Bayesian calibration of terrestrial ecosystem models: a study of advanced Markov chain Monte Carlo methods

Lu, Dan; Ricciuto, Daniel M.; Walker, Anthony P.; Safta, Cosmin; Munger, J. William

doi:10.5194/bg-14-4295-2017

Cited by 30 publications

(27 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additional fine tuning did not significantly affect the cost function value. More details about the QPSO algorithm and calibration approach are provided in Lu et al (2017).…”

Section: Model Optimization Algorithmmentioning

confidence: 99%

An Integrative Model for Soil Biogeochemistry and Methane Processes: I. Model Structure and Sensitivity Analysis

Ricciuto

Shi

et al. 2021

JGR Biogeosciences

Self Cite

View full text Add to dashboard Cite

Northern peatlands contain a vast pool of soil carbon that may be vulnerable to atmospheric release under changing environmental conditions, potentially causing a positive feedback to the climate system (Frolking et al., 2011;Nichols & Peteet, 2019;Yu, 2012). The magnitude of carbon emissions and their mechanistic responses to changing environments are elusive due to the complexity of hydrologic and biogeochemical processes in peatland systems (Blodau, 2002). Methane (CH 4 ) is one of the key carbon forms leaving peatlands under anaerobic conditions. Given the high radiative warming potential of CH 4 compared to CO 2 (Neubauer & Megonigal, 2015), it is critically important to accurately predict future CH 4 emissions from global peatlands. Peatlands are typically formed over millennial timescales due to organic carbon inputs, long-term water saturation and low temperatures and thus store major amounts of terrestrial carbon (Yu, 2012). It is expected that hydrological and biogeochemical dynamics play important roles affecting CH 4 fluxes from peatlands, but many of these key processes are missing in current Earth system models (Bohn et al., 2015). Therefore, to build a better predictive capacity for CH 4 dynamics in peatlands, it is necessary to fully consider the processes and environmental conditions controlling CH 4 processes, particularly the

show abstract

“…Additional fine tuning did not significantly affect the cost function value. More details about the QPSO algorithm and calibration approach are provided in Lu et al (2017).…”

Section: Model Optimization Algorithmmentioning

confidence: 99%

An Integrative Model for Soil Biogeochemistry and Methane Processes: I. Model Structure and Sensitivity Analysis

Ricciuto

Shi

et al. 2021

JGR Biogeosciences

Self Cite

View full text Add to dashboard Cite

show abstract

“…We probabilistically addressed the varying degrees of uncertainty (variety of species, drought scenarios, methods, and scales) of the three image-based modeling approaches (Table 1) using Bayesian statistical approaches (Link et al, 2002;Wang et al, 2009;Xu et al, 2015;Anderegg et al, 2016;Ryu et al, 2019). Hierarchical Bayesian approaches provide robust, testable predictions from the data in hand while allowing for probabilistic testing of false positives interacting with model complexity and probabilistic exploration of measurement error (Olejnik and Algina, 1983;Press, 2005;Lu et al, 2017). Both of these benefits are crucial when using different types of measurements to increase transparency in the analyses and inferences using explicit priors and credible intervals to better explain the variation in the data (Samanta et al, 2007;Ogle and Barber, 2008;Quaife and Cripps, 2016;King et al, 2019;Phillipson et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

