We showcase a hybrid modeling framework that embeds machine learning (ML) inference into the Geophysical Fluid Dynamics Laboratory Seamless System for Prediction and Earth System Research (SPEAR) climate model for online sea ice bias correction during a set of global fully coupled 1-year retrospective forecasts. We compare two hybrid versions of SPEAR to understand the importance of exposing ML models to coupled ice-atmosphere-ocean feedbacks before implementation into fully coupled simulations: HybridCPL (couple trained; with feedbacks) and HybridIO (ice ocean trained; without feedbacks). Relative to SPEAR, HybridCPL systematically reduces seasonal forecast errors in the Arctic and considerably reduces Antarctic errors for target months May to December, with >2× error reduction in 4- to 6-month lead forecasts of Antarctic winter sea ice extent. Meanwhile, HybridIO suffers from out-of-sample behavior that can trigger a chain of Southern Ocean feedbacks, leading to ice-free Antarctic summers. Our results emphasize that ML can demonstrably improve numerical sea ice prediction capabilities and that exposing ML models to coupled ice-atmosphere-ocean processes is essential for generalization in fully coupled simulations.
As spacecraft become more software-driven and interconnected, onboard flight software is an increasingly important security boundary. Popular flight software architectures often treat onboard components as trusted peers, simplifying integration while limiting internal isolation and access control. We analyze NASA's Core Flight Software (cFS) to examine how authority, identity, communication, observability, and persistence are distributed across onboard components. Using NASA's flight-representative NOS3 simulator, we validate these weaknesses through five experiments implemented with a malicious onboard component that abuses legitimate architectural privileges. We then compare cFS with other modular flight software frameworks to identify recurring trust assumptions and architectural weaknesses. Our results show that a single compromised component can exploit broadly shared authority in ways that are difficult to distinguish from legitimate behavior. We conclude with architectural implications and discuss mechanisms for strengthening internal trust boundaries in future flight software systems.
Despite progress in continental United States (CONUS) tornado forecasting, operational seasonal outlooks remain challenging due to intrinsic predictability limits. While previous studies link the El Niño–Southern Oscillation (ENSO) to spring tornado frequency, relying on single tropical indices misses crucial regional variability. Here, we present a novel hybrid dynamical-statistical framework that introduces a paradigm for long-lead predictability. Utilizing retrospective forecasts from NOAA/GFDL’s state-of-the-art SPEAR model, we demonstrate for the first time that a strategic selection of global sea surface temperature modes yields highly useful, skillful predictions with lead times approaching 4 months. Dominant predictable patterns are rooted in transitioning ENSO phases, the Pacific Meridional Mode, and the trend-associated Caribbean High. Unlike conventional predictors, our framework filters internal noise and captures high-wavenumber anomalies. These findings carry substantial implications for climate services, bridging the gap between weather forecasting and seasonal risk management.
The rapid expansion of solar and wind energy across the United States is increasing the demand for reliable seasonal forecasts to guide energy planning and grid operations. Here we show high seasonal prediction skill for springtime solar energy resources over the southwestern United States using a state-of-the-art seasonal forecast system. We identify two leading sources of solar predictability: a dominant component linked to primary El Niño-Southern Oscillation conditions, and a secondary component associated with El Niño-Southern Oscillation diversity and tropical Atlantic variability. Together, these components explain the spatial structure of forecast skill and its underlying physical drivers. State-level evaluations show robust solar prediction skill in high-resource states. The forecast system also captures the compound behavior of wind-solar variability and historical co-fluctuations associated with seasonal energy droughts and surpluses. These results provide a physically interpretable foundation for integrating seasonal forecasts into wind-solar energy planning and compound risk management. Spring solar resources can be reliably predicted across the United States Southwest, with skillful wind-solar variability forecasts and key climate drivers identified, according to a state-of-the-art seasonal forecast system and predictability analysis.
BACKGROUND:We previously identified three clinically relevant immune [i.e., low-tumor-infiltrating lymphocyte (TIL), high-TIL, and high-interferon-stimulated gene (ISG)] subtypes of luminal breast cancer using RNA sequencing in a cohort of women from Hong Kong. In this study, we evaluated whether standard, in situ immunohistochemistry (IHC), a more clinically accessible approach, could capture similar immune heterogeneity across intrinsic (PAM50) subtypes. METHODS:In 148 patients with invasive breast cancer, we performed dual IHC staining for eight markers (CD3, CD8, CD20, FOXP3, CD68, CD163, PDL1, and IDO1) with pan-cytokeratin and quantified immune-positive cells in stromal regions by digital image analysis. RESULTS:IHC marker expression strongly correlated with gene expression-based immune and intrinsic subtypes. Compared with luminal tumors classified as low-TIL, high-TIL tumors showed higher levels of CD3, CD8, CD20, PDL1, and IDO1. Similarly, high-ISG tumors expressed higher CD3, CD8, CD20, PDL1, and IDO1 than low-TIL tumors but were distinguished from high-TIL tumors by elevated CD163. Immune marker expression was generally higher in HER2-enriched and basal-like tumors than in luminal A and B tumors, though substantial variability existed within each subtype. Luminal B tumors expressed higher levels of CD3, CD8, CD20, and CD163 compared with luminal A tumors. Notably, immune infiltration was more strongly aligned with aggressive tumor features such as high grade, risk of recurrence scores, and TP53 mutations than with estrogen receptor/progesterone receptor expression. CONCLUSIONS:These findings confirm that subsets of luminal tumors harbor inflamed or macrophage-driven immune microenvironments. IMPACT:Parsimonious IHC panels can perform immune classification, supporting their potential utility in refining subtype stratification and informing clinical decision-making, particularly in resource-limited settings.
