798b0744b5
2a — Replace physically invalid summed-Mw seismic metric with correct
log10(ΣE) where E=10^(1.5·Mw+4.8) J (Kanamori 1977). Updates
usgs.py (new seismic_energy_per_bin()), scripts 02/07/08.
Impact: r(+15d) raw 0.310→0.081, peak r 0.469→0.139; sinusoidal
BF 27.5→0.75 (constant model now preferred).
2b — HP λ derivation: λ_5 = 1600×(365/5)^4 ≈ 4.54×10^10 (Ravn & Uhlig
2002 rescaling); detrend robustness figure (HP/Butterworth/rolling
mean), all show r(+15d)<0.04.
2c — Neff comparison: Bartlett 2923, Bretherton 769, Monte Carlo 594;
MC 95% CI [-0.002, +0.158] straddles zero → raw r not significant
under most conservative estimate.
2d — Magnitude threshold (M≥4.5/5.0/6.0): r(+15d) range 0.050–0.079
(Δ=0.029 < 0.05), peak stable at τ=−525 d; no aftershock bias.
Paper (main.tex, refs.bib): update all numbers, add Kanamori/Bartlett/
Ravn–Uhlig refs, new sections for 2b–2d with figures and Neff table.
PDF recompiled (27 pages).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
667 lines
25 KiB
Python
667 lines
25 KiB
Python
#!/usr/bin/env python3
|
||
"""
|
||
scripts/02_homola_replication.py
|
||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||
Reproduce the Homola et al. 2023 (JASTP 247, 106068) cross-correlation analysis
|
||
as literally as possible from the paper.
|
||
|
||
Method (from the paper)
|
||
-----------------------
|
||
* Data: NMDB pressure-corrected hourly neutron counts; USGS earthquake catalogue.
|
||
* CR index: Each station is normalised to its own long-term mean count rate
|
||
(so all stations become dimensionless relative-rate series around 1.0).
|
||
The global index is the mean across all stations that have valid data at
|
||
each 5-day bin.
|
||
* Seismic metric: sum of Mw for all M ≥ 4.0 events per 5-day bin (Homola
|
||
metric; zero for bins with no events).
|
||
* Binning: non-overlapping 5-day bins anchored at study_start.
|
||
* Correlation: Pearson r(τ) for τ ∈ [-1000, +1000] days (step = 5 days).
|
||
Convention: τ > 0 means CR leads seismic (at lag τ the seismic series is
|
||
advanced τ days into the future relative to the CR series).
|
||
* Significance: "naive" Pearson t-test p-value (treats bins as i.i.d. — this
|
||
is the inflated significance that Homola reports; reproduced here as the
|
||
baseline before phase-randomised correction).
|
||
|
||
Outputs
|
||
-------
|
||
results/figs/homola_replication.png — r(τ) plot with peak annotated
|
||
results/homola_replication.json — machine-readable summary
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import argparse
|
||
import json
|
||
import logging
|
||
import subprocess
|
||
import sys
|
||
from datetime import datetime, timezone
|
||
from pathlib import Path
|
||
from typing import Optional
|
||
|
||
import matplotlib
|
||
matplotlib.use("Agg")
|
||
import matplotlib.pyplot as plt
|
||
import matplotlib.ticker as mticker
|
||
import numpy as np
|
||
import pandas as pd
|
||
import scipy.stats
|
||
import yaml
|
||
|
||
PROJECT_ROOT = Path(__file__).resolve().parent.parent
|
||
sys.path.insert(0, str(PROJECT_ROOT / "src"))
|
||
|
||
from crq.ingest.nmdb import load_station, resample_daily
|
||
from crq.ingest.usgs import load_usgs, seismic_energy_per_bin
|
||
|
||
logging.basicConfig(
|
||
level=logging.INFO,
|
||
format="%(asctime)s %(levelname)-8s %(name)s — %(message)s",
|
||
datefmt="%Y-%m-%dT%H:%M:%S",
|
||
)
|
||
logger = logging.getLogger("crq.homola")
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Constants (matching Homola et al. 2023)
|
||
