def load_blkdata_wrf(wrfpath, wrffiles="wrfout*", date0=False, date1=False, quiet=True):
from okean.datasets import wrf
a = wrf.WRFData(wrfpath, wrffiles)
data = a.data(date0, date1, quiet)
out = cb.odict()
if not data:
print "no data found !"
return out
time = data["time"]
for it in range(len(time)):
# be sure time increases!
if out.keys() and time[it] <= out.keys()[-1]:
continue
out[time[it]] = {}
for k in data.keys():
if k in ("time",):
continue
elif k.startswith("INFO"):
out[time[it]][k] = data[k]
else:
out[time[it]][k] = data[k].data[it, ...]
# x and y should be the same, so:
if not out[time[it]].has_key("x"):
out[time[it]]["x"] = data[k].x
out[time[it]]["y"] = data[k].y
return out
def load_blkdata_gfs(gfspath, date0, date1=False, nforec=0, quiet=True):
"""
nforec=0 means not forecast, auto means default from config file
"""
from okean.datasets import gfs
a = gfs.GFSData(gfspath)
if nforec == 0:
data, miss = a.data_analysis(date0, date1, quiet=quiet)
else:
data, miss = a.data_forecast(date0, nforec, quiet=quiet)
out = cb.odict()
time = data.keys()
for t in time:
out[t] = {}
for k in data[t].keys():
if k.startswith("INFO"):
out[t][k] = data[t][k]
else:
out[t][k] = data[t][k].data
# x and y may change with time, but should be the same for each time (gfs file)
if not out[t].has_key("x"):
out[t]["x"] = data[t][k].x
out[t]["y"] = data[t][k].y
return out, miss
def load_blkdata_interim(interimpath,
date0=False,
date1=False,
quiet=True,
past=False):
'''
If past, then module interim_past will be used to extract data from
old interim server. Otherwise interim module will be used.
'''
if past:
from okean.datasets import interim_past as interim
else:
from okean.datasets import interim
a = interim.INTERIMData(interimpath)
data = a.data(date0, date1, quiet)
out = cb.odict()
time = data['time']
for it in range(len(time)):
out[time[it]] = {}
for k in data.keys():
if k in ('time', ): continue
elif k.startswith('INFO'):
out[time[it]][k] = data[k]
else:
out[time[it]][k] = data[k].data[it, ...]
# x and y should be the same, so:
if not out[time[it]].has_key('x'):
out[time[it]]['x'] = data[k].x
out[time[it]]['y'] = data[k].y
return out
def __dims(self):
'''Returns variable dimensions dict'''
d=cbt.odict()
if self._interface=='pycdf': names = self._nc.dimensions()
elif self._interface=='scientific': names = self._nc.dimensions
elif self._interface=='netcdf4': names = self._nc.dimensions
if self._interface in ('pycdf','scientific'):
for n in names: d[n]=self.parent.dims[n]
# In netcdf 4, dimensions are scoped such that they can be seen
# in all descendant groups. That is, dimensions can be shared
# between variables in different groups, if they are defined in
# a parent group.
elif self._interface == 'netcdf4':
for n in names:
p=self.parent
while True:
if p.dims.has_key(n):
d[n]=p.dims[n]
break
if p.parent: p=p.parent
else: break
return d
def __dims(self):
'''Returns variable dimensions dict'''
d = cbt.odict()
if self._interface == 'pycdf': names = self._nc.dimensions()
elif self._interface == 'scientific': names = self._nc.dimensions
elif self._interface == 'netcdf4': names = self._nc.dimensions
if self._interface in ('pycdf', 'scientific'):
for n in names:
d[n] = self.parent.dims[n]
# In netcdf 4, dimensions are scoped such that they can be seen
# in all descendant groups. That is, dimensions can be shared
# between variables in different groups, if they are defined in
# a parent group.
elif self._interface == 'netcdf4':
for n in names:
p = self.parent
while True:
if p.dims.has_key(n):
d[n] = p.dims[n]
break
if p.parent: p = p.parent
else: break
return d
def __dims(self):
'''Returns file dimensions dict'''
dims = cbt.odict()
if self._interface == 'pycdf':
for i in range(self._nc.inq_ndims()):
dims[self._nc.dim(i).inq()[0]] = self._nc.dim(i).inq()[1]
elif self._interface == 'scientific':
# get ordered dims:
if not self._ncdump_info is False:
odims = self._ncdump_info['dimensions']
else:
odims = False
if not odims: odims = self._nc.dimensions.keys()
for k in odims:
dims[k] = self._nc.dimensions[k]
elif self._interface == 'netcdf4':
# get ordered dims:
if not self._ncdump_info is False:
odims = self._ncdump_info['dimensions']
else:
odims = False
if not odims: odims = self._nc.dimensions.keys()
for k in odims:
dims[k] = len(self._nc.dimensions[k])
return dims
def load_blkdata_interim(interimpath, date0=False, date1=False, quiet=True, past=False):
"""
If past, then module interim_past will be used to extract data from
old interim server. Otherwise interim module will be used.
"""
if past:
from okean.datasets import interim_past as interim
else:
from okean.datasets import interim
a = interim.INTERIMData(interimpath)
data = a.data(date0, date1, quiet)
out = cb.odict()
time = data["time"]
for it in range(len(time)):
out[time[it]] = {}
for k in data.keys():
if k in ("time",):
continue
elif k.startswith("INFO"):
out[time[it]][k] = data[k]
else:
out[time[it]][k] = data[k].data[it, ...]
# x and y should be the same, so:
if not out[time[it]].has_key("x"):
out[time[it]]["x"] = data[k].x
out[time[it]]["y"] = data[k].y
return out
def load_blkdata_wrf(wrfpath,
wrffiles='wrfout*',
date0=False,
date1=False,
quiet=True):
from okean.datasets import wrf
a = wrf.WRFData(wrfpath, wrffiles)
data = a.data(date0, date1, quiet)
out = cb.odict()
if not data:
print('no data found !')
return out
time = data['time']
for it in range(len(time)):
# be sure time increases!
if out.keys() and time[it] <= out.keys()[-1]: continue
out[time[it]] = {}
for k in data.keys():
if k in ('time', ): continue
elif k.startswith('INFO'):
out[time[it]][k] = data[k]
else:
out[time[it]][k] = data[k].data[it, ...]
# x and y should be the same, so:
if not out[time[it]].has_key('x'):
out[time[it]]['x'] = data[k].x
out[time[it]]['y'] = data[k].y
return out
def load_blkdata_cfsr(cfsrpath, date0=False, date1=False, quiet=True):
from okean.datasets import cfsr
a = cfsr.CFSRData(cfsrpath)
data = a.data(date0, date1, quiet)
out = cb.odict()
time = data["time"]
for it in range(len(time)):
out[time[it]] = {}
for k in data.keys():
if k in ("time",):
continue
elif k.startswith("INFO"):
out[time[it]][k] = data[k]
else:
out[time[it]][k] = data[k].data[it, ...]
# x and y should be the same, so:
if not out[time[it]].has_key("x"):
out[time[it]]["x"] = data[k].x
out[time[it]]["y"] = data[k].y
return out
def load_blkdata_gfs(gfspath, date0, date1=False, nforec=0, quiet=True):
'''
nforec=0 means not forecast, auto means default from config file
'''
from okean.datasets import gfs
a = gfs.GFSData(gfspath)
if nforec == 0:
data, miss = a.data_analysis(date0, date1, quiet=quiet)
else:
data, miss = a.data_forecast(date0, nforec, quiet=quiet)
out = cb.odict()
time = data.keys()
for t in time:
out[t] = {}
for k in data[t].keys():
if k.startswith('INFO'):
out[t][k] = data[t][k]
else:
out[t][k] = data[t][k].data
# x and y may change with time, but should be the same for each time (gfs file)
if not out[t].has_key('x'):
out[t]['x'] = data[t][k].x
out[t]['y'] = data[t][k].y
return out, miss
def __groups(self):
out=cbt.odict()
if self._interface=='netcdf4':
gs=self._nc.groups
try: gnames=gs.keys()
except: gnames=()
for k in gnames: out[k]=Pyncgroup(nc=gs[k],name=k,parent=self)
else:
print ':: groups only implemented in interface ','netcdf4'
return out
def read_log(f):
out=cb.odict()
if os.path.isfile(f):
L=open(log).readlines()
for l in L:
tmp=l.split(' ',1)
sdate=tmp[0]
scontents=tmp[1].rstrip().split(' + ')
out[tmp[0]]=scontents
return out
def read_log(f):
out = cb.odict()
if os.path.isfile(f):
L = open(log).readlines()
for l in L:
tmp = l.split(' ', 1)
sdate = tmp[0]
scontents = tmp[1].rstrip().split(' + ')
out[tmp[0]] = scontents
return out
def __atts(self):
'''Returns netcdf ordered attributes, with types and values'''
att = cbt.odict()
if self._interface == 'pycdf':
for i in range(self._nc.inq_natts()):
a = self._nc.attr(i)
# ps: a.get() produces a segmentation fault for large
# strings, so:
L = a.inq_len()
if L >= 250: val = 'PYCDF ERROR: CANNOT READ !!!!'
else: attvalue = a.get()
# find nctype:
nctype = nctypes.type_pycdf2nc(a.inq_type())
att[a.inq_name()] = Pyncatt(self._nc,
a.inq_name(),
attvalue,
nctype=nctype,
ncversion=self.ncversion,
interface=self._interface)
elif self._interface == 'scientific':
a = self._nc.__dict__
# get ordered atts:
if not self._ncdump_info is False:
oatts = self._ncdump_info['vattributes'][self.varname]
else:
oatts = False
if oatts: keys = oatts
else: keys = a.keys()
for k in keys:
attvalue = a[k]
att[k] = Pyncatt(self._nc,
k,
attvalue,
ncversion=self.ncversion,
interface=self._interface)
elif self._interface == 'netcdf4':
aname = self._nc.ncattrs() # already ordered!
