Gaps and ways forward in atmospheric blocking and extreme weather research

Abstract

Atmospheric blocking often results in significant weather extremes and associated impacts such as heatwaves, droughts, wildfires, cold spells, and floods in mid-latitude regions. However, the physical processes behind blocking and associated extreme weather events are not well understood, hampering prediction and decision-making for mitigation and adaptation. Further, numerical models often struggle to simulate the frequency, duration, and geographic distribution of atmospheric blocking events. Here, we provide an overview of blocking-related weather extremes and impacts, as well as of our current understanding of the key physical processes. We identify knowledge gaps and current challenges and provide our perspective on potential ways forward.

Atmospheric blocking often causes weather extremes in the mid- to high latitudes, yet the physical processes underlying blocks are not well understood. This article provides an overview, identifies current knowledge gaps and challenges, and offers perspectives on potential ways forward.

Introduction

Atmospheric blocking often leads to high-impact weather extremes such as heatwaves, droughts, wildfires, cold outbreaks, and floods. Blocking associated extreme weather can have devastating impacts on ecosystems and society^1–4. However, our understanding of the complex thermodynamic and dynamical processes involved in blocking, as well as its potential responses to external forcings, remains incomplete. Persistent biases in the representation of blocking in numerical models^5–7 further complicate the prediction of blocking and weather extremes, as they statistically fall at the tails of a climatological distribution. The most recent IPCC report⁸ suggested an overall low confidence about future changes in the magnitude, frequency, and spatial distribution of large-scale atmospheric circulation patterns—including persistent and quasi-stationary blocking patterns. Such limitations hinder our ability and confidence in prediction of subseasonal to decadal timescales and projections for the coming decades, thereby impeding informed decisions and policy-making in relation to mitigation and adaptation. Here, we identify and discuss the progress that has been made during the past decade and highlight knowledge gaps, with the goal of stimulating discussion on the paths forward to 1) deepen our understanding of atmospheric blocking, 2) guide the prediction of blocking on synoptic to subseasonal time scales, 3) improve confidence in future projections of blocking, and 4) design and coordinate a community effort to identify key physical processes in both weather and climate models by leveraging the crosscutting nature of blocking and associated extreme weather.

Atmospheric blocking: impacts and mechanisms

Atmospheric blocking is a weather phenomenon in which a large and persistent high-pressure system, sometimes combined with a low-pressure region, disrupts the normal eastward flow of the jet stream. This disruption can lead to extreme weather events in mid-latitude regions. The important connection between atmospheric blocking and extreme weather events is highlighted here through two significant consecutive occurrences in early 2025 in the western United States. In January, 14 devastating wildfires struck the Los Angeles metropolitan area⁹, resulting in widespread ecological destruction, billions of dollars in economic losses, and significant impacts on human health and safety, including at least 30 deaths. More than 200,000 people were forced to evacuate and over 18,000 homes and structures were destroyed. As shown in Fig. 1a, this event was associated with an Omega-like blocking pattern that developed over the West Coast on January 8 and lasted roughly one week. The northern edge of the cyclonic anomaly, centered over the southwestern U.S., contributed to the Santa Ana Winds in the LA region. The strong wind gusts¹⁰ of 80 to 130 km/h, combined with warm and dry surface conditions producing an exceptionally high vapor pressure deficit of up to 1.6 kPa, all contributed to enhanced fire danger at this time. This weather pattern closely corresponded with the areas affected by the wildfires. Shortly after, on February 2, the omega-like pattern retrograded westward and amplified, evolving into a dipole block that persisted for a week over the central-north Pacific (Fig. 1b). A large amount of water vapor was transported along the southeastern flank of the associated cyclone, providing favorable conditions for a strong atmospheric river¹¹ that delivered extreme precipitation—over 11 trillion gallons—within a 7-day period to the northern West Coast of the U.S. These two events highlight the stark contrast in extremes associated with blocks in the same season: when blocking was positioned over the Western U.S. and coastlines, it exacerbated destructive wildfires, whereas when a block was located over the Pacific, its persistent flow pattern provided a conducive environment for a strong atmospheric river.

Fig. 1. Atmospheric blocking and associated extreme event examples in the western United States in 2025.

a 500-hPa geopotential height (Z500; contours) and surface vapor pressure deficit (VPD) on 8 January 2025 and active wildfire radiative power (shading) during January 2025; b Z500 (contour lines) and the total column water vapor (TCWV; shading) on 2 February 2025. Hatched areas denote regions where daily precipitation exceeds 10 mm.

Certain regions experience more pronounced impacts of compound extreme weather events that are associated with blocking patterns. In August 2024, an Omega block over the Ural Mountains led to a historical 10-day extreme precipitation event in northwestern India that contributed more than 60% of all-India rainfall¹². These blocking patterns not only lead to surface extreme weather events over land¹³ such as heatwaves^14,15 and extreme precipitation¹⁶, but they can also cause storm surge¹⁷ and marine heatwaves¹⁸ as the blocking system slowly propagates across the ocean-land interface. For example, from late June to early July 2021, a strong blocking pattern fueled by upstream diabatic forcing arising from precipitation within the atmosphere led to an unprecedented heatwave that shocked the Pacific Northwest and Canada in the following couple of days, causing over 1000 deaths as one of the most deadly extreme events in the region^4,19. Figure 2 illustrates the evolution of the marine heatwave in the Pacific Northwest in 2021, followed by the subsequent land heatwave. Around June 20, a strong anticyclone emerged over the central Pacific (180°−160°W), coinciding with positive sea surface temperature (SST) anomalies and the onset of the marine heatwave. Between June 27 and July 2, a notable land surface temperature anomaly occurred over the eastern Pacific and western North America (120°W-100°W), accompanied by an enhanced anticyclone and widespread land heatwave events. These patterns indicate that a coherent eastward-propagating atmospheric blocking pattern occurred at the same time as the sequence of a marine heatwave in the Northeast Pacific and an unprecedented land heatwave in the western U.S. and Canada. Further research is needed to determine whether this simultaneous occurrence is coincidental or driven by the same atmospheric blocking pattern.

