def BoxPlotByCluster(DataDirectory, OutDirectory, fname_prefix, raster_name, stream_order=1):
"""
Make a boxplot of the results of the clustering compared to the raster specified
by raster_name
"""
#df = pd.read_csv(OutDirectory+fname_prefix+'_profiles_clustered_SO{}.csv'.format(stream_order))
# read in the raster
raster_ext = '.bil'
this_raster = IO.ReadRasterArrayBlocks(DataDirectory+raster_name)
EPSG_string = IO.GetUTMEPSG(DataDirectory+raster_name)
NDV, xsize, ysize, GeoT, Projection, DataType = IO.GetGeoInfo(DataDirectory+raster_name)
CellSize,XMin,XMax,YMin,YMax = IO.GetUTMMaxMin(DataDirectory+raster_name)
pts = PT.LSDMap_PointData(OutDirectory+fname_prefix+'_profiles_clustered_SO{}.csv'.format(stream_order),data_type ='csv')
easting, northing = pts.GetUTMEastingNorthing(EPSG_string=EPSG_string)
cluster_id = pts.QueryData('cluster_id', PANDEX=True)
clusters = list(set(cluster_id))
# dict for the data
data = {k: [] for k in clusters}
for x, (i, j) in enumerate(zip(northing, easting)):
# convert to rows and cols
X_coordinate_shifted_origin = j - XMin;
Y_coordinate_shifted_origin = i - YMin;
col_point = int(X_coordinate_shifted_origin/CellSize);
row_point = (ysize - 1) - int(round(Y_coordinate_shifted_origin/CellSize));
# check for data at this cell
this_value = this_raster[row_point][col_point]
if not np.isnan(this_value):
if this_value < 10:
# get the cluster id
data[cluster_id[x]].append(this_value)
print(data)
# now make a boxplot
labels, these_data = [*zip(*data.items())] # 'transpose' items to parallel key, value lists
plt.boxplot(these_data)
plt.xticks(range(1, len(labels) + 1), labels)
plt.show()
def GetLithologyPercentages(DataDirectory, OutDirectory, fname_prefix, raster_name, stream_order=1):
"""
Get the percentage of the nodes in each cluster that drain each lithology
"""
from collections import Counter
# read in the raster
raster_ext = '.bil'
this_raster = IO.ReadRasterArrayBlocks(DataDirectory+raster_name)
EPSG_string = IO.GetUTMEPSG(DataDirectory+raster_name)
NDV, xsize, ysize, GeoT, Projection, DataType = IO.GetGeoInfo(DataDirectory+raster_name)
CellSize,XMin,XMax,YMin,YMax = IO.GetUTMMaxMin(DataDirectory+raster_name)
pts = PT.LSDMap_PointData(OutDirectory+fname_prefix+'_profiles_clustered_SO{}.csv'.format(stream_order),data_type ='csv')
easting, northing = pts.GetUTMEastingNorthing(EPSG_string=EPSG_string)
cluster_id = pts.QueryData('cluster_id', PANDEX=True)
clusters = list(set(cluster_id))
# dict for the data
data = {k: [] for k in clusters}
for x, (i, j) in enumerate(zip(northing, easting)):
# convert to rows and cols
X_coordinate_shifted_origin = j - XMin;
Y_coordinate_shifted_origin = i - YMin;
col_point = int(X_coordinate_shifted_origin/CellSize);
row_point = (ysize - 1) - int(round(Y_coordinate_shifted_origin/CellSize));
# check for data at this cell
this_value = this_raster[row_point][col_point]
if not np.isnan(this_value):
data[cluster_id[x]].append(this_value)
# you have the values. now what percentage are each?
for key, liths in data.items():
c = Counter(liths)
n_ndv = c[0.0]
print(c)
[print(x,": ",vals/len(liths) * 100) for x, vals in c.items()]
def MakeBoxPlotsKsnLithology(DataDirectory, fname_prefix, raster_name, theta=0.45, label_list=[]):
"""
Make boxplots of ksn compared to lithology raster. Lithology should have integer
values for the different rock types (rock type with 0 is excluded). Pass in list of
labels for the different units, which must be the same length as the number of lithology codes.
If none is passed then just use the integer values for labelling.
