for src_dataset in flist:
tStart = time.time()
dst_dataset = os.path.join(path_myd_processed,
'03mask_SouthKorea_MYD13A2', str(yr),
f'm_{os.path.basename(src_dataset)[2:]}')
matlab.check_make_dir(os.path.dirname(dst_dataset))
with rio.open(src_dataset) as src:
kwargs = src.meta.copy()
kwargs['transform'] = masksrc.transform
temp_dataset = os.path.join(os.path.dirname(dst_dataset),
'temp.tif')
resolution = 1.02308446206551E-02
dst_crs = masksrc.crs
with rio.open(temp_dataset, 'w+', **kwargs) as temp_dst:
for i in range(1, src.count + 1):
reproject(
source=rio.band(src, i),
destination=rio.band(temp_dst, i),
src_transform=src.transform,
src_crs=src.crs,
if t==3:
name = os.path.basenanme(cal_list[c])[36:70]
elif t==4:
name = os.path.basenanme(cal_list[c])[32:66]
else:
name = os.path.basenanme(cal_list[c])[31:65]
# Validation reslut
pred_val = rf_model.predict(val.drop([target[i]]), val[target[i]])
pred_val = np.exp(pre_val)
# save vali pred
fname = f'rf_{name}_val_ranger.csv'
pred_val = pd.DataFrame(pred_val)
matlab.check_make_dir(os.path.join(paht_loo, type_list[t],"/RF/",target[i]))
pred_val.to_csv(os.path.join(paht_loo, type_list[t],"/RF/",target[i], fname), sep=",")
print('Predicted val result is saved')
except:
pass
# LOO list
print (doy)
except:
pass
# doy
print (yr)
# yr
print (target[i])
# target
print (type_list[t])
# type
flist_temp = flist[k:k+14]
doy = os.path.basename(flist_temp[0])[13:16]
input_files = []
for m in range(0,14):
dst_dataset = os.path.join(tmpdirname.name, f"NDVI_{doy}_{m+1}.tif")
src_dataset = flist_temp[m]
print (os.path.join(path_myd_raw, src_dataset))
gdal_dataset = gdal.Open(os.path.join(path_myd_raw, src_dataset))
src_dataset = gdal_dataset.GetSubDatasets()[0][0]
cmd = ["gdal_translate", src_dataset, dst_dataset]
subprocess.call(cmd)
input_files.append(dst_dataset)
tStart = time.time()
dst_fname = os.path.join(path_myd_processed, '01mosaic', str(yr), f"MYD13A2_{yr}_{doy}.tif")
matlab.check_make_dir(os.path.dirname(dst_fname)) # debugging
pixel_type = 'Int16'
in_nodata_val = "-3000"
out_nodata_val = "-9999"
compression = "COMPRESS=LZW"
cmd = ["gdal_merge.py", "-n", in_nodata_val, "-a_nodata", out_nodata_val, "-ot", pixel_type]
cmd += ["-co", compression]
cmd += ["-o", dst_fname]
cmd += input_files
subprocess.call(cmd)
tmpdirname.cleanup()
tElapsed = time.time() - tStart
print (f'time taken : {tElapsed}')
print (os.path.basename(dst_fname))
print(doy)
band = masksrc.read(1)
maskarr = (band!=255)
shapes = []
for geometry, raster_value in features.shapes(band, mask=maskarr, transform=masksrc.transform):
shapes.append(json.loads(json.dumps(geometry)))
flist = glob.glob(os.path.join(path_mcd_processed, '01mosaic', "*.tif"))
flist.sort()
for src_dataset in flist:
tStart = time.time()
last_num = os.path.basename(src_dataset)[-8:] # b 2016.tif
print (src_dataset)
matlab.check_make_dir(os.path.join(path_mcd_processed, '02prj_GCS_WGS84')) # debugging
matlab.check_make_dir(os.path.join(path_mcd_processed, '03masked_N50W110S20E150')) # debugging