Fig. 2. Connecting land and marine heatwaves, soil moisture, and droughts through atmospheric blocking events.

a Hovmöller diagram of surface temperature anomalies and 500 hPa geopotential height anomalies (contours) averaged between 47°N and 52°N from 10 June to 5 July 2021. Shading represents sea surface temperature (SST) and land surface temperature (TS) anomalies; Hatches denote marine and land heatwave events; Green contours represent geopotential height at 500 hPa. Major blocking events are marked by stars. The second row shows Hovmöller diagrams of soil moisture anomalies, and standardized precipitation evapotranspiration index (SPEI) and 500 hPa geopotential height anomalies (contours) at 49°N for 15 June to 5 July 2021. The soil moisture is meridionally averaged from 40°N to 50°N; b surface soil moisture (shading); c root-zone soil moisture (shading); d SPEI (shading).

Blocking events in the Southern Hemisphere are also associated with various devastating extreme weather events. In the summer of 2013–2014, a record-breaking drought struck South America, extending from the southern tip of the continent to southern Brazil. At the same time, an extreme marine heatwave developed over the western South Atlantic, with sea surface temperature (SST) anomalies exceeding 3 °C. Figure 3a shows the Z500 anomaly pattern during this event, on 5 February 2014. A blocking high dominated the western South Atlantic Ocean, persisting from 15 January to 13 February, consistently contributing to increased land and sea surface temperatures, worsening dry conditions, and ultimately leading to both the terrestrial and marine heatwave^18,20. Blocking can also be linked to extreme precipitation. According to the Australian Bureau of Meteorology²¹, in March 2021, heavy rainfall and severe flooding affected a large portion of eastern Australia, including the capital, Sydney. This event was attributed to a blocking anticyclone over the Tasman Sea, between southern Australia and New Zealand (Fig. 3b). The persistent anticyclone structure intensified the easterly flow, advecting moist air over the east coast²², leading to the extreme precipitation.

Fig. 3. Atmospheric blocking and associated extreme event examples in the Southern Hemisphere.

a 500-hPa geopotential height (Z500) anomaly (contour lines), standardized precipitation evapotranspiration index (SPEI), and sea surface temperature (SST) anomaly on 5 February 2014; b Z500 anomaly and total precipitation (ERA5) on 21 March 2021.

Blocking is a complex phenomenon involving a range of physical processes in the climate system. Significant progress has been made over the past several decades that advanced our physical understanding as well as the numerical simulation, prediction, and climate projections of blocking and associated surface extremes^13,23. At the beginning of a recent US CLIVAR blocking and extreme weather workshop, researchers were provided with four distinct patterns of stagnating atmospheric circulation as shown in Fig. 4 and were asked “which of these is blocking?”. While one Omega-block pattern (panel c) was characterized/recognized by nearly all experts as a block, researchers held different, even opposing, views regarding the other three patterns (panels a, b, d), indicating that fundamental aspects of blocks differ from one type to another. In addition, slow-moving ridges in low latitudes may have different mechanisms compared to anticyclonic patterns in mid-latitudes. These blocking experts’ diverse opinions on whether those are blocks or not reflect the complex nature of atmospheric blocking patterns.

Fig. 4. Survey results from the US CLIVAR “Blocking and extreme weather workshop”, where the participants were asked “which of these is blocking?”.

a 500-hPa geopotential height (dam; contours) and 850 hPa temperature (°C; shading) for a + 129 h ECMWF Control Forecast initialized at 0000 UTC 12 July 2022 and valid at 0900 UTC 17 July 2022. Adapted from “ECMWF Control Forecast (ex-HRES)” by ECMWF, licensed under CC BY 4.0, available at https://charts.ecmwf.int/products/medium-z500-t850; b 500-hPa geopotential height (dam; contours) and its anomaly (standardized anomaly; shading) for a + 84 h GFS forecast initialized at 0600 UTC 29 May 2018 and valid at 1800 UTC 1 June 2018; c 250-hPa geopotential height (m; black contours) and geopotential height anomalies (standardized anomaly; blue and red contours) and wind speed (m/s; shading), respectively from Dr. Winters’s NPJ Phase Diagram—Deterministic GFS Forecast at https://www.atmos.albany.edu/facstaff/awinters/realtime/Deterministic_NPJPD.php; d Composite of 200-hPa streamfunction anomaly (see “Methods”). The statistics for the Yes or No votes are demonstrated at the top of each panel.

Despite the consensus that blocks involve quasi-stationary and persistent anticyclonic circulation anomalies, there are abundant flavors of blocks. The idea of blocking “flavors” came about from workshop discussions on the difficulty of identifying a unified blocking definition, as different blocks appear to be governed by a different balance of physical mechanisms, can vary in their impacts within a given region, and further, blocks in different seasons behave distinctly²⁴. Broadly speaking, the varying aspects of blocking exist also in its characteristics: block type²⁵, extent, magnitude, and persistence. The many flavors of blocks - the diversity of blocking - make it a challenging and multifaceted problem to model, understand, and predict. This warrants a discussion on the current knowledge of key physical processes involved in atmospheric blocking, including Rossby wave propagation and breaking, atmospheric background state, moist convection, cloud radiative feedback, and ocean-atmosphere-land interactions.

Why does blocking happen in the atmosphere, and what are the mechanisms for its existence, onset, maintenance, and decay? Linear Rossby wave dynamics suggest a diminishing energy dispersion²⁶ causes propagating Rossby waves to run into each other as a simplest possible starting point (although it may be oversimplified as observed blocks are highly nonlinear). An early explanation for the blocking phenomenon is that it is a coherent structure lasting longer than typical weather systems due to nonlinear resonance between topographic/thermal forcing and planetary waves^27,28 or the eddying westerly flow^29–31, although frequent blockings can be found in aquaplanet models that have no land or topography^32–34. Importantly, the peak distribution of blocking migrates with the latitude of the westerly jet throughout the seasonal cycle³⁵. The onset and maintenance of blocking by eddy forcing and/or the interaction between eddies traveling along the jet and the slow blocked vortex have been the central theme of contemporary blocking studies. Eddy straining^36,37, the selective absorption mechanism³⁸, and the multi-scale interaction model³⁹ are among the most promising paradigms for explaining blocking onset and maintenance. Blocking events are often associated with wave breaking and/or pinching of high (low) potential vorticity (PV) anomalies from their respective high- (low-) latitude PV pool (see Woollings et al.²³ for a review of the fundamental dynamics of blocking). Recently, studies have identified the diabatic heating in the warm conveyor belt (WCB) ahead of a block to be a leading order factor for the rapid intensification of a block during onset^{19, 25, 40–43} that can favor the maintenance of a block through the generation of more negative upper-tropospheric PV anomalies^7,44. The importance of the WCB for blocking onset may serve as a conduit to the oceanic flux, linking the onset and maintenance to the heat exchange between the atmosphere and the underlying boundaries^45–48.

