Sources Of Uncertainty In Models Research Articles

Uncertainty quantification of CO2 storage using Bayesian model averaging and polynomial chaos expansion

Carbon sequestration in oil reservoirs and deep saline formations may be accomplished by many different trapping mechanisms. Use of CO2 for Enhanced Oil Recovery (CO2-EOR) leads to CO2 in three distinct phases, including CO2 dissolved in oil, CO2 dissolved in water (aqueous) and/or supercritical CO2. We evaluated the total amount of stored CO2 as well as the amount of CO2 in each phase for an active CO2-EOR site in western Texas.Three-phase relative permeability and associated hysteresis are two major sources of model uncertainty. Both are difficult to measure and are usually predicted by interpolation models. Instead of using arbitrary interpolation models, we used a model-averaging method based on Bayesian inference to estimate integrated model uncertainty for 12 alternative models. Moreover, given the uncertainty of intrinsic rock properties including permeability and porosity, uncertainty quantification (UQ) of these parameters is also necessary for forecasting CO2 storage capacity. Thus, results of this study provide uncertainty based on both model and data uncertainty.Conventional Monte Carlo methods with geocellular simulations are computationally expensive. We applied a Polynomial Chaos Expansion (PCE) methodology instead, to reduce computational requirements while minimizing the loss of accuracy. Geostatistical techniques were applied to generate stochastic realizations based on well logs and seismic survey data.For the Texas case study, we developed a systematic approach to quantify overall uncertainty, including both model uncertainty and parameter uncertainty. The approach was applied to forecast results at three important time steps, the end of the 30-year CO2-EOR injection period, the end of the 20-year post EOR CO2 injection period, and the end of the 50-year monitoring period. Results suggest that oil solubility dominates CO2 trapping and aqueous solubility has the least relative importance with respect to trapping (storage). Predictions of model averaging preserved the general pattern and captured differences among alternative models. CO2 storage of the reference model was within one standard deviation of predictions of model averaging. Estimated relative error between forecasted CO2 storage and the reference model are 0.8%, 7.4%, and 6.1% at three selected time steps. By the end of simulation, estimated CO2 storage in five selected layers in oil, supercritical, and aqueous phases are 3.4 ± 0.3 million tonnes, 2.0 ± 0.25 million tonnes, and 0.24 ± 0.04 million tonnes, respectively.

Causes of variation among rice models in yield response to CO2 examined with Free-Air CO2 Enrichment and growth chamber experiments

The CO2 fertilization effect is a major source of uncertainty in crop models for future yield forecasts, but coordinated efforts to determine the mechanisms of this uncertainty have been lacking. Here, we studied causes of uncertainty among 16 crop models in predicting rice yield in response to elevated [CO2] (E-[CO2]) by comparison to free-air CO2 enrichment (FACE) and chamber experiments. The model ensemble reproduced the experimental results well. However, yield prediction in response to E-[CO2] varied significantly among the rice models. The variation was not random: models that overestimated at one experiment simulated greater yield enhancements at the others. The variation was not associated with model structure or magnitude of photosynthetic response to E-[CO2] but was significantly associated with the predictions of leaf area. This suggests that modelled secondary effects of E-[CO2] on morphological development, primarily leaf area, are the sources of model uncertainty. Rice morphological development is conservative to carbon acquisition. Uncertainty will be reduced by incorporating this conservative nature of the morphological response to E-[CO2] into the models. Nitrogen levels, particularly under limited situations, make the prediction more uncertain. Improving models to account for [CO2] × N interactions is necessary to better evaluate management practices under climate change.

