def cluster_ssh(sla,
lat,
lon,
nclusters,
distthres=3000,
returnall=False,
absmode=0,
distmode=0,
uncertmode=0,
printmsg=True,
calcsil=False):
# ---------------------------------------------
# Calculate Correlation, Covariance, and Distance Matrices
# ---------------------------------------------
ntime, nlat, nlon = sla.shape
srho, scov, sdist, okdata, okpts, coords2 = slutil.calc_matrices(
sla, lon5, lat5, return_all=True)
#npts = okdata.shape[1]
# -------------------------------
# Apply corrections based on mode
# -------------------------------
if absmode == 1: # Take Absolute Value of Correlation/Covariance
scov = np.abs(scov)
srho = np.abs(srho)
elif absmode == 2: # Use Anticorrelation, etc
scov *= -1
srho *= -1
# --------------------------
# Combine the Matrices
# --------------------------
a_fac = np.sqrt(
-distthres /
(2 * np.log(0.5))) # Calcuate so exp=0.5 when distance is 3000km
expterm = np.exp(-sdist / (2 * a_fac**2))
if distmode == 0: # Include distance and correlation
distance_matrix = 1 - expterm * srho
elif distmode == 1: # Just Include distance
distance_matrix = 1 - expterm
elif distmode == 2: # Just Include correlation
distance_matrix = 1 - srho
# --------------------------
# Do Clustering (scipy)
# --------------------------
cdist = squareform(distance_matrix, checks=False)
linked = linkage(cdist, 'weighted')
clusterout = fcluster(linked, nclusters, criterion='maxclust')
# --------------------
# Calculate Silhouette
# --------------------
if calcsil:
s_score, s, s_bycluster = slutil.calc_silhouette(
distance_matrix, clusterout, nclusters)
# fig,ax = plt.subplots(1,1)
# ax = slutil.plot_silhouette(clusterout,nclusters,s,ax1=ax)
# -------------------------
# Calculate the uncertainty
# -------------------------
uncertout = np.zeros(clusterout.shape)
for i in range(len(clusterout)):
covpt = scov[i, :] #
cid = clusterout[i] #
covin = covpt[np.where(clusterout == cid)]
covout = covpt[np.where(clusterout != cid)]
if uncertmode == 0:
uncertout[i] = np.mean(covin) / np.mean(covout)
elif uncertmode == 1:
uncertout[i] = np.median(covin) / np.median(covout)
# Apply rules from Thompson and Merrifield (Do this later)
# if uncert > 2, set to 2
# if uncert <0.5, set to 0
#uncertout[uncertout>2] = 2
#uncertout[uncertout<0.5] = 0
# ------------------------------
# Calculate Wk for gap statistic
# ------------------------------
Wk = np.zeros(nclusters)
for i in range(nclusters):
cid = i + 1
ids = np.where(clusterout == cid)[0]
dist_in = distance_matrix[ids[:, None], ids[
None, :]] # Get Pairwise Distances within cluster
dist_in = dist_in.sum(
) / 2 # Sum and divide by 2 (since pairs are replicated)
Wk[i] = dist_in
# -----------------------
# Replace into full array
# -----------------------
clustered = np.zeros(nlat * nlon) * np.nan
clustered[okpts] = clusterout
clustered = clustered.reshape(nlat, nlon)
cluster_count = []
for i in range(nclusters):
cid = i + 1
cnt = (clustered == cid).sum()
cluster_count.append(cnt)
if printmsg:
print("Found %i points in cluster %i" % (cnt, cid))
uncert = np.zeros(nlat * nlon) * np.nan
uncert[okpts] = uncertout
uncert = uncert.reshape(nlat, nlon)
if calcsil: # Return silhouette values
return clustered, uncert, cluster_count, Wk, s, s_bycluster
if returnall:
return clustered, uncert, cluster_count, Wk, srho, scov, sdist, distance_matrix
return clustered, uncert, cluster_count, Wk
#%% Calculate covariance matrix
sla = sla_lp
lon = lon5
lat = lat5
absmode = 0
distthres = 3000
distmode = 0
uncertmode = 0
# --------------------------------------------------------
# Calculate Correlation, Covariance, and Distance Matrices
# --------------------------------------------------------
ntime, nlat, nlon = sla.shape
srho, scov, sdist, okdata, okpts, coords2 = slutil.calc_matrices(
sla, lon, lat, return_all=True)
# -------------------------------
# Apply corrections based on mode
# -------------------------------
if absmode == 1: # Take Absolute Value of Correlation/Covariance
scov = np.abs(scov)
srho = np.abs(srho)
elif absmode == 2: # Use Anticorrelation, etc
scov *= -1
srho *= -1
# --------------------------
# Combine the Matrices
# --------------------------
a_fac = np.sqrt(
#snamelong = "Red Noise (Filtered)"
snamelong = "Distance Only"
#snamelong = "Distance Only"
#snamelong = "Characteristic Distance = %i km "% chardist
expname = "AVISO_DistanceOnly_nclusters%i_Filtered_" % (nclusters)
#expname = "AVISO_WhiteNoise_nclusters%i_Filtered_" % (nclusters)
#expname = "AVISO_WhiteNoise_chardist%i_nclusters%i_Filtered_" % (chardist,nclusters)
distmode = 0
absmode = 0
uncertmode = 0
# ------------------
# Calculate Matrices
# ------------------
ntime, nlat, nlon = varin.shape
srho, scov, sdist, okdata, okpts, coords2 = slutil.calc_matrices(
varin, lon5, lat5, return_all=True)
if absmode == 1:
scov = np.abs(scov)
srho = np.abs(srho)
elif absmode == 2:
scov *= -1
srho *= -1
# --------------------------
# Combine the Matrices
# --------------------------
distthres = 3000
# Default Distance Calculation
a_fac = np.sqrt(-distthres / (2 * np.log(0.5)))
expterm = np.exp(-sdist / (2 * a_fac**2))