Quantitative theories for the statistics of blocking in terms of size, frequency, duration, and intensity have begun to emerge only recently⁴⁹. A “traffic jam” theory⁵⁰ inspired by the quadratic relationship between wave activity flux and local wave activity has been proposed. In this analogy, the wave activity flux is the traffic, and the jet stream transporting wave activity is the “highway”. When the wave activity flux reaches the carrying capacity of the “highway”, the pathway for a wave to flux through becomes ‘blocked’ (similar to traffic congestion), leading to the formation of a block. This theory is derived from quantitative diagnostics⁵¹ of midlatitude wave activity-mean wind interactions, with the potential⁵² to be developed into the first predictive theory for the frequency distribution of blocking. Recent work found that a downstream reduction in flow capacity is associated with blocking events, as lane closures favor traffic jams⁵³. Other recent theories focused on an energetics approach^54,55 and a PV gradient approach⁵⁶. Blocking is studied by both weather and climate communities, presenting a unique opportunity to amend the “weather-climate schism”⁵⁷. Among the many theories proposed for blocking, there appears to be a difference between a weather-focused view and a climate-focused view: the former focuses on diagnosing and predicting lifecycles, including precursors, onset, growth, maintenance, and decay, whereas the latter focuses on statistics, eddy cascades, scaling, and projections. Critical assessments of each theory and a reconciliation of these two viewpoints are still lacking.

Current knowledge gaps across several fronts

Progress has been made over the last several decades in our understanding of atmospheric blocking, as well as improvements in the simulation and projections of atmospheric blocking. Nevertheless, there are still many knowledge gaps and opportunities for future research, which we discuss below.

Metrics and physical processes underlying atmospheric blocking

Past literature has intensively focused on seeking a unified or improved definition or metric to identify blocking. We suggest a transition away from seeking a unified definition and instead focusing on understanding the strengths and weaknesses of each definition and embracing the diversity of blocking. This requires acknowledging that different blocking types may be best captured by different blocking metrics, and the physical mechanisms involved may be regionally dependent. As schematically illustrated in Fig. 5, a strong blocking event can arise from multiple key physical processes and usually involves interactions across several components of the Earth system. In this artform depiction of the 2021 Pacific Northwest heatwave blocking pattern, we show many contributing factors to the onset and growth of the block from i) the jet stream feeding wave activity to the block; ii) cyclogenesis ahead of the block; iii) a warm conveyor belt (WCB) with strong cross isentropic motion; iv) warmer than normal SSTs at the WCB inflow; and v) divergent outflow expanding and shifting the ridge poleward. This demonstrates the complexities of a blocking event, including the role of air-sea interactions for blocks developing over and near an ocean. Understanding the uncertainties in blocking statistics associated with different contributing factors to a block and the role they play in the resulting surface extremes is crucial.

Fig. 5. A schematic summarizing different processes contributing to atmospheric blocking during the extreme Pacific Northwest heatwave event in June 2021.

The role of an upstream cyclone, tropical moisture transport, and local warmer sea surface temperatures (SSTs) in enhancing diabatic forcing of blocking precursors; In the bottom layer, SST anomalies on 26 June, 2021, (deviation from climatology of 1979-2020, from NOAA); In the middle layer: total precipitable water; In the upper layer: zonal wind speed at 200 hPa (color shading), horizontal wind streamfunction at 200 hPa (contours), and horizontal wind direction at 200 hPa (vectors); all from ERA5.

What are the unique aspects of various blocking metrics? Table 1 shows four classes of blocking diagnostics starting from the generic yet effective “persistent flow anomaly” definition⁵⁸, which searches for strong 500-hPa geopotential height anomalies. While almost all methods agree on temporal criteria (i.e., persistence of at least 5 days), they each focus on different aspects of the spatial patterns.

Table 1.

A list of process-oriented blocking diagnostics

Diagnostics	Feature(s) detected and strength(s)	Physical process(es)
Anomaly method	500 hPa height anomalies that persist at least for 5 days58; or based on PV anomalies64 ;Strength: simplest possible approach. Can capture omega blocks and deep troughs.	A persistent pattern for the eddies.
Absolute/reversal method	Absolute meridional gradient of geopotential height on a 500 hPa surface that exceeds certain threshold for at least 5 days72,115; Meridional gradient of potential temperature on the 2-PVU surface (B index)35,116,117. Strength: leveraging PV invertibility principle. Captures dipole blocks and omega blocks that have a reversal.	A weakening or reversal in the westerly mean flow; Rossby wave breaking.
Hybrid method	Positive anomalies and a reversal of an absolute gradient90,118 for at least 5 days; Strength: avoids misdetection by the above two approaches.	Both persistent flow anomalies and a reduction of the mean flow; Wave breaking.
Wave activity method	Local wave activity that exceeds a certain threshold for at least 5 days19,98,119,120. Strength: a full dynamical variable that follows a budget equation.	Eddy forcing, including dry and diabatic processes, and the blocking responses.

A list of blocking diagnostics, including the key feature(s) to detect, their strength(s), and the key physical processes involved.

Despite the fact that the anomaly method focuses on the eddies and the absolute/reversal method focuses on the mean flow, they are connected physically through the classical linearized non-acceleration theorem⁵⁹ (see also this textbook⁶⁰ as derived from its equation 10.95):

$$\frac{\partial }{\partial t}\left(\overline{{q^{\prime }}^2}+2ū\frac{\partial \overline{q}}{\partial y}\right)\approx D$$ 1

where $q$ is barotropic PV (including eddies $q^{\prime }$ and zonal-mean field $\overline{q}$) and $ū$ is the mean zonal wind measuring the jet strength. In the absence of diabatic processes $D$, a change in the anomaly part $\overline{{q^{\prime }}^2}$ is compensated by an opposite change in the mean flow part $2ū\frac{\partial \overline{q}}{\partial y}$ consisting of both the jet strength and the meridional PV gradient. Across the different methods, the anomaly method focuses on the first term, the absolute/reversal method focuses on the second term, while the hybrid method attempts to capture effects from both terms. Similar to the hybrid method, the wave activity method is a nonlinear extension of a quantity $\frac{\overline{{q^{\prime }}^2}}{2\partial \overline{q}/\partial y}$ combining the eddy term and the mean PV gradient term, and can also quantify the role of diabatic heating¹⁹. Through the lens of the physical processes, these seemingly different methods are connected by the shared non-acceleration theorem of the eddy-mean flow interactions.

