MIT University

Drought-induced canopy browning and recovery dynamics threaten ecosystem stability worldwide, with Australia serving as a representative case. This study examined the 2019–2020 drought and subsequent recovery in Eucalyptus forests across two bioregions of the Australian state of New South Wales (NSW): the North Coast and South Eastern Highlands. Canopy browning and recovery were quantified using a 2010–2022 Sentinel-2 time series of Normalized Burn Ratio (NBR), which has been previously identified as the most effective spectral index for detecting drought-related declines in canopy greenness, and were validated with field-measured canopy health. Artificial neural networks were used to link NBR z-scores with climatic (precipitation, temperature, potential evapotranspiration), topographic (Topographic Wetness Index, aspect, slope), soil, and vegetation variables. Lagged and cumulative precipitation and temperature emerged as the dominant drivers of canopy browning, while recovery was influenced by potential evapotranspiration and temperature. Regional contrasts underscored the role of local climate, topography, and vegetation composition in shaping drought impacts and post-drought recovery trajectories. • Drought-induced canopy browning (decline in spectral greenness) and subsequent recovery were driven by distinct environmental factors. • Extreme heat and rapid soil moisture loss caused immediate canopy browning. • Recovery required post-drought rain, moderate heat, and favorable topography. • Poorly recovered forests were highly sensitive to past climate and site factors.

Constructed floating wetlands (CFWs) provide a passive pathway for the retention and accumulation of nutrients and perfluoroalkyl and polyfluoroalkyl substances (PFAS) present in treated wastewater (TWW). This two-year field-scale study investigated the accumulation of PFAS and nutrients in three native wetland species in a lagoon setting. During the plant establishment period and despite measurable nutrient and PFAS uptake in plant tissues, their removal from the water was limited due to short residence times (∼10–12 h). The highest nutrient accumulation in shoots was observed in Phragmites australis (0.14 g N/m2/day and 0.02 g P/m2/day), which showed the highest biomass (12.1 g/m2/day). Phragmites australis also exhibited the highest PFAS and precursor accumulation in tissues. PFAS removal improved after plant establishment, and the daily removal rate for PFAS (based on PFAS accumulation in plant tissue and sediment) was 240 ng/m2/d. Overall, CFWs show promise as a passive, modular solution for PFAS reduction in TWW, particularly after root system development and biofilm formation. However, long-term monitoring across multiple growth and harvest cycles and assessing the contribution of other mechanisms (such as sorption to growth media and accumulation in biofilms) are required to evaluate the CFW system sustainability under varying environmental conditions.

The rapid proliferation of rooftop photovoltaics and behind-the-meter battery storage is creating systemic risks in distribution networks. This includes local network constraint violations, insufficient outage resilience, and enlarged cybersecurity attack surfaces, leaving millions of demand-side distributed energy resources (DERs) to operate without coordinated oversight. Existing solutions (dynamic operating envelopes, virtual power plants, peer-to-peer trading, and single-household energy management systems) each address only partial aspects of this challenge. Based on multi-agent reinforcement learning, this paper proposes a Decentralised Coordination Framework (DCF) that organises the energy dispatch into a three-tier hierarchical architecture. Specifically, it has a user layer executing Proximal Policy Optimisation (PPO) for local dispatch, a feeder layer in which a dynamically elected L1 leader applies MADDPG to coordinate community flexibility via a Virtual Aggregation Unit (VAU); and a cross-feeder layer where an L2 leader manages inter-community balancing and market interfaces. It integrates a directed acyclic graph (DAG)-based verifiable execution ledger, Paillier homomorphic encryption, and LLM-based anomaly detection to enhance security. Potential market participation pathway and revenue distribution mechanism are proposed to align with the Australian National Electricity Market. The DCF provides a scalable, market-ready foundation for commercial demandside DER deployment under high renewable penetration.

Poly (lactic acid) (PLA) is characterized by inherent flammability and limited mechanical performance, which restrict its engineering applications. In this study, ammonium polyphosphate (APP) was utilized as the primary flame retardant, while tannic acid (TA) and tea polyphenols (TP) acted as natural synergists. By combining these components with steam-exploded sugarcane fibers (SCF), we developed a fully bio-based intumescent flame-retardant PLA composite. The effects of polyphenol-assisted APP on thermal stability, flame retardancy, mechanical properties, and flame-retardant mechanisms were systematically investigated. The results indicate that polyphenols significantly enhance the charring ability of APP and improve the high-temperature thermal stability of the composites. Cone calorimetry demonstrates notable reductions in peak heat release rate and total heat release, along with an increased fire performance index. At a total flame-retardant loading of 13 wt%, the composites achieved a UL-94 V-0 rating with effective suppression of melt dripping. However, a moderate reduction in tensile strength and elongation at break was observed due to the incorporation of inorganic flame-retardant components. The addition of SCF partially mitigates the mechanical deterioration caused by the incorporation of flame-retardants. Char analysis indicates a condensed-phase-dominated mechanism, wherein polyphenols serve as efficient carbon sources and synergize with APP to form a dense, phosphorus-rich char layer. Notably, tea polyphenols exhibit slightly higher charring efficiency than tannic acid.

