SEARCH · Engineering Papers

Results for “Process sensitivity”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,477 records · Page 82

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Smart Pixels: In-pixel AI for on-sensor data filtering

We present a smart pixel prototype readout integrated circuit (ROIC) designed in CMOS 28 nm bulk process, with in-pixel implementation of an artificial intelligence (AI) / machine learning (ML) based data filtering algorithm designed as proof-of-principle for a Phase III upgrade at the Large Hadron Collider (LHC) pixel detector. The first version of the ROIC consists of two matrices of 256 smart pixels, each 25$\times$25 $\mu$m$^2$ in size. Each pixel consists of a charge-sensitive preamplifier with leakage current compensation and three auto-zero comparators for a 2-bit flash-type ADC. The frontend is capable of synchronously digitizing the sensor charge within 25 ns. Measurement results show an equivalent noise charge (ENC) of $\sim$30e$^-$ and a total dispersion of $\sim$100e$^-$ The second version of the ROIC uses a fully connected two-layer neural network (NN) to process information from a cluster of 256 pixels to determine if the pattern corresponds to highly desirable high-momentum particle tracks for selection and readout. The digital NN is embedded in-between analog signal processing regions of the 256 pixels without increasing the pixel size and is implemented as fully combinatorial digital logic to minimize power consumption and eliminate clock distribution, and is active only in the presence of an input signal. The total power consumption of the neural network is $\sim$ 300 $\mu$W. The NN performs momentum classification based on the generated cluster patterns and even with a modest momentum threshold, it is capable of 54.4% - 75.4% total data rejection, opening the possibility of using the pixel information at 40MHz for the trigger. The total power consumption of analog and digital functions per pixel is $\sim$ 6 $\mu$W per pixel, which corresponds to $\sim$ 1 W/cm$^2$ staying within the experimental constraints.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2026

Improving Chirped Fiber Bragg Grating Resolution for Position-Sensitive Sensors in Shock- and Detonation-Driven Experiments

Chirped fiber Bragg gratings (CFBGs) are robust diagnostic sensors that are widely used to track detonation-driven and shock wave propagation. CFBGs are inscribed with a linearly chirped periodic index of refraction changes that alter the Bragg wavelength along the length of the probe. The light return of each individual Bragg element is captured by a detector at a unique time to map the full reflected spectrum. The CFBG spectrum is measured with a dispersive Fourier transform of the reflected light that temporally stretches the spectrum to increase spatial resolution and make a one-to-one map of the wavelength on a time axis. Here, we propose an improvement of CFBG temporal resolution by incorporating two co-linear laser pulses with orthogonal polarization states and a 5 ns time offset. The two separate signals were split and tracked by two separate detectors. An oscilloscope captured good separation in the signals, and two separate spectrograms were generated and interleaved in the post-processing of the data. This novel technique doubled the CFBG temporal resolution and led to a doubled location resolution. As a proof-of-concept of this technique, the resolution improvement was compared between standard CFBG measurements and the two polarization states method on a position-sensitive CFBG sensor. CFBG resolution doubling will advance sensor capabilities and will have a direct impact on improving capture and analysis in dynamic, high-explosive experiments.

42 ENGINEERING↗

DOE OSTI2025

Performance evaluations of signed and unsigned noisy approximate quantum Fourier arithmetic

The Quantum Fourier Transform (QFT) grants competitive advantages, especially in resource usage and circuit approximation, for performing arithmetic operations on quantum computers, and offers a potential route toward a numerical quantum-computational paradigm. In this paper, we utilize efficient techniques to implement QFT-based integer addition and multiplications. These operations are fundamental to various quantum applications including Shor’s algorithm, weighted-sum optimization problems in data processing and machine learning, and quantum algorithms requiring inner products. We carry out performance evaluations of these implementations based on IBM’s superconducting-qubit architecture using different compatible noise models. We isolate the sensitivity of the component quantum circuits on both one-/two-qubit gate error rates, and the number of the arithmetic operands’ superposed integer states. We analyze performance and identify the most effective approximation depths for unsigned quantum addition and quantum multiplication within the given context. We then perform a similar analysis of signed addition and compare to the unsigned results. We observe significant dependency of the optimal approximation depth on the degree of machine noise and the number of superposed states in certain performance regimes. Finally, we elaborate on the algorithmic challenges—relevant to signed, unsigned, modular and non-modular versions—that could also be applied to current implementations of QFT-based subtraction, division, exponentiation, and their potential tensor extensions. Here, we analyze the performance trends in our results and speculate on possible future developments within this computational paradigm.

