Tuning Of Model Parameters Research Articles

Spatial prediction of organic carbon in German agricultural topsoil using machine learning algorithms

Abstract. As the largest terrestrial carbon pool, soil organic carbon (SOC) has the potential to influence and mitigate climate change; thus, SOC monitoring is of high importance in the frameworks of various international treaties. Therefore, high-resolution SOC maps are required. Machine learning (ML) offers new opportunities to develop these maps due to its ability to data mine large datasets. The aim of this study was to apply three algorithms commonly used in digital soil mapping – random forest (RF), boosted regression trees (BRT), and support vector machine for regression (SVR) – on the first German agricultural soil inventory to model the agricultural topsoil (0–30 cm) SOC content and develop a two-model approach to address the high variability in SOC in German agricultural soils. Model performance is often limited by the size and quality of the soil dataset available for calibration and validation. Therefore, the impact of enlarging the training dataset was tested by including data from the European Land Use/Cover Area frame Survey for agricultural sites in Germany. Nested cross-validation was implemented for model evaluation and parameter tuning. Grid search and the differential evolution algorithm were also applied to ensure that each algorithm was appropriately tuned . The SOC content of the German agricultural soil inventory was highly variable, ranging from 4 to 480 g kg−1. However, only 4 % of all soils contained more than 87 g kg−1 SOC and were considered organic or degraded organic soils. The results showed that SVR produced the best performance, with a root-mean-square error (RMSE) of 32 g kg−1 when the algorithms were trained on the full dataset. However, the average RMSE of all algorithms decreased by 34 % when mineral and organic soils were modelled separately, with the best result from SVR presenting an RMSE of 21 g kg−1. The model performance was enhanced by up to 1 % for mineral soils and by up to 2 % for organic soils. Despite the ability of machine learning algorithms, in general, and SVR, in particular, to model SOC on a national scale, the study showed that the most important aspect for improving the model performance was to separate the modelling of mineral and organic soils.

Analysis and Demonstration of Control Scheme for Multiple Operating Modes of Energy Storage Converters to Enhance Power Factor

Rectifiers are required by the devices connected to the distribution end of the electrical power networks for AC/DC conversion. The line current becomes non-sinusoidal when a capacitor with a significant value is used to mitigate the output voltage ripple. This type of converter emulates a non-resistive impedance to the grid, due to which a bend occurs in the shape of the line current, which results in high total harmonic distortion and a low power factor. For perceiving sinusoidal current, power factor correction techniques are required. A digital controller for parallel-connected buck-boost power factor correctors is presented in this article to maintain a constant output voltage and to deal with circulating currents amongst parallel-connected converters. The proposed digital supervisory controller also regulates the input and line currents to keep them sinusoidal according to the input supply voltage to maintain the high power factor of the system. In this paper, using the differential equations of a buck-boost converter, the duty cycle calculations are performed for both Continuous Conduction Mode (CCM) and Discontinuous Conduction Mode (DCM), which are responsible for providing a unity power factor. A supervisory controller encompasses a feed-forward control algorithm for tuning model parameters for eliminating the harmonics from the line current. The proposed scheme helps calculate duty cycles which provides a unity power factor and minimizes the circulating currents. The proposed method was simulated in MATLAB/Simulink and their digital-hardware validation testing was also performed using C2000 MCU Launchpad.

Machine learning-based improvement of MDS-CBC score brings platelets into the limelight to optimize smear review in the hematology laboratory

