In [1]:

    

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.colors as colors
from pysheds.grid import Grid
import timeit

%matplotlib inline


    

In [2]:

    

grid = Grid.from_raster('../data/n30w100_dir',
                        data_name='dir', input_type='ascii')

         #N    NE    E    SE    S    SW    W    NW
dirmap = (64,  128,  1,   2,    4,   8,    16,  32)


    

In [3]:

    

gridsize = grid.dir.size
gridshape = grid.dir.shape
gridbytes = grid.dir.nbytes

print(gridshape[0], ' x ', gridshape[1], ' grid')
print(gridsize, ' gridcells')
print(gridbytes / 1000000, 'MB in memory (stored as uint8)')


    

    




6000  x  6000  grid
36000000  gridcells
36.0 MB in memory (stored as uint8)

In [4]:

    

fdir = grid.view('dir')


    

In [5]:

    

%timeit grid.accumulation(fdir, dirmap=dirmap, pad_inplace=True, direction_name='dir')


    

    




15 s ± 700 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

In [6]:

    

pour_y, pour_x = np.unravel_index(np.argmax(grid.acc), gridshape)


    

In [7]:

    

%timeit grid.catchment(pour_x, pour_y, dirmap=dirmap, recursionlimit=15000, xytype='index')


    

    




4.55 s ± 114 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

In [8]:

    

catchment_cells = np.count_nonzero(grid.view('catch'))
print(catchment_cells, ' cells in catchment')


    

    




9792691  cells in catchment

In [9]:

    

D = np.logspace(1, 3.7782, 15).astype(int)


    

In [10]:

    

times = []

for i in D:
    fdir_sub = fdir[:i, :i]
    start = timeit.default_timer()
    grid.accumulation(fdir_sub, dirmap=dirmap, pad_inplace=True)
    stop = timeit.default_timer()
    dt = stop - start
    times.append(dt)


    

In [11]:

    

fig, ax = plt.subplots(figsize=(8,4))

ax.loglog(D**2, times, marker='o')
ax.loglog(D**2, D**2 / 1e6, linestyle='--', c='0.85')
ax.set_xlabel('Number of inputs')
ax.set_ylabel('Run time (s)')
ax.set_title('Flow accumulation run time vs. number of inputs')
ax.grid()
plt.tight_layout()
plt.savefig('accumulation_runtime.pdf')


    

In [12]:

    

grid.accumulation(fdir, dirmap=dirmap, pad_inplace=False)
pour_y, pour_x = np.unravel_index(np.argmax(grid.acc), gridshape)
grid.catchment(pour_x, pour_y, dirmap=dirmap, direction_name='dir', recursionlimit=15000, xytype='index')
grid.accumulation(direction_name='catch', dirmap=dirmap, pad_inplace=False)


    

In [13]:

    

D = np.logspace(2, np.log10(grid.acc.max()), 20).astype(int)


    

In [14]:

    

ix = [np.argmin(np.abs(grid.acc.ravel() - i)) for i in D]


    

In [15]:

    

times = []
sizes = []

for i in ix:
    pour_y, pour_x = np.unravel_index(i, fdir.shape)
    fdir_sub = fdir[:i, :i]
    start = timeit.default_timer()
    grid.catchment(pour_x, pour_y, fdir, dirmap=dirmap, recursionlimit=15000, xytype='index')
    stop = timeit.default_timer()
    dt = stop - start
    size = np.count_nonzero(grid.catch)
    times.append(dt)
    sizes.append(size)


    

In [16]:

    

times = np.asarray(times)
sizes = np.asarray(sizes)


    

In [17]:

    

s = pd.Series(times, index=sizes).sort_index()


    

In [18]:

    

fig, ax = plt.subplots(figsize=(8,4))

ax.loglog(s.index, s.values, marker='o')
ax.loglog(s.index[9:], s.index[9:] / 1e6, linestyle='--', c='0.85')
ax.set_xlabel('Number of inputs')
ax.set_ylabel('Run time (s)')
ax.set_title('Catchment delineation run time vs. number of inputs')
ax.set_ylim(0.05, 5)
ax.yaxis.set_ticklabels(['', '', '0.1', '1'])
ax.grid()
plt.tight_layout()
plt.savefig('catchment_runtime.pdf')


    

In [19]:

    

grid.accumulation(fdir, dirmap=dirmap, pad_inplace=False, direction_name='dir')


    

In [20]:

    

pour_y, pour_x = np.unravel_index(np.argmax(grid.acc), gridshape)


    

In [21]:

    

grid.catchment(pour_x, pour_y, dirmap=dirmap, recursionlimit=15000, xytype='index')


    

In [22]:

    

np.count_nonzero(grid.catch)


    

    
Out[22]:





9792691

In [ ]: