| 1 | #!/usr/bin/env python |
|---|
| 2 | # -*- coding: utf-8 -*- |
|---|
| 3 | |
|---|
| 4 | # import |
|---|
| 5 | import logging |
|---|
| 6 | |
|---|
| 7 | from pylab import * |
|---|
| 8 | from ReadRawFile import * |
|---|
| 9 | from datetime import date |
|---|
| 10 | import netCDF4 |
|---|
| 11 | from scipy import spatial |
|---|
| 12 | from mpl_toolkits.basemap import pyproj |
|---|
| 13 | import itertools as it |
|---|
| 14 | import time |
|---|
| 15 | |
|---|
| 16 | ##-------------------------------## |
|---|
| 17 | # Class for satellite name # |
|---|
| 18 | ##-------------------------------## |
|---|
| 19 | class SatName: |
|---|
| 20 | """ ### Defined a number for each satellite """ |
|---|
| 21 | def __init__(self): |
|---|
| 22 | self.satname = array(['AMSUBN15','AMSUBN16','MHSN18','MHSN19','MHSM01','MHSM02']) |
|---|
| 23 | self.satnum = arange(size(self.satname))+1 |
|---|
| 24 | |
|---|
| 25 | def get_satnum(self,name): |
|---|
| 26 | return concatenate([self.satnum[nn == self.satname] for nn in name]) |
|---|
| 27 | |
|---|
| 28 | def get_satname(self,num): |
|---|
| 29 | return concatenate([self.satname[nn == self.satnum] for nn in num]) |
|---|
| 30 | |
|---|
| 31 | ##-------------------------------## |
|---|
| 32 | # Correction for one orbit # |
|---|
| 33 | ##-------------------------------## |
|---|
| 34 | def applycorr2orb(fname,lonvec,latvec,ch=['All'],area=[-35.,35.,-180.,180.]): |
|---|
| 35 | """ ### Apply correction to AMSU or MHS orbit |
|---|
| 36 | ## Inputs |
|---|
| 37 | # fname = file name |
|---|
| 38 | # tstep = time resolution for new grid in minutes |
|---|
| 39 | # lonvec = longitude vecteur of new grid |
|---|
| 40 | # latvec = latitude vecteur of new grid |
|---|
| 41 | # ch = list of extrated channel ('B1','B2','B3','B4','B5' or 'All'), default 'All' |
|---|
| 42 | # area = limits [LS,LN,LW,LE] for extracted area, default is world (ie [-35.,35.,-180.,180.]) |
|---|
| 43 | # |
|---|
| 44 | # Outputs |
|---|
| 45 | # tbcorrm = brightness temperature of orbit in new grid |
|---|
| 46 | # trefm = ordinal time |
|---|
| 47 | # |
|---|
| 48 | """ |
|---|
| 49 | |
|---|
| 50 | t=time.time() |
|---|
| 51 | |
|---|
| 52 | logger = logging.getLogger() |
|---|
| 53 | |
|---|
| 54 | # ## Read orbit file |
|---|
| 55 | # print 'Read ', fname |
|---|
| 56 | lon, lat, dt, tb, fov = load_amsub_mhs(fname,ch,area) |
|---|
| 57 | |
|---|
| 58 | try: |
|---|
| 59 | # ## Class of each orbit points |
|---|
| 60 | # print 'Classification' |
|---|
| 61 | clz = find_class(lat,lon) |
|---|
| 62 | nozero = (clz != 0) |
|---|
| 63 | |
|---|
| 64 | except ValueError: |
|---|
| 65 | logger.error('Orbit is out of the area: {0}'.format(fname)) |
|---|
| 66 | return [], [], [] |
|---|
| 67 | |
|---|
| 68 | else: |
|---|
| 69 | # ## Coef of correction as a function of classification, fov and day of year |
|---|
| 70 | # print 'Correction' |
|---|
| 71 | coefz = find_corrcoef(dt[nozero],fov[nozero],clz[nozero],ch=ch) |
|---|
| 72 | tbcorr = array(tb)[:,nozero] - coefz |
|---|
| 73 | # ## Average on new grid |
|---|
| 74 | tbcorrm, tobsm = newgrid(lonvec,latvec,lon[nozero],lat[nozero],dt[nozero],tbcorr) |
|---|
| 75 | |
|---|
| 76 | logger.info('Deal {0} in: {1}'.format(fname, time.time()-t)) |