# ---------------------------------------------------------------------------
|
||
STUDY_START = "1976-01-01"
|
||
STUDY_END = "2019-12-31"
|
||
BIN_DAYS = 5
|
||
LAG_MIN_DAYS = -1000
|
||
LAG_MAX_DAYS = +1000
|
||
MIN_MAG = 4.0
|
||
SEED = 42
|
||
|
||
NMDB_COVERAGE_THRESHOLD = 0.60 # min valid hourly fraction per station-day
|
||
MIN_STATIONS_PER_BIN = 3 # global index → NaN if fewer stations valid
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Helpers
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def _get_git_sha() -> str:
|
||
try:
|
||
r = subprocess.run(
|
||
["git", "-C", str(PROJECT_ROOT), "rev-parse", "--short", "HEAD"],
|
||
capture_output=True, text=True, timeout=5,
|
||
)
|
||
return r.stdout.strip() if r.returncode == 0 else "unknown"
|
||
except Exception:
|
||
return "unknown"
|
||
|
||
|
||
def _load_station_names(config_path: Path) -> list[str]:
|
||
with config_path.open() as fh:
|
||
return list(yaml.safe_load(fh)["stations"].keys())
|
||
|
||
|
||
def _parse_args() -> argparse.Namespace:
|
||
p = argparse.ArgumentParser(
|
||
description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter,
|
||
)
|
||
p.add_argument("--study-start", default=STUDY_START)
|
||
p.add_argument("--study-end", default=STUDY_END)
|
||
p.add_argument("--bin-days", type=int, default=BIN_DAYS)
|
||
p.add_argument("--lag-min", type=int, default=LAG_MIN_DAYS)
|
||
p.add_argument("--lag-max", type=int, default=LAG_MAX_DAYS)
|
||
p.add_argument("--min-mag", type=float, default=MIN_MAG)
|
||
p.add_argument("--min-stations", type=int, default=MIN_STATIONS_PER_BIN)
|
||
p.add_argument("--nmdb-dir", type=Path, default=PROJECT_ROOT / "data" / "raw" / "nmdb")
|
||
p.add_argument("--usgs-dir", type=Path, default=PROJECT_ROOT / "data" / "raw" / "usgs")
|
||
p.add_argument("--output-dir", type=Path, default=PROJECT_ROOT / "results")
|
||
p.add_argument("--config", type=Path, default=PROJECT_ROOT / "config" / "stations.yaml")
|
||
p.add_argument("--seed", type=int, default=SEED)
|
||
return p.parse_args()
|
||
|
||
|
||
def _bin_series(
|
||
series: pd.Series,
|
||
study_start: str,
|
||
bin_days: int,
|
||
agg: str = "mean",
|
||
) -> pd.Series:
|
||
"""
|
||
Resample a daily Series to *bin_days*-day bins anchored exactly at
|
||
*study_start*.
|
||
|
||
Uses manual floor-division binning so the anchor date is always respected,
|
||
regardless of pandas version or calendar effects.
|
||
|
||
Parameters
|
||
----------
|
||
agg : ``"mean"`` (default) or ``"sum"``
|
||
Aggregation function. Use ``"mean"`` for normalised CR rates;
|
||
use ``"sum"`` for the seismic Mw total.
|
||
"""
|
||
t0 = pd.Timestamp(study_start)
|
||
days_from_origin = (series.index - t0).days
|
||
bin_number = days_from_origin // bin_days
|
||
bin_dates = t0 + pd.to_timedelta(bin_number * bin_days, unit="D")
|
||
grouped = series.groupby(bin_dates)
|
||
return grouped.sum() if agg == "sum" else grouped.mean()
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Step 1 — Global cosmic-ray index
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def build_cr_index(
|
||
station_names: list[str],
|
||
study_start: str,
|
||
study_end: str,
|
||
bin_days: int,
|
||
nmdb_dir: Path,
|
||
*,
|
||
coverage_threshold: float = NMDB_COVERAGE_THRESHOLD,
|
||
min_stations: int = MIN_STATIONS_PER_BIN,
|
||
) -> tuple[pd.Series, pd.DataFrame, int]:
|
||
"""
|
||
Load all available NMDB stations, normalise each to its mean count rate
|
||
over the study window, and return a *bin_days*-day-binned global index.
