-
Notifications
You must be signed in to change notification settings - Fork 0
/
BASED_inferenceB1.py
151 lines (118 loc) · 6.67 KB
/
BASED_inferenceB1.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
#%%
import pandas as pd
import xgboost as xgb
import pickle
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from multi_plotter import binscatter
import binsreg
B1 = pd.read_csv('data/B1_output.csv')
dicharge_rid = {62269900581: 356.765, 62269900571: 350.499, 62269900561: 341.238, 62269900551: 334.612,
62269900541: 320.473, 62269900531: 319.445, 62269900521: 322.209, 62269900511: 323.755,
62269900501: 328.199, 62269900491: 330.565, 62269900481: 351.444, 62269900471: 354.851,
62269900461: 356.663, 62269900451: 360.607, 62269900441: 361.981, 62269900431: 371.9,
62269900421: 372.763, 62269900411: 383.758, 62269900401: 385.937, 62269900391: 394.43,
62269900381: 397.365, 62269900371: 399.59}
discharge_series = pd.DataFrame(dicharge_rid, index=[0]).T
discharge_series.reset_index(inplace=True)
discharge_series.columns = ['reach_id', 'discharge_uncorrected']
discharge_series['reach_id'] = discharge_series['reach_id'].astype(int)
B1 = B1.merge(discharge_series, on='reach_id', how='left').dropna()
with open('data/inverted_discharge_params.pickle', 'rb') as f:
params = pickle.load(f)
def inverse_power_law(y, a, b):
return (y / a) ** (1 / b)
B1['corrected_discharge'] = inverse_power_law(B1['discharge_uncorrected'], *params)
B1['slope'] = B1['slope'] / 1000
guesswork = B1[['width', 'slope', 'corrected_discharge']].astype(float)
guesswork.columns = ['width', 'slope', 'discharge']
xgb_reg = xgb.XGBRegressor()
xgb_reg.load_model("data/based_us_sans_trampush_early_stopping_combat_overfitting.ubj")
B1['XGB_depth'] = xgb_reg.predict(guesswork)
B1['XGB_depth'] = B1['XGB_depth'].clip(lower=0)
df = B1
# Reflecting values when only ridge1 exists and ridge2 does not
mask_only_ridge1 = (~df['ridge1_elevation'].isna()) & (df['ridge2_elevation'].isna())
df.loc[mask_only_ridge1, 'ridge2_elevation'] = df.loc[mask_only_ridge1, 'ridge1_elevation']
df.loc[mask_only_ridge1, 'floodplain2_elevation'] = df.loc[mask_only_ridge1, 'floodplain1_elevation']
df.loc[mask_only_ridge1, 'ridge2_dist_along'] = -df.loc[mask_only_ridge1, 'ridge1_dist_along']
df.loc[mask_only_ridge1, 'floodplain2_dist_to_river_center'] = -df.loc[mask_only_ridge1, 'floodplain1_dist_to_river_center']
# Calculate slope for ridge1
ridge1_slope = df.apply(lambda row: (row['ridge1_elevation'] - row['floodplain1_elevation']) / abs(row['ridge1_dist_along']), axis=1)
df['ridge1_slope'] = ridge1_slope
# Since ridge2 data is mirrored from ridge1 when missing, we use the same calculation for ridge2_slope
df['ridge2_slope'] = df.apply(lambda row: (row['ridge2_elevation'] - row['floodplain2_elevation']) / abs(row['ridge2_dist_along']), axis=1)
# Calculate ridge1_height and ridge2_height
df['ridge1_height'] = df['ridge1_elevation'] - df['floodplain1_elevation']
df['ridge2_height'] = df['ridge2_elevation'] - df['floodplain2_elevation']
# Calculate ridge_height_mean
df['ridge_height_mean'] = (df['ridge1_height'] + df['ridge2_height']) / 2
# Calculate ridge_slope_mean
df['ridge_slope_mean'] = (df['ridge1_slope'] + df['ridge2_slope']) / 2
df['ridge_width'] = df['floodplain1_dist_to_river_center'] + df['floodplain2_dist_to_river_center']
# Calculate gamma values
df['gamma1'] = np.abs(df['ridge1_slope']) / df['slope']
df['gamma2'] = np.abs(df['ridge2_slope']) / df['slope']
# Calculate mean gamma
df['gamma_mean'] = df[['gamma1', 'gamma2']].mean(axis=1, skipna=True)
# Computing theta
# df['lambda'] = df['gamma_mean'] * df['superelevation_mean']
df['a_b_1'] = (df['ridge1_elevation'] - df['channel_elevation']) / (df['XGB_depth'])
df['a_b_2'] = (df['ridge2_elevation'] - df['channel_elevation']) / (df['XGB_depth'])
df['a_b'] = (df['a_b_1'] + df['a_b_2']) / 2
# Define conditions
conditions = [
df['a_b'] <= 5,
# (df['a_b'] > 2) & (df['a_b'] <= 2.5),
df['a_b'] > 5
]
choices = [
df['XGB_depth'], # If a_b is < 1, then depth is equal to XGB_depth
# (df['ridge1_elevation'] - (df['channel_elevation'] - df['XGB_depth'])) + (df['ridge2_elevation'] - (df['channel_elevation'] - df['XGB_depth'])) / 2, # If 1 <= a_b < 1.5
(df['ridge1_elevation'] - (df['channel_elevation'] - df['XGB_depth'])) + (df['ridge2_elevation'] - (df['channel_elevation'] - df['XGB_depth'])) / 2 # If a_b >= 1.5
]
# Apply conditions and choices to calculate corrected_depth
df['corrected_denominator'] = np.select(conditions, choices)
df['superelevation1'] = (df['ridge1_elevation'] - df['floodplain1_elevation']) / (df['corrected_denominator'])
df['superelevation2'] = (df['ridge2_elevation'] - df['floodplain2_elevation']) / (df['corrected_denominator'])
df['superelevation_mean'] = (df['superelevation1'] + df['superelevation2']) / 2
df['lambda'] = df['gamma_mean'] * df['superelevation_mean']
print(df['lambda'].describe())
df = df[df['lambda'] > 0]
df.to_csv('data/B1_output_corrected.csv', index=False)
df_est = binscatter(x='dist_out', y='lambda', data=df, ci=(3,3), noplot=True)
min_threshold = 0.0001
# Ensure 'lambda' and 'ci_l' values are above the threshold before calculating error bars
df_est['lambda'] = df_est['lambda'].clip(lower=min_threshold)
df_est['ci_l'] = df_est['ci_l'].clip(lower=min_threshold)
df_est['ci_r'] = df_est['ci_r'].clip(lower=min_threshold)
# Calculate the error bars in the original scale
df_est['error_lower'] = df_est['lambda'] - df_est['ci_l']
df_est['error_upper'] = df_est['ci_r'] - df_est['lambda']
# Ensure errors are positive
df_est['error_lower'] = np.abs(df_est['error_lower'])
df_est['error_upper'] = np.abs(df_est['error_upper'])
# Create an array with the lower and upper error margins
errors = np.array([df_est['error_lower'], df_est['error_upper']])
# Plot binned scatterplot
sns.scatterplot(x='dist_out', y='lambda', data=df_est, s=180, color='#26C6DA', alpha=0.7, edgecolor='black', marker='D', zorder=0)
# Use plt.errorbar to plot the error bars, passing the errors array directly
plt.errorbar(df_est['dist_out'], df_est['lambda'], yerr=errors, fmt='none', ecolor='k', elinewidth=1, capsize=3, alpha=0.5)
plt.xlabel('Distance along reach (km)')
plt.ylabel(r'$\Lambda$', rotation=0, labelpad=5)
plt.ylim(.01, 400) # Set limits in log scale
plt.yscale('log')
plt.gca().invert_xaxis() # Reverse the x-axis
# Add horizontal grid lines, light and dashed
plt.grid(True, which='major', linestyle='--', linewidth=0.5, color='grey')
plt.grid(True, which='minor', linestyle=':', linewidth=0.5, color='lightgrey', axis='y')
# Set y tick labels to non-scientific notation
plt.yticks([0.01, 0.1, 1, 10, 100], ['0.01', '0.1', '1', '10', '100'])
# Add minor tick marks
plt.minorticks_on()
plt.show()
# sns.scatterplot(x='dist_out', y='lambda', data=df, s=180, color='#26C6DA', alpha=0.7, edgecolor='black', marker='D', zorder=0)
# plt.yscale('log')
# plt.show()
# %%