As previously mentioned, rather than seeking a unified definition of blocking, we want to highlight the diversity of blocking events and suggest that different blocking flavors may be better captured by different indices. We recognize that blocking is a collective term that describes a large variety of stagnation patterns and should not be interpreted as an agreed atmospheric process. Moving beyond the issue of a unified definition allowed for the main focus of the workshop to be the advancement of our process-level physical understanding and of process-oriented diagnostics⁶¹ as it relates to the connections between atmospheric blocking and extreme weather in a changing climate, and the interactions with other components of the climate system. Pinheiro et al.⁶² performed an intercomparison of blocking detection methods, raising the issue of the algorithm biases. Chan et al.⁶³ performed a comprehensive comparison of some blocking indices from the perspective of surface hot extremes. The large array of blocking indices often gives quite different results due to their different technical details rather than conceptual differences. For example, the anomaly method of Schwierz et al.⁶⁴ relaxes the quasi-stationarity conditions typically used, so that the identified blocks are often more mobile and hence blocking occurrence is more focused on the storm tracks compared to the other methods.

To move forward, we may learn from past successful lessons: for example, in the investigation of jet stream variability, earlier works^65,66 used a wide range of variables such as sea level pressure, sea surface temperature, zonal winds, and geopotential height, to characterize the north-south movement of the jet stream. As synoptic-scale eddies covary with the jet, the difficult question was to what extent the change in eddies is responsible for the shift in the jet and to what extent the shift of the jet is responsible for the change in eddies—known as the eddy feedback problem. Through a simplified framework arising from the zonal momentum equation⁶⁷, a major breakthrough occurred following the discovery and quantification of key physical processes of eddy feedbacks. Similarly, we need more work addressing key physical processes that drive blocks and their associated extremes, and on quantifying any feedbacks between key physical processes in observations, such as wave breaking and the straining of synoptic eddies, and covarying blocks. We suggest a renaissance of interest in theoretical inquiry of blocking, such as using dry and moist quasi-geostrophic models and idealized general circulation models to provide mechanistic insights. We may seek to understand how each mechanistic insight informs us about the physical processes at play in nature, and, collectively, we can identify the most likely explanation for the blocks. Looking beyond Earth, it remains an open question whether terrestrial blocking events share fundamental similarities with blocking-like stagnation patterns observed in the atmospheres of other planets in our solar system, such as Mars and Jupiter.

The emerging role of diabatic heating

Diabatic processes with strong latent heat release have been identified as a factor that can influence the blocking lifecycle^19,25,41. Based on a Lagrangian framework, Pfahl et al.⁴¹ quantified the role of latent heat release by tracking the air masses that enter blocks. They demonstrated that during blocking events, nearly half of the air masses have experienced cross-isentropic transport, indicating the comparable role between diabatic processes and dry processes in generating and maintaining blocks. The formation and intensification of blocks are often preceded by strong latent heating in the upstream extratropical cyclogenesis region⁴⁴. By performing sensitivity experiments, they confirmed that downstream blocking events would be largely weakened or even disappear if upstream latent heating was eliminated. However, blocking has been identified in dry idealized general circulation models^32,33, and therefore dry dynamics alone can produce blocks or block-like patterns. In this regard, the relative importance of diabatic effects and how they may influence blocking in the future needs further study. Most Coupled Model Intercomparison Project (CMIP) models simulate less frequent or weaker blocking under future warming scenarios, alluding to the possibility that diabatic heating may not be the dominant force in shaping the change in blocking. However, due to the substantial bias in simulating key characteristics of blocks, we have low confidence in models correctly capturing the diabatic effects in the first place. For instance, WCB outflow may be underestimated in CMIP6 simulations⁷. In the future, such diabatic effects may become more important with the projected increase in moisture content to strengthen or weaken blocks depending on the type of the blocks²⁵, with a potential to bring about larger blocks⁴⁹, especially in summer. The existing literature has laid a solid foundation for understanding the moist dynamics of atmospheric blocking, but more work is needed to quantitatively assess the impacts of diabatic processes on the life cycles (weather perspective) and statistics (climate perspective) of atmospheric blocking. This includes, but is not limited to, the feedback role of cloud radiative effects (CRE) onto the formation and intensity of blocking⁶⁸, and how upstream moist convection and mesoscale convective systems (MCS) might impact blocks. The multi-scale nature of blocking could mean that failure to capture the mesoscale convection may cause a forecast bust. Moreover, we need more work comparing the effects of diabatic heating as captured by different methods (e.g., the recent wave activity method and the well-established Lagrangian tracking of air parcels⁴¹) to understand how different diagnostic methods collectively reveal distinct physical processes.

Deficiencies of model simulations of blocking

Despite the emergence of possible consensus on a reduced European blocking frequency in a warmer climate, it remains unclear why models tend to underestimate the frequency of the occurrence of blocking and exhibit substantial inter-model differences, and why some models better simulate the observed blocking climatology compared to others. For instance, models with higher horizontal resolution or well-resolved stratosphere do not necessarily represent blocks better^69,70 and the importance of better resolving ocean processes is not yet clear⁷¹. Mean state biases have been shown to influence blocking metrics that rely on reversals of a gradient⁷². Indeed, the blocking biases in CMIP models are reduced when focusing on anomalies relative to the observed mean state⁷², implying that modeled blocking events may not be as biased as previously thought, but rather it is how we’ve chosen to investigate them in the presence of biased mean state gradients. Biases still exist even when factoring out the mean state, but they appear to be much weaker; thus, the reversal methods are being contaminated by errors in mean state gradients, noting that the blocking biases themselves do not appear to explain the mean state biases⁷². When comparing model representations of blocking, both an AMIP and a CMIP GFDL model have similar biases despite one being forced by prescribed SSTs while the other had a fully interacting and evolving ocean⁷³, suggesting that the origin of biases is largely to be found in the atmospheric model component. In forecasting models, it is unclear how much forecast error is due to the intrinsic limits of predictability of blocks and how much is due to model biases. The model biases imply at least some contribution from model errors, but it is not clear which aspects/components of the models are the main contributors to forecast errors, the improvements of which will likely lead to significant improvements in blocking forecasts. Diabatic processes could be a key source of blocking error in models, although interestingly, blocking was found to have more sensitivity to orographic drag than moist physics parameters in one model study⁷⁴.