The carbon quality-temperature hypothesis does not consistently predict temperature sensitivity of soil organic matter mineralization in soils from two manipulative ecosystem experiments

The temperature sensitivity of soil organic matter (SOM) decomposition is a source of uncertainty in models of soil-climate feedbacks. However, empirical studies have given contradictory results concerning the temperature response of SOM fractions, even as the understanding of the chemical nature of SOM is evolving. The carbon-quality temperature (CQT) hypothesis states that more ‘recalcitrant’ SOM should have higher temperature sensitivity. Incubation studies have often shown a negative correlation between soil respiration rates and temperature sensitivity. However, there have been important exceptions to these results which challenge the assumption that older SOM is necessarily more chemically complex. We asked whether we would expect a universal relationship between temperature sensitivity and soil respiration given that SOM decomposition is influenced by factors other than chemical complexity. We examined temperature sensitivity in long-term incubations of four soils representing two biomes and two ecosystem-level manipulations. Soils from a manipulative climate experiment in Pacific Northwest grasslands demonstrated an increase in temperature sensitivity with incubation duration, but soil from a 20-year input manipulation study in a Northeastern forest showed no relationship of temperature sensitivity with either carbon depletion or incubation time. Furthermore, across all four soils, the temperature sensitivity of soil respiration was frequently inconsistent with indices of carbon quality and did not show a negative correlation with soil respiration rate. We conclude that the CQT hypothesis fails to universally capture the temperature sensitivity of SOM decomposition across environmental contexts, consistent with an emerging understanding of the multiplicity of factors that control soil carbon cycling.

Comparing pairwise‐additive and many‐body generalized Born models for acid/base calculations and protein design

Generalized Born (GB) solvent models are common in acid/base calculations and protein design. With GB, the interaction between a pair of solute atoms depends on the shape of the protein/solvent boundary and, therefore, the positions of all solute atoms, so that GB is a many-body potential. For compute-intensive applications, the model is often simplified further, by introducing a mean, native-like protein/solvent boundary, which removes the many-body property. We investigate a method for both acid/base calculations and protein design that uses Monte Carlo simulations in which side chains can explore rotamers, bind/release protons, or mutate. The fluctuating protein/solvent dielectric boundary is treated in a way that is numerically exact (within the GB framework), in contrast to a mean boundary. Its originality is that it captures the many-body character while retaining the residue-pairwise complexity given by a fixed boundary. The method is implemented in the Proteus protein design software. It yields a slight but systematic improvement for acid/base constants in nine proteins and a significant improvement for the computational design of three PDZ domains. It eliminates a source of model uncertainty, which will facilitate the analysis of other model limitations. © 2017 Wiley Periodicals, Inc.

Global evaluation of gross primary productivity in the JULES land surface model v3.4.1

Abstract. This study evaluates the ability of the JULES land surface model (LSM) to simulate gross primary productivity (GPP) on regional and global scales for 2001–2010. Model simulations, performed at various spatial resolutions and driven with a variety of meteorological datasets (WFDEI-GPCC, WFDEI-CRU and PRINCETON), were compared to the MODIS GPP product, spatially gridded estimates of upscaled GPP from the FLUXNET network (FLUXNET-MTE) and the CARDAMOM terrestrial carbon cycle analysis. Firstly, when JULES was driven with the WFDEI-GPCC dataset (at 0. 5° × 0. 5° spatial resolution), the annual average global GPP simulated by JULES for 2001–2010 was higher than the observation-based estimates (MODIS and FLUXNET-MTE), by 25 and 8 %, respectively, and CARDAMOM estimates by 23 %. JULES was able to simulate the standard deviation of monthly GPP fluxes compared to CARDAMOM and the observation-based estimates on global scales. Secondly, GPP simulated by JULES for various biomes (forests, grasslands and shrubs) on global and regional scales were compared. Differences among JULES, MODIS, FLUXNET-MTE and CARDAMOM on global scales were due to differences in simulated GPP in the tropics. Thirdly, it was shown that spatial resolution (0. 5° × 0. 5°, 1° × 1° and 2° × 2°) had little impact on simulated GPP on these large scales, with global GPP ranging from 140 to 142 PgC year−1. Finally, the sensitivity of JULES to meteorological driving data, a major source of model uncertainty, was examined. Estimates of annual average global GPP were higher when JULES was driven with the PRINCETON meteorological dataset than when driven with the WFDEI-GPCC dataset by 3 PgC year−1. On regional scales, differences between the two were observed, with the WFDEI-GPCC-driven model simulations estimating higher GPP in the tropics (5° N–5° S) and the PRINCETON-driven model simulations estimating higher GPP in the extratropics (30–60° N).

