""" createNSIDCclimos.py

July 10, 2014

Need to generate and SST and SIT dataset that is consistent

with NSIDC sea ice concentration. We are using the climos

from HadISST (and CCCma for thickness but used w/ HadISST).

"""

import copy as cp

import numpy as np

import numpy.ma as ma

import platform as platform

import cccmaplots as cplt

import cccmaNC as cnc

plt.close("all")

plt.ion()

testplots=True

testsicplots=False # just the sea-ice concentration plots

writefiles=True

dossts=False # else do SIT

# # # ########## Read NC data ###############

plat = platform.system()

if plat == 'Darwin': # means I'm on my mac

basepath = '/Volumes/MyPassport1TB/DATA/CanSISE/'

basepath2 = '/Users/kelly/CCCma/CanSISE/BoundaryConditionFiles/'

else: # on linux workstation in Vic

basepath = '/home/rkm/work/DATA/'

basepath2 = '/home/rkm/work/BCs/'

# read in HadISST SST & Thickness (and SIC for comparison)

# read in NSIDC SIC

timeper='2002-2011'

timeperc='1979-1989'

# NSIDC first

dset = 'NSIDC'

nsicfnamep = basepath2 + dset + '/nsidc_bt_128x64_1978m11_2011m12_sicn_' + timeper + 'climo.nc'

nsicfnamec = basepath2 + dset + '/nsidc_bt_128x64_1978m11_2011m12_sicn_' + timeperc + 'climo.nc'

nsicp = cnc.getNCvar(nsicfnamep,'SICN')

nsicc = cnc.getNCvar(nsicfnamec,'SICN')

# HadISST

dset='HadISST'

sstfnamep = basepath2 + dset + '/hadisst1.1_bc_128_64_1870_2013m03_gt_' + timeper + 'climo.nc'

sstfnamec = basepath2 + dset + '/hadisst1.1_bc_128_64_1870_2013m03_gt_' + timeperc + 'climo.nc'

sitfnamep = basepath2 + dset + '/hadisst1.1_bc_128_64_1870_2013m03_sic_' + timeper + 'climo.nc'

sitfnamec = basepath2 + dset + '/hadisst1.1_bc_128_64_1870_2013m03_sic_' + timeperc + 'climo.nc'

hsicfnamep = basepath2 + dset + '/hadisst1.1_bc_128_64_1870_2013m03_sicn_' + timeper + 'climo.nc'

hsicfnamec = basepath2 + dset + '/hadisst1.1_bc_128_64_1870_2013m03_sicn_' + timeperc + 'climo.nc'

hsicp = cnc.getNCvar(hsicfnamep,'SICN')

hsicc = cnc.getNCvar(hsicfnamec,'SICN')

cmap='red2blue_w20'

if testsicplots:

# compare sic

# looks like NSIDC has more ice in the control time period and ~less in the pert. so, greater trend

cplt.map_allmonths(nsicp-hsicp,lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='pert sic nsidc-hadisst',latlim=45)

cplt.map_allmonths(nsicc-hsicc,lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='ctl sic nsidc-hadisst',latlim=45)

cplt.map_allmonths((nsicp-nsicc),lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='pert-ctl sic nsidc',latlim=45)

cplt.map_allmonths((hsicp-hsicc),lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='pert-ctl sic hadisst',latlim=45)

cplt.map_allmonths((nsicp-nsicc) - (hsicp-hsicc),lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='pert-ctl sic nsidc-hadisst',latlim=45)

if dossts:

# ########### do SST ####################################

hsstp = cnc.getNCvar(sstfnamep,'GT')

hsstc = cnc.getNCvar(sstfnamec,'GT')

lat = cnc.getNCvar(sstfnamec,'lat')

lon = cnc.getNCvar(sstfnamec,'lon')

timefld = cnc.getNCvar(sstfnamec,'time')