BackgroundMyelodysplastic syndromes (MDS) are clonal hematopoietic diseases of the elderly characterized by chronic cytopenias, ineffective and dysplastic haematopoiesis, recurrent genetic abnormalities and increased risk of progression to acute myeloid leukemia. A challenge of routine laboratory Complete Blood Counts (CBC) is to correctly identify MDS patients while simultaneously avoiding excess smear reviews. To optimize smear review, the latest generations of hematology analyzers provide new cell population data (CPD) parameters with an increased ability to screen MDS, among which the previously described MDS-CBC Score, based on Absolute Neutrophil Count (ANC), structural neutrophil dispersion (Ne-WX) and mean corpuscular volume (MCV). Ne-WX is increased in the presence of hypogranulated/degranulated neutrophils, a hallmark of dysplasia in the context of MDS or chronic myelomonocytic leukemia. Ne-WX and MCV are CPD derived from leukocytes and red blood cells, therefore the MDS-CBC score does not include any platelet-derived CPD. We asked whether this score could be improved by adding the immature platelet fraction (IPF), a CPD used as a surrogate marker of dysplastic thrombopoiesis.MethodsHere, we studied a cohort of more than 500 individuals with cytopenias, including 168 MDS patients. In a first step, we used Breiman’s random forests algorithm, a machine-learning approach, to identify the most relevant parameters for MDS prediction. We then designed Classification And Regression Trees (CART) to evaluate, using resampling, the effect of model tuning parameters on performance and choose the “optimal” model across these parameters.ResultsUsing random forests algorithm, we identified Ne-WX and IPF as the strongest discriminatory predictors, explaining 37 and 33% of diagnoses respectively. To obtain “simplified” trees, which could be easily implemented into laboratory middlewares, we designed CART combining MDS-CBC score and IPF. Optimal results were obtained using a MDS-CBC score threshold equal to 0.23, and an IPF threshold equal to 3%.ConclusionsWe propose an extended MDS-CBC score, including CPD from the three myeloid lineages, to improve MDS diagnosis on routine laboratory CBCs and optimize smear reviews.

Machine learning and deep learning modeling and simulation for predicting PM2.5 concentrations

Particulate matter (PM) pollution greatly endanger human physical and mental health, and it is of great practical significance to predict PM concentrations accurately. This study measured one-year monitoring data of six main meteorological parameters and PM2.5 concentrations independently at two monitoring sites in central China's Hunan Province. These datasets were then employed to train, validate, and evaluate the proposed extreme gradient boosting (XGBoost) machine learning model and the fully connected neural network deep learning model, respectively. The performances of the two models were compared, analyzed, and optimized through model parameter tuning. The XGBoost model had better prediction ability with R2 higher than 0.761 in the complete test dataset. When the complete dataset was divided into stratified sub-sets by daytime-nighttime periods, the value of R2 increased to 0.856 in the nighttime test dataset. The feature importance and influential mechanism of meteorological variables on PM2.5 concentrations were analyzed and discussed.

A Deep Learning Model for Stroke Patients' Motor Function Prediction.

Deep learning models are effectively employed to transfer learning to adopt learning from other areas. This research utilizes several neural structures to interpret the electroencephalogram images (EEG) of brain-injured cases to plan operative imagery-computerized interface models for controlling left and right hand movements. This research proposed a model parameter tuning with less training time using transfer learning techniques. The precision of the proposed model is assessed by the aptitudes of motor imagery detection. The experiments depict that the best performance is attained with the incorporation of the proposed EEG-DenseNet and the transfer model. The prediction accuracy of the model reached 96.5% with reduced time computational cost. These high performance proves that the EEG-DenseNet model has high prospective for motor imagery brain-injured therapy systems. It also productively exhibited the effectiveness of transfer learning techniques for enhancing the accuracy of electroencephalogram brain-injured therapy models.

Dynamic QoS Prediction Algorithm Based on Kalman Filter Modification.

With the widespread adoption of service-oriented architectures (SOA), services with the same functionality but the different Quality of Service (QoS) are proliferating, which is challenging the ability of users to build high-quality services. It is often costly for users to evaluate the QoS of all feasible services; therefore, it is necessary to investigate QoS prediction algorithms to help users find services that meet their needs. In this paper, we propose a QoS prediction algorithm called the MFDK model, which is able to fill in historical sparse QoS values by a non-negative matrix decomposition algorithm and predict future QoS values by a deep neural network. In addition, this model uses a Kalman filter algorithm to correct the model prediction values with real-time QoS observations to reduce its prediction error. Through extensive simulation experiments on the WS-DREAM dataset, we analytically validate that the MFDK model has better prediction accuracy compared to the baseline model, and it can maintain good prediction results under different tensor densities and observation densities. We further demonstrate the rationality of our proposed model and its prediction performance through model ablation experiments and parameter tuning experiments.

DeepGraphh: AI-driven web service for graph-based quantitative structure-activity relationship analysis.