|
||
|
||
Returns
|
||
-------
|
||
cr_index : binned global CR series (NaN if < min_stations valid)
|
||
station_norm_daily: daily normalised series per station (for diagnostics)
|
||
effective_min : the min_stations threshold actually applied
|
||
"""
|
||
t0 = pd.Timestamp(study_start)
|
||
t1 = pd.Timestamp(study_end)
|
||
start_year, end_year = t0.year, t1.year
|
||
|
||
norm_cols: dict[str, pd.Series] = {}
|
||
|
||
for station in station_names:
|
||
hourly = load_station(station, start_year, end_year, nmdb_dir)
|
||
if hourly.empty:
|
||
continue
|
||
|
||
daily_df = resample_daily(hourly, station, coverage_threshold=coverage_threshold)
|
||
daily = daily_df[station].loc[study_start:study_end]
|
||
|
||
n_valid = int(daily.notna().sum())
|
||
if n_valid < 30:
|
||
continue
|
||
|
||
station_mean = daily.mean()
|
||
if not np.isfinite(station_mean) or station_mean <= 0:
|
||
logger.warning("station %s: non-positive mean — skipping", station)
|
||
continue
|
||
|
||
norm_cols[station] = daily / station_mean # dimensionless ~1.0
|
||
n_total = (t1 - t0).days + 1
|
||
coverage = 100 * n_valid / n_total
|
||
logger.info(
|
||
"station %-6s valid_days=%5d coverage=%5.1f%% mean=%.1f",
|
||
station, n_valid, coverage, station_mean,
|
||
)
|
||
|
||
n_loaded = len(norm_cols)
|
||
logger.info("Loaded %d stations with valid data.", n_loaded)
|
||
|
||
if n_loaded == 0:
|
||
raise RuntimeError(
|
||
"No valid station data found — run scripts/01_download_data.py first."
|
||
)
|
||
|
||
# Auto-adjust min_stations when data is sparse (e.g. subset test)
|
||
effective_min = min_stations
|
||
if n_loaded < min_stations:
|
||
effective_min = max(1, n_loaded - 1) if n_loaded > 1 else 1
|
||
logger.warning(
|
||
"Only %d stations loaded but min_stations=%d requested; "
|
||
"using effective_min=%d. Results with <%d stations are unreliable.",
|
||
n_loaded, min_stations, effective_min, min_stations,
|
||
)
|
||
|
||
station_norm_daily = pd.DataFrame(norm_cols)
|
||
|
||
# Global daily index
|
||
n_valid_per_day = station_norm_daily.notna().sum(axis=1)
|
||
global_daily = station_norm_daily.mean(axis=1) # NaN-safe
|
||
global_daily[n_valid_per_day < effective_min] = np.nan
|
||
|
||
# Bin to bin_days
|
||
cr_index = _bin_series(global_daily, study_start, bin_days)
|
||
cr_index.name = "cr_index"
|
||
|
||
n_valid_bins = int(cr_index.notna().sum())
|
||
logger.info(
|
||
"CR index: %d %d-day bins, %d non-NaN (%.1f%%)",
|
||
len(cr_index), bin_days, n_valid_bins,
|
||
100 * n_valid_bins / max(len(cr_index), 1),
|
||
)
|
||
return cr_index, station_norm_daily, effective_min
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Step 2 — Seismic metric
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def build_seismic_metric(
|
||
study_start: str,
|
||
study_end: str,
|
||
bin_days: int,
|
||
usgs_dir: Path,
|
||
*,
|
||
min_mag: float = MIN_MAG,
|
||
cr_index: Optional[pd.Series] = None,
|
||
) -> pd.Series:
|
||
"""
|
||
Sum of Mw for all M ≥ *min_mag* events per *bin_days*-day bin.
|
||
|
||
Aligned to *cr_index*'s index if supplied; otherwise built from scratch.
|
||
Bins with no events → 0.0.
|
||
"""
|
||
t0 = pd.Timestamp(study_start)
|
||
t1 = pd.Timestamp(study_end)
|
||
|
||
events = load_usgs(t0.year, t1.year, usgs_dir)
|
||
if events.empty:
|
||
raise RuntimeError("No USGS data — run scripts/01_download_data.py first.")