for a in aname:
attvalue = getattr(self._nc, a)
att[a] = Pyncatt(self._nc,
a,
attvalue,
ncversion=self.ncversion,
interface=self._interface)
return att
def __groups(self):
out = cbt.odict()
if self._interface == 'netcdf4':
gs = self._nc.groups
try:
gnames = gs.keys()
except:
gnames = ()
for k in gnames:
out[k] = Pyncgroup(nc=gs[k], name=k, parent=self)
else:
print ':: groups only implemented in interface ', 'netcdf4'
return out
def __vars(self):
'''Returns file variables'''
var=cbt.odict()
if self._interface=='pycdf':
names=[self._nc.var(i).inq_name() for i in range(self._nc.inq_nvars())]
elif self._interface in ('scientific','netcdf4'):
# get ordered names:
if not self._ncdump_info is False:
onames=self._ncdump_info['variables'].keys()
else: onames=False
if onames: names=onames
else: names=self._nc.variables.keys()
for vname in names: var[vname]=Pyncvar(self,vname)
return var
def list_files(cf,date1,date2,io,type,FA,nest=0,quiet=True):
files=cb.odict()
date=date1
while date<=date2:
f=opt.nameof(io,type,date,FA,nest,cf)
if os.path.isfile(f): files[date]=f
else: files[date]=False
date=dateu.next_date(date,1)
if not quiet:
for d in files.keys():
if files[d]:
f=files[d]
hs=cb.hsize(os.path.getsize(f))
print d,' ', f, ' ',hs[0],hs[1]
else:
print d,' no file'
return files
def load_blkdata_narr(date0, date1, quiet=True):
from okean.datasets import narr
a = narr.NARRData()
data, miss = a.data(date0, date1, quiet=quiet)
out = cb.odict()
time = data.keys()
for t in time:
out[t] = {}
for k in data[t].keys():
if k.startswith('INFO'):
out[t][k] = data[t][k]
else:
out[t][k] = data[t][k].data
if not out[t].has_key('x'):
out[t]['x'] = data[t][k].x
out[t]['y'] = data[t][k].y
return out, miss
def list_files(cf, date1, date2, io, type, FA, nest=0, quiet=True):
files = cb.odict()
date = date1
while date <= date2:
f = opt.nameof(io, type, date, FA, nest, cf)
if os.path.isfile(f): files[date] = f
else: files[date] = False
date = dateu.next_date(date, 1)
if not quiet:
for d in files.keys():
if files[d]:
f = files[d]
hs = cb.hsize(os.path.getsize(f))
print d, ' ', f, ' ', hs[0], hs[1]
else:
print d, ' no file'
return files
def __vars(self):
'''Returns file variables'''
var = cbt.odict()
if self._interface == 'pycdf':
names = [
self._nc.var(i).inq_name() for i in range(self._nc.inq_nvars())
]
elif self._interface in ('scientific', 'netcdf4'):
# get ordered names:
if not self._ncdump_info is False:
onames = self._ncdump_info['variables'].keys()
else:
onames = False
if onames: names = onames
else: names = self._nc.variables.keys()
for vname in names:
var[vname] = Pyncvar(self, vname)
return var
def load_blkdata_narr(date0,date1,quiet=True):
from okean.datasets import narr
a=narr.NARRData()
data,miss=a.data(date0,date1,quiet=quiet)
out=cb.odict()
time=data.keys()
for t in time:
out[t]={}
for k in data[t].keys():
if k.startswith('INFO'):
out[t][k]=data[t][k]
else:
out[t][k]=data[t][k].data
if not out[t].has_key('x'):
out[t]['x']=data[t][k].x
out[t]['y']=data[t][k].y
return out, miss
def __atts(self):
'''Returns netcdf ordered attributes, with types and values'''
att=cbt.odict()
if self._interface=='pycdf':
for i in range(self._nc.inq_natts()):
a=self._nc.attr(i)
# ps: a.get() produces a segmentation fault for large
# strings, so:
L= a.inq_len()
if L>=250: val='PYCDF ERROR: CANNOT READ !!!!'
else: attvalue=a.get()
# find nctype:
nctype=nctypes.type_pycdf2nc(a.inq_type())
att[a.inq_name()]=Pyncatt(self._nc,a.inq_name(),attvalue,
nctype=nctype,ncversion=self.ncversion,
interface=self._interface)
elif self._interface=='scientific':
a=self._nc.__dict__
# get ordered atts:
if not self._ncdump_info is False:
oatts=self._ncdump_info['vattributes'][self.varname]
else: oatts=False
if oatts: keys=oatts
else: keys= a.keys()
for k in keys:
attvalue=a[k]
att[k]=Pyncatt(self._nc,k,attvalue,ncversion=self.ncversion,
interface=self._interface)
elif self._interface=='netcdf4':
aname=self._nc.ncattrs() # already ordered!
for a in aname:
attvalue=getattr(self._nc,a)
att[a]=Pyncatt(self._nc,a,attvalue,ncversion=self.ncversion,
interface=self._interface)
return att
def data(self,date0,date1=False,xlim=False,ylim=False,quiet=True):
files,time=self.files(date0,date1)
res=cb.odict()
for i in range(len(files)):
if not quiet: print('|-> getting from file %s' % files[i])
try:
data=narr_file_data(files[i],xlim,ylim,quiet=quiet)
except:
if not quiet: print('CANNOT use data for %s' % time[i].isoformat())
continue
if data:
data['INFO_file'] = files[i]
res[time[i]]=data
else:
if not quiet: print('NO DATA for %s' % time[i].isoformat())
miss={}
return res,miss
def add_group(self,groupname):
'''
Creates netcdf groups
See:
http://netcdf4-python.googlecode.com/svn/trunk/docs/netCDF4-module.html
'''
if self._interface=='netcdf4':
self._nc.createGroup(groupname)
newgroup=Pyncgroup(groupname,self)
# update self.groups:
self.groups[groupname]=newgroup
# update ncdump info:
if not self._ncdump_info is False:
self._ncdump_info['groups'][groupname]=cbt.odict()
return newgroup
else:
print ':: add_group only implemented in interface ','netcdf4'
return False
def add_group(self, groupname):
'''
Creates netcdf groups
See:
http://netcdf4-python.googlecode.com/svn/trunk/docs/netCDF4-module.html
'''
if self._interface == 'netcdf4':
self._nc.createGroup(groupname)
newgroup = Pyncgroup(groupname, self)
# update self.groups:
self.groups[groupname] = newgroup
# update ncdump info:
if not self._ncdump_info is False:
self._ncdump_info['groups'][groupname] = cbt.odict()
return newgroup
else:
print ':: add_group only implemented in interface ', 'netcdf4'
return False
def load_blkdata_cfsr(cfsrpath, date0=False, date1=False, quiet=True):
from okean.datasets import cfsr
a = cfsr.CFSRData(cfsrpath)
data = a.data(date0, date1, quiet)
out = cb.odict()
time = data['time']
for it in range(len(time)):
out[time[it]] = {}
for k in data.keys():
if k in ('time', ): continue
elif k.startswith('INFO'):
out[time[it]][k] = data[k]
else:
out[time[it]][k] = data[k].data[it, ...]
# x and y should be the same, so:
if not out[time[it]].has_key('x'):
out[time[it]]['x'] = data[k].x
out[time[it]]['y'] = data[k].y
return out
def __dims(self):
'''Returns file dimensions dict'''
dims=cbt.odict()
if self._interface=='pycdf':
for i in range(self._nc.inq_ndims()):
dims[self._nc.dim(i).inq()[0]]=self._nc.dim(i).inq()[1]
elif self._interface=='scientific':
# get ordered dims:
if not self._ncdump_info is False:
odims=self._ncdump_info['dimensions']
else: odims=False
if not odims: odims=self._nc.dimensions.keys()
for k in odims: dims[k]=self._nc.dimensions[k]
elif self._interface=='netcdf4':
# get ordered dims:
if not self._ncdump_info is False:
odims=self._ncdump_info['dimensions']
else: odims=False
if not odims: odims=self._nc.dimensions.keys()
for k in odims: dims[k]=len(self._nc.dimensions[k])
return dims
def update_wind_blended2(fname, datapaths, **kargs):
'''
In days without blended data will try to use quikscat data
'''
from okean.datasets import quikscat
from okean.datasets import blended_wind
a = blended_wind.WINDData(datapaths[0])
b = quikscat.WINDData(datapaths[1])
time = netcdf.nctime(fname, 'time')
date0 = dts.next_date(time[0], -1)
date1 = dts.next_date(time[-1], +2)
data = a.data(date0, date1)