Artificial intelligence (AI)-based computational techniques allow rapid exploration of the chemical space. However, representation of the compounds into computational-compatible and detailed features is one of the crucial steps for quantitative structure-activity relationship (QSAR) analysis. Recently, graph-based methods are emerging as a powerful alternative to chemistry-restricted fingerprints or descriptors for modeling. Although graph-based modeling offers multiple advantages, its implementation demands in-depth domain knowledge and programming skills. Here we introduce deepGraphh, an end-to-end web service featuring a conglomerate of established graph-based methods for model generation for classification or regression tasks. The graphical user interface of deepGraphh supports highly configurable parameter support for model parameter tuning, model generation, cross-validation and testing of the user-supplied query molecules. deepGraphh supports four widely adopted methods for QSAR analysis, namely, graph convolution network, graph attention network, directed acyclic graph and Attentive FP. Comparative analysis revealed that deepGraphh supported methods are comparable to the descriptors-based machine learning techniques. Finally, we used deepGraphh models to predict the blood-brain barrier permeability of human and microbiome-generated metabolites. In summary, deepGraphh offers a one-stop web service for graph-based methods for chemoinformatics.

Data-Driven Model-Free Adaptive Displacement Control for Tap-Water-Driven Artificial Muscle and Parameter Design Using Virtual Reference Feedback Tuning

A McKibben artificial muscle has strong asymmetric hysteresis characteristics, which depend on the load applied to the muscle. Thus, designing a controller for high-performance displacement is difficult. In a previous study, model predictive control with a servomechanism combining an inverse optimization algorithm with adaptive model matching, and a data-driven model-free adaptive control (MFAC) were introduced. As a result, a high tracking control performance was achieved in both control methods. However, model-based and data-driven approaches require a highly accurate mathematical model and a large number of design parameters, making them time-consuming, respectively. To solve these problems, in the present study, a controller design that requires no precise mathematical model and less design parameter tuning with trial and error was developed by combining conventional MFAC and virtual reference feedback tuning, which is a data-driven control method. Experimental results indicated that important design parameters, such as the initial pseudo-gradient vector and weighting factor, can be readily obtained. Compared with conventional MFAC, higher tracking control performance without overshoot was achieved in transient response, while the same level of control performance was maintained in steady-state response.

A new control strategy of stopped-rotor aircraft based on manipulation model

PurposeThe purpose of this paper is to study the control strategy of transition mode of the stopped-rotor (SR) aircraft under the condition of redundant control and complex aerodynamic characteristics.Design/methodology/approachThis paper first proposes a transition strategy for the conversion between helicopter mode and fixed-wing mode. Then, aiming at the redundancy of the two control systems in the transition process, a control model is proposed, which greatly simplifies the control in conversion mode. Then, to facilitate the design of the control system, the Takagi-Sugeno model of the SR aircraft in transition mode is established. Finally, an explicit model tracking and tuning parameter stability augmentation control system is designed, so that the SR aircraft has a good stability during the transition process. Then, the outer loop control system of transition flight is designed.FindingsThe simulation results show that the control strategy proposed in this paper can realize the mode conversion well. It lays a solid foundation for the subsequent engineering flight test for the SR aircraft.Originality/valueThe work done in this paper provides ideas and methods for the flight control system design of SR aircraft in transition mode. The method of designing control model to solve the coordination of redundant control system is also applicable for other multimode aircraft, which provides a simple and convenient method for the multimode aircraft control.

Constraining Global Coronal Models with Multiple Independent Observables

Global coronal models seek to produce an accurate physical representation of the Sun’s atmosphere that can be used, for example, to drive space-weather models. Assessing their accuracy is a complex task, and there are multiple observational pathways to provide constraints and tune model parameters. Here, we combine several such independent constraints, defining a model-agnostic framework for standardized comparison. We require models to predict the distribution of coronal holes at the photosphere, and neutral line topology at the model’s outer boundary. We compare these predictions to extreme-ultraviolet (EUV) observations of coronal hole locations, white-light Carrington maps of the streamer belt, and the magnetic sector structure measured in situ by Parker Solar Probe and 1 au spacecraft. We study these metrics for potential field source surface (PFSS) models as a function of source surface height and magnetogram choice, as well as comparing to the more physical Wang–Sheeley–Arge (WSA) and the Magnetohydrodynamic Algorithm outside a Sphere (MAS) models. We find that simultaneous optimization of PFSS models to all three metrics is not currently possible, implying a trade-off between the quality of representation of coronal holes and streamer belt topology. WSA and MAS results show the additional physics that they include address this by flattening the streamer belt while maintaining coronal hole sizes, with MAS also improving coronal hole representation relative to WSA. We conclude that this framework is highly useful for inter- and intra-model comparisons. Integral to the framework is the standardization of observables required of each model, evaluating different model aspects.

