Emulation Model Research Articles

Three-Dimensional Variational Multi-Doppler Wind Retrieval over Complex Terrain

Abstract The interaction of airflow with complex terrain has the potential to significantly amplify extreme precipitation events and modify the structure and intensity of precipitating cloud systems. However, understanding and forecasting such events is challenging, in part due to the scarcity of direct in situ measurements. Doppler radar can provide the capability to monitor extreme rainfall events over land, but our understanding of airflow modulated by orographic interactions remains limited. The SAMURAI software is a three-dimensional variational data assimilation (3DVAR) technique that uses the finite element approach to retrieve kinematic and thermodynamic fields. The analysis has high fidelity to observations when retrieving flows over a flat surface, but the capability of imposing topography as a boundary constraint is not previously implemented. Here, we implement the immersed boundary method (IBM) as pseudo-observations at their native coordinates in SAMURAI to represent the topographic forcing and surface impermeability. In this technique, neither data interpolation onto a Cartesian grid nor explicit physical constraint integration during the cost function minimization is needed. Furthermore, the physical constraints are treated as pseudo-observations, offering the flexibility to adjust the strength of the boundary condition. A series of observing simulation sensitivity experiments (OSSEs) using a full-physics model and radar emulator simulating rainfall from Typhoon Chanthu (2021) over Taiwan are conducted to evaluate the retrieval accuracy and parameter settings. The OSSE results show that the strength of the IBM constraints can impact the overall wind retrievals. Analysis from real radar observations further demonstrates that the improved retrieval technique can advance scientific analyses for the underlying dynamics of orographic precipitation using radar observations.

Experimental Implementation of Retrospective Cost Adaptive Control for Suppressing Thermoacoustic Oscillations in a Rijke Tube

Thermoacoustic systems are self-oscillating due to the fact that a constant input (e.g., fuel rate in gas-turbine combustors) yields an asymptotically oscillatory response. This behavior arises from the interaction between combustion and acoustics, resulting in thermoacoustic oscillations. Under varying operating conditions, the dynamics of thermoacoustic systems may change dramatically, which may require that active suppression controllers can be tuned for specific working conditions. This work provides an experimental investigation of retrospective cost adaptive control for suppressing thermoacoustic oscillations under sampled-data control and varying system operating conditions. This approach is first applied to a Rijke-tube emulation model for hyperparameter selection and subsequently to an experimental Rijke-tube setup. Physical experiments are conducted to investigate the performance and robustness of the adaptive controller under varying operating conditions.

Bayesian weighting of climate models based on climate sensitivity

Using climate model ensembles containing members that exhibit very high climate sensitivities to increasing CO2 concentrations can result in biased projections. Various methods have been proposed to ameliorate this ‘hot model’ problem, such as model emulators or model culling. Here, we utilize Bayesian Model Averaging as a framework to address this problem without resorting to outright rejection of models from the ensemble. Taking advantage of multiple lines of evidence used to construct the best estimate of the earth’s climate sensitivity, the Bayesian Model Averaging framework produces an unbiased posterior probability distribution of model weights. The updated multi-model ensemble projects end-of-century global mean surface temperature increases of 2 oC for a low emissions scenario (SSP1-2.6) and 5 oC for a high emissions scenario (SSP5-8.5). These estimates are lower than those produced using a simple multi-model mean for the CMIP6 ensemble. The results are also similar to results from a model culling approach, but retain some weight on low-probability models, allowing for consideration of the possibility that the true value could lie at the extremes of the assessed distribution. Our results showcase Bayesian Model Averaging as a path forward to project future climate change that is commensurate with the available scientific evidence.