|
||
|
||
events = events.loc[study_start:study_end]
|
||
events = events[events["mag"] >= min_mag]
|
||
logger.info(
|
||
"Seismic: %d events M≥%.1f in %s–%s",
|
||
len(events), min_mag, study_start, study_end,
|
||
)
|
||
|
||
# Physically correct: E = 10^(1.5·Mw + 4.8) J per event, summed per bin,
|
||
# returned as log10(E_sum). See Kanamori (1977).
|
||
t0 = pd.Timestamp(study_start)
|
||
seismic = seismic_energy_per_bin(
|
||
events, study_start, study_end, bin_days, t0, min_mag=min_mag,
|
||
)
|
||
seismic = seismic.fillna(seismic.min()) # fill rare empty bins with floor
|
||
|
||
# Align to CR index bins
|
||
if cr_index is not None:
|
||
seismic = seismic.reindex(cr_index.index, fill_value=seismic.min())
|
||
|
||
nonnan_pct = 100 * seismic.notna().mean()
|
||
logger.info(
|
||
"Seismic metric (log10 E): %d bins, %.1f%% non-NaN, "
|
||
"range [%.1f, %.1f] log10(J)",
|
||
len(seismic), nonnan_pct, float(seismic.min()), float(seismic.max()),
|
||
)
|
||
return seismic
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Step 3 — Cross-correlation
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def pearson_lag_correlation(
|
||
x: pd.Series,
|
||
y: pd.Series,
|
||
lag_bins: np.ndarray,
|
||
) -> pd.DataFrame:
|
||
"""
|
||
Pearson r(τ) for each lag in *lag_bins* (bin units).
|
||
|
||
Convention: τ > 0 means x leads y.
|
||
At lag τ > 0: x[0..N-τ] vs y[τ..N]
|
||
At lag τ < 0: x[|τ|..N] vs y[0..N-|τ|]
|
||
|
||
Returns DataFrame: lag_bins, r, p_value, n_pairs.
|
||
"""
|
||
x_arr = np.asarray(x, dtype=float)
|
||
y_arr = np.asarray(y, dtype=float)
|
||
N = len(x_arr)
|
||
|
||
rs, ps, ns = [], [], []
|
||
for lag in lag_bins:
|
||
if lag >= 0:
|
||
end = N - lag
|
||
if end <= 0:
|
||
rs.append(np.nan); ps.append(np.nan); ns.append(0)
|
||
continue
|
||
xa = x_arr[:end]
|
||
ya = y_arr[lag : lag + end]
|
||
else:
|
||
abslag = -lag
|
||
end = N - abslag
|
||
if end <= 0:
|
||
rs.append(np.nan); ps.append(np.nan); ns.append(0)
|
||
continue
|
||
xa = x_arr[abslag : abslag + end]
|
||
ya = y_arr[:end]
|
||
|
||
valid = np.isfinite(xa) & np.isfinite(ya)
|
||
n = int(valid.sum())
|
||
ns.append(n)
|
||
if n < 10:
|
||
rs.append(np.nan); ps.append(np.nan)
|
||
continue
|
||
|
||
r, p = scipy.stats.pearsonr(xa[valid], ya[valid])
|
||
rs.append(float(r)); ps.append(float(p))
|
||
|
||
return pd.DataFrame({"lag_bins": lag_bins, "r": rs, "p_value": ps, "n_pairs": ns})
|
||
|
||
|
||
def naive_sigma(p_value: float) -> float:
|
||
"""Two-tailed p → Gaussian sigma z = Φ⁻¹(1 − p/2)."""
|
||
if not np.isfinite(p_value) or p_value <= 0:
|
||
return np.inf
|
||
return float(scipy.stats.norm.isf(p_value / 2.0))
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Step 4 — Figure
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def make_figure(
|
||
lag_df: pd.DataFrame,
|
||
bin_days: int,
|
||
peak_lag_days: float,
|
||
peak_r: float,
|
||
n_stations: int,
|
||
study_start: str,
|
||
study_end: str,
|
||
naive_p: float,
|
||
sigma: float,
|
||
cr_5d: pd.Series,
|
||
seismic_5d: pd.Series,
|
||
n_stations_eff: int,
|
||
r_at_15: float,
|
||
sigma_at_15: float,
|
||
) -> plt.Figure:
|
||
lags_days = lag_df["lag_bins"].values * bin_days
|
||
r_vals = lag_df["r"].values
|
||
|
||
n0 = int(lag_df.loc[lag_df["lag_bins"] == 0, "n_pairs"].iloc[0])
|
||
naive_ci95 = 1.96 / np.sqrt(n0) if n0 > 4 else 0.0
|
||
|
||
# Naive r thresholds for σ=3 and σ=6 (two-tailed, N=n0)
|
||
def _r_for_sigma(s: float, n: int) -> float:
|
||
"""r value corresponding to naive σ=s for sample size n."""
