# Create CWV bins

number_of_bins = 28 # default = 28

cwv_max = 70 # default = 70 (in mm)

cwv_min = 28 # default = 28 (in mm)

bin_width = 1.5 # default = 1.5

bin_center = np.arange((cwv_min+(bin_width/2)), (cwv_max-(bin_width/2))+bin_width, bin_width)

# Define precip threshold

precip_threshold = 0.5; # default 0.5 (in mm/hr)

# Define variables for binning

bin_index = np.zeros([number_of_bins,cwv.size])

precip_binned = np.empty([number_of_bins,cwv.size]) * np.nan

precip_counts = np.zeros([number_of_bins,cwv.size])

# In[264]:

# Bin the data by CWV value as specified above

for i in range (0,number_of_bins):

for j in range(0,cwv.size):

if cwv[j] > (cwv_min+(i*bin_width)):

bin_index[i,j] = 1

if cwv[j] > (cwv_min+(i*bin_width)+bin_width):

bin_index[i,j] = 0

if bin_index[i,j] == 1:

precip_binned[i,j] = precip[j]

else:

precip_binned[i,j] = np.nan

if precip_binned[i,j] >= precip_threshold:

precip_counts[i,j] = 1

if np.isnan(precip_binned[i,j]):

precip_counts[i,j] = np.nan

# In[265]:

# Create binned arrays

hist_cwv = np.empty([number_of_bins,1]) * np.nan

hist_precip_points = np.empty([number_of_bins,1]) * np.nan

pr_binned_mean = np.empty([number_of_bins,1]) * np.nan

pr_binned_var = np.empty([number_of_bins,1]) * np.nan

pr_binned_std = np.empty([number_of_bins,1]) * np.nan

pr_probability = np.empty([number_of_bins,1]) * np.nan

errorbar_hist_precip_points = np.empty([number_of_bins,1]) * np.nan

std_error_precip = np.empty([number_of_bins,1]) * np.nan

pdf_cwv = np.empty([number_of_bins,1]) * np.nan

pdf_precipitating_points = np.empty([number_of_bins,1]) * np.nan

# Fill binned arrays

hist_cwv = bin_index.sum(axis=1)

hist_precip_points = np.nansum(precip_counts,axis=1)

pr_binned_mean = np.nanmean(precip_binned,axis=1)

pr_binned_var = np.nanvar(precip_binned,axis=1)

pr_binned_std = np.nanstd(precip_binned,axis=1)

r = np.empty([1,number_of_bins]) * np.nan

r = np.sum(~np.isnan(precip_counts),axis=1)

pr_probability = np.nansum(precip_counts,axis=1)/r

pdf_cwv = hist_cwv/bin_width

pdf_precipitating_points = hist_precip_points/bin_width

for i in range(0,number_of_bins):

std_error_precip[i] = pr_binned_std[i]/math.sqrt(hist_cwv[i])

errorbar_hist_precip_points[i] = math.sqrt(hist_precip_points[i])/bin_width

# In[266]:

# create color map

# choose from maps here:

# http://matplotlib.org/examples/color/colormaps_reference.html

scatter_colors = cm.jet(np.linspace(0,1,number_of_bins+1,endpoint=True))

# scatter_colors = cm.plasma(numpy.linspace(0,1,number_of_bins+1,endpoint=True))

# In[267]:

axes_fontsize = 12 # size of font in all plots

legend_fontsize = 9

marker_size = 40 # size of markers in scatter plots

xtick_pad = 10 # padding between x tick labels and actual plot

# create figure canvas

fig = mp.figure(figsize=(8,2.5))

# create figure 1

ax1 = fig.add_subplot(131)

ax1.set_xlim(25,72)

ax1.set_ylim(0,6)

ax1.set_xticks([30,40,50,60,70])

ax1.set_yticks([0,1,2,3,4,5,6])

ax1.tick_params(labelsize=axes_fontsize)

ax1.tick_params(axis='x', pad=10)

error = [std_error_precip,std_error_precip]

ax1.errorbar(bin_center, pr_binned_mean, xerr=0, yerr=error, ls='none', color='black')

ax1.scatter(bin_center, pr_binned_mean, edgecolor='none', facecolor=scatter_colors, s=marker_size, clip_on=False, zorder=3)

ax1.set_ylabel('Precip (mm hr $^-$ $^1$)', fontsize=axes_fontsize)

ax1.set_xlabel('CWV (mm)', fontsize=axes_fontsize)

ax1.text(0.05, 0.95, sites[0], transform=ax1.transAxes, fontsize=12, verticalalignment='top')

ax1.text(0.05, 0.85, test, transform=ax1.transAxes, fontsize=12, verticalalignment='top')

#ax1.grid()

ax1.set_axisbelow(True)

# create figure 2 (probability pickup)

ax2 = fig.add_subplot(132)

ax2.set_xlim(25,72)

ax2.set_ylim(0,1)

ax2.set_xticks([30,40,50,60,70])

ax2.set_yticks([0.0,0.2,0.4,0.6,0.8,1.0])

ax2.tick_params(labelsize=axes_fontsize)

ax2.tick_params(axis='x', pad=xtick_pad)

ax2.scatter(bin_center, pr_probability, marker='d', s=marker_size, edgecolor='none', facecolor='steelblue', clip_on=False, zorder=3)

ax2.set_ylabel('Probability of Precip', fontsize=axes_fontsize)

ax2.set_xlabel('CWV (mm)', fontsize=axes_fontsize)

#ax2.grid()

ax2.set_axisbelow(True)

# create figure 3 (non-normalized PDF)

ax3 = fig.add_subplot(133)

ax3.set_yscale('log')

ax3.set_ylim(5e-1, 5e5)

ax3.set_xlim(25,72)

ax3.set_xticks([30,40,50,60,70])

ax3.set_yticks(10**np.array((0,1,2,3,4,5)))

ax3.tick_params(labelsize=axes_fontsize)

ax3.tick_params(axis='x', pad=xtick_pad)

# yscale is log scale, so throw out any zero values

pdf_precipitating_points[pdf_precipitating_points==0] = np.nan

error = [errorbar_hist_precip_points,errorbar_hist_precip_points]

ax3.errorbar(bin_center, pdf_precipitating_points, xerr=0, yerr=error, ls='none', color='black')

ax3.scatter(bin_center, pdf_precipitating_points, edgecolor='none', facecolor='b', s=marker_size, clip_on=False, zorder=3, label='precip $>$ 0.5 mm hr $^{\minus 1}$')

ax3.scatter(bin_center, pdf_cwv, marker='x', color='0', label='all')

ax3.set_ylabel('Freq Density', fontsize=axes_fontsize)

ax3.set_xlabel('CWV (mm)', fontsize=axes_fontsize)

#ax3.grid()

ax3.set_axisbelow(True)

# create legend

legend_handles, legend_labels = ax3.get_legend_handles_labels()

ax3.legend(legend_handles, legend_labels, loc='upper left', bbox_to_anchor=(0.1,0.95), fontsize=legend_fontsize, scatterpoints=1, handlelength=0, labelspacing=0, borderpad=0, borderaxespad=0, frameon=False)

# set layout to tight (so that space between figures is minimized)

mp.tight_layout()

# save figure

#mp.savefig('conv_diagnostics_example_kas_new.pdf', transparent=True, bbox_inches='tight')