Different flavors of blocks and extremes

While we use the term “blocking” to broadly coin a class of large-scale stagnation flow patterns in the atmosphere, each block is different, and their relationships to extreme weather shall also be assessed with respect to its own flavor. To demonstrate the complexity of blocking events and their distinct flavors as they relate to weather extremes, Fig. 6 shows that blocks may be associated with different surface conditions (hot vs cold, rainy vs dry), even at the same longitude and location. The analysis includes the top 100 events in boreal winter and summer, respectively, ranked by blocking intensity as measured by wave activity strength. These four types are distinguished by their associated near-surface thermal and hydrological characteristics. To illustrate the spatial patterns of each category, representative cases are also shown in the right panels. The first row shows a typical hot-dry block centered over western Asia in August 2010, associated with the devastating 2010 Russian heatwave and drought. Both high temperatures and suppressed precipitation occurred under the blocking high. The second case is a hot-rainy block over the North Pacific in August 2020, where strong positive precipitation and temperature anomalies appear southeast of the low-pressure system. The third panel presents a cold-dry block in December 1984 over the eastern Pacific, with cold and dry anomalies east of the blocking high, affecting North America and the U.S. West Coast. The final case is a cold-rainy block over eastern Siberia during the same period. It features a strong low-pressure center inducing extreme cold, while heavy precipitation occurs on the southeastern side of the trough. On average, blocks feature regional dry conditions immediately under the blocking anticyclone—the block’s eye, if we draw an analogy to a Hurricane’s eye. However, the blocked region also causes strong poleward transport of moisture and enhanced precipitation, both about 10 degrees upstream and downstream. Rainy blocks may be accompanied by enhanced precipitation on both sides, while dry blocks may be dry for most of the area. This longitudinal structure of blocking-induced precipitation suggests that blocks reshape the regional hydrological cycle by redistributing the precipitation processes. Beyond temperature and precipitation extremes, as demonstrated here, other flavors of blocks may include those associated with wildfires, droughts, storm surge, mesoscale convective systems, etc. More work is needed to capture the full spectrum of blocking flavors. This reinforces the need to embrace the complex nature of blocking, or blocking “flavors”, in our understanding of blocking mechanisms, modeling, and impacts.

Fig. 6. The complexity of blocking and extreme weather.

Top 100 blocking events (during the period 1980–2021) in boreal winter and summer and associated domain average temperature (T; in °C) and precipitation (P; in mm/day) anomalies (upper-left panel) in the Northern Hemisphere, distributions of zonal mean P anomalies for dry and rainy blocks (lower-left panel), and spatial patterns of typical events (right panels). Solid black contours indicate 500-hPa geopotential height (Z500; in m) anomalies. Colored shading shows T anomalies; crossed and dotted hatching represent dry and wet regions, respectively (threshold: P anomaly = ±2 mm/day). Composite P anomaly profiles are plotted relative to longitude, with 0 marking the blocking center; shaded bands denote the 30th to 70th percentile range.

Termination of blocks and heatwaves

Does the end of heatwaves terminate blocking events, or is it the other way around? The surface heating caused by heatwaves increases sensible heat fluxes and can induce diabatic heating in the mid-troposphere⁷⁵. This contributes to the maintenance or intensification of blocking events. Conversely, blocking systems tend to suppress convection and increase downward shortwave radiation, which further warms the surface and prolongs heatwave conditions. These processes create a potential feedback loop between heatwaves and blocking events, and the end of one type of event may potentially terminate the other type of event. While considerable research has focused on the formation and maintenance of heatwaves⁷⁶ and blocking events¹³ separately, the shared (or distinct) mechanisms that govern their terminations remain unclear. To address this knowledge gap, more work is needed to investigate the physical processes that contribute to the termination of blocks and heatwaves, respectively, and whether the termination of heatwaves leads to the termination of blocks and vice versa.

Land–atmosphere interactions and soil moisture

Land–atmosphere interactions exert significant impacts on atmospheric blocking and associated extreme weather events through exchanges of energy, moisture, and momentum, exacerbating stagnating atmospheric patterns. Figure 2b–d illustrates the evolution of a severe and expanding drought over the U.S., characterized by negative soil moisture anomalies and low values of the Standardized Precipitation Evapotranspiration Index (SPEI), followed by the 2021 Pacific Northwest blocking pattern. Throughout the period shown, both root-zone and surface soil moisture remained abnormally low. Around June 25, as an atmospheric blocking system approached the western U.S., as indicated by an enhanced anticyclone, the drought conditions intensified, with SPEI values dropping from around −0.5 in mid-June to below −1 in early July. This suggests that the slow movement of Rossby wave packets (RWPs) that left imprints on the soil moisture contributes to severe drought. For droughts in the central U.S., low soil moisture and high evaporation play a major role, but these droughts are sometimes initiated by a quasi-stationary RWP^77,78. The slow propagating RWP arises about a week prior to drought onset and originates from the western North Pacific, suggesting that atmospheric blocking played a substantial role in initiating the drought.

We envision that soil moisture, characterized by its diverse geographic distribution, could have significant impacts on atmospheric blocking mainly through two mechanisms: (1) exacerbating large-scale stationary atmospheric patterns that are favorable for blocking formation; (2) intensifying the land-atmosphere feedback that transforms the energy into more sensible heat flux and increases the temperature. Soil moisture has typically been considered separately from dynamical processes, while its influence on the atmospheric circulation and weather patterns beyond its immediate locations is largely overlooked. To address this knowledge gap, more work is needed to investigate the remote and local impacts of soil moisture on atmospheric blocking patterns and associated extreme weather such as heatwaves and droughts.