Computational models↗

DOE OSTI2026

Lead Isotope Fingerprinting of Nanoscale Mineral Particles via Single Particle Inductively Coupled Plasma Mass Spectrometry

Determining lead (Pb) isotopic ratios is of broad interest across chemical disciplines for source tracing and age determinations, but the technique is inherently limited when multiple sources of Pb are present in a single sample. Single particle methods offer a solution to directly resolve multiple distinct isotopic ratios within individual samples. In this study, single particle inductively coupled plasma-mass spectrometry (spICP-MS) was performed using time-of-flight (TOF) and multicollector (MC)-based platforms to measure the Pb isotopic composition of individual nanoscale particles. Distinct Pb isotope ratios were measured in particles from four powdered galena samples of different origins (average single-particle 206Pb/207Pb ratios ranging from 0.91 to 1.28) using both the TOF and the MC-based platforms. Differentiation of galena particles in a mixture (liberated grains and those hosted in a silicate matrix) was possible due to detecting their unique multielemental fingerprints. The differing behavior of the two galena samples during digestion was investigated; undigested galena particles trapped within silicate minerals were detected using spICP-MS, which has implications for Pb recovery in bulk-scale isotopic analysis. The multinuclide detection capability of the ICP-TOF-MS instrument allowed for the simultaneous detection of secondary constituent elements within the nanoparticles and the identification of multiple populations of isotopically distinct Pb-bearing particles in a copper ore sample, whereas the increased sensitivity of the MC instrument enabled quantification of 204Pb, allowing differentiation using multiple isotopic ratios. A single particle isotope ratio analysis module was developed within an open-source spICP-TOF-MS data processing platform, enabling its adoption across chemical disciplines.

Goodman, Aaron J. [University of Montreal, Quebec,↗

DOE OSTI2019

Quantification of Anionic Redox Chemistry in a Prototype Na-Rich Layered Oxide

Harnessing anionic redox reactions is of prime importance for boosting the capacity of sodium-ion batteries (NIBs). However, quantifying the cyclability of anionic redox reactions is still challenging. Herein, we conduct a qualitative and quantitative investigation of the cationic and anionic redox reactions of a prototype Na-rich layered oxide, namely, Na 3 RuO 4 , by a combination of bulk-sensitive X-ray absorption spectroscopy and full-range mapping of resonant inelastic X-ray scattering. We unequivocally reveal that both Ru cations and oxygen anions are involved in the charge compensation process of Na 3 RuO 4 . Ru redox is highly reversible over extended electrochemical cycles, while the cyclability of lattice oxygen redox gradually decreases with the retention of only 36% after 30 cycles, which is mainly responsible for the capacity fading of Na 3 RuO 4 . Finally, our findings provide deeper insights into the complex oxygen redox mechanism, which plays a decisive role for designing high-energy Na-rich electrode materials for NIBs.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2023

Dual Emission Bands of a 2D Perovskite Single Crystal with Charge Transfer State Characteristics

Several hybrid halide 2D-perovskite species emit light with an emergent and controversial broadband emission Stokes-shifted down from the narrow band emission. This work uncovers the sub- and above-bandgap emission and absorption characteristics of PEA 2 PbI 4 prepared with gap states introduced during single crystal growth. Here, gap states led to coexistent intrinsic and heterostructured electronic frameworks that are selectively accessible with ultraviolet (UV) and infrared (IR) light, respectively, resulting in the phenomenon of photoluminescence (PL) switching from narrowband green to broadband red. Electron-energy dependent cathodoluminescence shows a relative increase in the broadband red PL intensity as the electron penetration depth increases from 30 nm to 2 μm, confirming the heterostructured framework is formed in the bulk of the crystal. Excitation–emission power slope of 2.5 and up-conversion pump transient absorption (TA) spectra suggest that the IR up-conversion excitation with red photoluminescence, peaked at 655 nm, is a multiphoton process occurring in the heterostructured framework through a nonlinear optical response. The energetic pathways toward the dual emission bands are revealed by pump–probe transient absorption spectroscopy, showing energetically broad gap states with high sensitivity to an IR pump are upconverted and subsequently quickly relax from high to low energy levels within 4 ps. Furthermore, the up-conversion red PL demonstrates a linear polarization with magnetic field effects, thus affirming that the band-like heterostructured framework is crystallographically aligned with characteristics of spatially extended charge-transfer states.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2022

