Into the unknown: The role of post-fire soil erosion in the carbon cycle.

Wildfires directly emit 2.1 Pg carbon (C) to the atmosphere annually. The net effect of wildfires on the C cycle, however, involves many interacting source and sink processes beyond these emissions from combustion. Among those, the role of post-fire enhanced soil organic carbon (SOC) erosion as a C sink mechanism remains essentially unquantified. Wildfires can greatly enhance soil erosion due to the loss of protective vegetation cover and changes to soil structure and wettability. Post-fire SOC erosion acts as a C sink when off-site burial and stabilization of C eroded after a fire, together with the on-site recovery of SOC content, exceed the C losses during its post-fire transport. Here we synthesize published data on post-fire SOC erosion and evaluate its overall potential to act as longer-term C sink. To explore its quantitative importance, we also model its magnitude at continental scale using the 2017 wildfire season in Europe. Our estimations show that the C sink ability of SOC water erosion during the first post-fire year could account for around 13% of the C emissions produced by wildland fires. This indicates that post-fire SOC erosion is a quantitatively important process in the overall C balance of fires and highlights the need for more field data to further validate this initial assessment.

TorchMD-Net 2.0: Fast Neural Network Potentials for Molecular Simulations.

Achieving a balance between computational speed, prediction accuracy, and universal applicability in molecular simulations has been a persistent challenge. This paper presents substantial advancements in TorchMD-Net software, a pivotal step forward in the shift from conventional force fields to neural network-based potentials. The evolution of TorchMD-Net into a more comprehensive and versatile framework is highlighted, incorporating cutting-edge architectures such as TensorNet. This transformation is achieved through a modular design approach, encouraging customized applications within the scientific community. The most notable enhancement is a significant improvement in computational efficiency, achieving a very remarkable acceleration in the computation of energy and forces for TensorNet models, with performance gains ranging from 2× to 10× over previous, nonoptimized, iterations. Other enhancements include highly optimized neighbor search algorithms that support periodic boundary conditions and smooth integration with existing molecular dynamics frameworks. Additionally, the updated version introduces the capability to integrate physical priors, further enriching its application spectrum and utility in research. The software is available at https://github.com/torchmd/torchmd-net.

TorchMD-Net 2.0: Fast Neural Network Potentials for Molecular Simulations.

Achieving a balance between computational speed, prediction accuracy, and universal applicability in molecular simulations has been a persistent challenge. This paper presents substantial advancements in the TorchMD-Net software, a pivotal step forward in the shift from conventional force fields to neural network-based potentials. The evolution of TorchMD-Net into a more comprehensive and versatile framework is highlighted, incorporating cutting-edge architectures such as TensorNet. This transformation is achieved through a modular design approach, encouraging customized applications within the scientific community. The most notable enhancement is a significant improvement in computational efficiency, achieving a very remarkable acceleration in the computation of energy and forces for Tensor-Net models, with performance gains ranging from 2x to 10x over previous, non-optimized, iterations. Other enhancements include highly optimized neighbor search algorithms that support periodic boundary conditions and smooth integration with existing molecular dynamics frameworks. Additionally, the updated version introduces the capability to integrate physical priors, further enriching its application spectrum and utility in research. The software is available at https://github.com/torchmd/torchmd-net.

PlayMolecule Viewer: A Toolkit for the Visualization of Molecules and Other Data.

PlayMolecule Viewer is a web-based data visualization toolkit designed to streamline the exploration of data resulting from structural bioinformatics or computer-aided drug design efforts. By harnessing state-of-the-art web technologies such as WebAssembly, PlayMolecule Viewer integrates powerful Python libraries directly within the browser environment, which enhances its capabilities to manage multiple types of molecular data. With its intuitive interface, it allows users to easily upload, visualize, select, and manipulate molecular structures and associated data. The toolkit supports a wide range of common structural file formats and offers a variety of molecular representations to cater to different visualization needs. PlayMolecule Viewer is freely accessible at open.playmolecule.org, ensuring accessibility and availability to the scientific community and beyond.

Machine learning coarse-grained potentials of protein thermodynamics.

A generalized understanding of protein dynamics is an unsolved scientific problem, the solution of which is critical to the interpretation of the structure-function relationships that govern essential biological processes. Here, we approach this problem by constructing coarse-grained molecular potentials based on artificial neural networks and grounded in statistical mechanics. For training, we build a unique dataset of unbiased all-atom molecular dynamics simulations of approximately 9 ms for twelve different proteins with multiple secondary structure arrangements. The coarse-grained models are capable of accelerating the dynamics by more than three orders of magnitude while preserving the thermodynamics of the systems. Coarse-grained simulations identify relevant structural states in the ensemble with comparable energetics to the all-atom systems. Furthermore, we show that a single coarse-grained potential can integrate all twelve proteins and can capture experimental structural features of mutated proteins. These results indicate that machine learning coarse-grained potentials could provide a feasible approach to simulate and understand protein dynamics.

