#EDIT BOUNDS FOR AMO (AMOmid in O'Reilly 2016 is defined between 40 N and 60 N, Gulev. et al. 2013 defined between 30 N and 50 N)
for latbounds in latboundar:
fieldlagcorrs = np.zeros((len(lags), nlat, nlon))
fieldlagcorrs_lt = np.zeros((len(lags), nlat, nlon))
fieldlagcorrs_st = np.zeros((len(lags), nlat, nlon))
sstcorrs = MV.zeros((nlat,nlon))
sstpvals = MV.zeros((nlat,nlon))
fieldcorrs = MV.zeros((nlat, nlon))
fieldcorrs_lt = MV.zeros((nlat,nlon))
fieldcorrs_st = MV.zeros((nlat,nlon))
AMO, sstanom_globe_an, sstanom_na_an = calc_NA_globeanom(sst_an, latbounds, lats, lons, baseti, basetf)
NAfield2, fieldanom_globe_an, fieldanom_na_an = calc_NA_globeanom(field_an, latbounds, lats, lons, baseti, basetf)
AMO_lt = running_mean(AMO, N_map)
AMO_st = AMO[ci:-ci] - AMO_lt
scaler = StandardScaler()
AMOstd = scaler.fit_transform(AMO.reshape(-1,1))
AMOstd_lt = scaler.fit_transform(AMO_lt.reshape(-1,1))
AMOstd_st = scaler.fit_transform(AMO_st.reshape(-1,1))
#need to normalize in this manner in order to get pearson correlation coefficient from np.correlate
AMOstd2 = (AMO - np.mean(AMO)) / (np.std(AMO) * len(AMO))
AMOstd_lt2 = (AMO_lt - np.mean(AMO_lt)) / (np.std(AMO_lt) * len(AMO_lt))
AMOstd_st2 = (AMO_st - np.mean(AMO_st)) / (np.std(AMO_st) * len(AMO_st))
lags = np.arange(-lagmax, lagmax + 1)
field_lt = running_mean(fieldprime, N_map)
field_st = fieldprime[ci:-ci, :] - field_lt
nt_lt = field_lt.shape[0]
#EDIT BOUNDS FOR AMO (AMOmid in O'Reilly 2016 is defined between 40 N and 60 N, Gulev. et al. 2013 defined between 30 N and 50 N)
#latitude bounds for SST averaging
latboundar = np.array([[0, 60], [0, 20], [20, 45], [45, 60]])
#latboundar = np.array([[0,20],[20,45],[45,60]])
#latboundar = np.array([[0,60]])
for latbounds in latboundar:
AMO, sstanom_globe_an, sstanom_na_an = calc_NA_globeanom(
sst, latbounds, lats, lons, baseti, basetf)
NAfield2, fieldanom_globe_an, fieldanom_na_an = calc_NA_globeanom3D(
field, latbounds, lats, lons, baseti, basetf)
sstcorrs = MV.zeros((nlat, nlon))
fieldcorrs = MV.zeros((nplev, nlat * nlon))
fieldcorrs_lt = MV.zeros((nplev, nlat * nlon))
fieldcorrs_st = MV.zeros((nplev, nlat * nlon))
fieldlagcorrs = np.zeros((len(lags), nplev, nlat * nlon))
fieldlagcorrs_lt = np.zeros((len(lags), nplev, nlat * nlon))
fieldlagcorrs_st = np.zeros((len(lags), nplev, nlat * nlon))
AMO_lt = running_mean(AMO, N_map)
AMO_st = AMO[ci:-ci] - AMO_lt
#initial/final indices for base period
baseti = 0
basetf = 10
sst_an = an_ave(sst)
thf_an = an_ave(thf)
ps_an = an_ave(ps)
#detrend annual fields instead of monthly? prevents thf from blowing up for some reason...
if detr:
sst_an, params = detrend_separate(sst_an)
ps_an, params = detrend_separate(ps_an)
thf_an, params = detrend_separate(thf_an)
AMO, sstanom_globe_an, sstanom_na_an = calc_NA_globeanom(
sst_an, latbounds, lats, lons, baseti, basetf)
NAthf2, thfanom_globe_an, thfanom_na_an = calc_NA_globeanom(
thf_an, latbounds, lats, lons, baseti, basetf)
#thf blows up after detrending, need to mask values. Why doesn't this work for plotting later?
#thf = np.ma.array(thf, mask=np.bitwise_or(thf_mask, np.abs(thf) > 1e2))
#this seems to work for plotting, but have to mask nan values later
#thf[np.abs(thf) > 1e3] = np.nan
#sstanom_globe_an = signal.detrend(sstanom_globe_an)
#sstanom_na_an = signal.detrend(sstanom_na_an)
#AMO = signal.detrend(AMO)
#NAthf2 = signal.detrend(NAthf2)
#thfanom_globe_an = signal.detrend(thfanom_globe_an)