Abstract Rice‐wheat cropping systems underpin food security across Asia but also represent major sources of agricultural greenhouse gas (GHG) emissions. Replacing wheat with diversified crop species to create alternative rice‐based cropping systems may become an effective strategy for sustainable production. However, how such alternatives influence rice production and GHG emissions at large spatial scales remains poorly understood. Here, we integrated compiled field observations with Random Forest modeling frameworks to generate spatial estimates at 0.5° resolution across Asia. All optimized models were statistically significant ( p < 0.001) and showed good predictive performance, with testing R 2 of 0.83 for rice yield, 0.63 for methane emission, and 0.70 for nitrous oxide emission. Spatial estimations demonstrated that complete replacement of rice‐wheat with alternative rice‐based cropping systems was associated with a modest increase in rice yield of 1.3% and a reduction in methane emission of 8.5%, while leading to a slight increase in nitrous oxide emission of 1.4%. Across Asia, these transitions could boost rice production by 16.8 million tons and reduce carbon emission by 82.4 million tons. The resulting improvements were expected to generate a yield benefit of 6.4 billion dollars and a net benefit of 8.6 billion dollars. These economic returns exhibited a highly concentrated pattern, with India, China, Indonesia, Bangladesh, and Thailand accounting for roughly three‐quarters of the total benefits. Our findings highlight the great potential of alternative rice‐based cropping systems to enhance rice productivity, mitigate GHG emissions, and improve economic viability, providing a promising pathway for sustainable agricultural transformation.
Abstract The East/Japan Sea (EJS) is a semi‐enclosed marginal sea that has experienced rapid sea surface temperature (SST) warming exceeding 0.9°C since the 1980s, which has intensified the occurrence of marine heatwaves (MHWs). Its semi‐enclosed nature amplifies the influence of external climate forcing, making reliable projections of future SST and MHW changes essential for assessing ecological and socio‐economic impacts. Here, we investigate future SST and MHW changes using high‐resolution (1/8°) dynamical downscaling simulations based on the Regional Ocean Modeling System (ROMS), driven by seven Coupled Model Intercomparison Project Phase 6 (CMIP6) models under four Shared Socioeconomic Pathway scenarios (SSP1‐2.6, SSP2‐4.5, SSP3‐7.0, and SSP5‐8.5) for 1982–2100. The ROMS ensemble outperforms the CMIP6 ensemble in reproducing observed SST and MHW characteristics for the historical period of 1985–2014, particularly during winter, due to improved simulation of the transport of the Tsushima Warm Current (TWC) through the Korea Strait and associated regional circulations. Future projections (2071–2100) indicate that SST warming and MHWs in the central EJS will become stronger, longer‐lasting, and more spatially heterogeneous in ROMS, in contrast to the more uniform patterns projected by CMIP6, especially under high‐emission scenarios. This spatial heterogeneity is associated with intensified transport of the TWC and a strengthened East Korean Warm Current, which enhance heat advection along their pathways into the basin interior. These results highlight the added value of high‐resolution dynamical downscaling for understanding and preparing for future SST and MHW changes in the EJS, providing insights for regional climate impact assessment and adaptation planning.
Abstract Climate models consistently project a decline in diatom net primary production (NPP) under high‐emission scenarios in the highly productive subpolar North Atlantic. However, the magnitude of this decline remains highly uncertain and varies substantially across models, limiting our ability to anticipate its ecological and biogeochemical consequences. Several recent studies using individual Earth system models have identified a mechanistic relationship between the projected weakening of the Atlantic Meridional Overturning Circulation (AMOC) and the availability of for phytoplankton growth in the subpolar North Atlantic. These findings suggest that uncertainty in the future evolution of the AMOC may be a major driver of the spread in diatom NPP projections. Here, we analyze an ensemble of Earth system models to test this hypothesis and to better constrain projections of diatom NPP. We first show that a robust inter‐model relationship exists between projected changes in AMOC intensity and availability in the subpolar North Atlantic. We then apply an emergent‐constraint framework to estimate the future sensitivity of diatom NPP to nitrate changes, using satellite‐based reconstructions of historical NPP and nitrate variability. Combining these two approaches yields a constrained relationship between projected AMOC weakening and the associated decline in diatom NPP. Our results confirm that uncertainty in the magnitude of future AMOC weakening constitutes the dominant source of uncertainty in diatom NPP projections. Nevertheless, they suggest a smaller decrease in diatom NPP than previously anticipated for a plausible range of AMOC decline.