NNP/MM: Accelerating Molecular Dynamics Simulations with Machine Learning Potentials and Molecular Mechanics.

Machine learning potentials have emerged as a means to enhance the accuracy of biomolecular simulations. However, their application is constrained by the significant computational cost arising from the vast number of parameters compared with traditional molecular mechanics. To tackle this issue, we introduce an optimized implementation of the hybrid method (NNP/MM), which combines a neural network potential (NNP) and molecular mechanics (MM). This approach models a portion of the system, such as a small molecule, using NNP while employing MM for the remaining system to boost efficiency. By conducting molecular dynamics (MD) simulations on various protein-ligand complexes and metadynamics (MTD) simulations on a ligand, we showcase the capabilities of our implementation of NNP/MM. It has enabled us to increase the simulation speed by ∼5 times and achieve a combined sampling of 1 µs for each complex, marking the longest simulations ever reported for this class of simulations.

Accuracy of tropical peat and non-peat fire forecasts enhanced by simulating hydrology.

Soil moisture deficits and water table dynamics are major biophysical controls on peat and non-peat fires in Indonesia. Development of modern fire forecasting models in Indonesia is hampered by the lack of scalable hydrologic datasets or scalable hydrology models that can inform the fire forecasting models on soil hydrologic behaviour. Existing fire forecasting models in Indonesia use weather data-derived fire probability indices, which often do not adequately proxy the sub-surface hydrologic dynamics. Here we demonstrate that soil moisture and water table dynamics can be simulated successfully across tropical peatlands and non-peatland areas by using a process-based eco-hydrology model (ecosys) and publicly available data for weather, soil, and management. Inclusion of these modelled water table depth and soil moisture contents significantly improves the accuracy of a neural network model in predicting active fires at two-weekly time scale. This constitutes an important step towards devising an operational fire early warning system for Indonesia.

Membrane Composition and Raf[CRD]-Membrane Attachment Are Driving Forces for K-Ras4B Dimer Stability.

Ras proteins are membrane-anchored GTPases that regulate key cellular signaling networks. It has been recently shown that different anionic lipid types can affect the properties of Ras in terms of dimerization/clustering on the cell membrane. To understand the effects of anionic lipids on key spatiotemporal properties of dimeric K-Ras4B, we perform all-atom molecular dynamics simulations of the dimer K-Ras4B in the presence and absence of Raf[RBD/CRD] effectors on two model anionic lipid membranes: one containing 78% mol DOPC, 20% mol DOPS, and 2% mol PIP2 and another one with enhanced concentration of anionic lipids containing 50% mol DOPC, 40% mol DOPS, and 10% mol PIP2. Analysis of our results unveils the orientational space of dimeric K-Ras4B and shows that the stability of the dimer is enhanced on the membrane containing a high concentration of anionic lipids in the absence of Raf effectors. This enhanced stability is also observed in the presence of Raf[RBD/CRD] effectors although it is not influenced by the concentration of anionic lipids in the membrane, but rather on the ability of Raf[CRD] to anchor to the membrane. We generate dominant K-Ras4B conformations by Markov state modeling and yield the population of states according to the K-Ras4B orientation on the membrane. For the membrane containing anionic lipids, we observe correlations between the diffusion of K-Ras4B and PIP2 and anchoring of anionic lipids to the Raf[CRD] domain. We conclude that the presence of effectors with the Raf[CRD] domain anchoring on the membrane as well as the membrane composition both influence the conformational stability of the K-Ras4B dimer, enabling the preservation of crucial interface interactions.

PlayMolecule Glimpse: Understanding Protein-Ligand Property Predictions with Interpretable Neural Networks.

Deep learning has been successfully applied to structure-based protein-ligand affinity prediction, yet the black box nature of these models raises some questions. In a previous study, we presented KDEEP, a convolutional neural network that predicted the binding affinity of a given protein-ligand complex while reaching state-of-the-art performance. However, it was unclear what this model was learning. In this work, we present a new application to visualize the contribution of each input atom to the prediction made by the convolutional neural network, aiding in the interpretability of such predictions. The results suggest that KDEEP is able to learn meaningful chemistry signals from the data, but it has also exposed the inaccuracies of the current model, serving as a guideline for further optimization of our prediction tools.

