[Python] What is a proper way to renew matplotlib.pyplot figures?

I am creating multiple figures using a loop with same figure settings, but with different data points on every loop. My problem is that my code is occupying memory on every loop. How do I prevent this?

Below is the example of my code

for i in range(len(uniq_tidx)):
fig,ax = plt.subplots(2,1,figsize=(10,4.5))

utidx = int(uniq_tidx)

time_raw = int(itime[utidx])

tnow = tstd + timedelta(seconds=time_raw)

yr = tnow.year
mo = tnow.month
da = tnow.day
hr = tnow.hour

tfmt = "{0:4d}-{1:02d}-{2:02d}-{3:02d}Z".format(yr,mo,da,hr)

nidx = np.where((itidx == utidx) & (icbs >= cl_thres) & (isst < sst_thres))

cbs = np.unique(icbs[nidx])[0]
cos = np.unique(icos[nidx])[0]
sst = np.mean(isst[nidx])

lon_all = ilon[iloni[nidx]]
lat_all = ilat[ilati[nidx]]

tl = "{0:4d}-{1:02d}-{2:02d} {3:02d}Z\nCluster size = {4:7.2f} km\u00b2 ({5:4d} px), SST = {6:5.2f} K".format(yr,mo,da,hr,cos * qd2km2,cos,sst)

ofn = "cls.vs.sst.{0:}.{1:s}".format(tfmt,ofmt)
ofp = "{0:s}/{1:s}".format(odir,ofn)

print("\t\tofp : {0:s}".format(ofp))

tb_now = itb_2d [utidx]
sst_now = isst_2d[utidx]


# Tb
bmap = Basemap(**margs,ax=ax[0])
bmap.drawcoastlines()

tb_cntf = ax[0].contourf(lons,lats,tb_now,clvls,levels=tb_clvl,cmap="Spectral_r",norm=tb_cmp)
tb_cb = plt.colorbar(tb_cntf, location="bottom",ax=ax[0],pad=0.05,aspect=50)
tb_cb.set_label("Tb (K)")


# SST with scatter points
bmap = Basemap(**margs,ax=ax[1])
bmap.drawcoastlines()

lon_pts,lat_pts = bmap(lon_all,lat_all)

sst_cntf = ax[1].contourf(lons,lats,sst_now,clvls,levels=sst_clvl,cmap="Spectral_r",norm=sst_cmp)
ax[1].scatter(lon_pts,lat_pts,10,marker='x',color='k')

sst_cb = plt.colorbar(sst_cntf, location="bottom",ax=ax[1],pad=0.05,aspect=50)
sst_cb.set_label("SST (K)")


plt.subplots_adjust(left=0.024,right=0.992,bottom=0.08,top=0.942,hspace=0.2,wspace=0.2)

plt.suptitle("{0:s}".format(tl))
plt.savefig("{0:s}".format(ofp))

plt.close()



Above, I am creating fig and ax on every loop, and I am closing the figure by using plt.close(). I tried plt.cla() and matplotlib would complain with an error message of

RuntimeWarning: More than 20 figures have been opened. Figures created through the pyplot interface (`matplotlib.pyplot.figure`) are retained until explicitly closed and may consume too much memory. (To control this warning, see the rcParam `figure.max_open_warning`). Consider using `matplotlib.pyplot.close()`.


And the memory leak problem still persists.

With plt.clf(), I get an error message of

.
.
.
raise NotImplementedError('cannot remove artist')


What is a proper way of renewing or reusing a figure and axis object using loops?

Continue reading...