High-fidelity Simulations of the Near-Earth Object Search Performance of the Large Synoptic Survey Telescope

Abstract We perform high-fidelity simulations of a wide-field telescopic survey searching for Near-Earth Objects (NEOs) larger than 140 m, focusing on the observation and detection model, as well as detection efficiency and accuracy. As a test survey, we select the Large Synoptic Survey Telescope (LSST). We use its proposed pointings for a 10-year mission, and model the detection of NEOs in the fields. We discuss individual model parameters for magnitude losses, vignetting, fading, asteroid rotation and colors, fill factor, limiting magnitude, rate of motion, field shape and rotation, and survey patterns. We assess results in terms of the cumulative completeness of the detected population as a function of size and time. Additionally, we examine the sources of modeling uncertainty, and derive the overall NEO population completeness for the baseline LSST survey to be 55 ± 5% for NEOs with absolute magnitude brighter than 22. Including already discovered objects and ongoing surveys, the NEO completeness at the end of the LSST baseline survey should reach ∼77%.

Debates—Hypothesis testing in hydrology: A subsurface perspective

AbstractModels for flow in environmental systems are subject to uncertainty. Models can thus be interpreted as hypotheses on the validity of the underlying model assumptions. One important source of uncertainty in models for flow and transport processes in the subsurface is the model concept. While uncertain model parameters or forcing terms can be captured as random processes and random fields, this type of uncertainty cannot be included into a model in a straightforward manner. This is particularly true if established model descriptions of a given process are not known or are still being debated. In this contribution, we outline several examples of subsurface flow and transport modeling where uncertainty of the model concept plays an important role. We discuss the need for the development of methods and standards to deal with this type of uncertainty in model hypothesis testing.

On the relationship between climate sensitivity and modelling uncertainty

Climate model projections are used to investigate the potential impacts of climate change on future weather, agriculture, water resources, human health, the global economy, etc. However, climate projections have a broad range of associated uncertainties, and it is a challenge to take account of these uncertainties in impact studies and risk assessments. Knowing which uncertainties matter and which may be reduced via scientific research or political decisions can help policy-makers in making informed decisions, scientists in focusing their resources, and businesses in building resilience to uncertainties that cannot be avoided. On the global scale, the present political resistance or ability to move from agreements to significant action provides the largest uncertainty in climate projections, followed by the uncertainty associated with climate modelling itself. Here, we show that climate sensitivity is a very important source of model uncertainty over large parts of the globe not only for temperature, but also for precipitation and wind projections. Because ‘climate sensitivity’ is a collective term that encompasses a wide range of feedback mechanisms in the climate system, we may not know for a long time whether models with high or low climate sensitivities are more relevant for the twenty-first century projections. Nevertheless, investigations of climate impacts cannot wait. Here we argue that it is physically and statistically unsound to mix climate model with high and low climate sensitivities, and that the subset chosen for any impact study should depend on the question one is trying to answer.

Quantifying the anthropogenic dust emission from agricultural land use and desiccation of the Aral Sea in Central Asia

AbstractA regional dust model system is applied to quantify the anthropogenic dust emission in the post‐Soviet Central Asia from 2000 to 2014. Two physically based dust schemes suggest that a proportion of 18.3–32.8% of total dust emissions is contributed by agricultural land use and the desiccation of Aral Sea, whereas a simplified dust scheme yields higher estimates in the range of 49.7–56.5% depending on whether a static or dynamic preferential dust source function is used. The dust schemes also differ greatly in the spatial distribution of anthropogenic dust and the sensitivity to the use of land use intensity in separating natural and human‐made source areas, suggesting that the model representation of erosion threshold velocity, especially the role of vegetation, is a key source of model uncertainty in quantifying anthropogenic dust. The relative importance of agriculture and dried Aral Sea bed (Aralkum) differs greatly among the dust schemes. Despite the increased dust from the expansion of Aralkum, there is a negative trend in the anthropogenic dust proportion, indicating a shift of dust emission toward natural source areas. All dust schemes show a decrease in anthropogenic dust in response to land cover changes over agricultural lands.