# limit are... otherwise, quikscat interp will be very slow!
grd = netcdf.fatt(fname, 'grd_file')
import os
if not os.path.isfile(grd): grd = kargs['grd']
cond, inds = rt.grid_vicinity(grd,
data['x'],
data['y'],
margin=5,
rect=True,
retinds=True)
i1, i2, j1, j2 = inds
for d in data.keys():
if d == 'x': data[d] = data[d][i1:i2]
elif d == 'y': data[d] = data[d][j1:j2]
else: data[d] = data[d][j1:j2, i1:i2]
# check for missing days:
time0 = data.keys()
x0 = data['x']
y0 = data['y']
x0, y0 = np.meshgrid(x0, y0)
time0.remove('x')
time0.remove('y')
out = cb.odict()
out['x'] = x0
out['y'] = y0
info = ''
qs_ij_limits_done = False
for d in dts.drange(date0, date1):
found = 0
for t in time0:
if (t.year, t.month, t.day) == (d.year, d.month, d.day):
print('==> blended : ', t)
out[t] = data[t]
found = 1
if not found: # use quikscat:
print('==> quikscat : ', d.strftime('%Y-%m-%d'))
tmp = b.data(d, dts.next_date(d))
if not tmp.has_key('x'): continue
x, y = tmp['x'], tmp['y']
x, y = np.meshgrid(x, y)
# reduce qs data:
if not qs_ij_limits_done:
i1, i2, j1, j2 = calc.ij_limits(x, y,
[x0.min(), x0.max()],
[y0.min(), y0.max()])
qs_ij_limits_done = True
x = x[j1:j2, i1:i2]
y = y[j1:j2, i1:i2]
tmp[tmp.keys()[0]] = tmp[tmp.keys()[0]][j1:j2, i1:i2]
print(' griddata u')
u = calc.griddata(x, y, tmp[tmp.keys()[0]].real, x0, y0)
print(' griddata v')
v = calc.griddata(x, y, tmp[tmp.keys()[0]].imag, x0, y0)
out[tmp.keys()[0]] = u + 1.j * v
info += '#' + d.strftime('%Y%m%d')
new_wind_info = 'blended+quikscat at days: ' + info
update_wind(fname, out, new_wind_info, **kargs)
def data2roms(data,grd,sparams,**kargs):
'''
Interpolates data to roms 3D grid.
The dict data must contain the prognostic variables temp, salt, u,
v (3d) and ssh (zeta, 2d), as well as lon, lat (2d), depth (1d) and
time/date info: date (data date), date0 (reference date) and time
(difference between date and date0). The input data can be provided
by load_data.
Parameters
----------
data : dict with prognostic variables
grd : ROMS netcdf grid file
sparams : s-coordinates parameters, theta_s,theta_b, hc and NLevels
**kargs:
ij : axis for vertical interpolations (*i,j)
ij_ind : list of i or j index for vertical interpolation, all by
default (ij_ind=False)
horizAux : if True, the original data horizontally interpolated is
returned and can be used for next data2roms call with
this same karg
quiet : output messages flag (false by default)
proj : projection - False, name or basemap proj - lcc by default
if False, horizontal interpolations will use lonxlat instead of distances
interp_opts: options for griddata
rep_surf: repeat surface level (new upper level)
'''
ij = kargs.get('ij','j')
ij_ind = kargs.get('ij_ind',False)
horizAux = kargs.get('horizAux',False)
quiet = kargs.get('quiet',False)
proj = kargs.get('proj','lcc') # lonxlat to distance before
# horizontal interpolation
interp_opts = kargs.get('interp_opts',{})
rep_surf = kargs.get('rep_surf',True) # create a surface upper level
# before interpolation
if not quiet: print 'using grid %s' % grd
g=roms.Grid(grd)
xr,yr,hr,mr=g.vars('r')
xu,yu,hu,mu=g.vars('u')
xv,yv,hv,mv=g.vars('v')
ny,nx=hr.shape
nz=sparams[3]
if proj:
print 'projecting coordinates...'
if isinstance(proj,basestring):
lonc=(xr.max()+xr.min())/2.
latc=(yr.max()+yr.min())/2.
from mpl_toolkits.basemap import Basemap
proj=Basemap(projection=proj,width=1,height=1,resolution=None,
lon_0=lonc,lat_0=latc, lat_1=latc)
xr,yr=proj(xr,yr)
xu,yu=proj(xu,yu)
xv,yv=proj(xv,yv)
dlon,dlat=proj(data['lon'],data['lat'])
Rdz=1/100. # distance to depth ratio (300km vs 3000m)
distance=lambda x,y: np.append(0.,np.sqrt(np.diff(x)**2+np.diff(y)**2).cumsum())
else:
dlon,dlat=data['lon'],data['lat']
distance=calc.distance
# needed for s_levels and for rep_surf!
sshr=calc.griddata(dlon,dlat,data['ssh'],xr,yr,extrap=True,**interp_opts)
# repeat surface:
if rep_surf:
# copy data cos dont want to change the original dataset:
import copy
data=copy.deepcopy(data)
for vname in ['temp','salt','u','v','depth']:
if data[vname].ndim==1: # depth !
if np.ma.isMA(data[vname]): vstack=np.ma.hstack
else: vstack=np.hstack
else:
if np.ma.isMA(data[vname]): vstack=np.ma.vstack
else: vstack=np.vstack
if data['depth'][0]>data['depth'][1]: # surf at ind 0
data[vname]=vstack((data[vname][0][np.newaxis],data[vname]))
if vname=='depth': data[vname][0]=sshr.max()
else:
data[vname]=vstack((data[vname],data[vname][-1][np.newaxis]))
if vname=='depth': data[vname][-1]=sshr.max()
data['NZ']=data['NZ']+1
NX=data['NX']
NY=data['NY']
NZ=data['NZ']
if not quiet: print 'calc s levels...'
Zr = g.s_levels(sparams,sshr,hr,'r')
Zu = g.s_levels(sparams,sshr,hr,'u')
Zv = g.s_levels(sparams,sshr,hr,'v')
# interp horiz:
retHorizAux=horizAux is True
if horizAux in (True,False):
TEMP = np.ma.masked_all((NZ,ny,nx),data['temp'].dtype)
SALT = np.ma.masked_all((NZ,ny,nx),data['salt'].dtype)
U = np.ma.masked_all((NZ,ny,nx),data['u'].dtype)
V = np.ma.masked_all((NZ,ny,nx),data['v'].dtype)
if not quiet: print 'horizontal interpolation:'
for i in range(NZ):
if not quiet and i%10==0: print ' lev %d of %d' % (i,NZ)
#import pylab
#pylab.figure()
#pylab.pcolormesh(data['lon'],data['lat'],data['temp'][i,...])
try: TEMP[i,...] = calc.griddata(dlon,dlat,data['temp'][i,...],xr,yr,extrap=True,**interp_opts)
except: pass
try: SALT[i,...] = calc.griddata(dlon,dlat,data['salt'][i,...],xr,yr,extrap=True,**interp_opts)
except: pass
try: U[i,...] = calc.griddata(dlon,dlat,data['u'][i,...],xr,yr,extrap=True,**interp_opts)
except: pass
try: V[i,...] = calc.griddata(dlon,dlat,data['v'][i,...],xr,yr,extrap=True,**interp_opts)
except: pass
# rotate U,V:
if not quiet: print 'rotating U,V to grid angle'
angle=g.use('angle') # rad
U,V=calc.rot2d(U,V,angle)
U=rt.rho2uvp3d(U,'u')
V=rt.rho2uvp3d(V,'v')
horizAux={}
horizAux['TEMP'] = TEMP
horizAux['SALT'] = SALT
horizAux['U'] = U
horizAux['V'] = V
else:
TEMP = horizAux['TEMP']
SALT = horizAux['SALT']
U = horizAux['U']
V = horizAux['V']
# interp vert:
nxu=nx-1
nyv=ny-1
#> -----------------------------------------------------------------
useInd=not ij_ind is False
if ij_ind is False:
if ij=='j': ij_ind=range(ny)
elif ij=='i': ij_ind=range(nx)
else:
try: iter(ij_ind)
except: ij_ind=[ij_ind]
if ij=='j': ny=nyv=len(ij_ind)
elif ij=='i': nx=nxu=len(ij_ind)
# -----------------------------------------------------------------<
Temp = np.zeros((nz,ny ,nx ),data['temp'].dtype)
Salt = np.zeros((nz,ny ,nx ),data['salt'].dtype)
Uvel = np.zeros((nz,ny ,nxu),data['u'].dtype)
Vvel = np.zeros((nz,nyv,nx ),data['v'].dtype)
jslice=lambda x,ind: x[:,ind,:]
islice=lambda x,ind: x[:,:,ind]
ZZr = np.tile(data['depth'],(nx,ny,1)).T
ZZu = np.tile(data['depth'],(nxu,ny,1)).T
ZZv = np.tile(data['depth'],(nx,nyv,1)).T
if not useInd is False: #>------------------------------------------
if ij=='j':
slice=jslice
sshr=sshr[ij_ind,:]
hr =hr[ij_ind,:]
elif ij=='i':
slice=islice
sshr=sshr[:,ij_ind]
hr =hr[:,ij_ind]
Zr,Zu,Zv,TEMP,SALT,U,V=[slice(k,ij_ind) for k in [Zr,Zu,Zv,TEMP,SALT,U,V]]