# basically take these hadisst sst's and mod for NSIDC

# 1. if NSIDC SIC >=0.15 (ice) and HadISST SIC < 0.15 (no ice)

# SST should be set to freezing

# 2. if NSIDC SIC >=0.15 (ice) and HadISST SIC >= 0.15 (ice)

# use HadISST SST as is

# 3. if NSIDC SIC < 0.15 (no ice) and HadISST SIC >= 0.15 (ice)

# SST should be set to freezing if not already (easiest solution. will underestimate SST)

# 4. if NSIDC SIC < 0.15 (no ice) and HadISST SIC < 0.15 (no ice)

# use HadISST SST as is

#

# So, only change the SST field for case 1. and 3. Set to freezing for both

nsstp = cp.copy(hsstp)

nsstc = cp.copy(hsstc)

# case 1.

nsstp[np.logical_and(nsicp>=0.15, hsicp<0.15)] = 271.2

nsstc[np.logical_and(nsicc>=0.15, hsicc<0.15)] = 271.2

# case 3.

nsstp[np.logical_and(nsicp<0.15, hsicp>=0.15)] = 271.2

nsstc[np.logical_and(nsicc<0.15, hsicc>=0.15)] = 271.2

if testplots:

# compare original hadisst to new nsidcsst: each has very small regions of adjusted SST.

cplt.map_allmonths(nsstp-hsstp,lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='pert sst nsidc-hadisst',latlim=45)

cplt.map_allmonths(nsstc-hsstc,lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='ctl sst nsidc-hadisst',latlim=45)

# set up SST data for writing out

outflds = {}

outflds[timeperc] = nsstc

outflds[timeper] = nsstp

bcfield = 'GT'

bcunits = 'K'

bcdescrip = 'Ground Temperature'

printfld = 'gt'

else: # do sea ice thickness

deni = 913. # kg/m3

hsitp = cnc.getNCvar(sitfnamep,'SIC')

hsitc = cnc.getNCvar(sitfnamec,'SIC')

lat = cnc.getNCvar(sitfnamec,'lat')

lon = cnc.getNCvar(sitfnamec,'lon')

timefld = cnc.getNCvar(sitfnamec,'time')

""" Email from Mike Lazare 6/24/2014 entitled Re: sea ice thickness boundary condition

In general, a point is considered water only the sea-ice

concentration reaches 0.15 (15%). For concentrations less than this, any

sea-ice mass

is ignored essentially, even if it is non-zero. Typically, a

concentration of 0.15 maps to a sea-ice mass of 45 Kg/m2 (5 cms of ice),

so it is likely

that there will be non-zero mass present. The BC forcing data typically

tries to do consistency checks for mass vs concentration, so when

the concentration exceeds 0.15 the associated mass will typically be >

45 Kg/m2, and the model will interpolate between the previous value

(less than 45 Kg/m2) and the target (value >45 Kg/m2).

All this is a roundabout (and somewhat rambling) confirmation of your

hypothesis that it ignores the sea ice thickness. The ice present does

not contribute to albedo until the concentration exceeds 0.15;

thereafter both snow/ice and open water (ie leads) factor in.

"""

# basically take these hadisst thickness and mod for NSIDC

# Based on Mike Lazare's email, probably don't have to care too much about the

# tolerance for checking SIC b/c the model considers the grid cell to

# be open water only if <0.15 (and it doesn't count for albedo either)

# So actually just switch to 0.15, similar to SST tests above.

# 1. if NSIDC SIC >0 (ice) and HadISST SIC = 0 (no ice)

# SIT should be what?? 2m? No, 45kg/m3 (~5cm thick) based on the original

# way the hadisst thickness was created from some model climatology

# script is: to_gt_sic_sicn_job

# 2. if NSIDC SIC >0 (ice) and HadISST SIC > 0 (ice)

# use HadISST SIT as is

# 3. if NSIDC SIC = 0 (no ice) and HadISST SIC > 0 (ice)

# SIT should be set to zero

# 4. if NSIDC SIC = 0 (no ice) and HadISST SIC = 0 (no ice)

# use HadISST SIT as is (0)

#

# So, only change the SIT field for case 1. and 3.

# Should I use a tolerance for the zero check? yes, but what. @@

nsitp = cp.copy(hsitp)

nsitc = cp.copy(hsitc)