The global biosensors market is rapidly growing, driven by increasing demand for quick, affordable, and portable diagnostic tools across sectors, including healthcare, environmental monitoring, food safety, and biomedical research. The key to biosensor function is the biological recognition element (BRE), which determines the specificity, sensitivity, and reliability of the device. Traditional BREs, such as antibodies and enzymes, face significant limitations, including instability, high costs, and variability. Phage display technology offers a strong alternative, providing durable, stable, and highly specific peptides as BREs. However, it requires effective amplification and purification methods to produce high-quality peptide libraries. This study examines key factors affecting the amplification of filamentous M13 bacteriophage, highlighting the negative impact of high multiplicity of infection (MOI) caused by superinfection exclusion and reduced phage adsorption efficiency. Our findings indicate that the bacterial growth phase is the most important determinant of M13 amplification efficiency. Furthermore, post-infection PEG/NaCl precipitation followed by high-speed centrifugation significantly outperforms traditional filtration methods in purifying phages, maximizing recovery and viability. These findings present an optimized, reproducible, and scalable approach to M13 phage amplification, improving the effectiveness of phage display for developing advanced biorecognition elements. Ultimately, this research provides a foundational framework for more efficient biosensing and therapeutic applications, filling critical gaps in the current biosensor development landscape.

Graphs effectively model interactions in real-world applications such as social and trade networks, where Graph Neural Networks (GNNs) excel at tasks such as link prediction to enhance user experiences. Despite these benefits, users raise privacy concerns as user data can be exploited to improve GNN performance without consent. Accordingly, various graph unlearning methods have been developed. Prior work shows that comparing models before and after unlearning enables attackers to launch former membership inference attacks (FMIA) on unlearned data. However, the imprint of unlearned data left in the unlearned model itself remains underexplored, and existing membership inference methods mainly exploit overfitting, making them ineffective for identifying unlearned data. To address this, we conducted theoretical analysis and proposed an attack framework targeting unlearned GNNs by learning the distribution patterns of unlearned data to distinguish them from normal test data. Extensive experiments on four real-world datasets and GNN architectures confirm our framework's effectiveness and reveal significant vulnerabilities in current graph unlearning methods.

A major challenge in developing robust and generalizable Human Activity Recognition (HAR) systems for smart homes is the lack of large and diverse labeled datasets. Variations in home layouts, sensor configurations, and individual behaviors further exacerbate this issue. To address this, we leverage the idea of embodied AI agents—virtual agents that perceive and act within simulated environments guided by internal world models. We introduce AgentSense, a virtual data generation pipeline in which agents live out daily routines in simulated smart homes, with behavior guided by Large Language Models (LLMs). The LLM generates diverse synthetic personas and realistic routines grounded in the environment, which are then decomposed into fine-grained actions. These actions are executed in an extended version of the VirtualHome simulator, which we augment with virtual ambient sensors that record the agents’ activities. Our approach produces rich, privacy-preserving sensor data that reflects real-world diversity. We evaluate AgentSense on five real HAR datasets. Models pretrained on the generated data consistently outperform baselines, especially in low-resource settings. Furthermore, combining the generated virtual sensor data with a small amount of real data achieves performance comparable to training on full real-world datasets. These results highlight the potential of using LLM-guided embodied agents for scalable and cost-effective sensor data generation in HAR.

Low carbon dwelling improvements combine upgrades to thermal envelopes and appliances with renewable energy systems. Despite proven cost-effective options for emissions reductions and quality of life enhancements, widespread adoption remains elusive. Instead, such projects are often isolated, with uneven distribution across locations, housing contexts and socio-economic groups. Many literature and stakeholder-informed models of upscaling low carbon housing improvements are grounded in the logic that households can and will make optimising choices around long-term cost savings, and that markets of supply and demand for such home energy improvements are mature, functioning, and effective. In this paper, we demonstrate that, in reality, the achievement of low carbon home improvements is situational, social and contextual - shaped by the complexity of dwelling and householder characteristics and multiple market failures. Our study involved interviews with 161 Australian households to explore their past and planned home improvements. We focus on the socio-material dimensions of low carbon housing improvements and present an empirically-based model designed to more accurately reflect diverse societal capabilities. By combining elements of Social Practice Theories and the Capabilities Approach, we identified how social structures, norms and routines along with materials and shared understandings created and perpetuated inequalities in achieving energy-efficient, low carbon homes. Our framework suggests low carbon housing improvement policies and industry practices should address technical and affordability aspects while also considering the relational and emotional dimensions of home making. • Combines social practices and capabilities into one framework • Draws on a large dataset of householder interviews and home visits • Reveals the social construction and role of capabilities in home energy improvements • Contributes to a better understanding of inequality in home energy improvements • May help improve the uptake, equity and effectiveness of home energy improvements