Test Model Predictions Research Articles

An explainable deep learning platform for molecular discovery.

Deep learning approaches have been increasingly applied to the discovery of novel chemical compounds. These predictive approaches can accurately model compounds and increase true discovery rates, but they are typically black box in nature and do not generate specific chemical insights. Explainable deep learning aims to 'open up' the black box by providing generalizable and human-understandable reasoning for model predictions. These explanations can augment molecular discovery by identifying structural classes of compounds with desired activity in lieu of lone compounds. Additionally, these explanations can guide hypothesis generation and make searching large chemical spaces more efficient. Here we present an explainable deep learning platform that enables vast chemical spaces to be mined and the chemical substructures underlying predicted activity to be identified. The platform relies on Chemprop, a software package implementing graph neural networks as a deep learning model architecture. In contrast to similar approaches, graph neural networks have been shown to be state of the art for molecular property prediction. Focusing on discovering structural classes of antibiotics, this protocol provides guidelines for experimental data generation, model implementation and model explainability and evaluation. This protocol does not require coding proficiency or specialized hardware, and it can be executed in as little as 1-2 weeks, starting from data generation and ending in the testing of model predictions. The platform can be broadly applied to discover structural classes of other small molecules, including anticancer, antiviral and senolytic drugs, as well as to discover structural classes of inorganic molecules with desired physical and chemical properties.

Central bank asset purchases and lending: Impact on search frictions

We investigate the impact of two central bank policies, asset purchases and asset lending, on the search frictions in the government bond market in Japan. We build a search-theoretic model to explore the impact of a central bank’s securities lending facility (SLF) by introducing a central bank as a lender. We test model predictions using intraday data from an electronic platform for Japanese government bonds. First, we find large-scale asset purchases (LSAPs) increase order imbalance in the repurchase agreement (repo) market. Threshold analysis reveals that asset purchase amounts exceeding 0.18% of the outstanding, which corresponds to 38.98% of our sample, cause a significantly higher imbalance. Second, the SLF has a floor effect on the repo rate by affecting dealers’ choices between the repo market and the SLF. Third, the novel friction measures we test show that LSAPs and the SLF have opposite influences on bargaining power in the repo market.

Moral Hazard Incentives under Formal Insurance and Informal Insurance: Evidence from a Framed Field Experiment

Abstract This paper explores moral hazard incentives associated with formal and informal insurance. We develop a theoretical model of risky effort that incorporates formal insurance and informal risk sharing, and test model predictions through a real-effort experiment with smallholder farmers in rural Uganda. Consistent with the theory, we find evidence of moral hazard under informal insurance. We, however, do not find evidence of moral hazard under formal insurance in our experiment. Experiencing a bad outcome, however, makes the risk in production more salient to farmers, and they increase their insurance coverage. These results suggest there may be some space for expanding indemnity-based insurance to smallholder farmers through awareness programmes.

Quasar outflow deceleration or acceleration: predictions and a search

ABSTRACT Quasar winds can shock and sweep up ambient interstellar medium (ISM) gas, contributing to galactic quenching. We combine and extend past models of energy-conserving shock bubbles around quasars, investigate model implications from an observational standpoint, and test model predictions using new high-resolution spectroscopic observations of the broad absorption-line quasar SDSS J030000.56+004828.0 (J0300). Even with constant energy input from the wind, a bubble’s expansion decelerates over time as more ISM gas is swept up. Our new observations enable a direct search for this deceleration. We obtain the tightest reported 3σ limit on the average rest-frame deceleration (or acceleration) of a quasar outflow: |a| < 0.1 km s−1 yr−1 (<3 × 10−4 cm s−2) in the relatively low-velocity Ca ii outflow of J0300 over 9.65 rest-frame years. We can satisfy these limits with certain parameter choices in our model, but the large velocity range of the Ca ii absorption in J0300 rules out the hypothesis that such gas shares the velocity of the swept-up ISM gas in a self-similar shock bubble. We investigate the possibility of ram-pressure acceleration of preexisting ISM clouds and conclude that the velocity range seen in Ca ii in J0300 is potentially consistent with such an explanation. The Ca ii-absorbing gas clouds in J0300 have been inferred to have high densities by Choi et al., in which case they can only have been accelerated to their current speeds if they were originally at least an order of magnitude less dense than they are today.