Ba +2 ion trapping using organic submonolayer for ultra-low background neutrinoless double beta detector

If neutrinos are their own antiparticles the otherwise-forbidden nuclear reaction known as neutrinoless double beta decay can occur. The very long lifetime expected for these exceptional events makes its detection a daunting task. In order to conduct an almost background-free experiment, the NEXT collaboration is investigating novel synthetic molecular sensors that may capture the Ba dication produced in the decay of certain Xe isotopes in a high-pressure gas experiment. The use of such molecular detectors immobilized on surfaces must be explored in the ultra-dry environment of a xenon gas chamber. Here, using a combination of highly sensitive surface science techniques in ultra-high vacuum, we demonstrate the possibility of employing the so-called Fluorescent Bicolor Indicator as the molecular component of the sensor. We unravel the ion capture process for these molecular indicators immobilized on a surface and explain the origin of the emission fluorescence shift associated to the ion trapping.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2021

Climate-Mediated Changes to Linked Terrestrial and Marine Ecosystems across the Northeast Pacific Coastal Temperate Rainforest Margin

Coastal margins are important areas of materials flux that link terrestrial and marine ecosystems. Consequently, climate-mediated changes to coastal terrestrial ecosystems and hydrologic regimes have high potential to influence nearshore ocean chemistry and food web dynamics. Research from tightly coupled, high-flux coastal ecosystems can advance understanding of terrestrial–marine links and climate sensitivities more generally. In the present article, we use the northeast Pacific coastal temperate rainforest as a model system to evaluate such links. We focus on key above- and belowground production and hydrological transport processes that control the land-to-ocean flow of materials and their influence on nearshore marine ecosystems. We evaluate how these connections may be altered by global climate change and we identify knowledge gaps in our understanding of the source, transport, and fate of terrestrial materials along this coastal margin. Finally, we propose five priority research themes in this region that are relevant for understanding coastal ecosystem links more broadly.

temperate rainforest, coastal margin, climate chan↗

DOE OSTI2022

Desorption characteristics of selenium and tellurium thin films

The temperature-dependent desorption behavior of selenium and tellurium is investigated using a heated quartz crystal microbalance. Prior to heating the quartz crystal microbalance, selenium and tellurium films with varying thickness were deposited using thermal effusion cells in a molecular beam epitaxy system for subsequent determination of temperature-dependent mass loss of the deposited films. The desorption rate for tellurium was found to exhibit one sharp peak around 190°C, indicating the loss of the entire film irrespective of film thickness within a temperature window of 20°C, which was completely evaporated at 200°C. Similar experiments for selenium revealed that the thermal desorption took place via a two-stage process with a smaller portion of the material desorbing within an even narrower temperature window of 5°C at a much lower peak temperature of 65°C, while most selenium desorbed within a temperature range of 10°C around 90°C. This two-stage behavior indicated the presence of at least two chemically distinct selenium species or binding states. The direct and quantitative determination of the chalcogen desorption process provides important insights into the kinetics of chalcogenide-based film growth and is in addition of applied benefit to the research community in the area of Se/Te capping and decapping of air sensitive materials as it provides temperature ranges and rates at which full desorption is achieved. Our work furthermore points toward the need for a more detailed understanding of the chemical composition state of atomic and molecular beams supplied from thermal evaporation sources during growth.

36 MATERIALS SCIENCE↗

DOE OSTI2025

Integrated Methane Monitoring Platform Extension, Volume I: Final Technical Report

The IMMPE project, DE-FE0032284, was to enhance methane monitoring technologies and their applications across various natural gas asset classes. The scope included deploying advanced methane detection and monitoring technologies to identify and mitigate fugitive methane emissions, measuring emission rates, and assessing impacts. The findings included the successful mitigation of identified emissions and quantification of emission rates. A key outcome was the development of a comprehensive template and summary of recommendations for methane emissions monitoring, which is replicable for both upstream and downstream applications. Furthermore, the project emphasized the importance of education by providing training opportunities for technicians and regulators, thereby fostering awareness and promoting the adoption of cost-effective methane emissions monitoring and management techniques.