Tunable floating and grounded memristor emulator Model

ABSTRACT A Differential Difference Current Conveyor Transconductance Amplifier (DDCCTA) based resistor tunable memristor emulator has been proposed in this work. The emulator can be used in both grounded and floating circumstances. The proposed design incorporates only one active block along with few passive components. Moreover, the circuit can operate in both incremental and decremental modes, by simply changing the input ports. The circuit demonstrates all the characteristics of an ideal memristor up to 6 MHz. The proposed model has been simulated using TSMC 0.18m process parameter and occupies an area of 51 × 42.5 m2 chip-area, excluding capacitor. The circuit’s reliability has been verified by studying non-ideal, non-volatile, Monte-Carlo, process corner variations analysis. The circuit applicability has been tested through series/parallel combinations. To validate the experimental demonstration, AD844AN and CA3080 have been used to make a prototype, which shows good agreement with theoretical and simulation results.

Neural network-based emulation of interstellar medium models

Context. The interpretation of observations of atomic and molecular tracers in the galactic and extragalactic interstellar medium (ISM) requires comparisons with state-of-the-art astrophysical models to infer some physical conditions. Usually, ISM models are too timeconsuming for such inference procedures, as they call for numerous model evaluations. As a result, they are often replaced by an interpolation of a grid of precomputed models. Aims. We propose a new general method to derive faster, lighter, and more accurate approximations of the model from a grid of precomputed models for use in inference procedures. Methods. These emulators are defined with artificial neural networks (ANNs) with adapted architectures and are fitted using regression strategies instead of interpolation methods. The specificities inherent in ISM models need to be addressed to design and train adequate ANNs. Indeed, such models often predict numerous observables (e.g., line intensities) from just a few input physical parameters and can yield outliers due to numerical instabilities or physical bistabilities and multistabilities. We propose applying five strategies to address these characteristics: (1) an outlier removal procedure; (2) a clustering method that yields homogeneous subsets of lines that are simpler to predict with different ANNs; (3) a dimension reduction technique that enables us to adequately size the network architecture; (4) the physical inputs are augmented with a polynomial transform to ease the learning of nonlinearities; and (5) a dense architecture to ease the learning of simpler relations between line intensities and physical parameters. Results. We compare the proposed ANNs with four standard classes of interpolation methods, nearest-neighbor, linear, spline, and radial basis function (RBF), to emulate a representative ISM numerical model known as the Meudon PDR code. Combinations of the proposed strategies produce networks that outperform all interpolation methods in terms of accuracy by a factor of 2 in terms of the average error (reaching 4.5% on the Meudon PDR code) and a factor of 3 for the worst-case errors (33%). These networks are also 1000 times faster than accurate interpolation methods and require ten to forty times less memory. Conclusions. This work will enable efficient inferences on wide-field multiline observations of the ISM.

Exploring Randomly Wired Neural Networks for Climate Model Emulation

Abstract Exploring the climate impacts of various anthropogenic emissions scenarios is key to making informed decisions for climate change mitigation and adaptation. State-of-the-art Earth system models can provide detailed insight into these impacts but have a large associated computational cost on a per-scenario basis. This large computational burden has driven recent interest in developing cheap machine learning models for the task of climate model emulation. In this paper, we explore the efficacy of randomly wired neural networks for this task. We describe how they can be constructed and compare them with their standard feedforward counterparts using the ClimateBench dataset. Specifically, we replace the serially connected dense layers in multilayer perceptrons, convolutional neural networks, and convolutional long short-term memory networks with randomly wired dense layers and assess the impact on model performance for models with 1 million and 10 million parameters. We find that models with less-complex architectures see the greatest performance improvement with the addition of random wiring (up to 30.4% for multilayer perceptrons). Furthermore, of 24 different model architecture, parameter count, and prediction task combinations, only one had a statistically significant performance deficit in randomly wired networks relative to their standard counterparts, with 14 cases showing statistically significant improvement. We also find no significant difference in prediction speed between networks with standard feedforward dense layers and those with randomly wired layers. These findings indicate that randomly wired neural networks may be suitable direct replacements for traditional dense layers in many standard models. Significance Statement Modeling various greenhouse gas and aerosol emissions scenarios is important for both understanding climate change and making informed political and economic decisions. However, accomplishing this with large Earth system models is a complex and computationally expensive task. As such, data-driven machine learning models have risen in prevalence as cheap emulators of Earth system models. In this work, we explore a special type of machine learning model called randomly wired neural networks and find that they perform competitively for the task of climate model emulation. This indicates that future machine learning models for emulation may significantly benefit from using randomly wired neural networks as opposed to their more-standard counterparts.

