Sources Of Uncertainty In Models Research Articles

Uncertainty Management in Computational Simulations of Medical Devices1

Uncertainty Management in Computational Simulations of Medical Devices1

Effects of field plot size on prediction accuracy of aboveground biomass in airborne laser scanning-assisted inventories in tropical rain forests of Tanzania.

BackgroundAirborne laser scanning (ALS) has recently emerged as a promising tool to acquire auxiliary information for improving aboveground biomass (AGB) estimation in sample-based forest inventories. Under design-based and model-assisted inferential frameworks, the estimation relies on a model that relates the auxiliary ALS metrics to AGB estimated on ground plots. The size of the field plots has been identified as one source of model uncertainty because of the so-called boundary effects which increases with decreasing plot size. Recent research in tropical forests has aimed to quantify the boundary effects on model prediction accuracy, but evidence of the consequences for the final AGB estimates is lacking. In this study we analyzed the effect of field plot size on model prediction accuracy and its implication when used in a model-assisted inferential framework.ResultsThe results showed that the prediction accuracy of the model improved as the plot size increased. The adjusted R2 increased from 0.35 to 0.74 while the relative root mean square error decreased from 63.6 to 29.2%. Indicators of boundary effects were identified and confirmed to have significant effects on the model residuals. Variance estimates of model-assisted mean AGB relative to corresponding variance estimates of pure field-based AGB, decreased with increasing plot size in the range from 200 to 3000 m2. The variance ratio of field-based estimates relative to model-assisted variance ranged from 1.7 to 7.7.ConclusionsThis study showed that the relative improvement in precision of AGB estimation when increasing field-plot size, was greater for an ALS-assisted inventory compared to that of a pure field-based inventory.

A unified approach for process‐based hydrologic modeling: 1. Modeling concept

AbstractThis work advances a unified approach to process‐based hydrologic modeling to enable controlled and systematic evaluation of multiple model representations (hypotheses) of hydrologic processes and scaling behavior. Our approach, which we term the Structure for Unifying Multiple Modeling Alternatives (SUMMA), formulates a general set of conservation equations, providing the flexibility to experiment with different spatial representations, different flux parameterizations, different model parameter values, and different time stepping schemes. In this paper, we introduce the general approach used in SUMMA, detailing the spatial organization and model simplifications, and how different representations of multiple physical processes can be combined within a single modeling framework. We discuss how SUMMA can be used to systematically pursue the method of multiple working hypotheses in hydrology. In particular, we discuss how SUMMA can help tackle major hydrologic modeling challenges, including defining the appropriate complexity of a model, selecting among competing flux parameterizations, representing spatial variability across a hierarchy of scales, identifying potential improvements in computational efficiency and numerical accuracy as part of the numerical solver, and improving understanding of the various sources of model uncertainty.

Model-data assimilation of multiple phenological observations to constrain and predict leaf area index.

Our limited ability to accurately simulate leaf phenology is a leading source of uncertainty in models of ecosystem carbon cycling. We evaluate if continuously updating canopy state variables with observations is beneficial for predicting phenological events. We employed ensemble adjustment Kalman filter (EAKF) to update predictions of leaf area index (LAI) and leaf extension using tower-based photosynthetically active radiation (PAR) and moderate resolution imaging spectrometer (MODIS) data for 2002-2005 at Willow Creek, Wisconsin, USA, a mature, even-aged, northern hardwood, deciduous forest. The ecosystem demography model version 2 (ED2) was used as the prediction model, forced by offline climate data. EAKF successfully incorporated information from both the observations and model predictions weighted by their respective uncertainties. The resulting. estimate reproduced the observed leaf phenological cycle in the spring and the fall better than a parametric model prediction. These results indicate that during spring the observations contribute most in determining the correct bud-burst date, after which the model performs well, but accurately modeling fall leaf senesce requires continuous model updating from observations. While the predicted net ecosystem exchange (NEE) of CO2 precedes tower observations and unassimilated model predictions in the spring, overall the prediction follows observed NEE better than the model alone. Our results show state data assimilation successfully simulates the evolution of plant leaf phenology and improves model predictions of forest NEE.

