ERA5 / DFS annual diagnostics
Regional Heatwave Time-Series Diagnostics
Examine annual aggregation, reference-period anomalies, regional contrasts, and cross-metric structure before interpreting long-term change.
Loading annual regional series...
- Source
- ERA5 daily maximum 2 m temperature
- Coverage
- 1940-2025
- Detection
- Calendar-day P90, minimum 3 days
- Reference
- 1981-2010
- Frequency
- Detected events per region-year
- Event metrics
- Arithmetic mean across events in each region-year
- Zero encoding
- 0 denotes no detected event in that region-year
- Archive release
- 2026-01-29
Shared analytical scope
Define the time-series comparison
Annual time series
Detected-event frequency
Descriptive diagnostics
Series structure
| Series | N | Slope/decade | R² | SD | Lag-1 r |
|---|
- Peak year
- -
- Peak value
- -
- Latest percentile
- -
- Total events
- -
OLS and lag-1 correlation describe the selected annual series. They do not account for autocorrelated errors, field significance, or attribution.
Reference-aware distribution
Annual values by analysis period
| Period | Calendar years | Valid N | Events | Mean | SD | P10 | P50 | P90 |
|---|
For event-mean metrics, “event years only” excludes annual zeros that encode no detected event. Frequency always retains zero-event years.
IPCC WGI AR6 comparison
Regional annual-series diagnostics
| Region | Type | Valid N | Events | Reference mean | Recent mean | Absolute shift | Relative shift | Slope/decade | R² | Latest |
|---|
Select a regional row to make it the primary time series.
Cross-metric structure
Annual correlation matrix
Pairwise coefficients use overlapping annual values in the selected window. They are not detrended and are not adjusted for serial dependence or multiple comparisons.
Reproducible annual evidence
Selected-region annual table
| Year | Events | HWMS mean | HWD mean | HWMA mean | HWMM mean | sumofMd mean | Period |
|---|
Interpretation guardrails
What an annual chart can and cannot establish
Use Insights to inspect event-level quantiles, extremes, duration, area, and magnitude together.
For event-mean metrics, zero indicates no detected event in that region-year, not a measured zero magnitude.
Shared trends and serial dependence can produce strong annual correlations without causal dependence.