Ice-flow model emulator based on physics-informed deep learning

Abstract Convolutional neural networks (CNN) trained from high-order ice-flow model realisations have proven to be outstanding emulators in terms of fidelity and computational performance. However, the dependence on an ensemble of realisations of an instructor model renders this strategy difficult to generalise to a variety of ice-flow regimes found in the nature. To overcome this issue, we adopt the approach of physics-informed deep learning, which fuses traditional numerical solutions by finite differences/elements and deep-learning approaches. Here, we train a CNN to minimise the energy associated with high-order ice-flow equations within the time iterations of a glacier evolution model. As a result, our emulator is a promising alternative to traditional solvers thanks to its high computational efficiency (especially on GPU), its high fidelity to the original model, its simplified training (without requiring any data), its capability to handle a variety of ice-flow regimes and memorise previous solutions, and its relatively simple implementation. Embedded into the ‘Instructed Glacier Model’ (IGM) framework, the potential of the emulator is illustrated with three applications including a large-scale high-resolution (2400x4000) forward glacier evolution model, an inverse modelling case for data assimilation, and an ice shelf.

Unweighting multijet event generation using factorisation-aware neural networks

In this article we combine a recently proposed method for factorisation-aware matrix element surrogates with an unbiased unweighting algorithm. We show that employing a sophisticated neural network emulation of QCD multijet matrix elements based on dipole factorisation can lead to a drastic acceleration of unweighted event generation. We train neural networks for a selection of partonic channels contributing at the tree-level to Z+4,5Z+4,5 jets and t\bar{t}+3,4tt‾+3,4 jets production at the LHC which necessitates a generalisation of the dipole emulation model to include initial state partons as well as massive final state quarks. We also present first steps towards the emulation of colour-sampled amplitudes. We incorporate these emulations as fast and accurate surrogates in a two-stage rejection sampling algorithm within the SHERPA Monte Carlo that yields unbiased unweighted events suitable for phenomenological analyses and post-processing in experimental workflows, e.g. as input to a time-consuming detector simulation. For the computational cost of unweighted events we achieve a reduction by factors between 16 and 350 for the considered channels.

Democracy as a trust-building learning process: Organizational dilemmas in social movements

Combining agonistic pluralism and social movements literature with trust studies, I propose a conceptualization for how the organizational dilemma is tackled in social movements. Defined as a trust-building organizational learning process, I show the role-played by social trust—meaning, the construction of the relational boundaries of a shared goal without diluting the heterogeneity of self-identities and interests—as an organizational prerequisite for democratic organization of a political group. Empirically, I identify four alternative pathways to the (democratic) organizational dilemma: innovation through new organizational models; repetition of past experiences; reformulation of practices; and emulation of previous organizational models.

A Mythic France for Export: Auguste Genin, Cultural Brokers, and the Third Republic's Mission Civilisatrice in Porfirian Mexico