# -----------------------------------------------------------------<
if useInd: # then store distances for a possible bry file
dtype=Temp.dtype
distr=np.zeros((nz,ny, nx ),dtype)
distu=np.zeros((nz,ny, nxu),dtype)
distv=np.zeros((nz,nyv,nx ),dtype)
if not quiet: print 'vertical interpolation:'
if ij=='j':
for j in range(ny):
if not quiet and (ny<10 or (ny>=10 and j%10==0)): print ' j=%3d of %3d' % (j,ny)
ind=ij_ind[j]
dr=np.tile(distance(xr[ind,:],yr[ind,:]),(nz,1))
du=np.tile(distance(xu[ind,:],yu[ind,:]),(nz,1))
Dr=np.tile(distance(xr[ind,:],yr[ind,:]),(NZ,1))
Du=np.tile(distance(xu[ind,:],yu[ind,:]),(NZ,1))
if useInd:
distr[:,j,:]=dr;
distu[:,j,:]=du;
Temp[:,j,:] = calc.griddata(Rdz*Dr,ZZr[:,j,:],TEMP[:,j,:],Rdz*dr,Zr[:,j,:],extrap=True,**interp_opts)
Salt[:,j,:] = calc.griddata(Rdz*Dr,ZZr[:,j,:],SALT[:,j,:],Rdz*dr,Zr[:,j,:],extrap=True,**interp_opts)
if 0 and j%10==0:
print Dr.shape, ZZr[:,j,:].shape
import pylab as pl
pl.figure(1)
pl.clf()
pl.pcolormesh(Dr,ZZr[:,j,:],SALT[:,j,:])
pl.colorbar()
clim=pl.gci().get_clim()
pl.figure(2)
pl.clf()
pl.pcolormesh(dr,Zr[:,j,:],Salt[:,j,:])
pl.clim(clim)
pl.colorbar()
raw_input()
Uvel[:,j,:] = calc.griddata(Rdz*Du,ZZu[:,j,:],U[:,j,:], Rdz*du,Zu[:,j,:],extrap=True,**interp_opts)
if j<Vvel.shape[1]:
dv=np.tile(distance(xv[ind,:],yv[ind,:]),(nz,1))
Dv=np.tile(distance(xv[ind,:],yv[ind,:]),(NZ,1))
Vvel[:,j,:] = calc.griddata(Rdz*Dv,ZZv[:,j,:],V[:,j,:], Rdz*dv,Zv[:,j,:],extrap=True,**interp_opts)
if useInd:
distv[:,j,:]=dv
if np.any(np.isnan(Temp[:,j,:])): print 'found nan in temp',j
if np.any(np.isnan(Salt[:,j,:])): print 'found nan in salt',j
if np.any(np.isnan(Uvel[:,j,:])): print 'found nan in u',j
if j<Vvel.shape[1] and np.any(np.isnan(Vvel[:,j,:])): print 'found nan in v',j
elif ij=='i':
for i in range(nx):
if not quiet and (nx<10 or (nx>=10 and i%10==0)): print ' i=%3d of %3d' % (i,nx)
ind=ij_ind[i]
dr=np.tile(distance(xr[:,ind],yr[:,ind]),(nz,1))
dv=np.tile(distance(xv[:,ind],yv[:,ind]),(nz,1))
Dr=np.tile(distance(xr[:,ind],yr[:,ind]),(NZ,1))
Dv=np.tile(distance(xv[:,ind],yv[:,ind]),(NZ,1))
if useInd:
distr[:,:,i]=dr;
distv[:,:,i]=dv;
Temp[:,:,i] = calc.griddata(Rdz*Dr,ZZr[:,:,i],TEMP[:,:,i],Rdz*dr,Zr[:,:,i],extrap=True,**interp_opts)
Salt[:,:,i] = calc.griddata(Rdz*Dr,ZZr[:,:,i],SALT[:,:,i],Rdz*dr,Zr[:,:,i],extrap=True,**interp_opts)
Vvel[:,:,i] = calc.griddata(Rdz*Dv,ZZv[:,:,i],V[:,:,i], Rdz*dv,Zv[:,:,i],extrap=True,**interp_opts)
if i<Uvel.shape[2]:
du=np.tile(distance(xu[:,ind],yu[:,ind]),(nz,1))
Du=np.tile(distance(xu[:,ind],yu[:,ind]),(NZ,1))
Uvel[:,:,i] = calc.griddata(Rdz*Du,ZZu[:,:,i],U[:,:,i], Rdz*du,Zu[:,:,i],extrap=True,**interp_opts)
if useInd:
distu[:,:,i]=du
# uv bar:
if not quiet: print 'calc uvbar'
if useInd is False:
ubar,vbar=rt.uvbar(Uvel,Vvel,sshr,hr,sparams)
else: #>------------------------------------------------------------
sshu=calc.griddata(dlon,dlat,data['ssh'],xu,yu,extrap=True,**interp_opts)
sshv=calc.griddata(dlon,dlat,data['ssh'],xv,yv,extrap=True,**interp_opts)
if ij=='j':
sshu=sshu[ij_ind,:]
sshv=sshv[ij_ind,:]
hu =hu[ij_ind,:]
hv =hv[ij_ind,:]
elif ij=='i':
sshu=sshu[:,ij_ind]
sshv=sshv[:,ij_ind]
hu =hu[:,ij_ind]
hv =hv[:,ij_ind]
ubar=rt.barotropic(Uvel,sshu,hu,sparams)
vbar=rt.barotropic(Vvel,sshv,hv,sparams)
# -----------------------------------------------------------------<
Vars=cb.odict()
Vars['temp'] = Temp
Vars['salt'] = Salt
Vars['u'] = Uvel
Vars['v'] = Vvel
Vars['zeta'] = sshr
Vars['ubar'] = ubar
Vars['vbar'] = vbar
Vars['date'] = data['date']
if not useInd is False: #>------------------------------------------
Vars['depth'] = Zr
Vars['depthu'] = Zu
Vars['depthv'] = Zv
Vars['dist'] = distr
Vars['distu'] = distu
Vars['distv'] = distv
# -----------------------------------------------------------------<
if retHorizAux: return Vars, horizAux
else: return Vars
def data2roms(data, grd, sparams, **kargs):
'''
Interpolates data to roms 3D grid.
The dict data must contain the prognostic variables temp, salt, u,
v (3d) and ssh (zeta, 2d), as well as lon, lat (2d), depth (1d) and
time/date info: date (data date), date0 (reference date) and time
(difference between date and date0). The input data can be provided
by load_data.
Parameters
----------
data : dict with prognostic variables
grd : ROMS netcdf grid file
sparams : s-coordinates parameters, theta_s,theta_b, hc and NLevels
**kargs:
ij : axis for vertical interpolations (*i,j)
ij_ind : list of i or j index for vertical interpolation, all by
default (ij_ind=False)
horizAux : if True, the original data horizontally interpolated is
returned and can be used for next data2roms call with
this same karg
quiet : output messages flag (false by default)
proj : projection - False, name or basemap proj - lcc by default
if False, horizontal interpolations will use lonxlat instead of distances
interp_opts: options for griddata
rep_surf: repeat surface level (new upper level)
'''
ij = kargs.get('ij', 'j')
ij_ind = kargs.get('ij_ind', False)
horizAux = kargs.get('horizAux', False)
quiet = kargs.get('quiet', False)
proj = kargs.get('proj', 'lcc') # lonxlat to distance before
# horizontal interpolation
interp_opts = kargs.get('interp_opts', {})
rep_surf = kargs.get('rep_surf', True) # create a surface upper level
# before interpolation
if not quiet: print 'using grid %s' % grd
g = roms.Grid(grd)
xr, yr, hr, mr = g.vars('r')
xu, yu, hu, mu = g.vars('u')
xv, yv, hv, mv = g.vars('v')
ny, nx = hr.shape
nz = sparams[3]
if proj:
print 'projecting coordinates...'
if isinstance(proj, basestring):
lonc = (xr.max() + xr.min()) / 2.
latc = (yr.max() + yr.min()) / 2.
from mpl_toolkits.basemap import Basemap
proj = Basemap(projection=proj,
width=1,
height=1,
resolution=None,
lon_0=lonc,
lat_0=latc,
lat_1=latc)
xr, yr = proj(xr, yr)
xu, yu = proj(xu, yu)
xv, yv = proj(xv, yv)
dlon, dlat = proj(data['lon'], data['lat'])
Rdz = 1 / 100. # distance to depth ratio (300km vs 3000m)
distance = lambda x, y: np.append(
0.,
np.sqrt(np.diff(x)**2 + np.diff(y)**2).cumsum())
else:
dlon, dlat = data['lon'], data['lat']
distance = calc.distance
# needed for s_levels and for rep_surf!
sshr = calc.griddata(dlon,
dlat,
data['ssh'],
xr,
yr,
extrap=True,
**interp_opts)
# repeat surface:
if rep_surf:
# copy data cos dont want to change the original dataset:
import copy
data = copy.deepcopy(data)
for vname in ['temp', 'salt', 'u', 'v', 'depth']:
if data[vname].ndim == 1: # depth !
if np.ma.isMA(data[vname]): vstack = np.ma.hstack
else: vstack = np.hstack
else:
if np.ma.isMA(data[vname]): vstack = np.ma.vstack
else: vstack = np.vstack
if data['depth'][0] > data['depth'][1]: # surf at ind 0
data[vname] = vstack((data[vname][0][np.newaxis], data[vname]))
if vname == 'depth': data[vname][0] = sshr.max()
else:
data[vname] = vstack(
(data[vname], data[vname][-1][np.newaxis]))
if vname == 'depth': data[vname][-1] = sshr.max()
data['NZ'] = data['NZ'] + 1
NX = data['NX']
NY = data['NY']
NZ = data['NZ']
if not quiet: print 'calc s levels...'