Analysis of gap filling techniques for GRACE/GRACE-FO terrestrial water storage anomalies in Canada

Terrestrial water storage anomaly (TWSA) data derived from the Gravity Recovery and Climate Experiment (GRACE/GRACE-FO) have become indispensable for hydrological monitoring since the first mission launched in 2002. Evaluation of basin dynamics is enabled through TWSA integration with independent hydrological and climate datasets such as precipitation, evapotranspiration, and surface runoff. With over 20 years of observations, long-term statistical analysis reveals water storage trends, provided the 11-month gap between the two missions is filled. This study compares four methods for reconstructing GRACE-like TWSA over Canada, namely: extreme gradient boosting, artificial neural networks, automated machine learning, and projection onto convex sets. The GRACE mascon product released by the Jet Propulsion Laboratory is used over the study region. Nine sets of hydrological and climate parameters are used as predictors for the machine learning models, namely GRACE average seasonal signals, TWSA from the GLDAS Catchment Land Surface Model, precipitation, air temperature, glacier surface mass balance, ocean tides, the North Atlantic Oscillation, the Multivariate ENSO Index, and sea surface temperature. For each of the four methods, 20 % of the GRACE TWSA data (the ‘testing data’) is omitted from the model training, and the model is used to fill the artificial gap. Root mean square error (RMSE) is computed using the difference between testing data and model predictions. Normalized RMSE expresses the RMSE as a proportion of the range of the GRACE TWSA data. The automated machine learning algorithm performs best in terms of mean normalized root mean square error, with 6 % (2.4 cm equivalent water height RMSE), followed by projection onto convex sets, extreme gradient boosting, and artificial neural networks. Inclusion of glacier surface mass balance models derived from the GMAO Modern-Era Retrospective Analysis for Research and Applications Version 2 reanalysis and the Randolph Glacier Inventory improved the automated machine learning average root mean square error by 70 %. Results indicate that filling the gap between missions allows for more comprehensive analysis of terrestrial water storage changes, including basin-by-basin analysis, which is ongoing.

Perspectives on AI Architectures and Codesign for Earth System Predictability

Abstract Recently, the U.S. Department of Energy (DOE), Office of Science, Biological and Environmental Research (BER), and Advanced Scientific Computing Research (ASCR) programs organized and held the Artificial Intelligence for Earth System Predictability (AI4ESP) workshop series. From this workshop, a critical conclusion that the DOE BER and ASCR community came to is the requirement to develop a new paradigm for Earth system predictability focused on enabling artificial intelligence (AI) across the field, laboratory, modeling, and analysis activities, called model experimentation (ModEx). BER’s ModEx is an iterative approach that enables process models to generate hypotheses. The developed hypotheses inform field and laboratory efforts to collect measurement and observation data, which are subsequently used to parameterize, drive, and test model (e.g., process based) predictions. A total of 17 technical sessions were held in this AI4ESP workshop series. This paper discusses the topic of the AI Architectures and Codesign session and associated outcomes. The AI Architectures and Codesign session included two invited talks, two plenary discussion panels, and three breakout rooms that covered specific topics, including 1) DOE high-performance computing (HPC) systems, 2) cloud HPC systems, and 3) edge computing and Internet of Things (IoT). We also provide forward-looking ideas and perspectives on potential research in this codesign area that can be achieved by synergies with the other 16 session topics. These ideas include topics such as 1) reimagining codesign, 2) data acquisition to distribution, 3) heterogeneous HPC solutions for integration of AI/ML and other data analytics like uncertainty quantification with Earth system modeling and simulation, and 4) AI-enabled sensor integration into Earth system measurements and observations. Such perspectives are a distinguishing aspect of this paper. Significance Statement This study aims to provide perspectives on AI architectures and codesign approaches for Earth system predictability. Such visionary perspectives are essential because AI-enabled model-data integration has shown promise in improving predictions associated with climate change, perturbations, and extreme events. Our forward-looking ideas guide what is next in codesign to enhance Earth system models, observations, and theory using state-of-the-art and futuristic computational infrastructure.

Social media feedback and extreme opinion expression

On popular social media platforms such as Twitter, Facebook, Instagram, or Tiktok, the quantitative feedback received by content producers is asymmetric: counts of positive reactions such as ‘likes,’ or ‘retweets,’ are easily observed but similar counts of negative reactions are not directly available. We study how this design feature of social media platforms affects the expression of extreme opinions. Using simulations of a learning model, we compare two feedback environments that differ in terms of the availability of negative reaction counts. We find that expressed opinions are generally more extreme when negative reaction counts are not available than when they are. We rely on analyses of Twitter data and several online experiments to provide empirical support for key model assumptions and test model predictions. Our findings suggest that a simple design change might limit, under certain conditions, the expression of extreme opinions on social media.