Abstract: Historians of Mexico after its independence from Spain in 1821, especially during the era of rapid modernization under President Porfirio Díaz (1876–1911), speak generally of the strong French cultural influence on the middling and privileged classes in the Republic who viewed France as a model of emulation for national development. This afrancesamiento (Francophilia) is often taken for granted and few have questioned what "Frenchness" meant at a time when France itself was constructing a national identity, who were its greatest promoters in Mexico, and how did the Mexican experience compare with the Third Republic's (1870–1940) civilizing mission ( mission civilizatrice ) in France's formal colonies in Southeast Asia and Africa. The goal of this article is two-fold. The first is to advance the notion of certain members of the French immigrant community in Porfirian Mexico as cultural brokers, and the second is to place the French model of modernity to which these foreign residents and Mexican elites referred to in its domestic and larger imperial context. The influential Franco-Mexican Auguste Genin and his writings offer an illustrative case study through which to introduce and assess this. Particularly revealing is how Genin imagines the process by which the indigenous population of Mexico would be transformed into modern citizens and how this project—this imaginary construct—found a source of inspiration in the simultaneous French project to construct a French identity and a model ideology for overseas empire during the Third Republic.

Investigating factors in indoor transmission of respiratory disease through agent‐based modeling

AbstractThe transmission of respiratory diseases such as COVID‐19 is exacerbated in densely populated urban areas and crowded indoor settings. Despite the majority of transmissions occurring in such settings, controlling viral spread through individual‐level contacts indoors remains challenging. Experimental studies have investigated the transmission patterns of respiratory behaviors such as coughing or sneezing in controlled spatial environments. However, the effects of dynamic movement and spatial structures have been ignored, making it difficult to apply findings to urban policy and planning. To address this gap, we developed agent‐based simulations to investigate individual virus inhalation patterns across multiple scenarios in a symmetrical and formulaic indoor space. We conducted sensitivity analysis using regression emulator models to identify significant factors for viral transmission. Our results indicate positive associations with viral transmission in descending order of: (1) stay time; (2) encounter frequency; and (3) initial infected population; while negative associations are: (4) mask wearing; (5) distance to infected people; (6) nearest infected people's mask wearing; and (7) distance to entrance. We also found that narrow passages between obstacles increase virus transmission from breathing. Furthermore, we conducted a case study to investigate the potential of reducing the amount of individually inhaled virus by controlling behaviors and spatial environments. Our findings suggest that mask wearing and reduced stay time can substantially reduce transmission risk, while a large number of contacts and high grouping time result in the growth of the infected population at a certain threshold. These results provide guidance for decision makers to formulate guidelines for curbing the spread of respiratory diseases in indoor spaces.

Understanding pattern scaling errors across a range of emissions pathways

Abstract. The regional climate impacts of hypothetical future emissions scenarios can be estimated by combining Earth system model simulations with a linear pattern scaling model such as MESMER (Modular Earth System Model Emulator with spatially Resolved output), which uses estimated patterns of the local response per degree of global temperature change. Here we use the mean trend component of MESMER to emulate the regional pattern of the surface temperature response based on historical single-forcer and future Shared Socioeconomic Pathway (SSP) CMIP6 (Coupled Model Intercomparison Project Phase 6) simulations. Errors in the emulations for selected target scenarios (SSP1–1.9, SSP1–2.6, SSP2–4.5, SSP3–7.0, and SSP5–8.5) are decomposed into two components, namely (1) the differences in scaling patterns between scenarios as a consequence of varying combinations of external forcings and (2) the intrinsic time series differences between the local and global responses in the target scenario. The time series error is relatively small for high-emissions scenarios, contributing around 20 % of the total error, but is similar in magnitude to the pattern error for lower-emissions scenarios. This irreducible time series error limits the efficacy of linear pattern scaling for emulating strong mitigation pathways and reduces the dependence on the predictor pattern used. The results help guide the choice of predictor scenarios for simple climate models and where to target for the introduction of other dependent variables beyond global surface temperature into pattern scaling models.

Student empathy in standardized patient experiences: Applying concepts from a critical thinking emulation model.