Biophysically Informed Imaging Acquisition of Plant Water Status

Beverly

Ewers

2020

Front. For. Glob. Change

View full text Add to dashboard Cite

Vegetation controls carbon and water fluxes because of the fundamental tradeoff between carbon dioxide uptake and water loss occurring when stomata are open. Quantifying the rates of this exchange typically requires either intensive gas exchange or destructive harvesting of tissues and mass spectrometry analyses. Recent developments in high-throughput methods have enhanced our capacity to empirically test plant-environmental interactions. The vast integration characterizing satellite remote sensing methods masks organ-level physiological mechanisms limiting the predictive capability of current process models. Hence, more ground truth studies are necessary to determine the amount of mechanistic information needed to improve our understanding of forest, crop, and land management. Imaging methodologies, such as thermal and chlorophyll a fluorescence, are currently used to collect information for relevant traits such as water use, growth, and stress response. We tested these techniques during progressive drought across species with different susceptibility in controlled greenhouse conditions. We chose two highly represented tree species in North America: the gymnosperm Pinus ponderosa and the angiosperm Populus tremuloides. To better explore the whole drought response parameter space, we also tested a crop (Brassica rapa) and desert shrub (Artemisia tridentata). Thermal and fluorescence images of the canopy were coupled with leaf-level measurements as we performed three tests to predict drought response using (1) leaf temperature, (2) chlorophyll a fluorescence, and (3) the combination of the two. At 5 days of drought, leaf temperature increased 7 and 10%, accounting for 63 and 73% of the variation in stomatal conductance for both tree species, respectively. The fluorescence signal from images decreased ∼12% and ∼83% in moderately and severely droughted leaves respectively, reaching zero at mortality. Leaf water status was then predicted using a Bayesian approach that incorporated measurements' uncertainty and parsimony in the analysis of the parameters. Changes in canopy temperature provided confident predictions for the reductions of daily evapotranspiration at the onset of drought. Empirically combining thermal and fluorescence measurements improved predictions (R 2 = 0.81) of midday leaf water potential compared to univariate models. Our results represent an important step toward quantifying plant water status during drought using first principles that do not require species-specific information.

show abstract

“…Improving predictive understanding of Earth system variability and change requires datamodel integration. For example, Bilionis et al (2015) improved Community Land Model (CLM) prediction of crop productivity after model calibration; Müller et al (2015) improved the CLM prediction of methane emission after parameter optimization; and Fox et al (2009) and Lu et al (2017) improved the terrestrial ecosystem model predictive credibility of carbon fluxes after uncertainty quantification. However, data-model integration methods are usually computationally expensive involving a large ensemble of model simulations, which prohibits their application to complex Earth system models (ESMs) with lengthy simulation time.…”

Section: Introductionmentioning

confidence: 99%

Efficient surrogate modeling methods for large-scale Earth system models based on machine-learning techniques

Ricciuto

2019

Geosci. Model Dev.

Self Cite

View full text Add to dashboard Cite

Improving predictive understanding of Earth system variability and change requires data-model integration. Efficient data-model integration for complex models requires surrogate modeling to reduce model evaluation time. However, building a surrogate of a large-scale Earth system model (ESM) with many output variables is computationally intensive because it involves a large number of expensive ESM simulations. In this effort, we propose an efficient surrogate method capable of using a few ESM runs to build an accurate and fast-to-evaluate surrogate system of model outputs over large spatial and temporal domains. We first use singular value decomposition to reduce the output dimensions, and then use Bayesian optimization techniques to generate an accurate neural network surrogate model based on limited ESM simulation samples. Our machine learning based surrogate methods can build and evaluate a large surrogate system of many variables quickly. Thus, whenever the quantities of interest change such as a different objective function, a new site, and a longer simulation time, we can simply extract the information of interest from the surrogate system without rebuilding new surrogates, which significantly saves computational efforts. We apply the proposed method to a regional ecosystem model to approximate the relationship between 8 model parameters and 42660 carbon flux outputs. Results indicate that using only 20 model simulations, we can build an accurate surrogate system of the 42660 variables, where the consistency between the surrogate prediction and actual model simulation is 0.93 and the mean squared error is 0.02. This highly-accurate and fast-to-evaluate surrogate system will greatly enhance the computational efficiency in datamodel integration to improve predictions and advance our understanding of the Earth system.

show abstract

Bayesian calibration of terrestrial ecosystem models: a study of advanced Markov chain Monte Carlo methods

Cited by 30 publications

References 60 publications

An Integrative Model for Soil Biogeochemistry and Methane Processes: I. Model Structure and Sensitivity Analysis

An Integrative Model for Soil Biogeochemistry and Methane Processes: I. Model Structure and Sensitivity Analysis

Biophysically Informed Imaging Acquisition of Plant Water Status

Efficient surrogate modeling methods for large-scale Earth system models based on machine-learning techniques

Contact Info

Product

Resources

About