Response of Calamagrostis angustifolia to burn frequency and seasonality in the Sanjiang Plain wetlands (Northeast China).

Fire is an important disturbance in many wetlands, which are key carbon reservoirs at both regional and global scales. However, the effects of fire on wetland vegetation biomass and plant carbon dynamics are poorly understood. We carried out a burn experiment in a Calamagrostis angustifolia wetland in Sanjiang Plain (Northeast China), which is widespread wetland type in China and frequently exposed to fire. Using a series of replicated experimental annual burns over a three-year period (spring and autumn burns carried out one, two or three times over three consecutive years), together with a control unburned treatment, we assessed the effect of burn seasonality and frequency on aboveground biomass, stem density, and carbon content of aboveground plant parts and ground litter. We found that burning promoted plant growth and hence plant biomass in burned sites compared to the unburned control, with this effect being greatest after three consecutive burn years. Autumn burns promoted higher stem density and more total aboveground biomass than spring burns after three consecutive burn years. Burning increased stem density significantly, especially in twice and thrice burned plots, with stem densities in September over 2000 N/m2, which was much higher than in the control plots (987 ± 190 N/m2). Autumn burns had a larger effect than spring burns on total plant biomass and litter accumulated (e.g. 1236 ± 295 g/m2 after thrice autumn burns compared 796.2 ± 66.6 g/m2 after thrice spring burns), except after two burn treatments. With time since burning, total biomass loads increased in spring-burned plots, while autumn-burned plots showed the opposite trend, declining towards values found at unburned plots in year three. Our results suggest that, at short fire return intervals, autumn burns lead to a more pronounced increase in aboveground biomass and carbon accumulation than spring burns; however, the effects of spring burns on biomass and carbon accumulation are longer lasting than those observed for autumn burns.

Scientists' warning on extreme wildfire risks to water supply.

2020 is the year of wildfire records. California experienced its three largest fires early in its fire season. The Pantanal, the largest wetland on the planet, burned over 20% of its surface. More than 18 million hectares of forest and bushland burned during the 2019-2020 fire season in Australia, killing 33 people, destroying nearly 2500 homes, and endangering many endemic species. The direct cost of damages is being counted in dozens of billion dollars, but the indirect costs on water-related ecosystem services and benefits could be equally expensive, with impacts lasting for decades. In Australia, the extreme precipitation ("200 mm day -1 in several location") that interrupted the catastrophic wildfire season triggered a series of watershed effects from headwaters to areas downstream. The increased runoff and erosion from burned areas disrupted water supplies in several locations. These post-fire watershed hazards via source water contamination, flash floods, and mudslides can represent substantial, systemic long-term risks to drinking water production, aquatic life, and socio-economic activity. Scenarios similar to the recent event in Australia are now predicted to unfold in the Western USA. This is a new reality that societies will have to live with as uncharted fire activity, water crises, and widespread human footprint collide all-around of the world. Therefore, we advocate for a more proactive approach to wildfire-watershed risk governance in an effort to advance and protect water security. We also argue that there is no easy solution to reducing this risk and that investments in both green (i.e., natural) and grey (i.e., built) infrastructure will be necessary. Further, we propose strategies to combine modern data analytics with existing tools for use by water and land managers worldwide to leverage several decades worth of data and knowledge on post-fire hydrology.

Wildfire-Derived Pyrogenic Carbon Modulates Riverine Organic Matter and Biofilm Enzyme Activities in an In Situ Flume Experiment.

Wildfires produce large amounts of pyrogenic carbon (PyC), including charcoal, known for its chemical recalcitrance and sorption affinity for organic molecules. Wildfire-derived PyC can be transported to fluvial networks. Here it may alter the dissolved organic matter (DOM) concentration and composition as well as microbial biofilm functioning. Effects of PyC on carbon cycling in freshwater ecosystems remain poorly investigated. Employing in-stream flumes with a control versus treatment design (PyC pulse addition), we present evidence that field-aged PyC inputs to rivers can increase the dissolved organic carbon (DOC) concentration and alter the DOM composition. DOM fluorescence components were not affected by PyC. The in-stream DOM composition was altered due to leaching of pyrogenic DOM from PyC and possibly concurrent sorption of riverine DOM to PyC. Decreased DOM aromaticity indicated by a lower SUVA245 (-0.31 unit) and a higher pH (0.25 unit) was associated with changes in enzymatic activities in benthic biofilms, including a lower recalcitrance index (ß-glucosidase/phenol oxidase), suggesting preferential usage of recalcitrant over readily available DOM by biofilms. The deposition of particulate PyC onto biofilms may further modulate the impacts of PyC due to direct contact with the biofilm matrix. This study highlights the importance of PyC for in-stream biogeochemical organic matter cycling in fire-affected watersheds.

