Looking at high-frequency oscillation in nowcast green ssh at Point Atkinson 15 minute files.
In [1]:
import netCDF4 as nc
import matplotlib.pyplot as plt
import numpy as np
import os
import datetime
from salishsea_tools import nc_tools
%matplotlib inline
A sample nowcast-green output¶
In [2]:
f = nc.Dataset('/results/SalishSea/nowcast-green/08mar16/PointAtkinson.nc')
ssh = f.variables['sossheig'][:,0,0]
ts =nc_tools.timestamp(f,np.arange(ssh.shape[0]))
dates = [d.datetime for d in ts]
In [3]:
plt.plot(dates,ssh)
plt.title(dates[0].strftime('%Y-%m-%d'))
plt.ylabel('ssh [m]')
Out[3]:
<matplotlib.text.Text at 0x7efe427a3e48>
In [4]:
ssh[-10:]
Out[4]:
array([-0.17877117, 0.06571084, 0.06571084, 0.32868201, 0.32868201,
0.59840339, 0.59840339, 0.85108906, 0.85108906, 1.05518746], dtype=float32)
- Saw-tooth like behaviour and repeated values.
- Could this be because 15 minutes is not divisible by out 40 second time step?
In [5]:
rundir = '/ocean/nsoontie/MEOPAR/SalishSea/results/mixing_paper/holl_jul/'
times = [10,15,20]
fig,ax = plt.subplots(1,1,figsize=(15,5))
for time in times:
f = nc.Dataset(os.path.join(rundir,'PointAtkinson_{}.nc'.format(time)))
ssh = f.variables['sossheig'][:,0,0]
ts =nc_tools.timestamp(f,np.arange(ssh.shape[0]))
dates = [d.datetime for d in ts]
ax.plot(dates, ssh,label ='{} minute output'.format(time) )
ax.legend(loc=0)
ax.set_ylabel('ssh [m]')
print('{} minute output'.format(time))
print('last 8 ssh: ', ssh[-8:])
ax.set_xlim([datetime.datetime(2015,7,11),datetime.datetime(2015,7,12)])
10 minute output last 8 ssh: [ -8.49195659e-01 -7.13157475e-01 -5.74840605e-01 -4.34347630e-01 -2.90346920e-01 -1.45057648e-01 8.20399509e-05 1.43809989e-01] 15 minute output last 8 ssh: [-1.37539923 -1.37539923 -1.01599765 -1.01599765 -0.61179161 -0.61179161 -0.18381555 -0.18381555] 20 minute output last 8 ssh: [-1.79265356 -1.60702991 -1.40031064 -1.1705935 -0.9154433 -0.64399904 -0.36234728 -0.0724878 ]
Out[5]:
(735790.0, 735791.0)
- This saw-tooth behaviour is not present in the 10 minute and 20 minute outputs. So, I think we should switch to 10 minute output in nowcast-green. The behaviors is an artifact of
- Unfortunately, this run went unstable in Boundary Pass at 15 m depth after ~15000 time steps. Using
- 40s timestep
- kappa/mu=10
- background visc= 1e-5
- background diff=1e-6
- hollingsworth correction
- rn_bfri2 = 2e-3
- ln_loglayer=T
- rn_bfri2_max = 1.e-1
- rn_bfrz0=0.0015
- bathy2
- TS4 tides
It was in the middle of a spring tides. A run without hollingsworth correction was stable.
In [ ]: