%%bash
mkdir tor_data
cd tor_data
wget -q -nc https://www.spc.noaa.gov/wcm/data/1950-2024_actual_tornadoes.csv > /dev/null

import pandas as pd

tor = pd.read_csv("tor_data/1950-2024_actual_tornadoes.csv")
tor['datetime'] = tor['date'] + " " + tor['time']
tor['datetime'] = pd.to_datetime(tor['datetime'])
tor = tor.set_index('datetime')

tor_count = tor.resample('YS').size()
tor_count.name = 'count'

tor_count.plot()

ef1_plus = tor.copy()

ef1_plus_count = ef1_plus.resample('YS').size()
ef1_plus_count.name = 'count'

ef1_plus_count.plot()

from sklearn import preprocessing

counts1d = ef1_plus_count.values

# this reshapes the 1D data into a 2D shape required by sklearn
counts2d = counts1d.reshape(-1, 1)

scaler = preprocessing.MinMaxScaler().fit_transform(counts2d)

# get it back to 1D
minmax_counts = scaler.flatten()

print(counts1d)
print(minmax_counts)

counts1d = ef1_plus_count.values

# this reshapes the 1D data into a 2D shape required by sklearn
counts2d = counts1d.reshape(-1, 1)

scaler = preprocessing.StandardScaler().fit_transform(counts2d)

# get it back to 1D
std_counts = scaler.flatten()

print(counts1d)
print(std_counts)

import matplotlib.pyplot as plt

posneg = ['red' if x >= 0 else 'blue' for x in std_counts]

plt.bar(list(range(1950, 2025)), std_counts, color=posneg)

plt.title("Tornado Counts per Year (Z-score)")

plt.ylabel("count (standardized)")
plt.xlabel("year")

import numpy as np

def std_cat(std, bins):
    '''Takes a list of standardized numbers and categorizes the numbers based on
    the bins argument. The bin values must be increasing from left to
    right, and the first and last numbers define the upper edge of the
    first category (0; not inclusive) and the bottom edge of the last
    category ([len(bins) + 1]; inclusive)

    It then returns the transformed values as categories.

    E.g., [-5, 0, 5] with bins [-1, 1]
          returns [0, 1, 2]
    E.g., [-5.2, -1.1, 0.5, 1.1, 5.4] with bins [-2, -1, 1, 2]
          returns [0, 1, 2, 3, 4]
    '''

    return np.digitize(std, bins=bins)

# your function parameter `bin`
function_params = {'bins': [-1, 1]}

trans_cat = preprocessing.FunctionTransformer(std_cat,
                                              kw_args=function_params)
std_cat = trans_cat.transform(std_counts)

# your color -> category dictionary
color_lookup = {0: 'blue', 1: 'grey', 2: 'red'}

posneg = [color_lookup[val] for val in std_cat]

plt.bar(list(range(1950, 2025)), std_counts, color=posneg, alpha=0.75)

plt.title("Annual Tornado Counts (Z-score)")

# defines your category threshold lines
plt.hlines(1, 1950, 2024, color='r', ls='--')
plt.hlines(-1, 1950, 2024, color='b', ls='--')
plt.hlines(0, 1950, 2024, color='k', ls='-')

plt.ylabel("Count (Standardized)")
plt.xlabel("Year")

plt.xlim(1949.5, 2024.5)
plt.grid()

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.preprocessing import StandardScaler

# ── Sample tornado fatality data (replace with your own dataset) ──────────
data = {
    'Year': list(range(2000, 2025)),
    'Fatalities': [
        41, 40, 55, 54, 36, 38, 67, 81, 126, 21,
        45, 553, 70, 55, 47, 36, 17, 35, 10, 42,
        76, 103, 24, 84, 50
    ]
}
df = pd.DataFrame(data)

# ── Compute z-scores using sklearn StandardScaler ─────────────────────────
scaler = StandardScaler()
df['Z_Score'] = scaler.fit_transform(df[['Fatalities']]).flatten()

# ── Assign categories and colors based on z-score ranges ──────────────────
def assign_category(z):
    if z < -2:
        return '< -2', 'darkblue'
    elif z < -1:
        return '-2 to -1', 'steelblue'
    elif z < 0:
        return '-1 to 0', 'lightblue'
    elif z < 1:
        return '0 to 1', 'lightcoral'
    elif z < 2:
        return '1 to 2', 'tomato'
    else:
        return '> 2', 'darkred'

df[['Category', 'Color']] = df['Z_Score'].apply(
    lambda z: pd.Series(assign_category(z))
)

# ── Print summary ─────────────────────────────────────────────────────────
print(f"Mean Fatalities (μ): {scaler.mean_[0]:.2f}")
print(f"Std Dev (σ):         {scaler.scale_[0]:.2f}\n")
print(df[['Year', 'Fatalities', 'Z_Score', 'Category']].to_string(index=False))

# ── Plot ──────────────────────────────────────────────────────────────────
fig, ax = plt.subplots(figsize=(14, 6))
fig.patch.set_facecolor('#f8f8f8')
ax.set_facecolor('#f8f8f8')

bars = ax.bar(
    df['Year'].astype(str),
    df['Z_Score'],
    color=df['Color'],
    edgecolor='white',
    linewidth=0.6,
    width=0.75,
    zorder=3
)

# Reference lines at standard deviation boundaries
for yval, label in [(-2, '-2σ'), (-1, '-1σ'), (0, 'μ'), (1, '+1σ'), (2, '+2σ')]:
    ax.axhline(y=yval, color='gray', linewidth=0.7, linestyle='--', alpha=0.5, zorder=1)
    ax.text(len(df) - 0.3, yval + 0.08, label, fontsize=8, color='gray',
            ha='right', va='bottom')

# Labels and formatting
ax.set_xlabel('Year', fontsize=12, fontweight='bold', labelpad=10)
ax.set_ylabel('Z-Score', fontsize=12, fontweight='bold', labelpad=10)
ax.set_title('Tornado Fatalities — Categorical Z-Score by Year',
             fontsize=16, fontweight='bold', pad=16)

ax.tick_params(axis='x', rotation=45, labelsize=9)
ax.tick_params(axis='y', labelsize=10)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)

# Custom legend
from matplotlib.patches import Patch
legend_items = [
    Patch(facecolor='darkblue',   label='< -2  (Very Low)'),
    Patch(facecolor='steelblue',  label='-2 to -1  (Low)'),
    Patch(facecolor='lightblue',  label='-1 to 0  (Below Avg)'),
    Patch(facecolor='lightcoral', label='0 to 1  (Above Avg)'),
    Patch(facecolor='tomato',     label='1 to 2  (High)'),
    Patch(facecolor='darkred',    label='> 2  (Very High)'),
]
ax.legend(handles=legend_items, loc='upper left', fontsize=9,
          framealpha=0.9, edgecolor='#ddd', title='Z-Score Category',
          title_fontsize=10)

plt.tight_layout()

import seaborn as sns

tor = pd.read_csv("tor_data/1950-2024_actual_tornadoes.csv")

tor = tor[tor.mag >= 1]

cols = ['mag', 'fat', 'len', 'wid']

sns.pairplot(tor[cols], kind="scatter", diag_kind="hist")