# case 1.

nsitp[np.logical_and(nsicp>=0.15, hsicp<0.15)] = 45 # ~5cm thickness

nsitc[np.logical_and(nsicc>=0.15, hsicc<0.15)] = 45 # ~5cm thickness

# case 3.

nsitp[np.logical_and(nsicp<0.15, hsicp>=0.15)] = 0

nsitc[np.logical_and(nsicc<0.15, hsicc>=0.15)] = 0

if testplots:

# compare original hadisst SIT to new nsidcsit:

# if blue, nsidc has greater thickness than hadisst, which is where was set to 2m.

cplt.map_allmonths(nsitp-hsitp,lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='pert sit nsidc-hadisst',latlim=45)

cplt.map_allmonths(nsitc-hsitc,lat,lon,cmin=-.1,cmax=.1,cmap=cmap,type='nh',lmask=1,title='ctl sit nsidc-hadisst',latlim=45)

# set up SIT data for writing out

outflds = {}

outflds[timeperc] = nsitc

outflds[timeper] = nsitp

bcfield = 'SIC'

bcunits = 'kg/m2'

bcdescrip = 'sea ice thickness*density (kg/m2)'

printfld = 'sic'

############### Write the NetCDF file ####################

from netCDF4 import Dataset

if writefiles:

for thetimeper in outflds.keys():

print thetimeper

fldout = outflds[thetimeper]

# for some reason the long filename doesn't work, but a short filename does.

# some talk about that here, but I don't understand why?

# https://code.google.com/p/netcdf4-python/issues/detail?id=141

#outfile = 'nsidc_bt_128x64_1978m11_2011m12_' + printfld + '_' + thetimeper + 'climo.nc'

outfile = 'nsidc_' + printfld + '_' + thetimeper + '.nc'

outnc = Dataset(outfile,'w')

# create the dimensions

# not sure why these need python variables, they are never used@@

# I guess they are the keys in the dict of Dimensions...

outtime = outnc.createDimension('time', None)

outlat = outnc.createDimension('lat',len(lat))

outlon = outnc.createDimension('lon',len(lon))

# create variables

outtimes = outnc.createVariable('time','f8',('time',)) # f8 and d are the same dtype

outlats = outnc.createVariable('lat','d',('lat',))

outlons = outnc.createVariable('lon','d',('lon',))

outfld = outnc.createVariable(bcfield,'f4',('time','lat','lon',),fill_value=1.0e38)

# add attributes to variables

outfld.units = bcunits

outfld.long_name = bcdescrip

outfld.grid_type = 'gaussian'

outfld.missing_value = 1.0e38

outtimes.long_name = 'time'

outtimes.units = 'days since 1850-01-01 00:00:00'

outtimes.calendar = '365_day'

outtimes.standard_name = 'time'

outlats.units = 'degrees_north'

outlats.long_name = 'Latitude'

outlats.standard_name = 'latitude'

outlats.axis = 'Y'

outlons.units = 'degrees_east'

outlons.long_name = 'Longitude'

outlons.standard_name = 'longitude'

outlons.axis = 'X'

# global attributes

import time

outnc.title = 'original ' + bcfield + ' from hadisst1.1_bc_128_64_1870_2013m03_' + \

printfld + '_' + thetimeper + 'climo.nc, adjusted for NSIDC sea ice conc'

outnc.creation_date = time.ctime(time.time())

outnc.created_by = 'Kelly E. McCusker, CCCma / U. of Victoria'

# set the data to the variables: important to have [:]!

outtimes[:] = timefld

outlats[:] = lat

outlons[:] = lon

outfld[:] = fldout

outnc.close()