DGSA: A Matlab toolbox for distance-based generalized sensitivity analysis of geoscientific computer experiments

Sensitivity analysis plays an important role in geoscientific computer experiments, whether for forecasting, data assimilation or model calibration. In this paper we focus on an extension of a method of regionalized sensitivity analysis (RSA) to applications typical in the Earth Sciences. Such applications involve the building of large complex spatial models, the application of computationally extensive forward modeling codes and the integration of heterogeneous sources of model uncertainty. The aim of this paper is to be practical: 1) provide a Matlab code, 2) provide novel visualization methods to aid users in getting a better understanding in the sensitivity 3) provide a method based on kernel principal component analysis (KPCA) and self-organizing maps (SOM) to account for spatial uncertainty typical in Earth Science applications and 4) provide an illustration on a real field case where the above mentioned complexities present themselves. We present methods that extend the original RSA method in several ways. First we present the calculation of conditional effects, defined as the sensitivity of a parameter given a level of another parameters. Second, we show how this conditional effect can be used to choose nominal values or ranges to fix insensitive parameters aiming to minimally affect uncertainty in the response. Third, we develop a method based on KPCA and SOM to assign a rank to spatial models in order to calculate the sensitivity on spatial variability in the models. A large oil/gas reservoir case is used as illustration of these ideas.

Uncertainty of wheat water use: Simulated patterns and sensitivity to temperature and CO2

Projected global warming and population growth will reduce future water availability for agriculture. Thus, it is essential to increase the efficiency in using water to ensure crop productivity. Quantifying crop water use (WU; i.e. actual evapotranspiration) is a critical step towards this goal. Here, sixteen wheat simulation models were used to quantify sources of model uncertainty and to estimate the relative changes and variability between models for simulated WU, water use efficiency (WUE, WU per unit of grain dry mass produced), transpiration efficiency (Teff, transpiration per kg of unit of grain yield dry mass produced), grain yield, crop transpiration and soil evaporation at increased temperatures and elevated atmospheric carbon dioxide concentrations ([CO2]). The greatest uncertainty in simulating water use, potential evapotranspiration, crop transpiration and soil evaporation was due to differences in how crop transpiration was modelled and accounted for 50% of the total variability among models. The simulation results for the sensitivity to temperature indicated that crop WU will decline with increasing temperature due to reduced growing seasons. The uncertainties in simulated crop WU, and in particularly due to uncertainties in simulating crop transpiration, were greater under conditions of increased temperatures and with high temperatures in combination with elevated atmospheric [CO2] concentrations. Hence the simulation of crop WU, and in particularly crop transpiration under higher temperature, needs to be improved and evaluated with field measurements before models can be used to simulate climate change impacts on future crop water demand.

Bayesian Solutions for Handling Uncertainty in Survival Extrapolation.

Survival extrapolation using a single, best-fit model ignores 2 sources of model uncertainty: uncertainty in the true underlying distribution and uncertainty about the stability of the model parameters over time. Bayesian model averaging (BMA) has been used to account for the former, but it can also account for the latter. We investigated BMA using a published comparison of the Charnley and Spectron hip prostheses using the original 8-year follow-up registry data. A wide variety of alternative distributions were fitted. Two additional distributions were used to address uncertainty about parameter stability: optimistic and skeptical. The optimistic (skeptical) model represented the model distribution with the highest (lowest) estimated probabilities of survival but reestimated using, as prior information, the most optimistic (skeptical) parameter estimated for intermediate follow-up periods. Distributions were then averaged assuming the same posterior probabilities for the optimistic, skeptical, and noninformative models. Cost-effectiveness was compared using both the original 8-year and extended 16-year follow-up data. We found that all models obtained similar revision-free years during the observed period. In contrast, there was variability over the decision time horizon. Over the observed period, we detected considerable uncertainty in the shape parameter for Spectron. After BMA, Spectron was cost-effective at a threshold of £20,000 with 93% probability, whereas the best-fit model was 100%; by contrast, with a 16-year follow-up, it was 0%. This case study casts doubt on the ability of the single best-fit model selected by information criteria statistics to adequately capture model uncertainty. Under this scenario, BMA weighted by posterior probabilities better addressed model uncertainty. However, there is still value in regularly updating health economic models, even where decision uncertainty is low.