Lagrangian Mixing Models for Turbulent Combustion: Review and Prospects

Compared to other simulation approaches utilized for modeling turbulent reacting flows with detailed chemistry, probability density function (PDF) methods offer several advantages. This is because the changes in fluid composition due to convection and reaction processes can be treated exactly. PDF methods require however closure models for the mixing process representing the transport of the PDF owing to molecular diffusion. Over the years several mixing models with different degrees of complexity have been developed. A review of the main Lagrangian mixing models for turbulent combustion developed so far is presented in this work. This review includes models where the composition of a particle changes (i) essentially independently of the composition associated with the other particles, and (ii) through direct interaction with other particles. The main advantages and shortcomings of the mixing models reviewed are highlighted accordingly. Because Lagrangian Monte Carlo techniques are usually used for solving PDF transport equations, a particular emphasis is put on their corresponding particle implementation. The mixing models review is preceded by a section highlighting the mathematical formulation associated with the use of PDF methods for turbulent reacting flows. In the last part of the article both comparative results of mixing models performance and prospects for the mixing models are discussed. Despite the effort that has been devoted to the development of more capable mixing models, currently there is no mixing model presenting all desirable characteristics. Even more, there are significant differences in the results obtained when different mixing models are utilized, especially when higher order scalar statistics are accounted for. Therefore work still needs to be carried out in order to develop a mixing model satisfying all desirable characteristics expected from these models. Several avenues can be further explored in order to achieve this goal. These potential routes include those accounting for spatial scalar structures and both scalar length and turbulent frequency scales distributions. Other approaches based on competitive mixing, manifold-based features and Lagrangian coherent structures have also the potential to further improve upon existing mixing models. The development of a sound mixing model will allow eventually removing one of the largest sources of modeling uncertainty in PDF-based computations.

Regional modeling of surface‐atmosphere interactions and their impact on Great Lakes hydroclimate

AbstractLand and water surfaces play a critical role in hydroclimate by supplying moisture to the atmosphere, yet the ability of climate models to capture their feedbacks with the atmosphere relative to large‐scale transport is uncertain. To assess these land‐lake‐atmosphere feedbacks, we compare the controls on atmospheric moisture simulated by a regional climate model (RegCM) with observations and reanalysis products for the Great Lakes region. Three 23 year simulations, driven by one reanalysis product and two general circulation models, are performed. RegCM simulates wetter winters and drier summers than observed by up to 31 and 21%, respectively. Moisture advection exhibits similar biases, suggesting the contribution of external sources. Land surface fluxes account for nearly one third of summer precipitation according to two reanalysis products. RegCM underestimates reanalysis evapotranspiration by nearly 50%; however, the reanalyses overestimate measurements at three FLUXNET sites by up to a factor of 2, which may explain the model‐reanalysis differences. Neither RegCM nor the reanalyses capture the spatial variability in land evapotranspiration observed across the three FLUXNET sites, indicating a source of model uncertainty. In addition, RegCM underestimates the observed evapotranspiration response to its atmospheric drivers such as vapor pressure deficit and temperature. Over the lakes, one model member overestimates convective precipitation caused by enhanced evaporation under warm lake surface temperatures, highlighting the need for accurate representation of lake temperature in the surface boundary condition. We conclude that climate models, including those driving reanalyses, underestimate the observed surface‐atmosphere feedbacks and their influence on regional hydroclimate.

A sensitivity study of the WRF-Chem ozone forecast for Slovenia

Extensive efforts have been devoted in recent years to develop an air quality forecasting system for Slovenia. In the present study, the results of an experimental real-time WRF-Chem forecast are evaluated for the summertime conditions, when ozone exceedances were measured during three heat wave events. The main focus of the paper is the assessment of model sensitivity to the domain setup and anthropogenic emission dataset used for Slovenia. To reduce model errors due to these two important sources of model uncertainties, significant efforts are needed either by increasing the computational power (domain setup and resolution) or by constructing an improved emission dataset based on detailed information about emissions, often deficiently available. In the study, the impact of these two sources of model uncertainties on ozone daily maximum predictions over a geographically complex area is explored.