|---|
| 77 | |
|---|
| 78 | return tbcorrm, tobsm, fname.split('/')[3] |
|---|
| 79 | |
|---|
| 80 | ##----------------------------------## |
|---|
| 81 | # Read AMSU-B and MHS files # |
|---|
| 82 | ##----------------------------------## |
|---|
| 83 | def load_amsub_mhs(fname,ch,area): |
|---|
| 84 | """ ### Read AMSU-B and MHS files |
|---|
| 85 | ## Inputs |
|---|
| 86 | # fname = file name |
|---|
| 87 | # ch = list of extrated channel ('B1','B2','B3','B4','B5' or 'All'), default 'All' |
|---|
| 88 | # area = limits [LS,LN,LW,LE] for extracted area, default is world (ie [-35,35,-180.,180]) |
|---|
| 89 | # |
|---|
| 90 | ## Outputs |
|---|
| 91 | # lon = Longitude |
|---|
| 92 | # lat = Latitude |
|---|
| 93 | # yday = Day of year |
|---|
| 94 | # tms = UTC time in milliseconds |
|---|
| 95 | # tb = Brightness temperature for each channel |
|---|
| 96 | # err = 1 if reading error, 0 else |
|---|
| 97 | # |
|---|
| 98 | """ |
|---|
| 99 | |
|---|
| 100 | # Index of channel |
|---|
| 101 | ch=[cch.lower() for cch in ch] |
|---|
| 102 | if (ch[0] == 'all'): |
|---|
| 103 | ich = arange(5) |
|---|
| 104 | else: |
|---|
| 105 | ich = array([int(cch[1])-1 for cch in ch]) |
|---|
| 106 | |
|---|
| 107 | # Variables |
|---|
| 108 | dset_names = ['amb1c_latlon','amb1c_btemps','amb1c_scnlindy','amb1c_scnlintime','amb1c_scnlinyr'] |
|---|
| 109 | try: |
|---|
| 110 | headers, datasets = ReadRawFile(fname,'/home/mleduc/python/ReadRawFile/data/amsub-mhs_1c_header.txt','/home/mleduc/python/ReadRawFile/data/amsub-mhs_1c_record.txt', False,dset_names) |
|---|
| 111 | except: |
|---|
| 112 | logger.error('Error with {0}'.format(fname)) |
|---|
| 113 | # continue |
|---|
| 114 | return |
|---|
| 115 | |
|---|
| 116 | # Latitude |
|---|
| 117 | lat = float32(datasets['amb1c_latlon'][:,:,0]/1.E4) |
|---|
| 118 | # Longitude |
|---|
| 119 | lon = float32(datasets['amb1c_latlon'][:,:,1]/1.E4) |
|---|
| 120 | # Year |
|---|
| 121 | yyyy= datasets['amb1c_scnlinyr'] |
|---|
| 122 | # Day of Year |
|---|
| 123 | yday= datasets['amb1c_scnlindy'] |
|---|
| 124 | # Time in milliseconds |
|---|
| 125 | tms = datasets['amb1c_scnlintime'] |
|---|
| 126 | # Tb |
|---|
| 127 | tb = float32(datasets['amb1c_btemps'][:,:,ich]/1.E2) |
|---|
| 128 | # Remplace fill values -9999.99 by NaN |
|---|
| 129 | tb[(tb == -9999.99)] = NaN |
|---|
| 130 | |
|---|
| 131 | ### Extract Area |
|---|
| 132 | # Index in the defined area |
|---|
| 133 | LS, LN, LW, LE = area |
|---|
| 134 | zone = (lon>=LW) & (lon<=LE) & (lat>=LS) & (lat<=LN) |
|---|
| 135 | nscn, nfov, nch = tb.shape |
|---|
| 136 | # Extraction of Tb for each channel |
|---|
| 137 | tb = [(tb[:,:,ii])[zone] for ii in xrange(nch)] |
|---|
| 138 | # day of year and time |
|---|
| 139 | yyyy = transpose([yyyy]*nfov)[zone] |
|---|
| 140 | yday = transpose([yday]*nfov)[zone] |
|---|
| 141 | tms = transpose([tms]*nfov)[zone] |
|---|
| 142 | # Find location in fov |
|---|
| 143 | fov = array(range(nfov)*nscn).reshape(nscn,nfov)[zone]+1 |
|---|
| 144 | |
|---|
| 145 | ### Convert in python date/time |
|---|
| 146 | dd = num2date(yday) |
|---|
| 147 | tt = [ms2time(ms) for ms in tms] |
|---|
| 148 | dt = array([d.replace(year=y,hour=h,minute=m,second=s) for d,y,(h,m,s) in zip(dd,yyyy,tt)]) |