Data-driven and Machine Learning (ML)/Artificial Intelligence (AI) approaches

Conventional composite and regression analyses can only identify spatial patterns that are most correlated with blocking. As a data-driven approach, Breeden et al.⁷⁹ developed a Linear Inverse Model (LIM) framework, in which a nonlinear dynamical system can be approximated by the dynamics of a stable, linear system forced by stochastic white noise, to calculate the optimal, subseasonal precursors of boreal winter North Pacific blocking. This approach highlighted both tropical and extratropical precursors to blocking. Another LIM was used to identify a subseasonal forecast of opportunity for a cold air outbreak over the central United States in February 2021, identifying the important role of constructive interference between a sudden stratospheric warming event, ENSO, and the MJO⁸⁰. LIMs have been shown to be a useful data-driven approach in identifying constructive and destructive interference across modes of variability in subseasonal timescales^81,82, helpful in identifying conditions that lead to blocking events and weather extremes.

Explainable machine learning methods⁸³, including those capturing nonlinearities, have proven to be promising tools for identifying potential precursor patterns of extremes, and predicting future blocking occurrences in addition to providing physically-informed interpretations⁸⁴. By leveraging these methods, valuable insights can be gained to better understand the underlying mechanisms of atmospheric blocking. Convolutional Neural Networks (CNNs) are particularly effective at recognizing patterns in visual data, using the fundamentals of signal processing to learn how to detect informative features that recur in large datasets. Layer-Wise Relevance Propagation, an Explainable AI technique, can identify precursor patterns and geographic regions guiding the CNN’s predictions, providing insights into how the model interprets the data and identifies extreme events. There is a growing interest in using AI methods, such as explainable AI⁸⁵, to elucidate the relative contribution of various physical processes involved in atmospheric blocking events and their interactions in modulating extreme weather events. However, a challenge of using AI to advance blocking research is disentangling the comprehensive causal relationships between land-atmosphere interactions, large-scale atmospheric dynamics, including both background flow and weather systems, and extreme weather events, and further to quantify the strength of the climate system’s response to perturbations arising from land-atmosphere interactions.

We struggle to determine if atmospheric blocking events possess unique predictability compared with regular weather systems and, if they do, whether there are any precursor patterns that signal the onset of blocks and can improve our forecasting capabilities, but AI may help to advance this. To date, research is limited in identifying various precursor patterns within the midlatitudes that can be used to predict atmospheric blocking and associated extreme weather events, as well as quantifying how forecast skill with lead time has changed with such precursor patterns. We need more work utilizing data-driven and eXplainable Artificial Intelligence (XAI) methods to gain valuable insights into the mechanisms and predictability of atmospheric blocking, especially identifying any valuable precursor patterns that may vary depending on the flavor of the block. More work is needed utilizing simple idealized models, which can provide large training sets for XAI, to deepen our conceptual understanding of machine-learning results and blocking dynamics, optimizing ML frameworks to improve blocking prediction accuracy, and comparing identified precursor patterns across model hierarchies to obtain deeper physical insights. Further, it remains an open question whether deep learning-based forecasting systems can predict unseen extreme weather events, such as record-breaking blocking events, without training on them, as is often the case with limited observations.

Furthermore, traditional geoscience approaches face a bottleneck in bridging conceptual models derived from pencil-and-paper approaches with the vast volume of observational data. Emerging AI-based equation discovery methods^86,87 show great promise to identify key underlying dynamics in the presence of noise. We need more work to leverage these AI tools to discover equations underlying blocks from idealized models, and ultimately extend this to observations.

Predictability of blocking events

In ECMWF operational ensemble forecasts at late medium-range, forecasts in the Euro-Atlantic sector that were initialized during European blocking events resulted in the least skillful forecasts after 15 days among the four commonly defined weather regimes (positive and negative North Atlantic Oscillation, European blocking, and Atlantic ridge)⁸⁸. With a much longer reforecast dataset, poor skill arose more from predicting the onset of blocking, highlighting that such studies are sensitive to sample size⁸⁹. To put things in perspective, while there are problems in forecasting blocking events, there are challenges in forecasting other regimes, too. The differences in skill for blocking and the other regimes are often not substantial, with large forecast sampling required to adequately characterize their skill.

The difficulties in simulating blocks are also well known in climate models, which tend to underestimate blocking frequency climatologically and tend to exhibit substantial inter-model differences^{23,71,90–92}. A recent study⁹³ showed that these blocking biases are present in both CMIP6 models and the seasonal prediction model developed by GFDL across different lead times. The poor medium-range forecast skill for blocks has serious consequences. Missing the formation of a blocking pattern can result in failure to predict an extreme weather event and lead to a missed opportunity for advance warnings, and failing to simulate long-lived blocks can result in missing the extreme low or high temperatures and precipitation associated with their mature phase. As an example, this issue led to the prediction of a shorter than observed lifetime for the devastating 2010 heatwave in Russia, which was caused by a month-long blocking event⁹⁴. Furthermore, because blocks are linked to large-scale changes in the jet stream, inaccurate representation of blocking patterns in a model can degrade the forecast skill for other extreme-causing phenomena such as heatwaves, droughts, or cold spells. Through the climate perspective, atmospheric blocking is also closely linked to leading modes of variability such as the Madden-Julian Oscillation (MJO)⁹⁵, El Niño–Southern Oscillation (ENSO)⁹⁶, Northern Annular Mode (NAM)⁹⁷ and Baroclinic Annular Mode (BAM)⁹⁸, and more work is needed to understand how these natural modes of variability interfere and influence the predictability of blocks.

Recommendations for the path forward

Given the progress already made in the fundamental understanding, numerical simulation, and prediction of blocking, along with existing knowledge gaps and potential for further improvements, we make the following recommendations to prioritize research objectives and optimize resource allocations for improved understanding and greater predictive capability of blocking:

A renaissance of interest in the theoretical inquiry of blocking

With new observations leading the way and novel diagnostic tools recently developed, we are now better positioned than ever to identify the most likely physical explanation for atmospheric blocking. We recommend leveraging the vast amount of observational evidence alongside simple conceptual models, including those developed using pencil-and-paper approaches, to provide new mechanistic insights. We also recommend making the fundamental connections between blocking patterns and other related atmospheric phenomena, including mesoscale convective systems, synoptic storms, jet streams, atmospheric rivers, and Rossby wave breaking, as these interactions may play a critical role in the onset, maintenance, and decay of blocking events. Furthermore, it’s also important to understand whether the mechanisms of terrestrial atmospheric blocking can be applied to understand blocking-like patterns in the atmospheres of other planets, such as Mars and Jupiter.