dst_dataset02 = os.path.join(path_mcd_processed, '02prj_GCS_WGS84', f'GCS_EA_MCD12Q1_{last_num}') # c
dst_dataset03 = os.path.join(path_mcd_processed, '03masked_N50W110S20E150', f'm_MODIS_LC_500m_{last_num}') # d
dst_crs = 'EPSG:4326'
resolution = 5.11542231032757E-03 # same with maskfile resolution
with rio.open(src_dataset) as src:
transform, width, height = calculate_default_transform(
src.crs, dst_crs,
src.width, src.height, *src.bounds,
resolution=resolution)
kwargs = src.meta.copy()
kwargs.update({
sys.path.insert(0, project_path)
from Code.utils import matlab
import numpy as np
import glob
import time
import pandas as pd
### Setting path
data_base_dir = os.path.join('/data2', 'sehyun', 'Data')
path_grid_raw = os.path.join(data_base_dir, 'Raw', 'grid')
path_ea_goci = os.path.join(data_base_dir, 'Preprocessed_raw', 'EA_GOCI6km')
path_korea_cases = os.path.join(data_base_dir, 'Preprocessed_raw', 'Korea',
'cases')
matlab.check_make_dir(path_korea_cases)
path_station = os.path.join(data_base_dir, 'Preprocessed_raw', 'Station')
path_stn_jp = os.path.join(path_station, 'Station_JP')
path_stn_cn = os.path.join(path_station, 'Station_CN')
path_stn_kr = os.path.join(path_station, 'Station_KR')
tg = ['PM10', 'PM25']
## Load grid
mat = matlab.loadmat(os.path.join(path_grid_raw, 'grid_korea.mat'))
lat_kor, lon_kor = mat['lat_kor'], mat['lon_kor']
del mat
mat = matlab.loadmat(
os.path.join(path_stn_kr,
doy = matlab.datenum(temp[19:23] + temp[24:28]) - doy_000
print(f'Reading OMNO2d {yr}_{doy:03d}')
data = matlab.h5read(
read_fname,
'/HDFEOS/GRIDS/ColumnAmountNO2/Data Fields/ColumnAmountNO2TropCloudScreened'
)
data = np.float64(data.T)
# 720X1440
data[
data <=
-1.2676506e+30] = np.nan # Assign NaN value to pixel that is out of valid range
data = data * 3.7216e-17
data_yr[:, doy - 1] = data.ravel(order='F')
out_fname = os.path.join(path_write, f'OMNO2d_trop_CS_{yr}_DU.mat')
matlab.check_make_dir(os.path.dirname(out_fname))
data_yr[np.isnan(data_yr)] = -9999
matlab.savemat(out_fname, {'data_yr': data_yr})
tElapsed = time.time() - tStart
print(f'{tElapsed} second')
del data, data_yr
print('==========================================================')
### OMSO2e
print('OMSO2e')
for yr in YEARS:
tStart = time.time()
doy_000 = matlab.datenum(f'{yr}0000')
file_list = glob.glob(
os.path.join(path_read, 'L3_grid', 'OMSO2e', str(yr), '*.he5'))
file_list.sort()
### Setting path
data_base_dir = os.path.join('/data2', 'sehyun', 'Data')
path_myd_processed = os.path.join(data_base_dir, 'Preprocessed_raw', 'MODIS',
'MYD13A2')
YEARS = [2016]
for yr in YEARS:
path_read = os.path.join(path_myd_processed, '01mosaic', str(yr))
flist = glob.glob(os.path.join(path_read, "*.tif"))
flist.sort()
for src_dataset in flist:
tStart = time.time()
matlab.check_make_dir(
os.path.join(path_myd_processed, '02prj_GCS_WGS84',
str(yr))) # debugging
dst_dataset02 = os.path.join(path_myd_processed, '02prj_GCS_WGS84',
str(yr),
f'p_{os.path.basename(src_dataset)}') # c
dst_crs = 'EPSG:4326'
resolution = 1.02308446206551E-02 # same with maskfile resolution