|---|
| 149 | |
|---|
| 150 | return lon[zone], lat[zone], dt, tb, fov |
|---|
| 151 | |
|---|
| 152 | ##-----------------------------------------------------## |
|---|
| 153 | # Convert milliseconds in hour, minute, second # |
|---|
| 154 | ##-----------------------------------------------------## |
|---|
| 155 | def ms2time(ms): |
|---|
| 156 | """ ### Convert milliseconds in hour, minute, second |
|---|
| 157 | # Warning 1: lost millisecond because second is integer 2) no day |
|---|
| 158 | # Warning 2: no day |
|---|
| 159 | ## Inputs |
|---|
| 160 | # ms = millisecond |
|---|
| 161 | # |
|---|
| 162 | ## Output |
|---|
| 163 | # h = hour |
|---|
| 164 | # m = minute |
|---|
| 165 | # s = second |
|---|
| 166 | # |
|---|
| 167 | """ |
|---|
| 168 | s = ms/1000 |
|---|
| 169 | m, s = divmod(s,60) |
|---|
| 170 | h, m = divmod(m,60) |
|---|
| 171 | #d, h = divmod(h,24) |
|---|
| 172 | return h, m, s |
|---|
| 173 | |
|---|
| 174 | ##---------------------## |
|---|
| 175 | # Find classes # |
|---|
| 176 | ##---------------------## |
|---|
| 177 | def find_class(lat,lon): |
|---|
| 178 | """ ### Find class of each given (lat,lon) |
|---|
| 179 | ## Inputs |
|---|
| 180 | # lat = Latitude (degrees) |
|---|
| 181 | # lon = Longitude (degrees) |
|---|
| 182 | # |
|---|
| 183 | ## Output |
|---|
| 184 | # clpt = class of each point (same shape as lat and lon) |
|---|
| 185 | # |
|---|
| 186 | """ |
|---|
| 187 | |
|---|
| 188 | # Read the classification file |
|---|
| 189 | fic_class = '/homedata/mleduc/CorrTb/classifERAI_Trop.nc' |
|---|
| 190 | fileID = netCDF4.Dataset(fic_class) |
|---|
| 191 | lat_class = fileID.variables['lat'][:] |
|---|
| 192 | lon_class = fileID.variables['lon'][:] |
|---|
| 193 | classif = int16(fileID.variables['classif_tot'][:]) |
|---|
| 194 | fileID.close() |
|---|
| 195 | |
|---|
| 196 | # Latitude limits of classification (ie tropical band) |
|---|
| 197 | latstep = abs(diff(lat_class))[0] |
|---|
| 198 | latmin = lat_class.min()-latstep/2. |
|---|
| 199 | latmax = lat_class.max()+latstep/2. |
|---|
| 200 | # Only points in the area of classification |
|---|
| 201 | inclass = logical_and(lat>=latmin,lat<=latmax) |
|---|
| 202 | |
|---|
| 203 | # Geographic projection (lon/lat in degrees to x/y in m) |
|---|
| 204 | projG = pyproj.Proj("+proj=merc +ellps=WGS84 +datum=WGS84") |
|---|
| 205 | longd_class,latgd_class = meshgrid(lon_class,lat_class) |
|---|
| 206 | xclass,yclass = projG(longd_class,latgd_class) |
|---|
| 207 | xx,yy = projG(lon[inclass],lat[inclass]) |
|---|
| 208 | |
|---|
| 209 | # Find the nearest neighbour (euclidian distance) with a kd-tree |
|---|
| 210 | tree = spatial.cKDTree(zip(ravel(xclass),ravel(yclass))) |
|---|
| 211 | _, ind = tree.query(zip(xx,yy)) |
|---|
| 212 | clpt = zeros(shape(lat),dtype=int16) |
|---|
| 213 | clpt[inclass] = classif.flatten()[ind] |
|---|
| 214 | |
|---|
| 215 | return clpt |
|---|
| 216 | |
|---|
| 217 | ##-------------------------------------------## |
|---|
| 218 | # Find the coefficient of correction # |
|---|
| 219 | ##-------------------------------------------## |
|---|
| 220 | def find_corrcoef(dt,fov,cl,ch=['All']): |
|---|
| 221 | """ ### Find the coefficient of correction |
|---|
| 222 | ## Inputs |
|---|