Role of Surface Energy Exchange for Simulating Wind Turbine Inflow: A Case Study in the Southern Great Plains, USA

The Weather Research and Forecasting (WRF) model is used to investigate choice of land surface model (LSM) on the near surface wind profile, including heights reached by multi-megawatt (MW) wind turbines. Simulations of wind profiles and surface energy fluxes were made using five LSMs of varying degrees of sophistication in dealing with soil–plant–atmosphere feedbacks for the Department of Energy (DOE) Southern Great Plains (SGP) Atmospheric Radiation Measurement Program (ARM) Central Facility in Oklahoma, USA. Surface flux and wind profile measurements were available for validation. WRF was run for three, two-week periods covering varying canopy and meteorological conditions. The LSMs predicted a wide range of energy flux and wind shear magnitudes even during the cool autumn period when we expected less variability. Simulations of energy fluxes varied in accuracy by model sophistication, whereby LSMs with very simple or no soil–plant–atmosphere feedbacks were the least accurate; however, the most complex models did not consistently produce more accurate results. Errors in wind shear were also sensitive to LSM choice and were partially related to energy flux accuracy. The variability of LSM performance was relatively high suggesting that LSM representation of energy fluxes in WRF remains a large source of model uncertainty for simulating wind turbine inflow conditions.

Improving ecosystem productivity modeling through spatially explicit estimation of optimal light use efficiency

A common assumption of remote sensing-based light use efficiency (LUE) models for estimating vegetation gross primary productivity (GPP) is that plants in a biome matrix operate at their photosynthetic capacity under optimal climatic conditions. A prescribed constant biome maximum light use efficiency parameter (LUE max ) defines the maximum photosynthetic carbon conversion rate under these conditions and is a large source of model uncertainty. Here we used tower eddy covariance measurement-based carbon (CO 2 ) fluxes for spatial estimation of optimal LUE (LUE opt ) across North America. LUE opt was estimated at 62 Flux Network sites using tower daily carbon fluxes and meteorology, and satellite observed fractional photosynthetically active radiation from the Moderate Resolution Imaging Spectroradiometer. Ageostatistical model was fitted to 45 flux tower-derived LUE opt data points using independent geospatial environmental variables, including global plant traits, soil moisture, terrain aspect, land cover type, and percent tree cover, and validated at 17 independent tower sites. Estimated LUE opt shows large spatial variability within and among different land cover classes indicated from the sparse tower network. Leaf nitrogen content and soil moisture regime are major factors explaining LUE opt patterns. GPP derived from estimated LUE opt shows significant correlation improvement against tower GPP records (R2 = 76.9%; mean root-mean-square error (RMSE) = 257gCm-2yr-1), relative to alternative GPP estimates derived using biome-specific LUE max constants (R2 = 34.0%; RMSE = 439gCm-2yr-1). GPP determined from the LUE opt map also explains a 49.4% greater proportion of tower GPP variability at the independent validation sites and shows promise for improving understanding of LUE patterns and environmental controls and enhancing regional GPP monitoring from satellite remote sensing.

Multi-scale spatial sensitivity analysis of a model for economic appraisal of flood risk management policies

We demonstrate the use of sensitivity analysis to rank sources of uncertainty in models for economic appraisal of flood risk management policies, taking into account spatial scale issues. A methodology of multi-scale variance-based global sensitivity analysis is developed, and illustrated on the NOE model on the Orb River, France. The variability of the amount of expected annual flood avoided damages, and the associated sensitivity indices, are estimated over different spatial supports, ranging from small cells to the entire floodplain. Both uncertainty maps and sensitivity maps are produced to identify the key input variables in the NOE model at different spatial scales. Our results show that on small spatial supports, variance of the output indicator is mainly due to the water depth maps and the assets map (spatially distributed model inputs), while on large spatial supports, it is mainly due to the flood frequencies and depth–damage curves (non spatial inputs).

Assessing parametrization uncertainty associated with horizontal resolution in numerical weather prediction models.