with rio.open(src_dataset) as src:
transform, width, height = calculate_default_transform(
src.crs,
dst_crs,
src.width,
src.height,
*src.bounds,
resolution=resolution)
pred = pred.loc[:, features + [target[i]]]
pred.fillna(-9999, inplace=True)
name = f"{target[i]}_RTT_EA6km_{yr}_{doy:03d}_{utc:02d}"
# Prediction result
pred_cases = rf_model.predict(
pred[features].values
) # predict(rf_model, data = pred)
pred_cases = np.exp(pred_cases)
# save pred prediction
fname = f"rf_{name}.csv"
pred_cases = pd.DataFrame(pred_cases)
matlab.check_make_dir(
os.path.join(path_rtt, type_list[t], "RF_pred",
target[i]))
pred_cases.to_csv(os.path.join(path_rtt, type_list[t],
"RF_pred", target[i],
fname),
sep=",")
# print('Predicted prediction result is saved')
features = [
'AOD', 'AE', 'FMF', 'SSA', 'NDVI', 'RSDN',
'Precip', 'DEM', 'LCurban', 'Temp', 'Dew', 'RH',
'P_srf', 'MaxWS', 'PBLH', 'Visibility',
'stack1_maxWS', 'stack3_maxWS', 'stack5_maxWS',
'stack7_maxWS', 'DOY', 'PopDens', 'RoadDens'
]
# Validation result
data_base_dir = os.path.join('/data2', 'sehyun', 'Data')
path_mcd_processed = os.path.join(data_base_dir, 'Preprocessed_raw', 'MODIS',
'MCD12Q1')
class_name = [
"forest", "shrub", "savannas", "grass", "wetland", "crop", "urban", "snow",
"barren", "water"
]
YEARS = [2016]
for yr in YEARS:
tStart = time.time()
src_dataset = os.path.join(path_mcd_processed, '03masked_N50W110S20E150',
f'm_MODIS_LC_500m_{yr}.tif')
dst_dataset = os.path.join(path_mcd_processed, '01_reclassified',
f'reclass_MODIS_LC_500m_EA_{yr}.tif')
matlab.check_make_dir(os.path.dirname(dst_dataset)) # Debugging
matlab.check_make_dir(
os.path.join(path_mcd_processed, '02_LC_binary', str(yr))) # Debugging
with rio.open(src_dataset) as src:
band = src.read(1).copy()
band[(band >= 1) & (band < 6)] = 1
band[(band >= 6) & (band < 8)] = 2
band[(band >= 8) & (band < 10)] = 3
band[band == 10] = 4
band[band == 11] = 5
band[(band == 12) | (band == 14)] = 6
band[band == 13] = 7
band[band == 15] = 8
band[band == 16] = 9
band[(band == 17) | (band == src.meta['nodata'])] = 10
import scipy.io as sio
import numpy as np
import glob
import time
import pandas as pd
### Setting path
data_base_dir = os.path.join('/data2', 'sehyun', 'Data')
path_grid_raw = os.path.join(data_base_dir, 'Raw', 'grid')
path_ea_goci = os.path.join(data_base_dir, 'Preprocessed_raw', 'EA_GOCI6km')
path_rtt = os.path.join(data_base_dir, 'Preprocessed_raw', 'RTT') # path_save
path_nox_korea = os.path.join(data_base_dir, 'Preprocessed_raw', 'NOX03', 'Korea')
matlab.check_make_dir(path_nox_korea)
path_korea_cases = os.path.join(data_base_dir, 'Preprocessed_raw', 'Korea', 'cases')
path_station = os.path.join(data_base_dir, 'Preprocessed_raw', 'Station')
path_stn_jp = os.path.join(path_station, 'Station_JP')
path_stn_cn = os.path.join(path_station, 'Station_CN')
path_stn_kr = os.path.join(path_station, 'Station_KR')
path_data = '/share/irisnas5/Data/'
path= '/share/irisnas5/GEMS/PM/00_EA6km/'
target = ['PM10','PM25']
type_list = ['conc','time','time_conc']
