Most at-risk regions for near-term surprise extremes

Impacts resulting from extreme weather and climate events depend on a multitude of factors, such as event intensity, and the level of risk awareness and preparation of the affected region. Climate change exacerbates extreme temperatures and extreme rainfall in most land regions, yet, natural variability can regionally mask or amplify the forced trend for years or decades. In regions where the forced trend of a certain extreme has been masked for an extended period, local communities may be inexperienced with, unaware of, and unprepared for record extremes that exceed the most extreme event in history. Global mean temperatures are currently warming at a rate unprecedented in the observational record. The ratio of occurrence of daily surface temperature records since 1950 relative to the theoretically expected occurrence in a stationary climate, is now about 3–3.5 for hot records globally [Fischer et al. (2025)]. The forced signal in annual maximum daily precipitation rates lags behind the forced signal in temperature, with a global record occurrence ratio of about 1.5 [Fischer et al. (2025)]. To prepare for future record heat and rainfall, it is crucial to identify regions where the probability of breaking or shattering the local standing record is highest. Here, we aim to quantify which regions are exposed to particularly high risk of “surprise extremes” in temperature and rainfall, taking into account local extreme event history.

We quantify local record-breaking probabilities conditional on the standing record level and identify hotspots of high near-term record-breaking probability, using several statistical detection algorithms and extreme value theory. The forecast skill and robustness of our method is evaluated using several climate models, and applied to observations and reanalysis for real world record projections.

The global pattern of near-term conditional record-breaking probabilities shows clear signatures of the forced climate change pattern, with strong modifications attributable to natural variability. It is these modifications that are most relevant for near-term risks. The conditional probability of settinga new record is particularly high in years and regions where the forced trend has been masked by unforced natural variability for (multi-)decadal periods, leading to little to no detectable forced trend and absence of record breaking. Ironically, it is thus often regions where recent reminders of climate change were absent that deserve particular attention with regard to preparing for high impact record-breaking extremes. Quantitative estimates of conditional record-breaking probability are uncertain; the largest source of uncertainty lies in the separation of historical trends into forced response and internal variability. We evaluate different methods to estimate the forced response and to determine non-stationary extreme value distributions. While the exact conditional probability is uncertain, we find that hotspot regions where the conditional record probability is high can be robustly identified.