Model parameter uncertainty analysis for an annual field-scale P loss model

Phosphorous (P) fate and transport models are important tools for developing and evaluating conservation practices aimed at reducing P losses from agricultural fields. Because all models are simplifications of complex systems, there will exist an inherent amount of uncertainty associated with their predictions. It is therefore important that efforts be directed at identifying, quantifying, and communicating the different sources of model uncertainties. In this study, we conducted an uncertainty analysis with the Annual P Loss Estimator (APLE) model. Our analysis included calculating parameter uncertainties and confidence and prediction intervals for five internal regression equations in APLE. We also estimated uncertainties of the model input variables based on values reported in the literature. We then predicted P loss for a suite of fields under different management and climatic conditions while accounting for uncertainties in the model parameters and inputs and compared the relative contributions of these two sources of uncertainty to the overall uncertainty associated with predictions of P loss. Both the overall magnitude of the prediction uncertainties and the relative contributions of the two sources of uncertainty varied depending on management practices and field characteristics. This was due to differences in the number of model input variables and the uncertainties in the regression equations associated with each P loss pathway. Inspection of the uncertainties in the five regression equations brought attention to a previously unrecognized limitation with the equation used to partition surface-applied fertilizer P between leaching and runoff losses. As a result, an alternate equation was identified that provided similar predictions with much less uncertainty. Our results demonstrate how a thorough uncertainty and model residual analysis can be used to identify limitations with a model. Such insight can then be used to guide future data collection and model development and evaluation efforts.

Inter‐El Niño variability in CMIP5 models: Model deficiencies and future changes

AbstractInter‐El Niño variability, which represents the diversity in spatiotemporal evolution among El Niño events, has been identified using the first two leading modes of sea surface temperature anomalies along the equator. The first mode represents the transition from El Niño into La Niña and the second mode reveals El Niño's persistence through the following spring. Here we examine the ability of models to capture inter‐El Niño variability and predict future changes due to global warming using historical and Representative Concentration Pathway 8.5 simulations of phase 5 of the Coupled Model Intercomparison Project (CMIP5). Most CMIP5 models realistically reproduce the first mode, but three fifths of the models fail to capture the second mode, with considerable intermodel diversity. The 10 best models are therefore selected according to a measurement of pattern correlation coefficients and normalized root‐mean‐square errors with respect to reproduction of the second mode. The results aggregated from the best models indicate that the first “transition” mode will remain unchanged from the present climate to the future climate; in contrast, the second “persistence” mode changes stochastically across the CMIP5 models. Consequently, we conclude that El Niño's transition into La Niña is the most dominant characteristic of simulated inter‐El Niño variability and will remain unswayed under global warming conditions. Model deficiency in simulating El Niño's persistence is a key source of uncertainty in modeling inter‐El Niño variability, resulting in difficulty predicting how certain characteristics of El Niño events may change with global warming.