Abstract A July 2025 report from the U.S. Department of Energy (DOE) made the key claim that the stratosphere has warmed since 2000, contrary to model projections and inconsistent with the expected anthropogenic fingerprint. The DOE report provided no references or data to support this claim. Its basis appears to be an unpublished figure comparing simulated and observed tropical trends in the temperature of the lower stratosphere (TLS), a layer extending from roughly 15 to 20 km above Earth's surface. The observed tropical TLS trends shown in the unpublished DOE figure are incorrect. When the correct observational data sets are used, model tropical TLS trends are consistent with satellite observations, and a model‐predicted anthropogenic fingerprint is identifiable with high confidence in observed latitude‐altitude profiles of tropical temperature change. An earlier rebuttal of the DOE “no human fingerprint” claim considered global‐mean temperature changes only and did not perform a pattern‐based fingerprint study. Unless corrected or retracted, the DOE report will continue to misinform policymakers and the public about the reality of human effects on climate.
Abstract Synthetic aperture radar (SAR) has been the backbone of Arctic sea ice monitoring since the 1990s, yet nearly every operational product that assigns an ice class or geophysical state (melt onset dates, freeze‐up timing, ice‐type classifications) is derived by compressing continuous radar backscatter into discrete categorical labels. This Commentary argues that such compression is not merely a simplification but a systematic loss of physical information that increasingly compromises trend attribution and cross‐sensor synthesis as the Arctic transitions from a perennial to a seasonal ice regime. A semi‐empirical forward‐model demonstration shows that physically distinct ice states can produce indistinguishable C‐band backscatter yet separate cleanly at L‐band, confirming that single‐frequency categorical detection is fundamentally non‐unique. The multi‐frequency SAR constellation now in orbit (Sentinel‐1 and RCM at C‐band, NISAR at L‐ and S‐band) provides the measurement basis to move beyond labels toward physically based retrieval. A Bayesian inversion framework that outputs continuous state estimates with formal uncertainty bounds could preserve backward‐compatible categorical products for operational ice services while adding the process‐level traceability that climate science demands. Realizing this architecture requires coordinated investment in multi‐frequency forward models, designated Arctic supersites, and in situ campaigns targeting the snow microstructure variables that remain the largest source of retrieval ambiguity.
Abstract This study connects the dots between global extent of severe water scarcity (SWS) events and spikes in food price. The authors start with three staple crops (rice, wheat, and maize) using a empirical but crop specific characterization of severe water scarcity occurrence, which is defined based on the area of land cultivated with each studied crop that on a given year is under SWS conditions. Water scarcity is estimated based on 1, 3, and 12 months standardized precipitation evapotranspiration index during crop specific sensitive periods (SPs). SPs coincide with the 4 months prior to harvest. A SWS‐wheat price relationship model was developed for period 2001–2021 and tested for 1986–2000 and 2022–2024 with future projections calculated under climate change using climate model simulations from CMIP5 and CMIP6 up to 2100. The results show a marked increase in wheat prices driven by increasing SWS area, which is in turn a function of greenhouse gas emissions. The results indicate that the meanglobal mean temperature increase by 3°C (compared to 1951–1980) could lead to the tripling of the mean global wheat price compared to 2010 (after deflating).
Abstract Forest insect outbreaks are intensifying under climate change and global trade, yet their spatiotemporal dynamics remain poorly quantified. This lack of predictive understanding limits timely and effective management. The Japanese pine bast scale ( Matsucoccus matsumurae ), which has persisted in South Korea for more than 50 years despite ongoing control efforts, exemplifies this challenge. We developed the first national framework that converts raw monitoring data into spatially explicit forecasts of pest abundance and phenology. A nationwide pheromone‐trap network (164 sites in 2022 and 65 in 2023) provided standardized observations of male flight activity. Deep learning automated the counting of captured insects, and statistical modeling of biweekly captures yielded the timing of first emergence and peak flight. These data, combined with a 30 m host‐abundance map and 1 km environmental predictors, were used to train extreme gradient boosting (XGBoost) models to predict abundance and phenology across South Korea. The models explained 78% of abundance variance and up to 90% of phenological variance. They revealed strong spatial heterogeneity: captures ranged from 0 to >43,000 per trap, were concentrated along southern coasts, while emergence timing differed by up to 9 weeks nationwide. Warmer winters advanced emergence, and stable climates promoted higher abundance. Bootstrap ensembles quantified spatial prediction uncertainty, highlighting regions with low confidence that require intensified monitoring. Our results expose a highly dynamic and uneven infestation landscape that fixed, calendar‐based strategies cannot address. By linking large‐scale ecological observation with predictive modeling, this framework establishes a data‐driven basis for adaptive, climate‐resilient forest pest management.
Abstract Improved management of nitrogen could provide environmental benefits. Crop yields can be affected by reactive nitrogen emissions (NH 3 , NO X , and N 2 O) as they influence climate, ozone formation, and nitrogen deposition thus crop fertilization. Here we constructed a integrated framework to evaluate the benefits of improving agricultural nitrogen management for reducing reactive nitrogen emissions and their consequent impacts on crop yields through above‐mentioned pathways, with impacts monetized addressing dynamic food prices. Considering three levels of increasingly stringent nitrogen management scenarios from 2020 to 2050, we find that anthropogenic emissions of NH 3 , NO X , and N 2 O could be reduced by 44.5%, 61.1%, and 35.2%, respectively, by 2050 relative to 2020. With the most effective nitrogen management strategies, crop yields during 2040–2050 are projected to increase by up to 22.5%, 13.1%, 22.9%, and 16.6% for maize, rice, soybeans, and wheat, respectively, corresponding to average economic gains of 10.2, 11.8, 11.8, and 7.4 billion dollars. These results indicate that enhanced nitrogen management can substantially increase crop production, providing a strong rationale for integrating nitrogen emission limits into policy frameworks.
Abstract The 7th Rio declaration principle emphasizes global collaboration to protect and restore ecosystem health. While dryland ecosystem health (EH) remains critical to global sustainability agendas, its long‐term regional dynamics remain poorly constrained across the Asian drylands (AD), where land management decisions increasingly shape ecological outcomes. Here, we develop an integrated regional framework that combines vegetation dynamics, soil moisture, land use, and landscape spatial configuration to assess EH trajectories across AD from 2000 to 2020. We found that EH declined across AD until 2012 and partially recovered after 2013. Specifically, areas with poor EH increased by 2.1% (58 Mha), while healthy ecosystems declined by 0.08% and accounted for only 2% (7.9 Mha) of AD by 2020. These regional trends concealed pronounced spatial contrasts, with continued degradation across much of Central Asia and improvement in parts of Dryland East Asia. These divergent trends reflect interacting human land‐use and climate pressures. Land‐use change affected nearly 4.9% (42.87 Mha) of AD, largely driven by increasing pressure on rangelands and cropland expansion, and reductions in inland water and urbanization. Climatic conditions further contributed to these drivers, as post‐2013 warming coincided with increased soil moisture stress in Central Asia, while higher precipitation supported ecohealth recovery in East Asia drylands. These findings show that recent improvements in AD remain limited in extent and unevenly distributed, emphasizing that without region‐specific strategies integrating sustainable water resources management, rangeland stewardship, and climate adaptation, poor ecosystem health is likely to continue expanding across large portions of the Asian drylands.
Abstract Hot droughts are intensifying under anthropogenic warming, yet whether extreme heat systematically alters the triggering of abrupt drought‐to‐pluvial transitions and their downstream impacts remains poorly understood. Here, we identify global drought‐to‐pluvial abrupt transition (DPAT) events and isolate the influence of antecedent thermal conditions by comparing hot‐drought‐triggered DPATs (DH‐DPATs) with transitions preceded by droughts without concurrent heat. We find that hot drought related transitions display broader hotspots and substantially higher occurrence, with frequency increasing from 66.7% to 316.7% across the interquartile range and a higher conditional transition probability (global mean: 44.89% vs. 38.95%). They also transition faster, sustain longer post drought heavy rainfall, and produce stronger rainfall intensity. Mechanistically, although both event types are associated with enhanced moisture convergence, antecedent heat imposes an additional thermodynamic preconditioning characterized by deeper boundary layers and higher convective instability, making comparable moisture recovery more likely to trigger deep convection. These heat‐mediated transitions amplify ecosystem carbon losses, intensifying reductions in gross primary productivity and solar‐induced chlorophyll fluorescence by about 43.7% and 26.8%, respectively, with the strongest relative impacts in croplands. They also contribute to increasing population and managed‐land exposure. Under future warming, DPATs are projected to shift further toward hot‐drought‐dominated pathways, with increases in both the contribution and transition likelihood of DH‐DPATs. These results highlight a thermodynamic pathway through which hot droughts increase the likelihood and impacts of abrupt drought‐to‐pluvial transitions worldwide.
Abstract Air quality forecasts are essential to support decision‐making in urban agglomerations, where millions of people are exposed to high levels of pollution. However, these forecasts are often limited by uncertainties in anthropogenic emissions, especially in large urban agglomerations where no dedicated operational system exists. In this study, we present an observation‐based emission scaling approach aimed at improving operational forecasts. This method derives scaling factors (SF) from the ratio of observed‐to‐modeled concentrations over the previous week and applies them to anthropogenic emissions in the forecast model, assuming that in large urban agglomerations, biases between observed and modeled concentrations are primarily driven by uncertainties in anthropogenic emission inventories. The method also derives SF for anthropogenic volatile organic compound emissions based on modeled daytime O3 biases under the assumption of a NOx‐saturated regime. We implement this method in a chemistry‐transport model using a global anthropogenic emission inventory and apply it to São Paulo for two distinct periods (February–April 2023 and July–September 2024). The results show that forecasts of CO, NO2, O3 and SO2 concentrations are significantly improved within a few weeks. For PM2.5 and PM10, improvements are more limited by the influence of secondary aerosol formation and by pollution transport from outside of the agglomeration. Overall, our results demonstrate that observation‐based emission scaling provides an efficient and transferable methodological approach for improving operational air quality forecasts in urban agglomerations without requiring model‐specific developments.
Abstract Statistical earthquake forecasting models, such as the Epidemic‐type Aftershock Sequence (ETAS), embed decades of empirical knowledge and assumptions about earthquake triggering, clustering, and catalog completeness. In contrast, neural spatio‐temporal point process (STPP) models often treat seismicity as generic spatiotemporal data without such domain‐specific considerations. Using an earthquake catalog from the China Earthquake Networks Center, we investigate how seismology‐motivated inputs and training strategies affect the forecasting performance of the Deep Spatio‐Temporal Point Process (DeepSTPP). These domain‐driven designs include introducing magnitude as an input feature, incorporating an auxiliary training period, extending visible event history, and varying magnitude thresholds. We benchmark DeepSTPP against ETAS and a homogeneous Poisson process to assess relative strengths and limitations. Our results show that certain domain‐motivated configurations, such as including an auxiliary period, improve performance, particularly in forecasting immediate aftershocks. However, limitations in the model architecture, including short memory and attention dilution, restrict the benefits of event magnitude and long‐range history. The results also show that DeepSTPP remains inferior to ETAS in terms of overall spatiotemporal log‐likelihood, although it outperforms ETAS in temporal forecasting at lower magnitude thresholds, specifically for Mcut=1.4−2.8. This indicates that DeepSTPP does not yet perform as well as ETAS in the magnitude range most relevant to operational earthquake forecasting Mcut≥3.0. The gains observed at lower thresholds may primarily stem from DeepSTPP's flexibility in handling catalog incompleteness and artifacts. These findings highlight the value of integrating seismological practice into neural model design and point toward future architectures that can fully exploit earthquake catalogs.