Deep learning inversion of Rayleigh-wave dispersion curves with geological constraints for near-surface investigations

SUMMARY With the emergence of massive seismic data sets, surface wave methods using deep learning (DL) can effectively obtain shear wave velocity (Vs) structure for non-invasive near-surface investigations. Previous studies on DL inversion for deep geophysical investigation have a reference model to generate the training data set, while near-surface investigations have no model. Therefore, we systematically give a set of training data set generation processes. In the process, we use both prior information and the observed data to constrain the data set so that the DL inversion model can learn the local geological characteristics of the survey area. Because the space of inverted Vs models is constrained and thus narrowed, the inversion non-uniqueness can be reduced. Furthermore, the mean squared error, which is commonly used as loss function, may cause a poor fitting accuracy of phase velocities at high frequencies in near-surface applications. To make the fitting accuracy evenly in all frequency bands, we modify the loss function into a weighted mean squared relative error. We designed a convolutional neural network (CNN) to directly invert fundamental-mode Rayleigh-wave phase velocity for 1-D Vs models. To verify the feasibility and reliability of the proposed algorithm, we tested and compared it with the Levenberg–Marquardt (L-M) inversion and neighbourhood algorithm (NA) using field data from the Lawrence experiment (USA) and the Wuwei experiment (China). In both experiments, the inverted Vs models by CNN are consistent with the borehole information and are similar to that from existing methods after fine tuning of model parameters. The average root mean squares errors (RMSEs) of the CNN, NA and L-M methods are also similar, except in the Lawrence experiment, the RMSE of CNN is 17.33 m s−1 lower than previous studies using the L-M method. Moreover, the comparison of different loss functions for the Wuwei experiment indicates that the modified loss function can achieve higher accuracy than the traditional one. The proposed CNN is therefore ideally suited for rapid, repeated near-surface subsurface imaging and monitoring under similar geological settings.

Statistical shape modelling of the human mandible: 3D shape predictions based on external morphometric features

One of the main limitations in subject-centred design approach is represented by getting 3D models of the region of interest. Indeed, 3D reconstruction from imaging data (i.e., computed tomography scans) is expensive and exposes the subject to high radiation doses. Statistical Shape Models (SSMs) are mathematical models able to describe the variability associated to a population and allow predicting new shapes tuning model parameters. These parameters almost never have a physical meaning and so they cannot be directly related to morphometric features. In this study a gender-combined SSM model of the human mandible was setup, using Generalised Procrustes Analysis and Principal Component Analysis on a dataset of fifty mandibles. Twelve morphometric features, able to characterise the mandibular bone and readily collectable during external examinations, were recorded and correlated to SSM parameters by a multiple linear regression approach. Then a cross-validation procedure was performed on a control set to determine the combination of features able to minimise the average deviation between real and predicted shapes. Compactness of the SSM and main modes of deformations have been investigated and results consistent with previous works involving a higher number of shapes were found. A combination of five features was proved to characterise predicted shapes minimising the average error. As completion of the work, a male SSM was developed and performances compared with those of the combined SSM. The features-based model here proposed could represent a useful and easy-to-use tool for the generation of 3D customised models within a virtual interactive design environment.

Speech as a Biomarker for COVID-19 Detection Using Machine Learning.

The use of speech as a biomedical signal for diagnosing COVID-19 is investigated using statistical analysis of speech spectral features and classification algorithms based on machine learning. It is established that spectral features of speech, obtained by computing the short-time Fourier Transform (STFT), get altered in a statistical sense as a result of physiological changes. These spectral features are then used as input features to machine learning-based classification algorithms to classify them as coming from a COVID-19 positive individual or not. Speech samples from healthy as well as “asymptomatic” COVID-19 positive individuals have been used in this study. It is shown that the RMS error of statistical distribution fitting is higher in the case of speech samples of COVID-19 positive speech samples as compared to the speech samples of healthy individuals. Five state-of-the-art machine learning classification algorithms have also been analyzed, and the performance evaluation metrics of these algorithms are also presented. The tuning of machine learning model parameters is done so as to minimize the misclassification of COVID-19 positive individuals as being COVID-19 negative since the cost associated with this misclassification is higher than the opposite misclassification. The best performance in terms of the “recall” metric is observed for the Decision Forest algorithm which gives a recall value of 0.7892.