Studies with a hierarchy of models

Theoretical studies leveraging hierarchical modeling, ranging from simple two-layer channel models to fully coupled climate models. It is recommended that the hierarchy should include moist aquaplanet models and save diabatic heating terms to assess diabatic heating contributions, in view of the fact that a diabatic PV source can be an important building block for intensive blocking events. For each rung of the modeling hierarchy, systematic sensitivity experiments can be performed to examine: a) blocking and extreme definitions with tailored blocking indices to capture specific impacts, such as heatwaves and conducive environments for wildfires, and b) sensitivity to varying forcing, boundary conditions, and model parameterizations of physical processes. We suggest using climate models of increasing complexity and performing single forcing experiments, which allows for critical assessment of existing theories and a reconciling of weather and climate system perspectives. To seek the most likely mechanistic explanation for blocks, we recommend idealized modeling to develop testable, falsifiable, predictive theories for the statistics of blocking rather than theories for predicting a specific blocking event. While theories exist for the minimal understanding of the most fundamental aspects of the blocking phenomenon and for the statistical distribution of blocking, the pressing challenge is the lack of theories to explain the blocking response already emerging in climate model projections, and to explain the different flavors of blocking as they relate to weather extremes (e.g., hot vs cold, rainy vs dry, etc.) and blocking diversity based on the broad range of possible characteristics (e.g., blocking type and amplitude, regional differences, etc). In particular, we suggest pivoting towards moisture and cloud-related diabatic processes, orographic gravity-wave drag, and subgrid-scale processes using idealized models with various levels of complexity. Similar to efforts to understand complexity in other phenomena⁹⁹, we suggest investigating distinct climate responses and feedbacks for blocks of different flavors.

Use of existing diagnostic tools

We have abundant quantitative diagnostic tools (e.g., wave activity vs eddy kinetic energy, QGPV vs Ertel PV, PV vs moist PV, Eulerian vs Lagrangian). Rather than developing new diagnostic methods, already established tools could be used more effectively to study blocking-extreme connections. Generally speaking, blocking often occurs when there is a stagnation of atmospheric circulation. Blocking is not just one specific phenomenon, and so it seems unlikely to ever be well defined with just one definition. There is a range of mechanisms that contribute to blocking, and the balance of these is different among different regions, seasons, or even within these. Understanding how the range of mechanisms maps onto the range of metrics would be a useful target for the community. When planning a study, investigators should consider the types of events they are most interested in and tailor their metrics and mechanistic analyses to those. For example, a publicly available North American blocking database is expected to become available soon, with blocks as defined by two blocking metrics: one is reversal-based (used to capture dipole blocks) and the other anomaly-based (used to capture omega blocks). For each blocking event, the database will include blocking characteristics, synoptic conditions, and temperature and precipitation impacts (e.g., rate, phase, and flags for extremes) using a broad range of precipitation datasets (satellite, radar, and reanalysis), and more work is needed for other regions and seasons. For the diabatic influence, adopting a wider range of tools and blocking indices and comparing the results would help to deepen physical insights. The aim of the “which of these is blocking?” activity at the workshop was to test whether blocking experts agree on what is blocking in a qualitative sense. The results seemed clear—blocking covers a range of different patterns, so attempting to come up with one index that captures all aspects may be infeasible. Instead, our goal should be to understand the strengths of various diagnostic tools and put them under a more critical light. Instead of aiming to pick winners and losers among multiple indices, we suggest an intercomparison project to discern the relative advantages and disadvantages of different approaches for different blocking flavors. For example, it would be beneficial for the community to compare Eulerian, semi-Lagrangian, and Lagrangian metrics to collectively reveal and compare the distinct physical processes each method highlights.

New experimental methods to enhance our understanding of blocking and extreme weather

We are scoping a new Blocking—Extremes—Method Intercomparison Project (BEMIP) to identify the key factors that control the overall number, geographic distribution, intensity, and trend of atmospheric blocking and extreme weather events. As reviewed in Woollings et al²³ from a previous Blocking workshop, the largest discrepancies among climatologies come from the different methodologies of blocking indices and a lack of coordinated efforts to compare indices using the same datasets and adhering to the same protocols. This discrepancy has severely limited our ability to reach a consensus on blocking and extreme weather responses in the future. To understand the strengths and weaknesses of each definition, the proposed BEMIP should embrace the different flavors of blocks where different blocking types may be best captured by different blocking metrics, and also allows us to understand the uncertainties in blocking statistics associated with these differences in the metrics, and among various blocking flavors and associated surface extremes and their impacts such as droughts or heat-related illnesses. Similar obstacles in the Atmospheric River research community were overcome through the development of the Atmospheric River Tracking Method Intercomparison Project (ARTMIP)¹⁰⁰. We propose to model BEMIP after the success of the ARTMIP, but with a focus on blocking and extreme weather. The proposed BEMIP could provide a critical and much-needed opportunity for experts from both the weather and climate communities to work together to address the grand challenges regarding blocking prediction, diagnostics, and extreme weather impacts. Blocking patterns are associated with a wide range of surface extreme weather, such as heatwaves, droughts, wildfires, cold spells, severe thunderstorms, and extreme precipitation. Some of these connections may arise from the local influence of atmospheric blocks, while others may result from their remote effects. It is recommended to establish quantitative links between atmospheric blocking and a range of extreme weather events, considering both cases when blocks occur locally and when they are remote relative to the surface extremes. It is also recommended to examine the temporal and causal relationships between blocks and a range of extreme weather. For example, to assess whether the termination of blocks leads to the termination of heatwaves and droughts, or the other way around.

Data-driven and ML/AI approaches

We encourage more research on adopting data-driven approaches⁷⁹ and machine learning/AI approaches⁸⁵ to tackle the comprehensive causal relationships between atmospheric blocking, land-atmosphere-ocean interactions, and associated extreme weather events. We recommend using these novel tools to investigate the predictability of atmospheric blocking, including any potential precursor patterns, to disentangle the complex causal relationships, and to explore if AI models can predict (rare) blocking events without seeing them before. We further recommend leveraging numerical models and AI tools to discover key equations that may govern blocks from the data.