American Society of Biomechanics Journal of Biomechanics Award 2022: Computer models do not accurately predict human muscle passive muscle force and fiber length: Evaluating subject-specific modeling impact on musculoskeletal model predictions

Musculoskeletal models are valuable for studying and understanding the human body in a variety of clinical applications that include surgical planning, injury prevention, and prosthetic design. Subject-specific models have proven to be more accurate and useful compared to generic models. Nevertheless, it is important to validate all models when possible. To this end, gracilis muscle–tendon parameters were directly measured intraoperatively and used to test model predictions. The aim of this study was to evaluate the benefits and limitations of systematically incorporating subject-specific variables into muscle models used to predict passive force and fiber length. The results showed that incorporating subject-specific values generally reduced errors, although significant errors still existed. Optimization of the modeling parameter “tendon slack length” was explored in two cases: minimizing fiber length error and minimizing passive force error. The results showed that using all subject-specific values yielded the most favorable outcome in both models and optimization cases. However, the trade-off between fiber length error and passive force error will depend on the specific circumstances and research objectives due to significant individual errors. Notably, individual fiber length and passive force errors were as high as 20% and 37% respectively. Finally, the modeling parameter “tendon slack length” did not correlate with any real-world anatomical length.

Hydraulic segmentation explains differences in loss of branch conductance caused by fire.

The hydraulic death hypothesis suggests that fires kill trees by damaging the plant's hydraulic continuum in addition to stem cambium. A corollary to this hypothesis is that plants that survive fires possess 'pyrohydraulic' traits that prevent heat-induced embolism formation in the xylem and aid post-fire survival. We examine whether hydraulic segmentation within stem xylem may act as such a trait. To do so, we measured the percentage loss of conductance (PLC) and vulnerability to embolism axially along segments of branches exposed to heat plumes in two differing species, fire-tolerant Eucalyptus cladocalyx F. Muell and fire-sensitive Kiggelaria africana L., testing model predictions that fire-tolerant species would exhibit higher degrees of hydraulic segmentation (greater PLC in the distal parts of the branch than the basal) than fire-intolerant species (similar PLC between segments). Following exposure to a heat plume, K. africana suffered between 73 and 84% loss of conductance in all branch segments, whereas E. cladocalyx had 73% loss of conductance in whole branches, including the distal tips, falling to 29% in the most basal part of the branch. There was no evidence for differences in resistance segmentation between the species, and there was limited evidence for differences in distal vulnerability to embolism across the branches. Hydraulic segmentation in E. cladocalyx may enable it to resprout effectively post-fire with a functional hydraulic system. The lack of hydraulic segmentation in K. africana reveals the need to understand possible trade-offs associated with hydraulic segmentation in long-lived woody species with respect to drought and fire.

Global leaf‐trait mapping based on optimality theory

AbstractAimLeaf traits are central to plant function, and key variables in ecosystem models. However recently published global trait maps, made by applying statistical or machine‐learning techniques to large compilations of trait and environmental data, differ substantially from one another. This paper aims to demonstrate the potential of an alternative approach, based on eco‐evolutionary optimality theory, to yield predictions of spatio‐temporal patterns in leaf traits that can be independently evaluated.InnovationGlobal patterns of community‐mean specific leaf area (SLA) and photosynthetic capacity (Vcmax) are predicted from climate via existing optimality models. Then leaf nitrogen per unit area (Narea) and mass (Nmass) are inferred using their (previously derived) empirical relationships to SLA and Vcmax. Trait data are thus reserved for testing model predictions across sites. Temporal trends can also be predicted, as consequences of environmental change, and compared to those inferred from leaf‐level measurements and/or remote‐sensing methods, which are an increasingly important source of information on spatio‐temporal variation in plant traits.Main conclusionsModel predictions evaluated against site‐mean trait data from > 2,000 sites in the Plant Trait database yielded R2 = 73% for SLA, 38% for Nmass and 28% for Narea. Declining species‐level Nmass, and increasing community‐level SLA, have both been recently reported and were both correctly predicted. Leaf‐trait mapping via optimality theory holds promise for macroecological applications, including an improved understanding of community leaf‐trait responses to environmental change.