PyLEnM: A Machine Learning Framework for Long-Term Groundwater Contamination Monitoring Strategies.

In this study, we have developed a comprehensive machine learning (ML) framework for long-term groundwater contamination monitoring as the Python package PyLEnM (Python for Long-term Environmental Monitoring). PyLEnM aims to establish the seamless data-to-ML pipeline with various utility functions, such as quality assurance and quality control (QA/QC), coincident/colocated data identification, the automated ingestion and processing of publicly available spatial data layers, and novel data summarization/visualization. The key ML innovations include (1) time series/multianalyte clustering to find the well groups that have similar groundwater dynamics and to inform spatial interpolation and well optimization, (2) the automated model selection and parameter tuning, comparing multiple regression models for spatial interpolation, (3) the proxy-based spatial interpolation method by including spatial data layers or in situ measurable variables as predictors for contaminant concentrations and groundwater levels, and (4) the new well optimization algorithm to identify the most effective subset of wells for maintaining the spatial interpolation ability for long-term monitoring. We demonstrate our methodology using the monitoring data at the Savannah River Site F-Area. Through this open-source PyLEnM package, we aim to improve the transparency of data analytics at contaminated sites, empowering concerned citizens as well as improving public relations.

An efficient bounded-variable nonlinear least-squares algorithm for embedded MPC

This paper presents a novel approach to solve linear and nonlinear model predictive control (MPC) problems that requires small memory footprint and throughput and is particularly suitable when the model and/or controller parameters change at runtime. The contributions of the paper include: (i) a formulation of the nonlinear MPC problem as a bounded-variable nonlinear least-squares (BVNLS) problem, demonstrating that the use of an appropriate solver can outperform industry-standard solvers; (ii) an easily-implementable library-free BVNLS solver with a novel proof of global convergence; (iii) a matrix-free method for computing the products of vectors and Jacobians, required by BVNLS; (iv) an efficient method for updating sparse QR factors when using active-set methods to solve sparse BVNLS problems. Thanks to explicitly parameterizing the optimization algorithm in terms of the model and MPC tuning parameters, the resulting approach is inherently and immediately adaptive to any changes in the MPC formulation. The same algorithmic framework can cope with linear, nonlinear, and adaptive MPC variants based on a broad class of prediction models and sum-of-squares cost functions.

Adaptive model-based predictive control for cross-phase cavity pressure control in injection molding

Injection molding is a manufacturing process for the production of plastic parts. The traditional process control switches between an injection velocity control and a pressure control. If the switchover is not optimal, quality issues occur during production. Additionally, the setup of processes with the traditional process control requires a high setup effort. The objective of cross-phase process control strategies is to reduce both weaknesses by controlling only one process variable during molding. Disadvantages of available cross-phase control approaches are related to the use of multiple empirical models in the controller which have to be adapted manually to the used machine settings. This contribution proposes a novel adaptive cross-phase cavity pressure control concept based on only one nonlinear time-variant model for the entire process. The adaptive approach reduces the setup effort and compensates inaccuracies during modeling, thus avoiding the need to adjust the controller model when machine settings change. Herefore, the controller model is adapted to the time-varying process dynamics by use of an Unscented Kalman Filter which estimates all process states along with an additional lumped parameter. The time-variant model is then used in a Model-based Predictive Controller to track a given cavity pressure reference. For the adaptive concept to operate properly, the estimated parameter must be positive which is ensured by an introduced model tuning parameter (MTP) influencing the convergence behavior of the parameter estimation. The MTP allows not only to taylor the convergence behavior of the parameter estimation for safe operation and good performance but also to rapidly adapt the control topology for new processing conditions, e.g. different machines or part geometries. Tuning guidlines for the MTP are provided, resulting in a setup procedure of greatly reduced effort resulting in an easy-to-follow setup procedure. The controller concept is validated on two different industrial injection molding machines and for two different part geometries showing good reference tracking capabilities with a steady-state error ⩽4bar.