YEARS = [2016]
## Load grid
print(accuracy[0, 1])
pred_pred = rf_model.predict(plots_pred[features])
pred_pred = np.exp(pred_pred)
if (accuracy[0, 3] < max_accuracy):
max_accuracy = accuracy[0, 3]
ss = s
print(rf_model.feature_importances_)
name = f"{target[i]}_dataset"
# save variable importance
feature_imp = pd.Series(
rf_model.feature_importances_).sort_values(ascending=False)
fname = f"rf_{name}_imp_ranger.csv"
matlab.check_make_dir(os.path.join(path_rtt, "RF/", target[i]))
feature_imp.to_csv(os.path.join(path_rtt, "RF/", target[i], fname),
sep=",")
fname = f"rf_{name}_model_ranger.pickle"
with open(os.path.join(path_rtt, "RF", target[i], fname),
'wb') as f:
pickle.dump(rf_model, f)
print(f'RF model is saved with {ss} of num.random.splits')
parameter = rf_model.get_params(
) #f'RF model is saved with {ss} of num.random.splits'
pd.DataFrame(parameter).to_csv(os.path.join(
path_rtt, "RF/", target[i], f"rf_{name}_parameter_ranger.csv"),
sep=",")
flist_temp = flist[k:k+14]
yr = os.path.basename(flist_temp[0])[9:13]
input_files = []
for m in range(0,14):
fname = flist_temp[m]
dst_dataset = os.path.join(tmpdirname.name, f"LC_{yr}_{m+1}.tif")
gdal_dataset = gdal.Open(os.path.join(path_mcd_raw, fname))
src_dataset = gdal_dataset.GetSubDatasets()[0][0]
subprocess.call(["gdal_translate", src_dataset, dst_dataset])
input_files.append(dst_dataset)
# Mosaic
tStart = time.time()
matlab.check_make_dir(os.path.join(path_mcd_processed, '01mosaic')) # debugging
dst_fname = os.path.join(path_mcd_processed, '01mosaic', f"EA_MCD12Q1_mosaic_{yr}.tif")
pixel_type = 'Int16'
in_nodata_val = "255"
out_nodata_val = "-9999"
compression = "COMPRESS=LZW"
cmd = ["gdal_merge.py", "-n", in_nodata_val, "-a_nodata", out_nodata_val, "-ot", pixel_type]
cmd += ["-co", compression]
cmd += ["-o", dst_fname]
cmd += input_files
subprocess.call(cmd)
tmpdirname.cleanup()
tElapsed = time.time() - tStart
print (f'time taken : {tElapsed}')
data[:,32] = Wcos.ravel(order='F')
data[:,33] = Wsin.ravel(order='F')
data[:,34] = AP3h.ravel(order='F')
# ancillary data
data[:,35] = np.sin((doy-112)*2*np.pi/365.25) # DOY
data[:,36] = popDens.ravel(order='F') # Population Density
data[:,37] = roadDens.ravel(order='F') # Road Density
data[:,38:52]=EA_emis
data[:,52:62]=LC_ratio
# additional variables
data[:,62]=omno2d_trop.ravel(order='F')
data[np.isnan(data)] = -9999
matlab.check_make_dir(os.path.join(path_ea_goci, 'cases_csv', str(yr)))
if yr>=2018: # BESS 없음
data[:,9]= 0
header_temp = header[:9]+header[10:]
else:
header_temp = header
print ('data shape :', data.shape)
tmp_df = pd.DataFrame(data, columns=header_temp)
tmp_df.to_csv(os.path.join(path_ea_goci, 'cases_csv',str(yr),f'cases_EA6km_{yr}_{doy:03d}_{utc:02d}.csv'))
del tmp_df
data[data==-9999] = np.nan
data_tbl = pd.DataFrame(data,columns=header_temp)
print (data_tbl.to_dict('list').keys())
header_temp = np.array(data_tbl.columns, dtype=h5py.string_dtype(encoding='utf-8'))
matlab.savemat(os.path.join(path_ea_goci, 'cases_mat', str(yr), f'cases_EA6km_{yr}_{doy:03d}_{utc:02d}.mat'),