Testing Model Predictions of Depth of Air-Shower Maximum and Signals in Surface Detectors using Hybrid Data of the Pierre Auger Observatory

We present a method for testing the predictions of hadronic interaction models and improving their consistency with observed two-dimensional distributions of the depth of shower maximum, X max, and signal at the ground level as a function of zenith angle. The method relies on the assumption that the mass composition is the same at all zenith angles, while the atmospheric shower development and attenuation depend on composition in a correlated way. In the present work, for each of the three leading LHC-tuned hadronic interaction models, we allow a global shift ∆ X max of the predicted shower maximum, which is the same for every mass and energy, and a rescaling R Had of the hadronic component at the ground level which is constant with the zenith angle. We apply the analysis to 2297 events reconstructed with both the fluorescence and surface detectors of the Pierre Auger Observatory with energies 1018.5−19.0 eV and zenith angles below 60°. Given the modeling assumptions made in this analysis, the best fit reaches its optimum value when shifting the X max predictions of hadronic interaction models to deeper values and increasing the hadronic signal. This change in the predicted X max scale alleviates the previously identified model deficit in the hadronic signal (commonly called the muon puzzle) but does not fully remove it. Because of the size of the adjustments ∆ X max and R Had and the large number of events in the sample, the statistical significance of need for these adjustments is large, greater than 5σstat, even for the combination of the systematic experimental shifts within 1σsys that are the most favorable for the models.

Temperature-dependent compressive stress-strain behaviors of alkali-activated slag-based ultra-high strength concrete

In this work, the uniaxial compressive behaviors of alkali-activated slag-based ultra-high strength concrete (AAS-UHSC) subjected to different temperatures were investigated, with an emphasis on the temperature-dependent stress-strain relation. To this end, 104 cylinder specimens with three steel fiber contents (0.5%, 1.0%, and 1.5%) and two water-to-binder (W/B) ratios (0.20 and 0.25) were tested upon the monotonic and cyclic compression performed at ambient temperature (20 ℃) and four elevated temperatures (200, 400, 600, and 800 ℃). The test results showed that the mechanical properties of AAS-UHSC in terms of elastic modulus, ultimate strength, energy dissipation, and post-peak softening can be effectively improved with the fiber inclusion, regardless of the W/B ratio. Moreover, AAS-UHSC exhibited a quite different plastic strain evolution law featured as two discontinuous stages due to the damage localization and possessed better energy dissipation capacity compared to OPC-based concretes. After being exposed to elevated temperatures, a consistent decline in compressive strength was observed as the temperature increased, with residual strength at 800 ℃ remaining around 10% of the ultimate compressive strength at 20 ℃. Based on the test results, a unified empirical temperature-dependent model was proposed to capture the main features of the constitutive compressive stress-strain relations of AAS-UHSC. The fairly good comparisons between test results and model predictions in different loading scenarios demonstrated a compromise between the accuracy and the generality of the model and contributed a beneficial addition to the understanding of nonlinear responses of AAS-UHSC.

Probabilistic safety assessment of the burst strength of corroded pipelines of different steel grades with calibrated strength models

This paper proposes an approach to assess the structural reliability of corroded pipes with calibrated strength models that reduce the subjectivity in model selection and incorporates their uncertainty. The approach calibrates and assesses the structural safety of pipes based on 25 burst strength models for corroded pipelines that are categorized into three classes of steel grades, i.e., low (X42 or less to X56), medium (X60 to X70) and high strength (X80 to X120). Statistical parameters such as the absolute mean error and standard deviation are adopted to analyse and compare the strength models’ performance against test results. Then, model uncertainties factors are derived and probabilistically characterized for each strength model. The importance of this methodology is demonstrated using a case study of one corroded pipe in each category of steel grade. The top 10 models from each steel pipe category are used for structural reliability evaluation. The First Order Reliability method is then employed for reliability and sensitivity assessments. Finally, partial safety factors are estimated by the best model in each steel category and compared with available codes. The proposed approach is believed to improve upon the existing methods by adopting the most accurate strength model and accounting for model uncertainties for different steel grades derived from test data and model predictions.