Zr = g.s_levels(sparams, sshr, hr, 'r')
Zu = g.s_levels(sparams, sshr, hr, 'u')
Zv = g.s_levels(sparams, sshr, hr, 'v')
# interp horiz:
retHorizAux = horizAux is True
if horizAux in (True, False):
TEMP = np.ma.masked_all((NZ, ny, nx), data['temp'].dtype)
SALT = np.ma.masked_all((NZ, ny, nx), data['salt'].dtype)
U = np.ma.masked_all((NZ, ny, nx), data['u'].dtype)
V = np.ma.masked_all((NZ, ny, nx), data['v'].dtype)
if not quiet: print 'horizontal interpolation:'
for i in range(NZ):
if not quiet and i % 10 == 0: print ' lev %d of %d' % (i, NZ)
#import pylab
#pylab.figure()
#pylab.pcolormesh(data['lon'],data['lat'],data['temp'][i,...])
try:
TEMP[i, ...] = calc.griddata(dlon,
dlat,
data['temp'][i, ...],
xr,
yr,
extrap=True,
**interp_opts)
except:
pass
try:
SALT[i, ...] = calc.griddata(dlon,
dlat,
data['salt'][i, ...],
xr,
yr,
extrap=True,
**interp_opts)
except:
pass
try:
U[i, ...] = calc.griddata(dlon,
dlat,
data['u'][i, ...],
xr,
yr,
extrap=True,
**interp_opts)
except:
pass
try:
V[i, ...] = calc.griddata(dlon,
dlat,
data['v'][i, ...],
xr,
yr,
extrap=True,
**interp_opts)
except:
pass
# rotate U,V:
if not quiet: print 'rotating U,V to grid angle'
angle = g.use('angle') # rad
U, V = calc.rot2d(U, V, angle)
U = rt.rho2uvp3d(U, 'u')
V = rt.rho2uvp3d(V, 'v')
horizAux = {}
horizAux['TEMP'] = TEMP
horizAux['SALT'] = SALT
horizAux['U'] = U
horizAux['V'] = V
else:
TEMP = horizAux['TEMP']
SALT = horizAux['SALT']
U = horizAux['U']
V = horizAux['V']
# interp vert:
nxu = nx - 1
nyv = ny - 1
#> -----------------------------------------------------------------
useInd = not ij_ind is False
if ij_ind is False:
if ij == 'j': ij_ind = range(ny)
elif ij == 'i': ij_ind = range(nx)
else:
try:
iter(ij_ind)
except:
ij_ind = [ij_ind]
if ij == 'j': ny = nyv = len(ij_ind)
elif ij == 'i': nx = nxu = len(ij_ind)
# -----------------------------------------------------------------<
Temp = np.zeros((nz, ny, nx), data['temp'].dtype)
Salt = np.zeros((nz, ny, nx), data['salt'].dtype)
Uvel = np.zeros((nz, ny, nxu), data['u'].dtype)
Vvel = np.zeros((nz, nyv, nx), data['v'].dtype)
jslice = lambda x, ind: x[:, ind, :]
islice = lambda x, ind: x[:, :, ind]
ZZr = np.tile(data['depth'], (nx, ny, 1)).T
ZZu = np.tile(data['depth'], (nxu, ny, 1)).T
ZZv = np.tile(data['depth'], (nx, nyv, 1)).T
if not useInd is False: #>------------------------------------------
if ij == 'j':
slice = jslice
sshr = sshr[ij_ind, :]
hr = hr[ij_ind, :]
elif ij == 'i':
slice = islice
sshr = sshr[:, ij_ind]
hr = hr[:, ij_ind]
Zr, Zu, Zv, TEMP, SALT, U, V = [
slice(k, ij_ind) for k in [Zr, Zu, Zv, TEMP, SALT, U, V]
]
# -----------------------------------------------------------------<
if useInd: # then store distances for a possible bry file
dtype = Temp.dtype
distr = np.zeros((nz, ny, nx), dtype)
distu = np.zeros((nz, ny, nxu), dtype)
distv = np.zeros((nz, nyv, nx), dtype)
if not quiet: print 'vertical interpolation:'
if ij == 'j':
for j in range(ny):
if not quiet and (ny < 10 or (ny >= 10 and j % 10 == 0)):
print ' j=%3d of %3d' % (j, ny)
ind = ij_ind[j]
dr = np.tile(distance(xr[ind, :], yr[ind, :]), (nz, 1))
du = np.tile(distance(xu[ind, :], yu[ind, :]), (nz, 1))
Dr = np.tile(distance(xr[ind, :], yr[ind, :]), (NZ, 1))
Du = np.tile(distance(xu[ind, :], yu[ind, :]), (NZ, 1))
if useInd:
distr[:, j, :] = dr
distu[:, j, :] = du
Temp[:, j, :] = calc.griddata(Rdz * Dr,
ZZr[:, j, :],
TEMP[:, j, :],
Rdz * dr,
Zr[:, j, :],
extrap=True,
**interp_opts)
Salt[:, j, :] = calc.griddata(Rdz * Dr,
ZZr[:, j, :],
SALT[:, j, :],
Rdz * dr,
Zr[:, j, :],
extrap=True,
**interp_opts)
if 0 and j % 10 == 0:
print Dr.shape, ZZr[:, j, :].shape
import pylab as pl
pl.figure(1)
pl.clf()
pl.pcolormesh(Dr, ZZr[:, j, :], SALT[:, j, :])
pl.colorbar()
clim = pl.gci().get_clim()
pl.figure(2)
pl.clf()
pl.pcolormesh(dr, Zr[:, j, :], Salt[:, j, :])
pl.clim(clim)
pl.colorbar()
raw_input()
Uvel[:, j, :] = calc.griddata(Rdz * Du,
ZZu[:, j, :],
U[:, j, :],
Rdz * du,
Zu[:, j, :],
extrap=True,
**interp_opts)
if j < Vvel.shape[1]:
dv = np.tile(distance(xv[ind, :], yv[ind, :]), (nz, 1))
Dv = np.tile(distance(xv[ind, :], yv[ind, :]), (NZ, 1))
Vvel[:, j, :] = calc.griddata(Rdz * Dv,
ZZv[:, j, :],
V[:, j, :],
Rdz * dv,
Zv[:, j, :],
extrap=True,
**interp_opts)
if useInd:
distv[:, j, :] = dv
if np.any(np.isnan(Temp[:, j, :])): print 'found nan in temp', j
if np.any(np.isnan(Salt[:, j, :])): print 'found nan in salt', j
if np.any(np.isnan(Uvel[:, j, :])): print 'found nan in u', j
if j < Vvel.shape[1] and np.any(np.isnan(Vvel[:, j, :])):
print 'found nan in v', j
elif ij == 'i':
for i in range(nx):
if not quiet and (nx < 10 or (nx >= 10 and i % 10 == 0)):
print ' i=%3d of %3d' % (i, nx)
ind = ij_ind[i]
dr = np.tile(distance(xr[:, ind], yr[:, ind]), (nz, 1))
dv = np.tile(distance(xv[:, ind], yv[:, ind]), (nz, 1))
Dr = np.tile(distance(xr[:, ind], yr[:, ind]), (NZ, 1))
Dv = np.tile(distance(xv[:, ind], yv[:, ind]), (NZ, 1))
if useInd:
distr[:, :, i] = dr
distv[:, :, i] = dv
Temp[:, :, i] = calc.griddata(Rdz * Dr,
ZZr[:, :, i],
TEMP[:, :, i],
Rdz * dr,
Zr[:, :, i],
extrap=True,
**interp_opts)
Salt[:, :, i] = calc.griddata(Rdz * Dr,
ZZr[:, :, i],
SALT[:, :, i],
Rdz * dr,
Zr[:, :, i],
extrap=True,
**interp_opts)
Vvel[:, :, i] = calc.griddata(Rdz * Dv,
ZZv[:, :, i],
V[:, :, i],
Rdz * dv,
Zv[:, :, i],
extrap=True,
**interp_opts)
if i < Uvel.shape[2]:
du = np.tile(distance(xu[:, ind], yu[:, ind]), (nz, 1))
Du = np.tile(distance(xu[:, ind], yu[:, ind]), (NZ, 1))
Uvel[:, :, i] = calc.griddata(Rdz * Du,
ZZu[:, :, i],
U[:, :, i],
Rdz * du,
Zu[:, :, i],
extrap=True,
**interp_opts)
if useInd:
distu[:, :, i] = du
# uv bar:
if not quiet: print 'calc uvbar'
if useInd is False:
ubar, vbar = rt.uvbar(Uvel, Vvel, sshr, hr, sparams)
else: #>------------------------------------------------------------
sshu = calc.griddata(dlon,
dlat,
data['ssh'],
xu,
yu,
extrap=True,
**interp_opts)
sshv = calc.griddata(dlon,
dlat,
data['ssh'],
xv,
yv,
extrap=True,
**interp_opts)
if ij == 'j':
sshu = sshu[ij_ind, :]
sshv = sshv[ij_ind, :]
hu = hu[ij_ind, :]
hv = hv[ij_ind, :]
elif ij == 'i':
sshu = sshu[:, ij_ind]
sshv = sshv[:, ij_ind]
hu = hu[:, ij_ind]
hv = hv[:, ij_ind]
ubar = rt.barotropic(Uvel, sshu, hu, sparams)
vbar = rt.barotropic(Vvel, sshv, hv, sparams)
# -----------------------------------------------------------------<
Vars = cb.odict()
Vars['temp'] = Temp
Vars['salt'] = Salt
Vars['u'] = Uvel
Vars['v'] = Vvel
Vars['zeta'] = sshr
Vars['ubar'] = ubar
Vars['vbar'] = vbar
Vars['date'] = data['date']
if not useInd is False: #>------------------------------------------
Vars['depth'] = Zr
Vars['depthu'] = Zu
Vars['depthv'] = Zv
Vars['dist'] = distr
Vars['distu'] = distu
Vars['distv'] = distv
# -----------------------------------------------------------------<
if retHorizAux: return Vars, horizAux
else: return Vars
def data2roms(data, grd, sparams, **kargs):
'''
Interpolates data to roms 3D grid.
The dict data must contain the prognostic variables temp, salt, u,
v (3d) and ssh (zeta, 2d), as well as lon, lat (2d), depth (1d) and
time/date info: date (data date), date0 (reference date) and time
(difference between date and date0). The input data can be provided
by load_data.