|
||
p = 2.0 * scipy.stats.norm.sf(s)
|
||
t_crit = scipy.stats.t.isf(p / 2.0, df=n - 2)
|
||
return float(t_crit / np.sqrt(t_crit**2 + n - 2))
|
||
|
||
r_sigma3 = _r_for_sigma(3.0, n0) if n0 > 4 else np.nan
|
||
r_sigma6 = _r_for_sigma(6.0, n0) if n0 > 4 else np.nan
|
||
|
||
fig, axes = plt.subplots(
|
||
3, 1, figsize=(14, 12),
|
||
gridspec_kw={"height_ratios": [3, 1.4, 1.4]},
|
||
)
|
||
|
||
# ── Panel 1: r(τ) ──────────────────────────────────────────────────────
|
||
ax = axes[0]
|
||
ax.axhline(0, color="k", lw=0.8)
|
||
ax.axhline( naive_ci95, color="C0", lw=0.8, ls="--",
|
||
label=f"±{naive_ci95:.3f} naive 95 % CI (n={n0:,})")
|
||
ax.axhline(-naive_ci95, color="C0", lw=0.8, ls="--")
|
||
|
||
if np.isfinite(r_sigma3):
|
||
ax.axhline( r_sigma3, color="C2", lw=0.9, ls=":", alpha=0.7,
|
||
label=f"Naive 3σ (r={r_sigma3:.3f})")
|
||
ax.axhline(-r_sigma3, color="C2", lw=0.9, ls=":", alpha=0.7)
|
||
if np.isfinite(r_sigma6):
|
||
ax.axhline( r_sigma6, color="C1", lw=0.9, ls="-.", alpha=0.8,
|
||
label=f"Naive 6σ (r={r_sigma6:.3f}) ← Homola claim")
|
||
ax.axhline(-r_sigma6, color="C1", lw=0.9, ls="-.", alpha=0.8)
|
||
|
||
ax.axvline(0, color="k", lw=0.6, alpha=0.35)
|
||
|
||
ax.plot(lags_days, r_vals, color="k", lw=1.1, zorder=3)
|
||
|
||
# Mark the τ=+15d bin (Homola claimed signal)
|
||
lag15_bins = 15 // bin_days
|
||
row15 = lag_df[lag_df["lag_bins"] == lag15_bins]
|
||
if len(row15):
|
||
r15 = float(row15["r"].iloc[0])
|
||
if np.isfinite(r15):
|
||
s15_str = f"{sigma_at_15:.1f}σ" if np.isfinite(sigma_at_15) and sigma_at_15 < 100 else ">8σ"
|
||
ax.scatter([15], [r15], color="darkorange", s=60, zorder=6,
|
||
label=f"τ = +15 d: r = {r15:+.4f} ({s15_str} naive)")
|
||
ax.axvline(15, color="darkorange", lw=1.0, ls=":", alpha=0.6)
|
||
|
||
# Mark the overall peak
|
||
ax.axvline(peak_lag_days, color="crimson", lw=1.3, ls="--",
|
||
label=f"Peak τ = {peak_lag_days:+.0f} d, r = {peak_r:+.4f}")
|
||
ax.scatter([peak_lag_days], [peak_r], color="crimson", s=50, zorder=5)
|
||
|
||
r_clean = r_vals[np.isfinite(r_vals)]
|
||
ylo = min(-0.15, r_clean.min() * 1.15) if len(r_clean) else -0.15
|
||
yhi = max( 0.15, r_clean.max() * 1.15) if len(r_clean) else 0.15
|
||
ax.set_xlim(lags_days[0], lags_days[-1])
|
||
ax.set_ylim(ylo, yhi)
|
||
ax.set_xlabel("Lag τ (days) [τ > 0: CR leads seismic]")
|
||
ax.set_ylabel("Pearson r")
|
||
ax.set_title(
|
||
f"Homola et al. 2023 — replication baseline | "
|
||
f"{n_stations} NMDB stations | "
|
||
f"{study_start[:4]}–{study_end[:4]} | "
|
||
f"5-day bins | M≥{MIN_MAG:.1f}",
|
||
fontsize=10,
|
||
)
|
||
ax.legend(fontsize=8.5, loc="upper right", ncol=2)
|
||
ax.grid(True, alpha=0.22)
|
||
|
||
sig_str = f"{sigma:.1f}σ" if np.isfinite(sigma) and sigma < 100 else ">8σ"
|
||
info = (
|
||
f"Peak: r = {peak_r:+.5f} at τ = {peak_lag_days:+.0f} d ({sig_str} naive)\n"