02 PETROLEUM↗

DOE OSTI2021

BAWLD-CH 4 : a comprehensive dataset of methane fluxes from boreal and arctic ecosystems

Methane (CH 4 ) emissions from the boreal and arctic region are globally significant and highly sensitive to climate change. There is currently a wide range in estimates of high-latitude annual CH 4 fluxes, where estimates based on land cover inventories and empirical CH 4 flux data or process models (bottom-up approaches) generally are greater than atmospheric inversions (top-down approaches). A limitation of bottom-up approaches has been the lack of harmonization between inventories of site-level CH 4 flux data and the land cover classes present in high-latitude spatial datasets. Here we present a comprehensive dataset of small-scale, surface CH 4 flux data from 540 terrestrial sites (wetland and non-wetland) and 1247 aquatic sites (lakes and ponds), compiled from 189 studies. The Boreal–Arctic Wetland and Lake Methane Dataset (BAWLD-CH 4 ) was constructed in parallel with a compatible land cover dataset, sharing the same land cover classes to enable refined bottom-up assessments. BAWLD-CH 4 includes information on site-level CH 4 fluxes but also on study design (measurement method, timing, and frequency) and site characteristics (vegetation, climate, hydrology, soil, and sediment types, permafrost conditions, lake size and depth, and our determination of land cover class). The different land cover classes had distinct CH 4 fluxes, resulting from definitions that were either based on or co-varied with key environmental controls. Fluxes of CH 4 from terrestrial ecosystems were primarily influenced by water table position, soil temperature, and vegetation composition, while CH 4 fluxes from aquatic ecosystems were primarily influenced by water temperature, lake size, and lake genesis. Models could explain more of the between-site variability in CH 4 fluxes for terrestrial than aquatic ecosystems, likely due to both less precise assessments of lake CH 4 fluxes and fewer consistently reported lake site characteristics. Analysis of BAWLD-CH 4 identified both land cover classes and regions within the boreal and arctic domain, where future studies should be focused, alongside methodological approaches. Overall, BAWLD-CH 4 provides a comprehensive dataset of CH 4 emissions from high-latitude ecosystems that are useful for identifying research opportunities, for comparison against new field data, and model parameterization or validation.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2022

The molecular mechanism of load adaptation by branched actin networks

Branched actin networks are self-assembling molecular motors that move biological membranes and drive many important cellular processes, including phagocytosis, endocytosis, and pseudopod protrusion. When confronted with opposing forces, the growth rate of these networks slows and their density increases, but the stoichiometry of key components does not change. The molecular mechanisms governing this force response are not well understood, so we used single-molecule imaging and AFM cantilever deflection to measure how applied forces affect each step in branched actin network assembly. Although load forces are observed to increase the density of growing filaments, we find that they actually decrease the rate of filament nucleation due to inhibitory interactions between actin filament ends and nucleation promoting factors. The force-induced increase in network density turns out to result from an exponential drop in the rate constant that governs filament capping. The force dependence of filament capping matches that of filament elongation and can be explained by expanding Brownian Ratchet theory to cover both processes. We tested a key prediction of this expanded theory by measuring the force-dependent activity of engineered capping protein variants and found that increasing the size of the capping protein increases its sensitivity to applied forces. In summary, we find that Brownian Ratchets underlie not only the ability of growing actin filaments to generate force but also the ability of branched actin networks to adapt their architecture to changing loads.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2024

Nuclear Data Sensitivity/Uncertainty Studies of Tantalum-Reflected Systems

Tantalum (Ta) is a refractory metal with high melting point and high corrosion resistance. This makes it useful as a mold material for plutonium casting operations. Unfortunately, there are few criticality benchmarks with high nuclear data sensitivities. Given that, it is desirable to design new experiments to provide additional validation data for nuclear criticality safety, which is the objective of the Thales project. Multiple application models were used to represent process conditions (either normal conditions or credible upset conditions) both at Los Alamos National Laboratory (LANL) and Savannah River Site (SRS). A total of 40 application models were obtained. Given this, it was useful to down-select the number of needed application models. This work documents the down-selection based on several nuclear data sensitivity/uncertainty metrics. In addition to showing the result of the specific down-selection used for the Thales project, this work discusses potential ways to represent a large number of application models with a smaller number of proposed experiment designs.