Clinician empathy can improve patient outcomes, but the literature is scant on patient-based, student-led experiences to demonstrate the projection of empathy in patient interactions. (1) Develop a learning guide for observable behaviors communicating emotional and cognitive empathy and (2) determine whether the learning guide can be used as a rubric for assessing empathy in a standardized patient experience. Eleven standardized patients assessed 80 D3 students using a 4-point interval scale on 19 behavioral criteria in four domains: Initiation (four criteria); Health History and Caries Risk (four criteria); Treatment Planning (six criteria); and Communication Skills (five criteria). Standardized patients also provided qualitative feedback. Standardized patients completed all 1520 interval scales on the rubric and 94% of 320 open-ended entries. Students performed well. Of the 1520 criterion interval scales, 1242 (81.7%) were rated "excellent." Wilcoxon signed-rank tests revealed Initiation scores (Mean [M]=3.82, Standard deviation [SD]=0.28) and Treatment Planning scores (M=3.82, SD=0.36) were significantly higher than Health History Scores (M=3.75, SD=0.34; p<0.05). Qualitative feedback also was overwhelmingly positive for Treatment Planning and more equivocal for Health History. The emulation model for students to demonstrate observable aspects of empathy is viable as both a learning guide and evaluation rubric in a standardized patient format. The next steps include the development of a succinct skillset for reinforcement in the patient setting and continued discussion on what best captures core observable aspects of empathy.

TIMBER v0.1: a conceptual framework for emulating temperature responses to tree cover change

Abstract. Land cover changes have been proposed to play a significant role, alongside emission reductions, in achieving the temperature goals agreed upon under the Paris Agreement. Such changes carry both global implications, pertaining to the biogeochemical effects of land cover change and thus the global carbon budget, and regional or local implications, pertaining to the biogeophysical effects arising within the immediate area of land cover change. Biogeophysical effects of land cover change are of high relevance to national policy and decision makers, and accounting for them is essential for effective deployment of land cover practices that optimise between global and regional impacts. To this end, Earth system model (ESM) outputs that isolate the biogeophysical responses of climate to land cover changes are key in informing impact assessments and supporting scenario development exercises. However, generating multiple such ESM outputs in a manner that allows comprehensive exploration of all plausible land cover scenarios is computationally untenable. This study proposes a framework to explore in an agile manner the local biogeophysical responses of climate under customised tree cover change scenarios by means of a computationally inexpensive emulator, the Tree cover change clIMate Biophysical responses EmulatoR (TIMBER) v0.1. The emulator is novel in that it solely represents the biogeophysical responses of climate to tree cover changes, and it can be used as either a standalone device or as a supplement to existing climate model emulators that represent the climate responses from greenhouse gas (GHG) or global mean temperature (GMT) forcings. We start off by modelling local minimum, mean, and maximum surface temperature responses to tree cover changes by means of a month- and Earth system model (ESM)-specific generalised additive model (GAM) trained over the whole globe; 2 m air temperature responses are then diagnosed from the modelled minimum and maximum surface temperature responses using observationally derived relationships. Such a two-step procedure accounts for the different physical representations of surface temperature responses to tree cover changes under different ESMs whilst respecting a definition of 2 m air temperature that is more consistent across ESMs and with observational datasets. In exploring new tree cover change scenarios, we employ a parametric bootstrap sampling method to generate multiple possible temperature responses, such that the parametric uncertainty within the GAM is also quantified. The output of the final emulator is demonstrated for the Shared Socioeconomic Pathway (SSP) 1-2.6 and 3-7.0 scenarios. Relevant temperature responses are identified as those displaying a clear signal in relation to their surrounding parametric uncertainty, calculated as the signal-to-noise ratio between the sample set mean and sample set variability. The emulator framework developed in this study thus provides a first step towards bridging the information gap surrounding biogeophysical implications of land cover changes, allowing for smarter land use decision making.

A Framework for Deep Learning Emulation of Numerical Models With a Case Study in Satellite Remote Sensing.