|
||
f"τ = +15 d: r = {r_at_15:+.5f} "
|
||
+ (f"({sigma_at_15:.1f}σ naive)" if np.isfinite(sigma_at_15) and sigma_at_15 < 100 else "")
|
||
+ f"\nn = {n0:,} valid bin-pairs at τ=0\n"
|
||
f"⚠ Naive — ignores autocorr., solar cycle, lag-scan penalty"
|
||
)
|
||
ax.text(
|
||
0.014, 0.975, info, transform=ax.transAxes,
|
||
va="top", ha="left", fontsize=8.5,
|
||
bbox=dict(boxstyle="round,pad=0.35", fc="lightyellow", alpha=0.92),
|
||
)
|
||
|
||
# ── Panel 2: CR index ──────────────────────────────────────────────────
|
||
ax2 = axes[1]
|
||
cr_dev = (cr_5d - 1.0) * 100 # % deviation from station-normalised mean
|
||
ax2.plot(cr_dev.index, cr_dev.values, lw=0.7, color="navy", alpha=0.85)
|
||
ax2.axhline(0, color="k", lw=0.5)
|
||
ax2.set_ylabel("CR deviation (%)")
|
||
ax2.set_title(
|
||
f"Global CR index ({n_stations} stations, each normalised to their long-term mean)",
|
||
fontsize=9,
|
||
)
|
||
ax2.grid(True, alpha=0.18)
|
||
ax2.set_xlim(cr_5d.index[0], cr_5d.index[-1])
|
||
|
||
# ── Panel 3: seismic metric ────────────────────────────────────────────
|
||
ax3 = axes[2]
|
||
ax3.fill_between(seismic_5d.index, seismic_5d.values, color="firebrick", alpha=0.6, lw=0)
|
||
ax3.set_ylabel(f"log₁₀(Σ E) [log₁₀ J] (M≥{MIN_MAG:.1f})")
|
||
ax3.set_title(
|
||
f"Global seismic metric (log₁₀ of summed energy per 5-day bin, "
|
||
f"E=10^(1.5·Mw+4.8) J)",
|
||
fontsize=9,
|
||
)
|
||
ax3.grid(True, alpha=0.18)
|
||
ax3.set_xlim(seismic_5d.index[0], seismic_5d.index[-1])
|
||
|
||
fig.tight_layout(pad=1.2)
|
||
return fig
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Main
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def main() -> None:
|
||
args = _parse_args()
|
||
np.random.seed(args.seed)
|
||
git_sha = _get_git_sha()
|
||
|
||
logger.info("Homola replication | git=%s seed=%d", git_sha, args.seed)
|
||
logger.info("Study window: %s – %s", args.study_start, args.study_end)
|
||
logger.info("Bin: %d days | Lags: %+d … %+d days",
|
||
args.bin_days, args.lag_min, args.lag_max)
|
||
|
||
args.output_dir.mkdir(parents=True, exist_ok=True)
|
||
figs_dir = args.output_dir / "figs"
|
||
figs_dir.mkdir(parents=True, exist_ok=True)
|
||
|
||
# ── 1. CR index
|
||
station_names = _load_station_names(args.config)
|
||
cr_5d, station_norm_daily, effective_min = build_cr_index(
|
||
station_names,
|
||
args.study_start, args.study_end,
|
||
args.bin_days, args.nmdb_dir,
|
||
coverage_threshold=NMDB_COVERAGE_THRESHOLD,
|
||
min_stations=args.min_stations,
|
||
)
|
||
n_stations = station_norm_daily.shape[1]
|
||
|
||
# ── 2. Seismic metric
|
||
seismic_5d = build_seismic_metric(
|
||
args.study_start, args.study_end,
|
||
args.bin_days, args.usgs_dir,
|
||
min_mag=args.min_mag,
|
||
cr_index=cr_5d,
|
||
)
|
||
|
||
# ── 3. Data quality assessment
|
||