Parameters
----------
data : dict with prognostic variables
grd : ROMS netcdf grid file
sparams : s-coordinates parameters, theta_s,theta_b, hc and NLevels
**kargs:
ij : axis for vertical interpolations (*i,j)
ij_ind : list of i or j index for vertical interpolation, all by
default (ij_ind=False)
horizAux : if True, the original data horizontally interpolated is
returned and can be used for next data2roms call with
this same karg
quiet : output messages flag (false by default)
'''
ij = 'j'
ij_ind = False
horizAux = False
quiet = False
if 'ij' in kargs.keys(): ij = kargs['ij']
if 'ij_ind' in kargs.keys(): ij_ind = kargs['ij_ind']
if 'horizAux' in kargs.keys(): horizAux = kargs['horizAux']
if 'quiet' in kargs.keys(): quiet = kargs['quiet']
if not quiet: print 'using grid %s' % grd
g = roms.Grid(grd)
xr, yr, hr, mr = g.vars('r')
xu, yu, hu, mu = g.vars('u')
xv, yv, hv, mv = g.vars('v')
ny, nx = hr.shape
nz = sparams[3]
NX = data['NX']
NY = data['NY']
NZ = data['NZ']
if not quiet: print 'calc s levels...'
sshr = calc.griddata(data['lon'],
data['lat'],
data['ssh'],
xr,
yr,
extrap=True)
Zr = g.s_levels(sparams, sshr, hr, 'r')
Zu = g.s_levels(sparams, sshr, hr, 'u')
Zv = g.s_levels(sparams, sshr, hr, 'v')
# interp horiz:
retHorizAux = horizAux is True
if horizAux in (True, False):
TEMP = np.ma.masked_all((NZ, ny, nx), data['temp'].dtype)
SALT = np.ma.masked_all((NZ, ny, nx), data['salt'].dtype)
U = np.ma.masked_all((NZ, ny, nx), data['u'].dtype)
V = np.ma.masked_all((NZ, ny, nx), data['v'].dtype)
if not quiet: print 'horizontal interpolation:'
for i in range(NZ):
if not quiet and i % 10 == 0: print ' lev %d of %d' % (i, NZ)
#import pylab
#pylab.figure()
#pylab.pcolormesh(data['lon'],data['lat'],data['temp'][i,...])
try:
TEMP[i, ...] = calc.griddata(data['lon'],
data['lat'],
data['temp'][i, ...],
xr,
yr,
extrap=True)
except:
pass
try:
SALT[i, ...] = calc.griddata(data['lon'],
data['lat'],
data['salt'][i, ...],
xr,
yr,
extrap=True)
except:
pass
try:
U[i, ...] = calc.griddata(data['lon'],
data['lat'],
data['u'][i, ...],
xr,
yr,
extrap=True)
except:
pass
try:
V[i, ...] = calc.griddata(data['lon'],
data['lat'],
data['v'][i, ...],
xr,
yr,
extrap=True)
except:
pass
# rotate U,V:
if not quiet: print 'rotating U,V to grid angle'
angle = g.use('angle') # rad
U, V = calc.rot2d(U, V, angle)
U = rt.rho2uvp3d(U, 'u')
V = rt.rho2uvp3d(V, 'v')
horizAux = {}
horizAux['TEMP'] = TEMP
horizAux['SALT'] = SALT
horizAux['U'] = U
horizAux['V'] = V
else:
TEMP = horizAux['TEMP']
SALT = horizAux['SALT']
U = horizAux['U']
V = horizAux['V']
# interp vert:
nxu = nx - 1
nyv = ny - 1
#> -----------------------------------------------------------------
useInd = not ij_ind is False
if ij_ind is False:
if ij == 'j': ij_ind = range(ny)
elif ij == 'i': ij_ind = range(nx)
else:
try:
iter(ij_ind)
except:
ij_ind = [ij_ind]
if ij == 'j': ny = nyv = len(ij_ind)
elif ij == 'i': nx = nxu = len(ij_ind)
# -----------------------------------------------------------------<
Temp = np.zeros((nz, ny, nx), data['temp'].dtype)
Salt = np.zeros((nz, ny, nx), data['salt'].dtype)
Uvel = np.zeros((nz, ny, nxu), data['u'].dtype)
Vvel = np.zeros((nz, nyv, nx), data['v'].dtype)
jslice = lambda x, ind: x[:, ind, :]
islice = lambda x, ind: x[:, :, ind]
ZZr = np.tile(data['depth'], (nx, ny, 1)).T
ZZu = np.tile(data['depth'], (nxu, ny, 1)).T
ZZv = np.tile(data['depth'], (nx, nyv, 1)).T
if not useInd is False: #>------------------------------------------
if ij == 'j':
slice = jslice
sshr = sshr[ij_ind, :]
hr = hr[ij_ind, :]
elif ij == 'i':
slice = islice
sshr = sshr[:, ij_ind]
hr = hr[:, ij_ind]
Zr, Zu, Zv, TEMP, SALT, U, V = [
slice(k, ij_ind) for k in [Zr, Zu, Zv, TEMP, SALT, U, V]
]
# -----------------------------------------------------------------<
if useInd: # then store distances for a possible bry file
dtype = Temp.dtype
distr = np.zeros((nz, ny, nx), dtype)
distu = np.zeros((nz, ny, nxu), dtype)
distv = np.zeros((nz, nyv, nx), dtype)
if not quiet: print 'vertical interpolation:'
if ij == 'j':
for j in range(ny):
if not quiet and (ny < 10 or (ny >= 10 and j % 10 == 0)):
print ' j=%3d of %3d' % (j, ny)
ind = ij_ind[j]
dr = np.tile(calc.distance(xr[ind, :], yr[ind, :]), (nz, 1))
du = np.tile(calc.distance(xu[ind, :], yu[ind, :]), (nz, 1))
Dr = np.tile(calc.distance(xr[ind, :], yr[ind, :]), (NZ, 1))
Du = np.tile(calc.distance(xu[ind, :], yu[ind, :]), (NZ, 1))
if useInd:
distr[:, j, :] = dr
distu[:, j, :] = du
Temp[:, j, :] = calc.griddata(Dr,
ZZr[:, j, :],
TEMP[:, j, :],
dr,
Zr[:, j, :],
extrap=True)
Salt[:, j, :] = calc.griddata(Dr,
ZZr[:, j, :],
SALT[:, j, :],
dr,
Zr[:, j, :],
extrap=True)
if 0 and j % 10 == 0:
print Dr.shape, ZZr[:, j, :].shape
import pylab as pl
pl.figure(1)
pl.clf()
pl.pcolormesh(Dr, ZZr[:, j, :], SALT[:, j, :])
pl.colorbar()
clim = pl.gci().get_clim()
pl.figure(2)
pl.clf()
pl.pcolormesh(dr, Zr[:, j, :], Salt[:, j, :])
pl.clim(clim)
pl.colorbar()
raw_input()
Uvel[:, j, :] = calc.griddata(Du,
ZZu[:, j, :],
U[:, j, :],
du,
Zu[:, j, :],
extrap=True)
if j < Vvel.shape[1]:
dv = np.tile(calc.distance(xv[ind, :], yv[ind, :]), (nz, 1))
Dv = np.tile(calc.distance(xv[ind, :], yv[ind, :]), (NZ, 1))
Vvel[:, j, :] = calc.griddata(Dv,
ZZv[:, j, :],
V[:, j, :],
dv,
Zv[:, j, :],
extrap=True)
if useInd:
distv[:, j, :] = dv
if np.any(np.isnan(Temp[:, j, :])): print 'found nan in temp', j
if np.any(np.isnan(Salt[:, j, :])): print 'found nan in salt', j
if np.any(np.isnan(Uvel[:, j, :])): print 'found nan in u', j
if j < Vvel.shape[1] and np.any(np.isnan(Vvel[:, j, :])):
print 'found nan in v', j
elif ij == 'i':
for i in range(nx):
if not quiet and (nx < 10 or (nx >= 10 and i % 10 == 0)):
print ' i=%3d of %3d' % (i, nx)
ind = ij_ind[i]
dr = np.tile(calc.distance(xr[:, ind], yr[:, ind]), (nz, 1))
dv = np.tile(calc.distance(xv[:, ind], yv[:, ind]), (nz, 1))
Dr = np.tile(calc.distance(xr[:, ind], yr[:, ind]), (NZ, 1))
Dv = np.tile(calc.distance(xv[:, ind], yv[:, ind]), (NZ, 1))
if useInd:
distr[:, :, i] = dr
distv[:, :, i] = dv
Temp[:, :, i] = calc.griddata(Dr,
ZZr[:, :, i],
TEMP[:, :, i],
dr,
Zr[:, :, i],
extrap=True)
Salt[:, :, i] = calc.griddata(Dr,
ZZr[:, :, i],
SALT[:, :, i],
dr,
Zr[:, :, i],
extrap=True)
Vvel[:, :, i] = calc.griddata(Dv,
ZZv[:, :, i],
V[:, :, i],
dv,
Zv[:, :, i],
extrap=True)
if i < Uvel.shape[2]:
du = np.tile(calc.distance(xu[:, ind], yu[:, ind]), (nz, 1))
Du = np.tile(calc.distance(xu[:, ind], yu[:, ind]), (NZ, 1))
Uvel[:, :, i] = calc.griddata(Du,
ZZu[:, :, i],
U[:, :, i],
du,
Zu[:, :, i],
extrap=True)
if useInd:
distu[:, :, i] = du
# uv bar:
if not quiet: print 'calc uvbar'
if useInd is False:
ubar, vbar = rt.uvbar(Uvel, Vvel, sshr, hr, sparams)
else: #>------------------------------------------------------------
sshu = calc.griddata(data['lon'],
data['lat'],
data['ssh'],
xu,
yu,
extrap=True)
sshv = calc.griddata(data['lon'],
data['lat'],
data['ssh'],
xv,
yv,
extrap=True)
if ij == 'j':
sshu = sshu[ij_ind, :]
sshv = sshv[ij_ind, :]
hu = hu[ij_ind, :]
hv = hv[ij_ind, :]
elif ij == 'i':
sshu = sshu[:, ij_ind]
sshv = sshv[:, ij_ind]
hu = hu[:, ij_ind]
hv = hv[:, ij_ind]
ubar = rt.barotropic(Uvel, sshu, hu, sparams)
vbar = rt.barotropic(Vvel, sshv, hv, sparams)
# -----------------------------------------------------------------<
Vars = cb.odict()
Vars['temp'] = Temp
Vars['salt'] = Salt
Vars['u'] = Uvel
Vars['v'] = Vvel
Vars['zeta'] = sshr
Vars['ubar'] = ubar
Vars['vbar'] = vbar
Vars['date'] = data['date']
if not useInd is False: #>------------------------------------------
Vars['depth'] = Zr
Vars['depthu'] = Zu
Vars['depthv'] = Zv
Vars['dist'] = distr
Vars['distu'] = distu
Vars['distv'] = distv
# -----------------------------------------------------------------<
if retHorizAux: return Vars, horizAux
else: return Vars
def update_wind_blended2(fname, datapaths, **kargs):
"""
In days without blended data will try to use quikscat data
"""
from okean.datasets import quikscat
from okean.datasets import blended_wind
a = blended_wind.WINDData(datapaths[0])
b = quikscat.WINDData(datapaths[1])
time = netcdf.nctime(fname, "time")
date0 = dts.next_date(time[0], -1)
date1 = dts.next_date(time[-1], +2)
data = a.data(date0, date1)