The impact of in-situ stress and outcrop-based fracture geometry on hydraulic aperture and upscaled permeability in fractured reservoirs

Aperture has a controlling impact on porosity and permeability and is a source of uncertainty in modeling of naturally fractured reservoirs. This uncertainty results from difficulties in accurately quantifying aperture in the subsurface and from a limited fundamental understanding of the mechanical and diagenetic processes that control aperture. In the absence of cement bridges and high pore pressure, fractures in the subsurface are generally considered to be closed. However, experimental work, outcrop analyses and subsurface data show that some fractures remain open, and that aperture varies even along a single fracture. However, most fracture flow models consider constant apertures for fractures. We create a stress-dependent heterogeneous aperture by combining Finite Element modeling of discrete fracture networks with an empirical aperture model. Using a modeling approach that considers fractures explicitly, we quantify equivalent permeability, i.e. combined matrix and stress-dependent fracture flow. Fracture networks extracted from a large outcropping pavement form the basis of these models. The results show that the angle between fracture strike and σ1 has a controlling impact on aperture and permeability, where hydraulic opening is maximum for an angle of 15°. At this angle, the fracture experiences a minor amount of shear displacement that allows the fracture to remain open even when fluid pressure is lower than the local normal stress. Averaging the heterogeneous aperture to scale up permeability probably results in an underestimation of flow, indicating the need to incorporate full aperture distributions rather than simplified aperture models in reservoir-scale flow models.

The treatment of mixing in core helium burning models – II. Constraints from cluster star counts

The treatment of convective boundaries during core helium burning is a fundamental problem in stellar evolution calculations. In Paper~I we showed that new asteroseismic observations of these stars imply they have either very large convective cores or semiconvection/partially mixed zones that trap g-modes. We probe this mixing by inferring the relative lifetimes of asymptotic giant branch (AGB) and horizontal branch (HB) from $R_2$, the observed ratio of these stars in recent HST photometry of 48 Galactic globular clusters. Our new determinations of $R_2$ are more self-consistent than those of previous studies and our overall calculation of $R_2 = 0.117 \pm 0.005$ is the most statistically robust now available. We also establish that the luminosity difference between the HB and the AGB clump is $\Delta \log{L}_\text{HB}^\text{AGB} = 0.455 \pm 0.012$. Our results accord with earlier findings that standard models predict a lower $R_2$ than is observed. We demonstrate that the dominant sources of uncertainty in models are the prescription for mixing and the stochastic effects that can result from its numerical treatment. The luminosity probability density functions that we derive from observations feature a sharp peak near the AGB clump. This constitutes a strong new argument against core breathing pulses, which broaden the predicted width of the peak. We conclude that the two mixing schemes that can match the asteroseismology are capable of matching globular cluster observations, but only if (i) core breathing pulses are avoided in models with a semiconvection/partially mixed zone, or (ii) that models with large convective cores have a particular depth of mixing beneath the Schwarzschild boundary during subsequent early-AGB `gravonuclear' convection.

Objective Calibration of Regional Climate Models: Application over Europe and North America

Abstract An important source of model uncertainty in climate models arises from unconfined model parameters in physical parameterizations. These parameters are commonly estimated on the basis of manual adjustments (expert tuning), which carries the risk of overtuning the parameters for a specific climate region or time period. This issue is particularly germane in the case of regional climate models (RCMs), which are often developed and used in one or a few geographical regions only. This study addresses the role of objective parameter calibration in this context. Using a previously developed objective calibration methodology, an RCM is calibrated over two regions (Europe and North America) and is used to investigate the transferability of the results. A total of eight different model parameters are calibrated, using a metamodel to account for parameter interactions. The study demonstrates that the calibration is effective in reducing model biases in both domains. For Europe, this concerns in particular a pronounced reduction of the summer warm bias and the associated overestimation of interannual temperature variability that have persisted through previous expert tuning efforts and are common in many global and regional climate models. The key process responsible for this improvement is an increased hydraulic conductivity. Higher hydraulic conductivity increases the water availability at the land surface and leads to increased evaporative cooling, stronger low cloud formation, and associated reduced incoming shortwave radiation. The calibrated parameter values are found to be almost identical for both domains; that is, the parameter calibration is transferable between the two regions. This is a promising result and indicates that models may be more universal than previously considered.

The role of residence time in diagnostic models of global carbon storage capacity: model decomposition based on a traceable scheme.