Vertically extended and asymmetric CN emission in the Elias 2-27 protoplanetary disk

Context. Cyanide (CN) emission is expected to originate in the upper layers of protoplanetary disks, tracing UV-irradiated regions. This hypothesis, however, has been observationally tested only in a handful of disks. Elias 2-27 is a young star that hosts an extended, bright, and inclined disk of dust and gas. The inclination and extreme flaring of the disk make Elias 2-27 an ideal target to study the vertical distribution of molecules, particularly CN. Aims. Our aim is to directly trace the emission of CN in the disk around Elias 2-27 and compare it to previously published CO isotopolog data of the system. The two tracers can be combined and used to constrain the physical and chemical properties of the disk. Through this analysis we can test model predictions of CN emission and compare observations of CN in other objects to the massive, highly flared, asymmetric, and likely gravitationally unstable protoplanetary disk around Elias 2-27. Methods. We analyzed CN N = 3–2 emission in two different transitions J = 7/2–5/2 and J = 5/2–3/2, for which we detect two hyperfine group transitions. The vertical location of CN emission was traced directly from the channel maps, following geometrical methods that had been previously used to analyze the CO emission of Elias 2-27. Simple analytical models were used to parameterize the vertical profile of each molecule and study the extent of each tracer. From the radial intensity profiles we computed radial profiles of column density and optical depth. Results. We show that the vertical location of CN and CO isotopologs in Elias 2-27 is layered and consistent with predictions from thermochemical models. A north-south asymmetry in the radial extent of the CN emission is detected, which is likely due to shadowing on the north side of the disk. Combining the information from the peak brightness temperature and vertical structure radial profiles, we find that the CN emission is mostly optically thin and constrained vertically to a thin slab at z/r ~ 0.5. A column density of 1014 cm−2 is measured in the inner disk, which for the north side decreases to 1012 cm−2 and for the south side to 1013 cm−2 in the outer regions. Conclusions. In Elias 2-27, CN traces a vertically elevated region above the midplane, very similar to that traced by 12CO. The inferred CN column densities, low optical depth (τ ≤ 1), and location near the disk surface are consistent with thermo-chemical disk models in which CN formation is initiated by the reaction of N with UV-pumped H2. The observed north–south asymmetry may be caused by either ongoing infall or by a warped inner disk. This study highlights the importance of tracing the vertical location of various molecules to constrain the disk physical conditions.

Too hot to hunt: Mechanistic predictions of thermal refuge from cat predation risk

AbstractMany threatened species depend on climatic microrefugia, but places with harsh climates for predators may also play a refugial role. Feral cats threaten many native species in arid Australia. Although cats can persist in regions with no free water, their abundance should depend on the availability of microclimates that protect them from harsh environmental conditions. We developed a biophysical model of feral cat heat stress and used it to explore how behavior and microhabitat features influence water requirements and activity. Tests of model predictions against fine‐scale GPS and microclimate data highlight the importance of refuges, particularly rabbit burrows. Continent‐wide simulations show large but temporally varying areas of the arid zone that would be lethal for cats without access to deep or shaded burrows. Our approach can identify locations that may act as natural refuges for native prey, and where habitat management strategies may be effective in controlling cat abundance.

Learning to simulate high energy particle collisions from unlabeled data

In many scientific fields which rely on statistical inference, simulations are often used to map from theoretical models to experimental data, allowing scientists to test model predictions against experimental results. Experimental data is often reconstructed from indirect measurements causing the aggregate transformation from theoretical models to experimental data to be poorly-described analytically. Instead, numerical simulations are used at great computational cost. We introduce Optimal-Transport-based Unfolding and Simulation (OTUS), a fast simulator based on unsupervised machine-learning that is capable of predicting experimental data from theoretical models. Without the aid of current simulation information, OTUS trains a probabilistic autoencoder to transform directly between theoretical models and experimental data. Identifying the probabilistic autoencoder’s latent space with the space of theoretical models causes the decoder network to become a fast, predictive simulator with the potential to replace current, computationally-costly simulators. Here, we provide proof-of-principle results on two particle physics examples, Z-boson and top-quark decays, but stress that OTUS can be widely applied to other fields.

A New Approach to Bicarbonate Addition During Hemodialysis: Testing Model Predictions in a Patient Cohort