# limit are... otherwise, quikscat interp will be very slow!
grd = netcdf.fatt(fname, "grd_file")
import os
if not os.path.isfile(grd):
grd = kargs["grd"]
cond, inds = rt.grid_vicinity(grd, data["x"], data["y"], margin=5, rect=True, retinds=True)
i1, i2, j1, j2 = inds
for d in data.keys():
if d == "x":
data[d] = data[d][i1:i2]
elif d == "y":
data[d] = data[d][j1:j2]
else:
data[d] = data[d][j1:j2, i1:i2]
# check for missing days:
time0 = data.keys()
x0 = data["x"]
y0 = data["y"]
x0, y0 = np.meshgrid(x0, y0)
time0.remove("x")
time0.remove("y")
out = cb.odict()
out["x"] = x0
out["y"] = y0
info = ""
qs_ij_limits_done = False
for d in dts.drange(date0, date1):
found = 0
for t in time0:
if (t.year, t.month, t.day) == (d.year, d.month, d.day):
print "==> blended : ", t
out[t] = data[t]
found = 1
if not found: # use quikscat:
print "==> quikscat : ", d.strftime("%Y-%m-%d")
tmp = b.data(d, dts.next_date(d))
if not tmp.has_key("x"):
continue
x, y = tmp["x"], tmp["y"]
x, y = np.meshgrid(x, y)
# reduce qs data:
if not qs_ij_limits_done:
i1, i2, j1, j2 = calc.ij_limits(x, y, [x0.min(), x0.max()], [y0.min(), y0.max()])
qs_ij_limits_done = True
x = x[j1:j2, i1:i2]
y = y[j1:j2, i1:i2]
tmp[tmp.keys()[0]] = tmp[tmp.keys()[0]][j1:j2, i1:i2]
print " griddata u"
u = calc.griddata(x, y, tmp[tmp.keys()[0]].real, x0, y0)
print " griddata v"
v = calc.griddata(x, y, tmp[tmp.keys()[0]].imag, x0, y0)
out[tmp.keys()[0]] = u + 1.0j * v
info += "#" + d.strftime("%Y%m%d")
new_wind_info = "blended+quikscat at days: " + info
update_wind(fname, out, new_wind_info, **kargs)
def data2roms(data,grd,sparams,**kargs):
'''
Interpolates data to roms 3D grid.
The dict data must contain the prognostic variables temp, salt, u,
v (3d) and ssh (zeta, 2d), as well as lon, lat (2d), depth (1d) and
time/date info: date (data date), date0 (reference date) and time
(difference between date and date0). The input data can be provided
by load_data.
Parameters
----------
data : dict with prognostic variables
grd : ROMS netcdf grid file
sparams : s-coordinates parameters, theta_s,theta_b, hc and NLevels
**kargs:
ij : axis for vertical interpolations (*i,j)
ij_ind : list of i or j index for vertical interpolation, all by
default (ij_ind=False)
horizAux : if True, the original data horizontally interpolated is
returned and can be used for next data2roms call with
this same karg
quiet : output messages flag (false by default)
'''
ij='j'
ij_ind=False
horizAux=False
quiet=False
if 'ij' in kargs.keys(): ij = kargs['ij']
if 'ij_ind' in kargs.keys(): ij_ind = kargs['ij_ind']
if 'horizAux' in kargs.keys(): horizAux = kargs['horizAux']
if 'quiet' in kargs.keys(): quiet = kargs['quiet']
if not quiet: print 'using grid %s' % grd
g=roms.Grid(grd)
xr,yr,hr,mr=g.vars('r')
xu,yu,hu,mu=g.vars('u')
xv,yv,hv,mv=g.vars('v')
ny,nx=hr.shape
nz=sparams[3]
NX=data['NX']
NY=data['NY']
NZ=data['NZ']
if not quiet: print 'calc s levels...'
sshr=calc.griddata(data['lon'],data['lat'],data['ssh'],xr,yr,extrap=True)
Zr = g.s_levels(sparams,sshr,hr,'r')
Zu = g.s_levels(sparams,sshr,hr,'u')
Zv = g.s_levels(sparams,sshr,hr,'v')
# interp horiz:
retHorizAux=horizAux is True
if horizAux in (True,False):
TEMP = np.ma.masked_all((NZ,ny,nx),data['temp'].dtype)
SALT = np.ma.masked_all((NZ,ny,nx),data['salt'].dtype)
U = np.ma.masked_all((NZ,ny,nx),data['u'].dtype)
V = np.ma.masked_all((NZ,ny,nx),data['v'].dtype)
if not quiet: print 'horizontal interpolation:'
for i in range(NZ):
if not quiet and i%10==0: print ' lev %d of %d' % (i,NZ)
#import pylab
#pylab.figure()
#pylab.pcolormesh(data['lon'],data['lat'],data['temp'][i,...])
try: TEMP[i,...] = calc.griddata(data['lon'],data['lat'],data['temp'][i,...],xr,yr,extrap=True)
except: pass
try: SALT[i,...] = calc.griddata(data['lon'],data['lat'],data['salt'][i,...],xr,yr,extrap=True)
except: pass
try: U[i,...] = calc.griddata(data['lon'],data['lat'],data['u'][i,...],xr,yr,extrap=True)
except: pass
try: V[i,...] = calc.griddata(data['lon'],data['lat'],data['v'][i,...],xr,yr,extrap=True)
except: pass
# rotate U,V:
if not quiet: print 'rotating U,V to grid angle'
angle=g.use('angle') # rad
U,V=calc.rot2d(U,V,angle)
U=rt.rho2uvp3d(U,'u')
V=rt.rho2uvp3d(V,'v')
horizAux={}
horizAux['TEMP'] = TEMP
horizAux['SALT'] = SALT
horizAux['U'] = U
horizAux['V'] = V
else:
TEMP = horizAux['TEMP']
SALT = horizAux['SALT']
U = horizAux['U']
V = horizAux['V']
# interp vert:
nxu=nx-1
nyv=ny-1
#> -----------------------------------------------------------------
useInd=not ij_ind is False
if ij_ind is False:
if ij=='j': ij_ind=range(ny)
elif ij=='i': ij_ind=range(nx)
else:
try: iter(ij_ind)
except: ij_ind=[ij_ind]
if ij=='j': ny=nyv=len(ij_ind)
elif ij=='i': nx=nxu=len(ij_ind)
# -----------------------------------------------------------------<
Temp = np.zeros((nz,ny ,nx ),data['temp'].dtype)
Salt = np.zeros((nz,ny ,nx ),data['salt'].dtype)
Uvel = np.zeros((nz,ny ,nxu),data['u'].dtype)
Vvel = np.zeros((nz,nyv,nx ),data['v'].dtype)
jslice=lambda x,ind: x[:,ind,:]
islice=lambda x,ind: x[:,:,ind]
ZZr = np.tile(data['depth'],(nx,ny,1)).T
ZZu = np.tile(data['depth'],(nxu,ny,1)).T
ZZv = np.tile(data['depth'],(nx,nyv,1)).T
if not useInd is False: #>------------------------------------------
if ij=='j':
slice=jslice
sshr=sshr[ij_ind,:]
hr =hr[ij_ind,:]
elif ij=='i':
slice=islice
sshr=sshr[:,ij_ind]
hr =hr[:,ij_ind]
Zr,Zu,Zv,TEMP,SALT,U,V=[slice(k,ij_ind) for k in [Zr,Zu,Zv,TEMP,SALT,U,V]]
# -----------------------------------------------------------------<
if useInd: # then store distances for a possible bry file
dtype=Temp.dtype
distr=np.zeros((nz,ny, nx ),dtype)
distu=np.zeros((nz,ny, nxu),dtype)
distv=np.zeros((nz,nyv,nx ),dtype)
if not quiet: print 'vertical interpolation:'
if ij=='j':
for j in range(ny):
if not quiet and (ny<10 or (ny>=10 and j%10==0)): print ' j=%3d of %3d' % (j,ny)
ind=ij_ind[j]
dr=np.tile(calc.distance(xr[ind,:],yr[ind,:]),(nz,1))
du=np.tile(calc.distance(xu[ind,:],yu[ind,:]),(nz,1))
Dr=np.tile(calc.distance(xr[ind,:],yr[ind,:]),(NZ,1))
Du=np.tile(calc.distance(xu[ind,:],yu[ind,:]),(NZ,1))
if useInd:
distr[:,j,:]=dr;
distu[:,j,:]=du;
Temp[:,j,:] = calc.griddata(Dr,ZZr[:,j,:],TEMP[:,j,:],dr,Zr[:,j,:],extrap=True)
Salt[:,j,:] = calc.griddata(Dr,ZZr[:,j,:],SALT[:,j,:],dr,Zr[:,j,:],extrap=True)
if 0 and j%10==0:
print Dr.shape, ZZr[:,j,:].shape
import pylab as pl
pl.figure(1)
pl.clf()
pl.pcolormesh(Dr,ZZr[:,j,:],SALT[:,j,:])
pl.colorbar()
clim=pl.gci().get_clim()