42 ENGINEERING↗

DOE OSTI2024

Emerging Jets Search, Triton Server Deployment, and Track Quality Development: Machine Learning Applications in High Energy Physics

Machine learning is becoming prevalent in high energy physics, with numerous applications in physics analyses and event reconstruction showing great improvements compared to traditional computing methods. This thesis studies three projects which each propose new avenues for machine learning applications within the high energy physics CMS experiment located at CERN. In the first project, a search for a dark matter signal called “emerging jets” is performed, using graph neural networks to greatly increase sensitivity to the signal’s signature within the data. The result of this dark matter search sets the most stringent exclusion limits to date on theoretical emerging jet models. Motivated by inefficiencies encountered when processing the emerging jet graph neural network at Fermi National Accelerator Laboratory’s computing centers, the second project re-optimizes the computing centers for machine learning inference. This re-optimization uses NVIDIA Triton Inference Servers to process users’ analysis code heterogeneously, therefore achieving high processing throughput and decreasing user time-to-insight. The last project focuses on an upgrade to the CMS experiment’s real-time event selection system which improves physics object reconstruction under harsh processing conditions. A boosted decision tree is used to quickly and efficiently quantify a reconstructed particle’s “track quality” in order to remove particle tracks reconstructed erroneously. In summary, this thesis will not only present examples of how high energy physics can greatly benefit by leveraging machine learning techniques for physics analysis and reconstruction, but will also provide guidance on how the field can prepare for the inevitable increase in machine learning applications.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2025

Effect of Networking Density on the Patterning Performance of Molecular Layer Deposited Alucone Electron Beam/EUV Resists

One of the main barriers to continued device scaling in the era of extreme ultraviolet (EUV) lithography is the need for improved photoresist chemistries to address challenges such as poor EUV sensitivity, inadequate etch resistance, and pattern collapse. Metal-organic photoresists are a promising class of materials that can address many of these challenges, and among them, resists deposited via hybrid molecular layer deposition (MLD) have attracted interest for their unique advantages in thickness control, chemical homogeneity, and compatibility with vacuum processing. However, despite many successful demonstrations of patterning, little is known about how the molecular design of hybrid MLD resists affects their lithographic performance. In this work, we study the effect of the network structure, a common feature among all hybrid MLD resists, via a series of aluminum alkoxide ("alucone") negative tone resists with varying networking density. Their patterning mechanism is investigated via electron beam lithography (EBL)─a common proxy for EUV─and compared to their EUV-induced reactions studied via flood exposure and in situ characterization. In conclusion, we show that the resist with the least networking density demonstrates the best sensitivity and resolution, with the ability to resolve dense line/space gratings as small as 14 nm half pitch via EBL.

electron beam lithography↗

DOE OSTI2024

Smart Pixels: In-pixel AI for on-sensor data filtering

We present a smart pixel prototype readout integrated circuit (ROIC) designed in CMOS 28 nm bulk process, with in-pixel implementation of an artificial intelligence (AI) / machine learning (ML) based data filtering algorithm designed as proof-of-principle for a Phase III upgrade at the Large Hadron Collider (LHC) pixel detector. The first version of the ROIC consists of two matrices of 256 smart pixels, each 25$\times$25 µm\textsuperscript{2} in size. Each pixel consists of a charge-sensitive preamplifier with leakage current compensation and three auto-zero comparators for a 2-bit flash-type ADC. The frontend is capable of synchronously digitizing the sensor charge within 25 ns. Measurement results show an equivalent noise charge (ENC) of $\sim$30e\textsuperscript{-} and a total dispersion of $\sim$100e\textsuperscript{-} The second version of the ROIC uses a fully connected two-layer neural network (NN) to process information from a cluster of 256 pixels to determine if the pattern corresponds to highly desirable high-momentum particle tracks for selection and readout. The digital NN is embedded in-between analog signal processing regions of the 256 pixels without increasing the pixel size and is implemented as fully combinatorial digital logic to minimize power consumption and eliminate clock distribution, and is active only in the presence of an input signal. The total power consumption of the neural network is $\sim$ 300 $\mu$W. The NN performs momentum classification based on the generated cluster patterns and even with a modest momentum threshold, it is capable of 54.4\% – 75.4\% total data rejection, opening the possibility of using the pixel information at 40MHz for the trigger. The total power consumption of analog and digital functions per pixel is $\sim$ 6 $\mu$W per pixel, which corresponds to $\sim$ 1 W/cm\textsuperscript{2} staying within the experimental constraints.