Boreal forest soil carbon fluxes one year after a wildfire: Effects of burn severity and management.

The extreme 2018 hot drought that affected central and northern Europe led to the worst wildfire season in Sweden in over a century. The Ljusdal fire complex, the largest area burnt that year (8995 ha), offered a rare opportunity to quantify the combined impacts of wildfire and post-fire management on Scandinavian boreal forests. We present chamber measurements of soil CO2 and CH4  fluxes, soil microclimate and nutrient content from five Pinus sylvestris sites for the first growing season after the fire. We analysed the effects of three factors on forest soils: burn severity, salvage-logging and stand age. None of these caused significant differences in soil CH4 uptake. Soil respiration, however, declined significantly after a high-severity fire (complete tree mortality) but not after a low-severity fire (no tree mortality), despite substantial losses of the organic layer. Tree root respiration is thus key in determining post-fire soil CO2 emissions and may benefit, along with heterotrophic respiration, from the nutrient pulse after a low-severity fire. Salvage-logging after a high-severity fire had no significant effects on soil carbon fluxes, microclimate or nutrient content compared with leaving the dead trees standing, although differences are expected to emerge in the long term. In contrast, the impact of stand age was substantial: a young burnt stand experienced more extreme microclimate, lower soil nutrient supply and significantly lower soil respiration than a mature burnt stand, due to a thinner organic layer and the decade-long effects of a previous clear-cut and soil scarification. Disturbance history and burn severity are, therefore, important factors for predicting changes in the boreal forest carbon sink after wildfires. The presented short-term effects and ongoing monitoring will provide essential information for sustainable management strategies in response to the increasing risk of wildfire.

Designing tools to predict and mitigate impacts on water quality following the Australian 2019/2020 wildfires: Insights from Sydney's largest water supply catchment.

The 2019/2020 Australian bushfires (or wildfires) burned the largest forested area in Australia's recorded history, with major socio-economic and environmental consequences. Among the largest fires was the 280 000 ha Green Wattle Creek Fire, which burned large forested areas of the Warragamba catchment. This protected catchment provides critical ecosystem services for Lake Burragorang, one of Australia's largest urban supply reservoirs delivering ~85% of the water used in Greater Sydney. Water New South Wales (WaterNSW) is the utility responsible for managing water quality in Lake Burragorang. Its postfire risk assessment, done in collaboration with researchers in Australia, the UK, and United States, involved (i) identifying pyrogenic contaminants in ash and soil; (ii) quantifying ash loads and contaminant concentrations across the burned area; and (iii) estimating the probability and quantity of soil, ash, and associated contaminant entrainment for different rainfall scenarios. The work included refining the capabilities of the new WEPPcloud-WATAR-AU model (Water Erosion Prediction Project cloud-Wildfire Ash Transport And Risk-Australia) for predicting sediment, ash, and contaminant transport, aided by outcomes from previous collaborative postfire research in the catchment. Approximately two weeks after the Green Wattle Creek Fire was contained, an extreme rainfall event (~276 mm in 72 h) caused extensive ash and sediment delivery into the reservoir. The risk assessment informed on-ground monitoring and operational mitigation measures (deployment of debris-catching booms and adjustment of the water supply system configuration), ensuring the continuity of safe water supply to Sydney. WEPPcloud-WATAR-AU outputs can prioritize recovery interventions for managing water quality risks by quantifying contaminants on the hillslopes, anticipating water contamination risk, and identifying areas with high susceptibility to ash and sediment transport. This collaborative interaction among scientists and water managers, aimed also at refining model capabilities and outputs to meet managers' needs, exemplifies the successful outcomes that can be achieved at the interface of industry and science. Integr Environ Assess Manag 2021;17:1151-1161. © 2021 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

TorchMD: A Deep Learning Framework for Molecular Simulations.

Molecular dynamics simulations provide a mechanistic description of molecules by relying on empirical potentials. The quality and transferability of such potentials can be improved leveraging data-driven models derived with machine learning approaches. Here, we present TorchMD, a framework for molecular simulations with mixed classical and machine learning potentials. All force computations including bond, angle, dihedral, Lennard-Jones, and Coulomb interactions are expressed as PyTorch arrays and operations. Moreover, TorchMD enables learning and simulating neural network potentials. We validate it using standard Amber all-atom simulations, learning an ab initio potential, performing an end-to-end training, and finally learning and simulating a coarse-grained model for protein folding. We believe that TorchMD provides a useful tool set to support molecular simulations of machine learning potentials. Code and data are freely available at github.com/torchmd.