As a key factor that determines carbon storage capacity, residence time (τE) is not well constrained in terrestrial biosphere models. This factor is recognized as an important source of model uncertainty. In this study, to understand how τE influences terrestrial carbon storage prediction in diagnostic models, we introduced a model decomposition scheme in the Boreal Ecosystem Productivity Simulator (BEPS) and then compared it with a prognostic model. The result showed that τE ranged from 32.7 to 158.2 years. The baseline residence time (τ′E) was stable for each biome, ranging from 12 to 53.7 years for forest biomes and 4.2 to 5.3 years for non-forest biomes. The spatiotemporal variations in τE were mainly determined by the environmental scalar (ξ). By comparing models, we found that the BEPS uses a more detailed pool construction but rougher parameterization for carbon allocation and decomposition. With respect to ξ comparison, the global difference in the temperature scalar (ξt) averaged 0.045, whereas the moisture scalar (ξw) had a much larger variation, with an average of 0.312. We propose that further evaluations and improvements in τ′E and ξw predictions are essential to reduce the uncertainties in predicting carbon storage by the BEPS and similar diagnostic models.

Modeling methane and nitrous oxide emissions from direct‐seeded rice systems

AbstractProcess‐based modeling of CH4 and N2O emissions from rice fields is a practical tool for conducting greenhouse gas inventories and estimating mitigation potentials of alternative practices at the scale of management and policy making. However, the accuracy of these models in simulating CH4 and N2O emissions in direct‐seeded rice systems under various management practices remains a question. We empirically evaluated the denitrification‐decomposition model for estimating CH4 and N2O fluxes in California rice systems. Five and nine site‐year combinations were used for calibration and validation, respectively. The model was parameterized for two cultivars, M206 and Koshihikari, and able to simulate 30% and 78% of the variation in measured yields, respectively. Overall, modeled and observed seasonal CH4 emissions were similar (R2 = 0.85), but there was poor correspondence in fallow period CH4 emissions and in seasonal and fallow period N2O emissions. Furthermore, management effects on seasonal CH4 emissions were highly variable and not well represented by the model (0.2–465% absolute relative deviation). Specifically, simulated CH4 emissions were oversensitive to fertilizer N rate but lacked sensitivity to the type of seeding system (dry seeding versus water seeding) and prior fallow period straw management. Additionally, N2O emissions were oversensitive to fertilizer N rate and field drainage. Sensitivity analysis showed that CH4 emissions were highly sensitive to changes in the root to total plant biomass ratio, suggesting that it is a significant source of model uncertainty. These findings have implications for model‐directed field research that could improve model representation of paddy soils for application at larger spatial scales.

Predicting aboveground biomass with LANDIS-II: A global and temporal analysis of parameter sensitivity

Forest landscape models (FLMs) have become a valuable tool for projecting broad-scale forest dynamics, but incomplete knowledge about model behavior can make parameterization challenging and outcomes unreliable. FLMs generally model forest growth as a set of interacting processes, and, consequently, predictions can be influenced by process or parameter uncertainty. A sensitivity analysis can potentially help identify sources of uncertainty, but if it does not use global measures of sensitivity nor consider that sensitivity in a process-based model is likely time-dependent, results could be misleading. Our aim was to evaluate the sensitivity of nine key parameters when predicting live aboveground biomass (AGB) with the widely used FLM, LANDIS-II. To fully explore parameter interactions and nonlinear model behavior, we selected a range of parameter values based on LANDIS-II applications in North America that was considerably wider than in previous local sensitivity analyses. Our results showed commonalities with previous studies, which concluded the maximum allowable biomass and maximum annual net primary productivity specified for a species were most influential when predicting AGB. In contrast to earlier work, we also clearly demonstrated how relative importance was time-dependent for all but the least important parameters. Interactions between parameters and with simulation duration generated substantial variability in AGB and number of cohorts established. Results will improve future calibration efforts and may offer insight into opportunities for possible model refinements. This study also suggests, however, that parameters which cannot be calibrated based on empirical data will continue to be a major source of model uncertainty.