Parpillon, Benjamin↗

DOE OSTI2024

Spectroscopic performance of Low-Gain Avalanche Diodes for different types of radiation

LGADs (Low-Gain Avalanche Diodes or Detectors) are a type of silicon Avalanche Photo-Diodes originally developed for the fast detection of minimum ionizing particles in high-energy physics experiments. Thanks to their fast timing performance, the LGAD paradigm enables detectors to accurately measure minimum ionizing particles with a timing resolution of a few tens of picoseconds. Such a performance is due to a thin substrate and the presence of a moderate signal gain. This internal gain of a few tens is enough to compensate for the reduced charge deposition in the thinner substrate and the noise of fast read-out systems. While LGADs are optimized for the detection of minimum ionizing particles for high-energy particle detectors, it is critical to study their performance for the detection of different types of particle, such as X-rays, gamma-rays, or alphas. In this paper, we evaluate the gain of three types of LGADs: two devices with different geometries and doping profiles fabricated by Brookhaven National Laboratory, and one fabricated by Hamamatsu Photonics with a different process. Since the gain in LGADs depends on the bias voltage applied to the sensor, pulse-height spectra have been acquired for bias voltages spanning from the depletion voltage up to the breakdown voltage. Finally, the signal-to-noise ratio of the generated signals and the shape of their spectra allow us to probe the underlying physics of the multiplication process.

47 OTHER INSTRUMENTATION↗

DOE OSTI2024

DeepPhenoMem V1.0: deep learning modelling of canopy greenness dynamics accounting for multi-variate meteorological memory effects on vegetation phenology

Abstract. Vegetation phenology plays a key role in controlling the seasonality of ecosystem processes that modulate carbon, water and energy fluxes between the biosphere and atmosphere. Accurate modelling of vegetation phenology in the interplay of Earth's surface and the atmosphere is thus crucial to understand how the coupled system will respond to and shape climatic changes. Phenology is controlled by meteorological conditions at different timescales: on the one hand, changes in key meteorological variables (temperature, water, radiation) can have immediate effects on the vegetation development; on the other hand, phenological changes can be driven by past environmental conditions, known as memory effects. However, the processes governing meteorological memory effects on phenology are not completely understood, resulting in their limited performance of vegetation phenology represented in land surface models. A deep learning model, specifically a long short-term memory network (LSTM), has the potential to capture and model the meteorological memory effects on vegetation phenology. Here, we apply the LSTM to model the vegetation phenology using meteorological drivers and high-temporal-resolution canopy greenness observations through digital repeat photography by the PhenoCam network. We compare a multiple linear regression model, a no-memory-effect LSTM model and a full-memory-effect LSTM model to predict the whole seasonal greenness trajectory and the corresponding phenological transition dates across 50 sites and 317 site years during 2009–2018, covering deciduous broadleaf forests, evergreen needleleaf forests and grasslands. Results show that the deep learning model outperforms the multiple linear regression model, and the full-memory-effect LSTM model performs better than the no-memory-effect model for all three plant function types (median R2 of 0.878, 0.957 and 0.955 for broadleaf forests, evergreen needleleaf forests and grasslands). We also find that the full-memory-effect LSTM model is capable of predicting the seasonal dynamic variations of canopy greenness and reproducing trends in shifting phenological transition dates. We also performed a sensitivity analysis of the full-memory-effect LSTM model to assess its plausibility, revealing its coherence with established knowledge of vegetation phenology sensitivity to meteorological conditions, particularly changes in temperature. Our study highlights that (1) multi-variate meteorological memory effects play a crucial role in vegetation phenology, and (2) deep learning opens up new avenues for improving the representation of vegetation phenological processes in land surface models via a hybrid modelling approach.

Geology↗