Abstract Wildfires are complex events that are challenging to model. Researchers and land managers often rely on fire indices as a proxy of fire risk in operational forecasts and future projections. However, in the scientific literature, the choice of which fire index to use is often arbitrary and not tied to specific fire behaviors. Here, we provide a comprehensive evaluation of 21 hydrometeorological variables and fire indices based on their ability to represent daily ignition probability and spread rate in the western United States. Hydrometeorological variables like saturation vapor pressure (E_S) and vapor pressure deficit (VPD) perform the best and most consistently at representing fire ignition probability across regions and vegetation types. Water‐balance‐based indices like 1,000‐hr dead fuel moisture (FM1000) and the Keetch‐Byram Drought Index (KBDI) are best at representing ignition probability in forests. In contrast, indices like the Hot‐Dry‐Windy index (HDW) best represent fire spread rates, especially for non‐forest fires, while indices from the National Fire Danger Rating System are best for forest fire spread. Partitioning the skill of variables/indices into contributions from capturing seasonality and anomalies, ignitions are distributed based on the seasonal cycles, while seasonality and anomalies are comparably important for spread rate. We demonstrate that computational complexity does not always result in better fire variables/indices, and that distinct sets of variables/indices are needed to represent different aspects of fire behavior, emphasizing the importance of making careful and justifiable selections of fire variables/indices.
Abstract Rivers are central to the global hydrological cycle, supporting ecosystems and human water use. River discharge, one of the best observed hydrological variables, is expected to change under anthropogenic warming with potentially devastating consequences. We assess global changes in river flow by routing grid‐cell runoff from 18 CMIP6 and 25 ISIMIP3b simulations along the river network using the routing model mRM, combining online (fully‐coupled CMIP6) and offline (bias‐adjusted ISIMIP3b) approaches. Annual river flow projections, climatological seasonal cycles, and discharge trends are comparable between the two approaches and biases with respect to observations are consistent in magnitude with the literature. Therefore, we pool all 43 simulations into one large multi‐model ensemble covering 250 years (1850–2099). The projections show changes in river discharge in response to warming which are largely consistent with previous studies. In particular, at 3.0°C global warming, annual mean and extreme river discharge are projected to decrease in the Mediterranean, Central America, large parts of South America, and Southern Africa; and increase in Central Africa, South Asia, and high northern latitudes. Seasonal mean flow changes mirror annual projections, except for drying in high northern latitudes during boreal summer. Regarding changes to the continental water balance, the simulations indicate that while total global drainage from rivers to the ocean remains roughly constant with increasing warming, regional differences intensify which may affect ocean freshening. Notably, in many world regions at least 66% of simulations agree on the direction of change relative to the pre‐industrial baseline already at current global warming levels.