Numerical models based on physics represent the state of the art in Earth system modeling and comprise our best tools for generating insights and predictions. Despite rapid growth in computational power, the perceived need for higher model resolutions overwhelms the latest generation computers, reducing the ability of modelers to generate simulations for understanding parameter sensitivities and characterizing variability and uncertainty. Thus, surrogate models are often developed to capture the essential attributes of the full-blown numerical models. Recent successes of machine learning methods, especially deep learning (DL), across many disciplines offer the possibility that complex nonlinear connectionist representations may be able to capture the underlying complex structures and nonlinear processes in Earth systems. A difficult test for DL-based emulation, which refers to function approximation of numerical models, is to understand whether they can be comparable to traditional forms of surrogate models in terms of computational efficiency while simultaneously reproducing model results in a credible manner. A DL emulation that passes this test may be expected to perform even better than simple models with respect to capturing complex processes and spatiotemporal dependencies. Here, we examine, with a case study in satellite-based remote sensing, the hypothesis that DL approaches can credibly represent the simulations from a surrogate model with comparable computational efficiency. Our results are encouraging in that the DL emulation reproduces the results with acceptable accuracy and often even faster performance. We discuss the broader implications of our results in light of the pace of improvements in high-performance implementations of DL and the growing desire for higher resolution simulations in the Earth sciences.

Resistorless Floating/Grounded Memristor Emulator Model With Electronic Tunability

The realization of memristor devices with specific operating parameters is a challenging task, which drives innovation in emulation circuits. This article offers a simple and novel resistorless memristor emulator design with an electronic controllability feature. The proposed model is suitable for both floating and grounded memristor applications and can operate in incremental/decremental modes effectively without switches. The proposed model comprises a voltage differencing current conveyor (VDCC) and an operational transconductance amplifier (OTA) as active components and requires a single capacitor as its passive element. This model emulates the memristors pinched hysteresis loop in current-voltage (I-V) plane. Simulation is carried out for a 180 nm CMOS technology with a supply voltage of 10.9 V. The proposed design exhibits non-linear behavior for a wide range of input voltage and operates well up to 8 MHz. The reliability and robustness of the proposed design have been verified for short-term non-volatility, temperature variation, Monte-Carlo and process corner analysis. The experimental verifications are carried out on a breadboard using ICs AD844 and CA3080.

Deep Learning Regional Climate Model Emulators: A Comparison of Two Downscaling Training Frameworks

AbstractRegional climate models (RCMs) have a high computational cost due to their higher spatial resolution compared to global climate models (GCMs). Therefore, various downscaling approaches have been developed as a surrogate for the dynamical downscaling of GCMs. This study assesses the potential of using a cost‐efficient machine learning alternative to dynamical downscaling by using the example case study of emulating surface mass balance (SMB) over the Antarctic Peninsula. More specifically, we determine the impact of the training framework by comparing two training scenarios: (a) a perfect and (b) an imperfect model framework. In the perfect model framework, the RCM‐emulator learns only the downscaling function; therefore, it was trained with upscaled RCM (UPRCM) features at GCM resolution. This emulator accurately reproduced SMB when evaluated on UPRCM, but its predictions on GCM data conserved RCM‐GCM inconsistencies and led to underestimation. In the imperfect model framework, the RCM‐emulator was trained with GCM features and downscaled the GCM while exposed to RCM‐GCM inconsistencies. This emulator predicted SMB close to the truth, showing it learned the underlying inconsistencies and dynamics. Our results suggest that a deep learning RCM‐emulator can learn the proper GCM to RCM downscaling function while working directly with GCM data. Furthermore, the RCM‐emulator presents a significant computational gain compared to an RCM simulation. We conclude that machine learning emulators can be applied to produce fast and fine‐scaled predictions of RCM simulations from GCM data.