n_cr_bins = int(cr_5d.notna().sum())
|
||
study_days = (pd.Timestamp(args.study_end) - pd.Timestamp(args.study_start)).days
|
||
expect_bins = study_days // args.bin_days
|
||
coverage_pct = 100.0 * n_cr_bins / max(expect_bins, 1)
|
||
|
||
if coverage_pct < 50:
|
||
logger.warning(
|
||
"⚠ CR data coverage = %.1f%% of study window. "
|
||
"Results are not meaningful with less than half the data. "
|
||
"Run: python scripts/01_download_data.py (full run, no --subset).",
|
||
coverage_pct,
|
||
)
|
||
|
||
# ── 4. Cross-correlation
|
||
lag_bins = np.arange(
|
||
args.lag_min // args.bin_days,
|
||
args.lag_max // args.bin_days + 1,
|
||
dtype=int,
|
||
)
|
||
logger.info("Computing Pearson r(τ) for %d lags …", len(lag_bins))
|
||
lag_df = pearson_lag_correlation(cr_5d, seismic_5d, lag_bins)
|
||
|
||
# ── 5. Peak statistics
|
||
valid_mask = lag_df["r"].notna()
|
||
if not valid_mask.any():
|
||
raise RuntimeError(
|
||
"All correlation values are NaN — data is insufficient for analysis."
|
||
)
|
||
|
||
peak_idx = lag_df.loc[valid_mask, "r"].abs().idxmax()
|
||
peak_row = lag_df.loc[peak_idx]
|
||
peak_lag_bins = int(peak_row["lag_bins"])
|
||
peak_lag_days = peak_lag_bins * args.bin_days
|
||
peak_r = float(peak_row["r"])
|
||
naive_p = float(peak_row["p_value"])
|
||
n_at_peak = int(peak_row["n_pairs"])
|
||
sigma = naive_sigma(naive_p)
|
||
|
||
# Correlation specifically at the Homola ±15-day claim
|
||
lag15_bins = 15 // args.bin_days # = 3 bins
|
||
row15 = lag_df[lag_df["lag_bins"] == lag15_bins]
|
||
r_at_15 = float(row15["r"].iloc[0]) if len(row15) else np.nan
|
||
p_at_15 = float(row15["p_value"].iloc[0]) if len(row15) else np.nan
|
||
sigma_at_15 = naive_sigma(p_at_15) if np.isfinite(p_at_15) else np.nan
|
||
|
||
logger.info("─" * 55)
|
||
logger.info("Peak r = %+.5f at τ = %+d days (%+d bins)",
|
||
peak_r, peak_lag_days, peak_lag_bins)
|
||
logger.info("Naive p = %.3e (%.1fσ)", naive_p,
|
||
sigma if np.isfinite(sigma) else 999)
|
||
logger.info("n pairs at peak = %d", n_at_peak)
|
||
logger.info("r at τ = +15 d = %+.5f (naive p = %.2e, %.1fσ)",
|
||
r_at_15, p_at_15 if np.isfinite(p_at_15) else 1,
|
||
sigma_at_15 if np.isfinite(sigma_at_15) else 0)
|
||
logger.info("─" * 55)
|
||
logger.info(
|
||
"NOTE: naive significance ignores autocorrelation, the shared solar-"
|
||
"cycle trend, and scanning over %d lags — expect gross over-significance.",
|
||
len(lag_bins),
|
||
)
|
||
|
||
# ── 6. Figure
|
||
fig = make_figure(
|
||
lag_df, args.bin_days, peak_lag_days, peak_r,
|
||
n_stations, args.study_start, args.study_end,
|
||
naive_p, sigma, cr_5d, seismic_5d,
|
||
n_stations_eff=effective_min,
|
||
r_at_15=r_at_15,
|
||
sigma_at_15=sigma_at_15,
|
||
)
|
||
fig_path = figs_dir / "homola_replication.png"
|
||
fig.savefig(fig_path, dpi=150, bbox_inches="tight")
|
||
plt.close(fig)
|