Abstract Understanding of the spatiotemporal dynamics of vegetation carbon sequestration capacity (VCSC) in Southwest China's contiguous karst region is critical for regional carbon cycling. Using MOD17A3 data from 2001 to 2023, we analyzed spatial heterogeneity in vegetation net primary productivity (NPP) and applied Geodetector to evaluate nonlinear influencing mechanisms across 32 factors. The results showed that the multi‐year average NPP in the southwest karst was 725 g C m −2 yr −1 , exhibiting a southward‐increasing gradient. Recovery areas expanded at a rate of 3.21 g C m −2 yr −1 , covering 79.3% of the region. Comparative analysis revealed that Engineering Governance Areas (EGA) significantly outperformed Natural Recovery Areas (NRA) in enhancing VCSC. In EGA, the mean recovery rate (3.67 g C m −2 yr −1 ), significantly improved areas (51.3%), extremely stable area (71.7%), and future sustainable recovery (72.0%) were 1.38, 1.23, 1.22, and 1.18 times higher than those in NRA, respectively. However, natural factors played a greater role than anthropogenic activities in explaining the spatial heterogeneity of NPP. Environmental interactions exhibited both nonlinear and bivariate enhancement, with combined effects exceeding those of individual factors. The dominant interactions were AET ∩ ELE (77.7%) for the whole region, AET ∩ PRS (61.8%) in EGA, and AET ∩ ELE (87.3%) in NRA. The optimal environmental thresholds for vegetation growth were AET of 1,244–1,397 mm, precipitation of 2,025–2,259 mm, temperature of 18.5–20.8°C, and elevation of 1,094–1,438 m. Integrating restoration strategies with these environmental thresholds helps identify priority zones, providing a scientific basis for the precision management of karst desertification and supporting China's “dual carbon” goals.
Abstract Climate change driven shifts in temperature, precipitation patterns, and atmospheric CO 2 concentrations threaten both productivity and nutritional quality of maize, a staple crop that underpins daily energy and protein intake for millions of households in Africa and beyond. While several studies focus on yield, impacts on maize grain nutrient composition remain less assessed. Here, we use the Agricultural Production Systems simulator (APSIM) crop model to estimate climate impacts on maize yield and grain protein levels across 16 West African countries under two contrasting shared socioeconomic pathways (SSP1‐RCP2.6 and SSP3‐RCP7.0). Our model results show spatially heterogeneous climate impacts across the region, with a mean reduction of −5.7% (range: −43.7%–49.7%) for yield and −4.8% (range: −42.6%–54.3%) for protein content under the high‐emission scenario by mid‐century. Spatially, protein concentration shows an inverse trend with yield: areas that experience yield and protein content gains show reductions in protein concentration, and vice versa. Across the region, protein concentration shows a mean increase of 4.6% (range: −4.9%–19.3%). Northern regions exhibit larger percentage yield increases, driven by higher projected precipitation and comparatively low baseline yields. By jointly evaluating productivity and nutritional quality at subcontinental scale, we demonstrate that yield‐only assessments may underestimate future food system vulnerability and thus a potential trade‐off between yield and protein concentration. Since adaptation measures that aim to stabilize or increase maize grain yield may jeopardize improvements in grain quality, further research on region‐specific measures that optimize grain yield and quality are warranted and discussed in this study.
Abstract Permafrost thermal state, representing the stored “cold energy” in the ground, is crucial for assessing permafrost changes. However, the conventional metric, mean annual ground temperature (MAGT), has inherent limitations because it overlooks the thermodynamic contribution of ground ice. To address this, we developed an alternative metric based on enthalpy change (Δ H ′), which integrates ground temperature and ice content. Using a process‐based model, we designed controlled warming experiments to evaluate the performance of the two metrics. The results show that permafrost traditionally classified as “unstable” (MAGT between −0.5 and 0°C) can persist for 30–100 years under an air warming rate of 0.4°C per decade, largely due to variations in ground ice content. Compared with MAGT, Δ H ′ provides a more uniform scale and higher discernibility in indicating potential permafrost duration, and is therefore more suitable for representing permafrost thermal state. Furthermore, using high‐resolution gridded data sets, we mapped permafrost thermal state across the Qinghai‐Tibet Plateau (QTP) and quantified the spatial discrepancy between the two metrics. The results indicate that permafrost on the western QTP is less stable than suggested by MAGT, whereas permafrost in the southeast and around lakes in the endorheic basins is more stable than MAGT estimates. Explainable machine‐learning analyses reveal that precipitation, soil coarse fraction, and solar radiation are the primary factors controlling the spatial heterogeneity of this discrepancy. This study highlights the dominant role of ground ice in regulating permafrost thermal state and calls for greater attention to ground ice in future permafrost projections.