Predictability of blocking events

Evaluation of how the leading blocking theories can be used to predict blocks (S2S) and understand the projected changes in blocks (process-oriented diagnostics) is needed. We recommend a community effort to utilize the existing S2S datasets, such as the North American Multi-Model Ensemble¹⁰¹ and the Subseasonal Experiment (Sub-X)¹⁰², to build a blocking diagnostics and prediction testbed, for the purpose of evaluating different metrics for blocking detection/identification, testing the predictive blocking theories, and developing process-level diagnostics of forecast errors related to blocking. The forecast skill from models of varying complexity should be compared to determine what processes improve—or hinder—the predictability of blocking. For example, the blocking skill achieved by a low-order linear stochastic model⁵³ could be compared to the blocking skill of more complex numerical models.

Studies of blocking in the Southern Hemisphere

We encourage more research on investigating blocks and extremes in the Southern Hemisphere, which is currently understudied, to identify similarities and differences in the physical processes compared with those in the Northern Hemisphere, and to provide a more global perspective.

Methods

Data

ECMWF Reanalysis v5 (ERA5) is the primary dataset used in this study. It is the fifth-generation global atmospheric reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF), providing hourly data from 1940 to the present, covering both surface and various pressure levels¹⁰³. The key variables used in this study include 500 hPa geopotential height, u- and v-wind components at 500 hPa, total column water vapor, 2-meter air temperature, and 2-meter dewpoint temperature. The data are retrieved at a horizontal resolution of 0.25°, and daily values are computed from the hourly records. The NCEP Global Forecast System (GFS)¹⁰⁴ analysis and forecast grids are on a 0.25 by 0.25 global latitude longitude grid.

Daily 0.1° precipitation data are from the NASA Global Precipitation Measurement Mission (GPM)¹⁰⁵. Surface and root-zone soil moisture (SM) data, also at 0.1° and daily resolution, are taken from version 4.1a of Global Land Evaporation Amsterdam Model (GLEAM)¹⁰⁶. Sea surface temperature (SST) data are sourced from the Optimum Interpolation Sea Surface Temperature (OISST), which provides daily values at a 0.25° resolution¹⁰⁷.

The active fire data is from the Collection 6.1 Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Aqua and Terra satellites¹⁰⁸, retrieved from the NASA Fire Information for Resource Management System (FIRMS). Each hotspot or active fire detection represents the center of a pixel (approximately 1 km) identified as containing one or more fires or other thermal anomalies. The Fire Radiative Power (FRP) represents the total radiative energy emitted by a fire within a pixel, expressed in megawatts (MW).

To quantify drought severity, we use the Standardized Precipitation Evapotranspiration Index (SPEI), which quantifies the difference between precipitation and evapotranspiration. SPEI can be computed on a daily resolution, with positive values indicating wetter-than-normal conditions and increasingly negative values indicating drier conditions. A value below −0.5 is commonly used as the threshold for identifying drought events. In this study, daily SPEI data at a 0.25° resolution are retrieved from a published dataset¹⁰⁹.

Vapor Pressure Deficit

Vapor Pressure Deficit (VPD) is a critical metric to assess atmospheric dryness. It quantifies the difference between the saturation vapor pressure and the actual vapor pressure, representing how ‘thirsty’ the air is. Higher VPD values indicate drier air, which can suppress precipitation and increase plant water stress. Due to the limited availability of in-situ observations, daily average VPD (in KPa) can be estimated using air temperature and dewpoint temperature based on Teten’s equation^110,111:

$$e\left(T\right)=0.611exp\left[A\left(\frac{T}{T+B}\right)\right]$$ 2

where $e$ is the vapor pressure and $T$ is the temperature for calculating water vapor pressure. Specifically, air temperature is used for estimating saturated $e$, while dewpoint temperature is used for estimating actual $e$. The parameters $A$ and $B$ are constants and empirically determined as $A=17.27,B=237.15$ if T≥0 °C and $A=21.87,B=265.5$ if T < 0 °C. VPD is then calculated as the difference between the saturated and actual $e$.

Composite of 200-hPa streamfunction anomaly

The composite of 200-hPa streamfunction anomaly was produced using daily mean zonal and meridional wind data from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) Reanalysis I dataset were interpolated to 2° × 2° spatial resolution¹¹², and then smoothed with a 7-day running mean. The daily climatology from 1980 to 2014 was determined from each 7-day running mean field as the time-mean plus the first four harmonics of the annual cycle. Anomalies were then calculated as differences from this climatology. Streamfunction anomalies were subsequently calculated from these global wind anomalies using SPHEREPACK routines¹¹³ available through NCAR.

Blocking type classification

Temperature and precipitation anomalies are first computed as deviations from the daily climatology for each day of the year. Blocks are identified with the local wave activity framework (details in refs. ^50, 98). For each blocking event, the domain-averaged surface temperature and precipitation anomalies over these blocked grids are used to represent the event’s hot/cold and dry/rainy conditions. Based on the signs of these anomalies, blocking events are classified into four types: hot-dry, hot-rainy, cold-dry, and cold-rainy.

Author contributions

L.W. conceived the original idea of the perspective article and, with the valuable input from coauthors, wrote the first draft. Y.H. performed the analysis and with inputs from L.W. produced Figs. 1–3 and 5–6. S.L. performed the analysis and produced Fig. 4. All co-authors L.W., J.L., M.B., G.C., S.H., V.N., I.R.S., T.W., Y.H., and S.L. contributed to improvements and the revisions of the manuscript.

Peer review

Peer review information

Nature Communications thanks Stefano Tibaldi and the other, anonymous, reviewers for their contribution to the peer review of this work.

Data availability

All datasets used in this paper are open sourced without limitations. The ERA5 datasets are available at ECMWF. The active fire data can be retrieved from FIRMS. The NCEP GFS datasets can be retrieved from Geoscience Data Exchange (GDEX) (https://gdex.ucar.edu/datasets/d084001/). The NCEP–NCAR Reanalysis I dataset was interpolated to 2° × 2° spatial resolution and can be retrieved from NOAA physical science laboratory (PSL) (https://psl.noaa.gov/data/gridded/data.ncep.reanalysis.html). Precipitation data are from NASA GPM; soil moisture datasets are available at GLEAM; SST data is from the NOAA OISST (https://www.ncei.noaa.gov/products/optimum-interpolation-sst). The SPEI data can be retrieved from the public repository (10.5281/zenodo.8060268). Data to reproduce the figures of this paper can be downloaded from the open repository¹¹⁴.

Code availability

The code used in this paper is available and has been archived¹¹⁴.

Competing interests

The authors declare no competing interests.