Teleconnection explorer

Data. Precip = ERA5 daily (Open-Meteo archive), a small areal mean of grid cells over the resort — a reanalysis proxy that under-resolves peak orographic snowfall but captures the large-scale precip anomaly we composite. Southern-Hemisphere resorts use the austral winter (Apr–Oct) and its seasons (AMJ…ASO). MJO: RMM (Wheeler–Hendon, 1974–2020) or OMI (ERA5, 1950–2023). ENSO = CPC RONI (Relative ONI — Niño-3.4 minus the tropical-mean SST anomaly, which removes tropical-wide warming; e.g. 2023-24 is “Niño,” not “Strong”). ENSO is classified per winter, not per day: each winter takes one category from its core-winter-mean RONI (DJF for N. Hem, JJA for S. Hem), applied to all its days. The SST peak (ND–DJF) leads the downstream response by weeks, so a decaying strong event still “owns” its late-winter (JFM/FMA) days — 1997-98 stays a super-El-Niño winter through March. QBO = CPC 50 hPa, PDO = NCEI ERSSTv5, SOI = CPC, PNA/AO/NAO = CPC daily. WPO (West Pacific) and EPO (EP/NP) are CPC monthly patterns (labeled “mo” — one value per month, coarser than the daily modes; EPO is null in months where it isn’t a leading mode). These are Northern-Hemisphere indices — for S. Hem resorts they still apply (global MJO/ENSO), but the extratropical modes (PNA/AO/NAO/WPO/EPO) describe the NH and are less relevant there. 1950–2023.

Reading it. “Precip anomaly” is in standard deviations vs. climatology. The index responses (PNA/AO/NAO) are shown in σ relative to the current selection’s mean — active-MJO winters carry a background offset (e.g. +PNA), so centering on the selection exposes the phase-to-phase contrast (the actual teleconnection signal) instead of that offset. “Precip % of avg” is the precip departure over a rolling 7-day (weekly) or 30-day (monthly) window vs. the day-of-year normal — deseasonalized, so it isolates the factor’s effect rather than the annual cycle (0 = normal, +30% = a third above average). A solid outline = 95% significant (autocorrelation-adjusted; the % metrics deflate the sample by the window length); dashed/faded = fewer than 15 days, treat as noise. Two-way splits and stacked filters thin the sample fast — watch the day count and significance, not cell color alone. Correlated factors (e.g. ENSO × PDO, ENSO × SOI) can be near-empty.

PNA as mediator. Set rows = MJO phase, cell = PNA to see the MJO force a downstream PNA response; then rows = PNA, cell = precip to see PNA drive the resort’s precip; then filter by PNA to test whether it absorbs the MJO→precip signal. (Jet/500 hPa and AR-day responses need ERA5 upper-air data — not yet wired in.)