pl.figure(2)
pl.clf()
pl.pcolormesh(dr,Zr[:,j,:],Salt[:,j,:])
pl.clim(clim)
pl.colorbar()
raw_input()
Uvel[:,j,:] = calc.griddata(Du,ZZu[:,j,:],U[:,j,:], du,Zu[:,j,:],extrap=True)
if j<Vvel.shape[1]:
dv=np.tile(calc.distance(xv[ind,:],yv[ind,:]),(nz,1))
Dv=np.tile(calc.distance(xv[ind,:],yv[ind,:]),(NZ,1))
Vvel[:,j,:] = calc.griddata(Dv,ZZv[:,j,:],V[:,j,:], dv,Zv[:,j,:],extrap=True)
if useInd:
distv[:,j,:]=dv
if np.any(np.isnan(Temp[:,j,:])): print 'found nan in temp',j
if np.any(np.isnan(Salt[:,j,:])): print 'found nan in salt',j
if np.any(np.isnan(Uvel[:,j,:])): print 'found nan in u',j
if j<Vvel.shape[1] and np.any(np.isnan(Vvel[:,j,:])): print 'found nan in v',j
elif ij=='i':
for i in range(nx):
if not quiet and (nx<10 or (nx>=10 and i%10==0)): print ' i=%3d of %3d' % (i,nx)
ind=ij_ind[i]
dr=np.tile(calc.distance(xr[:,ind],yr[:,ind]),(nz,1))
dv=np.tile(calc.distance(xv[:,ind],yv[:,ind]),(nz,1))
Dr=np.tile(calc.distance(xr[:,ind],yr[:,ind]),(NZ,1))
Dv=np.tile(calc.distance(xv[:,ind],yv[:,ind]),(NZ,1))
if useInd:
distr[:,:,i]=dr;
distv[:,:,i]=dv;
Temp[:,:,i] = calc.griddata(Dr,ZZr[:,:,i],TEMP[:,:,i],dr,Zr[:,:,i],extrap=True)
Salt[:,:,i] = calc.griddata(Dr,ZZr[:,:,i],SALT[:,:,i],dr,Zr[:,:,i],extrap=True)
Vvel[:,:,i] = calc.griddata(Dv,ZZv[:,:,i],V[:,:,i], dv,Zv[:,:,i],extrap=True)
if i<Uvel.shape[2]:
du=np.tile(calc.distance(xu[:,ind],yu[:,ind]),(nz,1))
Du=np.tile(calc.distance(xu[:,ind],yu[:,ind]),(NZ,1))
Uvel[:,:,i] = calc.griddata(Du,ZZu[:,:,i],U[:,:,i], du,Zu[:,:,i],extrap=True)
if useInd:
distu[:,:,i]=du
# uv bar:
if not quiet: print 'calc uvbar'
if useInd is False:
ubar,vbar=rt.uvbar(Uvel,Vvel,sshr,hr,sparams)
else: #>------------------------------------------------------------
sshu=calc.griddata(data['lon'],data['lat'],data['ssh'],xu,yu,extrap=True)
sshv=calc.griddata(data['lon'],data['lat'],data['ssh'],xv,yv,extrap=True)
if ij=='j':
sshu=sshu[ij_ind,:]
sshv=sshv[ij_ind,:]
hu =hu[ij_ind,:]
hv =hv[ij_ind,:]
elif ij=='i':
sshu=sshu[:,ij_ind]
sshv=sshv[:,ij_ind]
hu =hu[:,ij_ind]
hv =hv[:,ij_ind]
ubar=rt.barotropic(Uvel,sshu,hu,sparams)
vbar=rt.barotropic(Vvel,sshv,hv,sparams)
# -----------------------------------------------------------------<
Vars=cb.odict()
Vars['temp'] = Temp
Vars['salt'] = Salt
Vars['u'] = Uvel
Vars['v'] = Vvel
Vars['zeta'] = sshr
Vars['ubar'] = ubar
Vars['vbar'] = vbar
Vars['date'] = data['date']
if not useInd is False: #>------------------------------------------
Vars['depth'] = Zr
Vars['depthu'] = Zu
Vars['depthv'] = Zv
Vars['dist'] = distr
Vars['distu'] = distu
Vars['distv'] = distv
# -----------------------------------------------------------------<
if retHorizAux: return Vars, horizAux
else: return Vars
def add_var(self, varname, vartype, dimnames, **kargs):
'''
Creates netcdf variable
Ex: add_var('temp','float32',('lon','lat','z'))
About compression:
Compression kargs options (available for netcdf4 interface)
zlib, default True, turn on compression
lsd (least_significant_digit), default is None: lossy
compression with lsd precision
complevel, 1..9, copression level, default 4
About vartype:
i) netcdf type names:
for netcdf3:
byte,char,short,int,float,double
for netcdf4:
+ int64, ushort, uint,uint64, string
ii) Numeric type code: fd1silbwuc
iii) Numpy dtype or type name
kargs are required when vartype is a netcdf type name
and the interface is scientific, see type_2numeric
and when interface is netcdf4, see type_2dtype
No need for kargs when:
1. interface = pycdf
2. interface is scientific and vartype is numpy type name or
dtype, or vartipe is numeric typecode
3. interface is netcdf4 and vartype is numpy type name or dtype
'''
if self._interface == 'pycdf':
vartype = nctypes.type_2pycdf(vartype)
self._nc.def_var(varname, vartype, dimnames)
elif self._interface == 'scientific':
vartype = type_2numeric(vartype, **kargs)
self._nc.createVariable(varname, vartype, dimnames)
elif self._interface == 'netcdf4':
vartype = nctypes.type_2numpy(vartype, **kargs)
# NOTA: ocorre erro ao criar ex int64 em ncver 3. Pode fazer sentido
# converter para int? ie, qd o tipo nao e suportado pela ncver!?
zlib = kargs.get('zlib', True)
lsd = kargs.get('lsd', None)
complevel = kargs.get('complevel', 4)
fill_value = kargs.get('fill_value', None)
self._nc.createVariable(varname,
vartype,
dimnames,
zlib=zlib,
complevel=complevel,
least_significant_digit=lsd,
fill_value=fill_value)
newvar = Pyncvar(self, varname)
# update self.vars and self.varnames:
self.vars[varname] = newvar
# update ncdump_info:
if not self._ncdump_info is False:
self._ncdump_info['variables'][varname] = cbt.odict()
return newvar
def add_var(self,varname,vartype,dimnames,**kargs):
'''
Creates netcdf variable
Ex: add_var('temp','float32',('lon','lat','z'))
About compression:
Compression kargs options (available for netcdf4 interface)
gzip, default True: gzip compression
lsd (least_significant_digit), default is None: lossy
compression with lsd precision
About vartype:
i) netcdf type names:
for netcdf3:
byte,char,short,int,float,double
for netcdf4:
+ int64, ushort, uint,uint64, string
ii) Numeric type code: fd1silbwuc
iii) Numpy dtype or type name
kargs are required when vartype is a netcdf type name
and the interface is scientific, see type_2numeric
and when interface is netcdf4, see type_2dtype
No need for kargs when:
1. interface = pycdf
2. interface is scientific and vartype is numpy type name or
dtype, or vartipe is numeric typecode
3. interface is netcdf4 and vartype is numpy type name or dtype
'''
if self._interface=='pycdf':
vartype=nctypes.type_2pycdf(vartype)
self._nc.def_var(varname,vartype,dimnames)
elif self._interface=='scientific':
vartype=type_2numeric(vartype,**kargs)
self._nc.createVariable(varname,vartype,dimnames)
elif self._interface=='netcdf4':
vartype=nctypes.type_2numpy(vartype,**kargs)
# NOTA: ocorre erro ao criar ex int64 em ncver 3. Pode fazer sentido
# converter para int? ie, qd o tipo nao e suportado pela ncver!?
zlib=True
lsd=None
fill_value=None
if 'gzip' in kargs: zlib = kargs['gzip']
if 'lsd' in kargs: lsd = kargs['lsd']
if 'fill_value' in kargs: fill_value = kargs['fill_value']
self._nc.createVariable(varname,vartype,dimnames,zlib=zlib,
least_significant_digit=lsd,fill_value=fill_value)
newvar=Pyncvar(self,varname)
# update self.vars and self.varnames:
self.vars[varname]=newvar
# update ncdump_info:
if not self._ncdump_info is False:
self._ncdump_info['variables'][varname]=cbt.odict()
return newvar