A cluster plugin method for selecting the GLM lasso tuning parameters in models for unbalanced panel data

New methods are discussed for estimating the population-averaged (PA) coefficients of some variables of interest in a sparse, high-dimensional generalized linear model (GLM) in an unbalanced panel-data setting with random effects. A cluster plugin for GLM lassos with unbalanced panel data is proposed. It is proven that a lasso that uses the new cluster-plugin method can outperform a lasso using the cross-sectional plugin, when the covariates have non-zero within-panel covariances. The proposed cluster plugin for GLMs extends the literature on Neyman-orthogonal moment conditions and provides estimators for the PA coefficients in sparse, high-dimensional logit, Poisson, and linear models when the data come from unbalanced panels. The results of the Monte Carlo simulations show that the implemented estimators perform well in finite samples. The simulations also show that a GLM lasso using the proposed cluster-plugin method produces more accurate covariate selection than a GLM lasso using the cross-sectional plugin method, when the covariates have non-zero within-panel covariances. Easy-to-use Stata commands are available for the proposed methods.

Evaluation of sensitive physical parameter combinations for determining the uncertainty of fire simulations and predictions in China

AbstractFires are fundamental and natural phenomena that affect terrestrial ecosystems and the global climate change. However, uncertainties in fire modeling still exist. It is important to improve the ability to simulate fires by tuning model parameters. In this study, a sensitivity analysis (SA) approach based on the conditional nonlinear optimal perturbation related to parameters (CNOP‐P) is employed to find the most sensitive parameter subset. First, the maximum uncertainty in modeling fire is estimated, and it is found that the degrees of uncertainty in modeling fire are different in different regions of China. The extents of the uncertainties in modeling fire in northeastern China and northern China with arid and semiarid climate conditions are greatest, and the magnitude of the uncertainty in southern China is the smallest. The uncertainty in modeling fires in northern China with a semihumid climate condition is between the above values. Second, we find that the most sensitive parameter combination with the number of elements determined by the SA method based on the CNOP‐P approach are different from the top five parameters combination via a sensitivity test using a traditional method (such as the one‐at‐a‐time [OAT] method). The most sensitive parameter combination includes not only the parameters in the fire module (e.g., ) but also the parameters that could cause variations in biomass and soil moisture. The prediction skill of fire by reducing the errors of the sensitive parameter combination using the SA method based on the CNOP‐P method are higher than those using the OAT method. The above results suggest that not only the parameters in the fire module but also the parameters of other physical processes (e.g., biomass and soil moisture) should be corrected and calibrated to improve simulation and prediction of fire.

An ensemble method to generate high-resolution gridded population data for China from digital footprint and ancillary geospatial data

Fine-scale population datasets are essential to many health and development applications. Quite a few population estimate approaches have been proposed and multiple gridded population datasets have been produced. However, it is still a challenge to accurately estimate daily and even hourly population dynamics. In this study, we present an ensemble learning approach to tackle the challenge through integrating a digital footprint dataset and multiple geospatial ancillary datasets to estimate population dynamics. More specifically, we used the geographically weighted regression model to integrate two aspatial tree-based learning models and generated preliminary hourly and daily gridded population estimates. We then adjusted the fine-scale population estimates based on the county-level estimates and their nonlinear relationship with the grid-level covariates. After sufficient model training and parameter tuning, we produced a series 0.01-degree gridded population maps (FinePop) of China for 2018, including a nationwide daily-average map and provincial hourly-average maps. The FinePop is more accurate than the WorldPop and LandScan datasets, as suggested by the highest R2 (0.72) obtained from the comparison against township-level population census data. The root mean squared error of the township population density estimates for FinePop, WorldPop, and LandScan are 3162, 3327, and 3423, respectively. The FinePop also shows its advantages in unraveling transportation networks and the diurnal-nocturnal population migration patterns in both small and large cities.

Comparison of machine learning algorithms for emulation of a gridded hydrological model given spatially explicit inputs

This study compares the performance of several machine learning algorithms in reproducing the spatial and temporal outputs of the process-based, hydrological model, ParFlow.CLM. Emulators or surrogate models are often used to reduce complexity and simulation times of complex models, and have typically been applied to evaluate parameter sensitivity or for model parameter tuning, without explicit treatment of variation resulting from spatially explicit inputs to the model. Here we present a case study in which we evaluate candidate machine learning algorithms for suitability emulating model outputs given spatially explicit inputs. We find that among random forest, gaussian process, k-nearest neighbors, and deep neural networks, the random forest algorithm performs the best on small training sets, is not as sensitive to hyperparameters chosen for the machine learning model, and can be trained quickly. Although deep neural networks were hypothesized to be able to better capture the potential nonlinear interactions in ParFlow.CLM, they also required more training data and much more refined tuning of hyperparameters to achieve the potential benefits of the algorithm.