Abstract Nitrogen pollution has severely impacted inland and coastal waters, contributing to eutrophication, harmful algal blooms, and drinking water contamination. Reducing nitrogen inputs is widely expected to improve water quality, but watershed responses to changing nitrogen surplus are often nonlinear and delayed, complicating expectations of recovery. Here, we analyze long‐term trends in nitrogen surplus and riverine nitrogen loads across 490 US watersheds from 1990 to 2017 to examine how changes in inputs translate into changes in load. By jointly classifying watersheds based on the direction of surplus and load trends, we identify four response regimes that capture the diversity of nitrogen dynamics across US landscapes. Using a multivariate classification framework, we show that watershed responses are shaped by interactions among nitrogen source type, hydrologic connectivity, climate constraints, and legacy nitrogen storage. Livestock density is associated with increasing nitrogen surplus, while declining atmospheric deposition is associated with decreasing surplus and loads. Groundwater‐dominated watersheds exhibit increasing nitrogen loads despite declining surplus, reflecting the long memory of subsurface nitrogen storage. Connectivity modulates these responses: tile‐drained watersheds exhibit tight coupling between surplus and load, whereas arid, low‐connectivity systems show declining loads despite increasing surplus. These results show that spatial patterns in riverine loads cannot be used to infer recovery trajectories following reductions in nitrogen inputs, and effective nitrogen management requires accounting for watershed‐specific response regimes.
Abstract Compound events, arising from the combinations of multiple interacting underlying physical drivers, have gained increasing attention over the past decade with a wide range of applications, including floods, droughts, and wildfires. This growing body of research has advanced both the conceptual understanding and statistical modeling of compound events, often emphasizing the role of dependence between underlying drivers. Building on this foundation, we recognize the need to further describe hydrological and climate phenomena not only through the overall tendency of contributing physical drivers to interact but also by considering the individual combinations of drivers leading to compound events, regardless of whether these drivers are statistically dependent. Here, we introduce the concepts of structural and transient compoundness as system properties of compound events. Structural compoundness quantifies the overall contribution of each driver to the generative mechanism of the compound variable, while transient compoundness captures the role of each driver in the individual realization of the compound variable. This will ultimately improve risk assessment and mitigation strategies. Indeed, by applying these concepts to two types of compound events, river discharge at confluences and hot and dry events, we show that structural and transient compoundness can inform modeling choices and enhance system understanding, with both the dependence and magnitude of the drivers playing critical roles in shaping compound events. Incorporating structural and transient compoundness enhances scientific modeling and, at the same time, improves risk assessment, guiding adaptation strategies for persistent (structural) and temporary (transient) risks.
Abstract The US Western Interconnection is facing unprecedented challenges in the form of less predictable peak energy demand, increasingly diverse generating resources, and fast‐growing loads due to the onset of artificial intelligence, hyperscale computing, and electrification. Projecting where future generation may be developed is critical to maintaining a robust and resilient electric grid under this mounting uncertainty and variability. Using an integrated multisectoral, multiscale modeling framework that links a human‐Earth systems model, an hourly load model, a geospatial power plant siting model, and an hourly grid operations model, we evaluate the power plant landscape evolution under eight alternative futures between 2020 and 2055. These futures represent a wide but plausible range of atmospheric conditions, emissions constraints, and economic, technological, and population growth assumptions. We find that local‐level development can vary substantially both by generation type and capacity buildout across these futures. Specific regions of the Western Interconnection are projected to see large amounts of capacity development regardless of the future scenario given local siting drivers. We additionally determine that projected power plant locations are more heavily influenced by the cost to interconnect to the electric grid than the revenue driven by locational energy prices.