| 223 | # dt = Python date |
|---|
| 224 | # fov = Location in the swath |
|---|
| 225 | # cl = Class |
|---|
| 226 | # ch = list of extrated channel ('B1','B2','B3','B4','B5' or 'All'), default 'All' |
|---|
| 227 | # |
|---|
| 228 | ## Outputs |
|---|
| 229 | # coef = Coefficient of correction |
|---|
| 230 | # |
|---|
| 231 | """ |
|---|
| 232 | # Index of channel |
|---|
| 233 | ch=[cch.lower() for cch in ch] |
|---|
| 234 | if (ch[0] == 'all'): |
|---|
| 235 | ich = arange(5) |
|---|
| 236 | else: |
|---|
| 237 | ich = array([int(cch[1])-1 for cch in ch]) |
|---|
| 238 | nch = ich.size |
|---|
| 239 | |
|---|
| 240 | # Read the classification file |
|---|
| 241 | fic_corr = '/homedata/mleduc/CorrTb/corrTb_AMSUB_MHS_L1C.nc' |
|---|
| 242 | fileID = netCDF4.Dataset(fic_corr) |
|---|
| 243 | day_corr = fileID.variables['day'][:] |
|---|
| 244 | coef_corr = fileID.variables['coefcorr'][:,:,:,ich]#[:] |
|---|
| 245 | fileID.close() |
|---|
| 246 | |
|---|
| 247 | # Find the day of year |
|---|
| 248 | yday = array([d.timetuple().tm_yday for d in dt]) |
|---|
| 249 | # Find the day of correction |
|---|
| 250 | nday = day_corr.size |
|---|
| 251 | daystep = diff(day_corr)[0] |
|---|
| 252 | icorr = list(flatten([[ii]*daystep for ii in xrange(nday)])) + [0]*6 |
|---|
| 253 | iday = array(icorr)[yday-1] |
|---|
| 254 | |
|---|
| 255 | # Find correction |
|---|
| 256 | noclass = (cl == 0) |
|---|
| 257 | icl = array(cl-1,dtype=int) |
|---|
| 258 | ifov = array(fov-1,dtype=int) |
|---|
| 259 | coef = [] |
|---|
| 260 | for ii in ich: |
|---|
| 261 | dd = zeros(cl.shape)+NaN |
|---|
| 262 | dd[~noclass] = coef_corr[iday[~noclass],icl[~noclass],ifov[~noclass],ii] |
|---|
| 263 | coef.append(dd) |
|---|
| 264 | |
|---|
| 265 | return float32(coef) |
|---|
| 266 | |
|---|
| 267 | ##---------------------------## |
|---|
| 268 | # New grid and time # |
|---|
| 269 | ##---------------------------## |
|---|
| 270 | def newgrid(lonvec,latvec,lon,lat,dt,tb): |
|---|
| 271 | """ ### Find the coefficient of correction |
|---|
| 272 | ## Inputs |
|---|
| 273 | # |
|---|
| 274 | # |
|---|
| 275 | ## Outputs |
|---|
| 276 | # |
|---|
| 277 | # |
|---|
| 278 | """ |
|---|
| 279 | |
|---|
| 280 | # from date to ordinal |
|---|
| 281 | dtord = date2num(dt) |
|---|
| 282 | |
|---|
| 283 | ### Find index of each pixel in new grid |
|---|
| 284 | # Geographic projection (lon/lat in degrees to x/y in m) |
|---|
| 285 | projG = pyproj.Proj("+proj=merc +ellps=WGS84 +datum=WGS84") |
|---|
| 286 | longd, latgd = meshgrid(lonvec,latvec) |
|---|
| 287 | xgd, ygd = projG(longd,latgd) |
|---|
| 288 | xx, yy = projG(lon,lat) |
|---|
| 289 | # Find the nearest neighbour (euclidian distance) with a kd-tree |
|---|
| 290 | tree = spatial.cKDTree(zip(ravel(xgd),ravel(ygd))) |
|---|
| 291 | _, indll = tree.query(zip(xx,yy)) |
|---|
| 292 | |
|---|
| 293 | # Mean |
|---|
| 294 | nch = tb.shape[0] |
|---|
| 295 | tbm = zeros((nch,lonvec.size*latvec.size),dtype=float32)+NaN |
|---|
| 296 | dtm = zeros((lonvec.size*latvec.size))+NaN |
|---|
| 297 | for ii in unique(indll): |
|---|
| 298 | tbm[:,ii] = nanmean(tb[:,(indll==ii)],axis=1) |
|---|
| 299 | dtm[ii] = nanmean(dtord[(indll==ii)]) |
|---|
| 300 | |
|---|
| 301 | return tbm, dtm |
|---|