"""
from scipy import stats
# read in the raster
raster_ext = '.bil'
this_raster = IO.ReadRasterArrayBlocks(DataDirectory+raster_name)
#EPSG_string = IO.GetUTMEPSG(DataDirectory+raster_name)
EPSG_string='epsg:32611'
print(EPSG_string)
NDV, xsize, ysize, GeoT, Projection, DataType = IO.GetGeoInfo(DataDirectory+raster_name)
CellSize,XMin,XMax,YMin,YMax = IO.GetUTMMaxMin(DataDirectory+raster_name)
pts = PT.LSDMap_PointData(DataDirectory+fname_prefix+'_ksn.csv',data_type ='csv')
print(pts)
easting, northing = pts.GetUTMEastingNorthing(EPSG_string=EPSG_string)
ksn = pts.QueryData('ksn', PANDEX=True)
#print(ksn)
# get the unique values in the raster
raster_values = np.unique(this_raster)
raster_values = raster_values[1:]
# dict for the data
data = {k: [] for k in raster_values}
for x, (i, j) in enumerate(zip(northing, easting)):
# convert to rows and cols
X_coordinate_shifted_origin = j - XMin;
Y_coordinate_shifted_origin = i - YMin;
col_point = int(X_coordinate_shifted_origin/CellSize);
row_point = (ysize - 1) - int(round(Y_coordinate_shifted_origin/CellSize));
# check for data at this cell
this_raster_value = this_raster[row_point][col_point]
if not np.isnan(this_raster_value) and this_raster_value != 0:
data[this_raster_value].append(ksn[x])
# set up a figure
fig,ax = plt.subplots(nrows=1,ncols=1, figsize=(5,5), sharex=True, sharey=True)
labels, dict = [*zip(*data.items())] # 'transpose' items to parallel key, value lists
print(label_list)
if label_list:
labels = label_list
print(labels)
box = plt.boxplot(dict, patch_artist=True)
plt.xticks(range(1, len(labels) + 1), labels)
plt.ylabel('$k_{sn}$', fontsize=14)
# get the medians for plotting as an upper label
medians = []
print("========SOME KSN STATISTICS=========")
for key, value in data.items():
print("Key {}, median ksn = {}".format(key, np.median(value)))
medians.append(np.median(value))
print("Key {}, IQR = {}".format(key, stats.iqr(value)))
print("========================================")
pos = np.arange(len(labels)) + 1
upperLabels = [str(np.round(s, 2)) for s in medians]
# change the colours for each lithology
colors=['#60609fff', '#fdbb7fff', '#935353ff', '#f07b72ff']
for patch, color in zip(box['boxes'], colors):
patch.set_facecolor(color)
patch.set_alpha(0.9)
patch.set_edgecolor('k')
for cap in box['caps']:
cap.set(color='k')
for wh in box['whiskers']:
wh.set(color='k')
for med in box['medians']:
med.set(color='k')
for flier, color in zip(box['fliers'], colors):
flier.set_markeredgecolor(color)
flier.set_markerfacecolor(color)
flier.set_markersize(2)
# for tick, label in zip(range(len(labels)), ax.get_xticklabels()):
# k = tick % 2
# ax.text(pos[tick], top - (top*0.05), upperLabels[tick],
# horizontalalignment='center', color=colors[k])
ax.grid(color='0.8', linestyle='--', which='major', zorder=1)
plt.title('Boxplots of $k_{sn}$ by lithology', fontsize=14)
plt.savefig(DataDirectory+fname_prefix+'_boxplot_lith_ksn.png', dpi=300, transparent=True)
plt.clf()
# Do some stats, yo
# KS test to see if we can distinguish the distributions at a confidence level of p = 0.05
# https://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.stats.kstest.html
# relief
keys = list(data.keys())
values = list(data.values())
k=0
for i in range(len(keys)-1):
for j in range(i+1, len(keys)):
print("KS test between {} and {}".format(keys[i], keys[j]))
d, p = stats.ks_2samp(values[i], values[j])
print(d, p)
k += 1
def PlotLithologyWithClusters(DataDirectory,
OutDirectory,
fname_prefix,
stream_order=1,
shapefile_name='geol.shp',
geol_field='geol'):
"""
Make a hillshade of the raster with the channels coloured by the cluster
value. Rasterise a geology shapefile and drape on the top.
Uses the LSDPlottingTools libraries. https://github.com/LSDtopotools/LSDMappingTools
Args:
stream_order: the stream order of the profiles that you are analysing
shapefile_name: name of the lithology shapefile
geol_field: the field of the shapefile that has the lithology information
Author: FJC
"""
import LSDPlottingTools as LSDP
from LSDMapFigure.PlottingRaster import MapFigure
df = pd.read_csv(DataDirectory + fname_prefix + '_all_tribs.csv')
cluster_df = pd.read_csv(
OutDirectory + fname_prefix +
'_profiles_clustered_SO{}.csv'.format(stream_order))
# set figure sizes based on format
fig_width_inches = 4.92126
# some raster names
raster_ext = '.bil'
BackgroundRasterName = fname_prefix + raster_ext
HSName = fname_prefix + '_hs' + raster_ext
if not os.path.isfile(DataDirectory + HSName):
print("Making a hillshade for you")
# make a hillshade
BM.GetHillshade(DataDirectory + BackgroundRasterName,
DataDirectory + HSName)
# create the map figure
MF = MapFigure(HSName, DataDirectory, coord_type="UTM")
res = IO.GetUTMMaxMin(DataDirectory + BackgroundRasterName)[0]
#rasterise the shapefile
new_shp, geol_dict = VT.geologic_maps_modify_shapefile(
DataDirectory + shapefile_name, geol_field)
lith_raster = VT.Rasterize_geologic_maps_pythonic(new_shp, res, geol_field)
MF.add_drape_image(lith_raster,
"",
colourmap=plt.cm.jet,
alpha=0.4,
show_colourbar=False,
discrete_cmap=True,
cbar_type=int,
mask_value=0)
clusters = cluster_df.cluster_id.unique()
for cl in clusters:
# plot the whole channel network in black
ChannelPoints = LSDP.LSDMap_PointData(df,
data_type="pandas",
PANDEX=True)
MF.add_point_data(ChannelPoints,
show_colourbar="False",
unicolor='0.9',
manual_size=2,
zorder=1,
alpha=0.5)
# plot the clustered profiles in the correct colour
this_df = cluster_df[cluster_df.cluster_id == cl]
this_colour = str(this_df.colour.unique()[0])
ClusteredPoints = LSDP.LSDMap_PointData(this_df,
data_type="pandas",
PANDEX=True)
MF.add_point_data(ClusteredPoints,
show_colourbar="False",
zorder=100,
unicolor=this_colour,
manual_size=3)
MF.save_fig(fig_width_inches=fig_width_inches,
FigFileName=OutDirectory + fname_prefix +
'_lith_clusters_SO{}.png'.format(stream_order),
FigFormat='png',
Fig_dpi=300,
fixed_cbar_characters=6,
adjust_cbar_characters=False,
transparent=True) # Save the figure