Developing a Methodology for Model Intercomparison and Its Application to Improve Simulated Streamflow by Land Surface Models

Abstract Runoff generated by land surface models (LSMs) is extensively used to predict future river discharge under global warming. However, the structural bias of LSMs, the precipitation bias of the climate model, and other factors could cause the runoff to be biased. A model intercomparison study can help understand LSM behavior. Traditional model intercomparison can discover output variation and evaluate performance, but explaining the reason for the difference is challenging. This study developed a novel method to identify the reasons for disparities and suggest improvements. Consequently, we investigated the impacts of model settings by adopting the settings of another model in one model until it can mimic similar features in its output. Hence, we developed a process called the “emulation model.” We employed two LSMs [Simple Biosphere with Urban Canopy (SiBUC) and Meteorological Research Institute Simple Biosphere model (MRI-SiB)] in the Thai River basin. SiBUC produced a higher surface runoff than MRI-SiB, and the development of the MRI-SiB emulation revealed the cause of this variation. The differences in runoff characteristics affected streamflow estimation. For instance, the SiBUC peak discharge was faster and higher than observed in the dry year. Conversely, there was a tendency to underestimate the flow estimated by MRI-SiB runoff during the transition from dry to wet seasons. Incorporating other model settings can alleviate the shortcomings of each model. Overall, the proposed method can identify the strengths and weaknesses of a model and enhance the reproducibility of the hydrological characteristics of the observed discharge in the basin. Significance Statement This study aims to develop a new methodology for model intercomparison to identify the reasons for model output variation. Understanding why models behave differently is important to enhancing the reliability of model prediction. Our findings guide what affects disparities in land surface model runoff-based streamflow estimation, which will help reduce the uncertainty of future flood and drought predictions.

Feasibility of neural network metamodels for emulation and sensitivity analysis of radionuclide transport models

In this paper we compare the outputs of neural network metamodels with numerical solutions of differential equation models in modeling cesium-137 transportation in sand. Convolutional neural networks (CNNs) were trained with differential equation simulation results. Training sets of various sizes (from 5120 to 163,840) were used. First order and total order Sobol methods were applied to both models in order to test the feasibility of neural network metamodels for sensitivity analysis of a radionuclide transport model. Convolutional neural networks were found to be capable of emulating the differential equation models with high accuracy when the training set size was 40,960 or higher. Neural network metamodels also gave similar results compared with the numerical solutions of the partial differential equation model in sensitivity analysis.

Galaxy clustering from the bottom up: a streaming model emulator I

ABSTRACT In this series of papers, we present a simulation-based model for the non-linear clustering of galaxies based on separate modelling of clustering in real space and velocity statistics. In the first paper, we present an emulator for the real-space correlation function of galaxies, whereas the emulator of the real-to-redshift space mapping based on velocity statistics is presented in the second paper. Here, we show that a neural network emulator for real-space galaxy clustering trained on data extracted from the dark quest suite of N-body simulations achieves sub-per cent accuracies on scales 1 &amp;lt; r &amp;lt; 30 $h^{-1} \, \mathrm{Mpc}$, and better than 3 per cent on scales r &amp;lt; 1 $h^{-1}\, \mathrm{Mpc}$ in predicting the clustering of dark-matter haloes with number density 10−3.5$(h^{-1}\, \mathrm{Mpc})^{-3}$, close to that of SDSS LOWZ-like galaxies. The halo emulator can be combined with a galaxy–halo connection model to predict the galaxy correlation function through the halo model. We demonstrate that we accurately recover the cosmological and galaxy–halo connection parameters when galaxy clustering depends only on the mass of the galaxies’ host halos. Furthermore, the constraining power in σ8 increases by about a factor of 2 when including scales smaller than 5 $h^{-1} \, \mathrm{Mpc}$. However, when mass is not the only property responsible for galaxy clustering, as observed in hydrodynamical or semi-analytic models of galaxy formation, our emulator gives biased constraints on σ8. This bias disappears when small scales (r &amp;lt; 10 $h^{-1}\, \mathrm{Mpc}$) are excluded from the analysis. This shows that a vanilla halo model could introduce biases into the analysis of future data sets.