||
logger.info("Saved: %s", fig_path)
|
||
|
||
# ── 7. JSON log
|
||
results = {
|
||
"script": "scripts/02_homola_replication.py",
|
||
"git_sha": git_sha,
|
||
"seed": args.seed,
|
||
"timestamp_utc": datetime.now(timezone.utc).isoformat(),
|
||
"study_window": [args.study_start, args.study_end],
|
||
"bin_days": args.bin_days,
|
||
"lag_range_days": [args.lag_min, args.lag_max],
|
||
"min_magnitude": args.min_mag,
|
||
"nmdb_coverage_threshold": NMDB_COVERAGE_THRESHOLD,
|
||
"min_stations_requested": args.min_stations,
|
||
"min_stations_effective": effective_min,
|
||
"n_stations_used": n_stations,
|
||
"n_cr_bins_total": int(len(cr_5d)),
|
||
"n_cr_bins_valid": n_cr_bins,
|
||
"cr_data_coverage_pct": round(coverage_pct, 2),
|
||
"peak_lag_days": peak_lag_days,
|
||
"peak_lag_bins": peak_lag_bins,
|
||
"peak_r": round(peak_r, 6),
|
||
"naive_p_value": float(f"{naive_p:.6e}"),
|
||
"naive_sigma": round(sigma, 3) if np.isfinite(sigma) and sigma < 1e6 else None,
|
||
"n_pairs_at_peak": n_at_peak,
|
||
"r_at_tau_15d": round(r_at_15, 6) if np.isfinite(r_at_15) else None,
|
||
"p_at_tau_15d": float(f"{p_at_15:.6e}") if np.isfinite(p_at_15) else None,
|
||
"sigma_at_tau_15d": round(sigma_at_15, 3) if np.isfinite(sigma_at_15) else None,
|
||
"n_lags_scanned": len(lag_bins),
|
||
"note_significance": (
|
||
"Naive p from Pearson t-test treating 5-day bins as i.i.d. "
|
||
"Not corrected for: (1) temporal autocorrelation, "
|
||
"(2) shared solar-cycle trend, (3) scan over multiple lags. "
|
||
"Expected to be grossly over-significant — corrected tests in phase 2."
|
||
),
|
||
}
|
||
json_path = args.output_dir / "homola_replication.json"
|
||
json_path.write_text(json.dumps(results, indent=2))
|
||
logger.info("Saved: %s", json_path)
|
||
|
||
# ── Human summary
|
||
print()
|
||
print("=" * 62)
|
||
print(" HOMOLA ET AL. 2023 — REPLICATION SUMMARY")
|
||
print("=" * 62)
|
||
print(f" Stations used : {n_stations}")
|
||
print(f" CR data coverage : {coverage_pct:.1f}% of study window")
|
||
print(f" Valid 5-day bins (CR) : {n_cr_bins:,}")
|
||
print()
|
||
print(f" Peak |r| : {abs(peak_r):.5f} (r = {peak_r:+.5f})")
|
||
print(f" Peak lag τ : {peak_lag_days:+d} days")
|
||
print(f" Naive p-value at peak : {naive_p:.3e}")
|
||
sstr = f"{sigma:.2f}σ" if np.isfinite(sigma) and sigma < 1e6 else ">8σ"
|
||
print(f" Naive sigma at peak : {sstr}")
|
||
print()
|
||
s15str = f"{sigma_at_15:.2f}σ" if np.isfinite(sigma_at_15) else "N/A"
|
||
print(f" r at τ = +15 d (claim) : {r_at_15:+.5f} ({s15str})")
|
||
print()
|
||
if coverage_pct < 50:
|
||
print(" ⚠ WARNING: data coverage < 50% — results not meaningful.")
|
||
print(" python scripts/01_download_data.py (full run)")
|
||
else:
|
||
print(" ✓ Data coverage sufficient.")
|
||
print("=" * 62)
|
||
|
||
|
||
if __name__ == "__main__":
|
||
main()
|