In this study, we present a new protocol for kidney replacement therapy (hemodialysis), based on an explicitly solvable mathematical model. With current protocols, the high and constant level of bath bicarbonate (HCO₃^&#x2212;) used to prevent metabolic acidosis leads to very rapid delivery of HCO₃^&#x2212; into the patient during the first part of the therapy. This rapid alkalinization elicits a robust buffer response that, paradoxically, consumes more HCO₃^&#x2212; than is added during the remainder of the treatment. In previous studies, we developed an analytical model that allows one to quantify these events and tested alternative protocols manipulating the rate of rise in blood bicarbonate concentration (HCO₃^&#x2212;). The protocol tested in this paper enforces a more gradual increase in blood HCO₃^&#x2212;, by means of a model-based staircase adjustment of bath HCO₃^&#x2212;. Model equations predict a reduction of buffer response and rate of organic acid production. These predictions are tested in 20 stable outpatients receiving hemodialysis. We find that the proposed protocol achieves the desired profile of blood HCO₃^&#x2212; with good accuracy and reduces the total buffer response by 1/3 and the rate of lactic acid production by at least 1/4, as compared to conventional therapy. Although more studies are needed, we believe that our work will pave the way for a more rational approach to correction of acidosis during hemodialysis. Article Highlights <list list-type="bullet"><list-item><label>&#x2022;</label> Our study tests an analytic model designed to enforce a more gradual rate of bicarbonate delivery during hemodialysis. </list-item><list-item><label>&#x2022;</label> Using our model, we show that we can reduce the excessive buffer response and lactic acid production that occur with the conventional approach. </list-item><list-item><label>&#x2022;</label> We demonstrate that our model can provide a rational approach to bicarbonate addition during treatment. </list-item> </list>

In Search of the Preference Reversal Zone.

A preference reversal is observed when a preference for a larger-later (LL) reward over a smaller-sooner (SS) reward reverses as both rewards come closer in time. Preference reversals are common in everyday life and in the laboratory and are often claimed to support hyperbolic delay-discounting models which, in their simplest form, can model reversals with only one free parameter. However, it is not clear if the temporal location of preference reversals can be predicted a priori. Studies testing model predictions have not found support for them, but they overlooked the well-documented effect of reinforcer magnitude on discounting rate. Therefore, we directly tested hyperbolic and exponential model predictions in a pre-registered study by assessing individual discount rates for two reinforcer magnitudes. We then made individualized predictions about pairs of choices between which preference reversals should occur. With 107 participants, we found (1) little evidence that hyperbolic and exponential models could predict the temporal location of preference reversals, (2) some evidence that hyperbolic models had better predictive performance than exponential models, and (3) in contrast to many previous studies, that exponential models generally produced superior fits to the observed data than hyperbolic models.

Catch, Restrict, and Release: The Real Story of Bank Bailouts

Bank bailouts are not the one-shot events commonly described in the literature. These bailouts are instead dynamic processes in which regulators catch financially distressed banks; restrict their activities over time; and release the banks from restrictions at sufficiently healthy capital ratios. The catch-restrict-release approach is a global phenomenon, which we document using hand-collected data on capital injection and debt guarantee bailouts in the European Union (EU) over 2008-2014. We present a dynamic theoretical model of socially-optimizing regulators engaging in catch-restrict-release capital injection and debt guarantee bailouts, and empirically test model predictions. Observed EU bailouts are qualitatively consistent with optimizing behavior.

Seasonal dynamics of the macrophyte test species Myriophyllum spicatum over two years in experimental ditches for population modeling application in risk assessment.

Myriophyllum spicatum is a sediment‐rooted, aquatic macrophyte growing submerged, with a wide geographical distribution and high ecological relevance in freshwater ecosystems. It is used in testing and risk assessment for pesticides in water and sediment. Population models enable effects measured under laboratory conditions to be extrapolated to effects expected in the field with time‐variable environmental factors including exposure. These models are a promising tool in higher‐tier risk assessments. However, there is a lack of data on the seasonal dynamics of M. spicatum, which is needed to test model predictions of typical population dynamics in the field. To generate such data, a two‐year study was set up in outdoor experimental systems from May 2017 to May 2019. The growth of M. spicatum was monitored in 0.2025 m2 plant baskets installed in an experimental ditch. Parameters monitored included biomass (fresh weight [FW] and dry weight [DW]), shoot length, seasonal short‐term growth rates of shoots, relevant environmental parameters, and weather data. The results showed a clear seasonal pattern of biomass and shoot length and their variability. M. spicatum reached a maximum total shoot length (TSL) of 279 m m−2 and a maximum standing crop above‐ground DW of 262 g m−2. Periodical growth rates reached up to 0.072, 0.095, and 0.085 day−1 for total length, FW, and DW, respectively. Multivariate regression revealed that pH (as a surrogate for the availability of carbon species) and water temperature could explain a significant proportion of the variability in M. spicatum growth rates (p < 0.05). This study has provided an ecologically relevant data set on seasonal population dynamics representative of shallow freshwater ecosystems, which can be used to test and refine population models for use in chemical risk assessment and ecosystem management. Integr Environ Assess Manag 2022;18:1375–1386. © 2021 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).