Fires prime terrestrial organic carbon for riverine export to the global oceans.

Black carbon (BC) is a recalcitrant form of organic carbon (OC) produced by landscape fires. BC is an important component of the global carbon cycle because, compared to unburned biogenic OC, it is selectively conserved in terrestrial and oceanic pools. Here we show that the dissolved BC (DBC) content of dissolved OC (DOC) is twice greater in major (sub)tropical and high-latitude rivers than in major temperate rivers, with further significant differences between biomes. We estimate that rivers export 18 ± 4 Tg DBC year-1 globally and that, including particulate BC fluxes, total riverine export amounts to 43 ± 15 Tg BC year-1 (12 ± 5% of the OC flux). While rivers export ~1% of the OC sequestered by terrestrial vegetation, our estimates suggest that 34 ± 26% of the BC produced by landscape fires has an oceanic fate. Biogeochemical models require modification to account for the unique dynamics of BC and to predict the response of recalcitrant OC export to changing environmental conditions.

AdaptiveBandit: A Multi-armed Bandit Framework for Adaptive Sampling in Molecular Simulations.

Sampling from the equilibrium distribution has always been a major problem in molecular simulations due to the very high dimensionality of the conformational space. Over several decades, many approaches have been used to overcome the problem. In particular, we focus on unbiased simulation methods such as parallel and adaptive sampling. Here, we recast adaptive sampling schemes on the basis of multi-armed bandits and develop a novel adaptive sampling algorithm under this framework, AdaptiveBandit. We test it on multiple simplified potentials and in a protein folding scenario. We find that this framework performs similarly to or better than previous methods in every type of test potential. Furthermore, it provides a novel framework to develop new sampling algorithms with better asymptotic characteristics.

Water repellency reduces soil CO2 efflux upon rewetting.

Carbon dioxide (CO2) efflux from soil represents one of the biggest ecosystem carbon (C) fluxes and high-magnitude pulses caused by rainfall make a substantial contribution to the overall C emissions. It is widely accepted that the drier the soil, the larger the CO2 pulses will be, but this notion has never been tested for water-repellent soils. Soil water repellency (SWR) is a common feature of many soils and is especially prominent after dry periods or fires. An important unanswered question is to what degree SWR affects common assumptions about soil CO2 dynamics. To address this, our study investigates, for the first time, the effect of SWR on the CO2 pulse upon wetting for water-repellent soils from recently burned forest sites. CO2 efflux measurements in response to simulated wetting were conducted both under laboratory and in situ conditions. Experiments were conducted on severely and extremely water-repellent soils, with a wettable scenario simulated by adding a wetting agent to the water. CO2 efflux upon rewetting was significantly lower in the water-repellent scenarios. Under laboratory conditions, CO2 pulse was up to four times lower under the water-repellent scenario as a result of limited wetting, with 70% of applied water draining rapidly via preferential flow paths, leaving much of the soil dry. We suggest that the predominant cause of the lower CO2 pulse in water-repellent soils was the smaller volume of pores in which the CO2 was replaced by infiltrating water, compared to wettable soil. This study shows that SWR should be considered as an important factor when measuring or predicting the CO2 flush upon rewetting of dry soils. Although this study focused mainly on short-term effects of rewetting on CO2 fluxes, the overall implications of SWR on physical changes in soil conditions can be long lasting, with overall larger consequences for C dynamics.

A Scalable Molecular Force Field Parameterization Method Based on Density Functional Theory and Quantum-Level Machine Learning.

Fast and accurate molecular force field (FF) parameterization is still an unsolved problem. Accurate FF are not generally available for all molecules, like novel druglike molecules. While methods based on quantum mechanics (QM) exist to parameterize them with better accuracy, they are computationally expensive and slow, which limits applicability to a small number of molecules. Here, we present an automated FF parameterization method which can utilize either density functional theory (DFT) calculations or approximate QM energies produced by different neural network potentials (NNPs), to obtain improved parameters for molecules. We demonstrate that for the case of torchani-ANI-1x NNP, we can parameterize small molecules in a fraction of time compared with an equivalent parameterization using DFT QM calculations while producing more accurate parameters than FF (GAFF2). We expect our method to be of critical importance in computational structure-based drug discovery (SBDD). The current version is available at PlayMolecule ( www.playmolecule.org ) and implemented in HTMD, allowing to parameterize molecules with different QM and NNP options.