The need to represent uncertainty resulting from model error in ensemble weather prediction systems has spawned a variety of ad hoc stochastic algorithms based on plausible assumptions about sub-grid-scale variability. Currently, few studies have been carried out to prove the veracity of such schemes and it seems likely that some implementations of stochastic parametrization are misrepresentations of the true source of model uncertainty. This paper describes an attempt to quantify the uncertainty in physical parametrization tendencies in the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecasting System with respect to horizontal resolution deficiency. High-resolution truth forecasts are compared with matching target forecasts at much lower resolution after coarse-graining to a common spatial and temporal resolution. In this way, model error is defined and its probability distribution function is examined as a function of tendency magnitude. It is found that the temperature tendency error associated with convection parametrization and explicit water phase changes behaves like a Poisson process for which the variance grows in proportion to the mean, which suggests that the assumptions underpinning the Craig and Cohen statistical model of convection might also apply to parametrized convection. By contrast, radiation temperature tendency errors have a very different relationship to their mean value. These findings suggest that the ECMWF stochastic perturbed parametrization tendency scheme could be improved since it assumes that the standard deviation of the tendency error is proportional to the mean. Using our finding that the variance error is proportional to the mean, a prototype stochastic parametrization scheme is devised for convective and large-scale condensation temperature tendencies and tested within the ECMWF Ensemble Prediction System. Significant impact on forecast skill is shown, implying its potential for further development.

A probabilistic model of gastroenteritis risks associated with consumption of street food salads in Kumasi, Ghana: Evaluation of methods to estimate pathogen dose from water, produce or food quality

With a rapidly growing urban population in Kumasi, Ghana, the consumption of street food is increasing. Raw salads, which often accompany street food dishes, are typically composed of perishable vegetables that are grown in close proximity to the city using poor quality water for irrigation. This study assessed the risk of gastroenteritis illness (caused by rotavirus, norovirus and Ascaris lumbricoides) associated with the consumption of street food salads using Quantitative Microbial Risk Assessment (QMRA). Three different risk assessment models were constructed, based on availability of microbial concentrations: 1) Water — starting from irrigation water quality, 2) Produce — starting from the quality of produce at market, and 3) Street — using microbial quality of street food salad. In the absence of viral concentrations, published ratios between faecal coliforms and viruses were used to estimate the quality of water, produce and salad, and annual disease burdens were determined. Rotavirus dominated the estimates of annual disease burden (~10−3Disability Adjusted Life Years per person per year (DALYs pppy)), although norovirus also exceeded the 10−4DALY threshold for both Produce and Street models. The Water model ignored other on-farm and post-harvest sources of contamination and consistently produced lower estimates of risk; it likely underestimates disease burden and therefore is not recommended. Required log reductions of up to 5.3 (95th percentile) for rotavirus were estimated for the Street model, demonstrating that significant interventions are required to protect the health and safety of street food consumers in Kumasi. Estimates of virus concentrations were a significant source of model uncertainty and more data on pathogen concentrations is needed to refine QMRA estimates of disease burden.

Sources of uncertainty in projections of twenty-first century westerly wind changes over the Amundsen Sea, West Antarctica, in CMIP5 climate models

The influence of changes in winds over the Amundsen Sea has been shown to be a potentially key mechanism in explaining rapid loss of ice from major glaciers in West Antarctica, which is having a significant impact on global sea level. Here, Coupled Model Intercomparison Project Phase 5 (CMIP5) climate model data are used to assess twenty-first century projections in westerly winds over the Amundsen Sea (U AS ). The importance of model uncertainty and internal climate variability in RCP4.5 and RCP8.5 scenario projections are quantified and potential sources of model uncertainty are considered. For the decade 2090–2099 the CMIP5 models show an ensemble mean twenty-first century response in annual mean U AS of 0.3 and 0.7 m s−1 following the RCP4.5 and RCP8.5 scenarios respectively. However, as a consequence of large internal climate variability over the Amundsen Sea, it takes until around 2030 (2065) for the RCP8.5 response to exceed one (two) standard deviation(s) of decadal internal variability. In all scenarios and seasons the model uncertainty is large. However the present-day climatological zonal wind bias over the whole South Pacific, which is important for tropical teleconnections, is strongly related to inter-model differences in projected change in U AS (more skilful models show larger U AS increases). This relationship is significant in winter (r = −0.56) and spring (r = −0.65), when the influence of the tropics on the Amundsen Sea region is known to be important. Horizontal grid spacing and present day sea ice extent are not significant sources of inter-model spread.

Severe accident source terms for a sodium-cooled fast reactor

In order to support the demonstration of a risk-informed approach to the design optimization of a sodium-cooled fast reactor (SFR), it was necessary to make realistic estimates of the consequences of severe accident scenarios. This paper describes the database, models, and assumptions used to estimate the offsite consequences of characteristic severe accident scenarios. As required for comparison with the NRC’s technology neutral framework limit curve, the offsite dose at one mile from the plant boundary is calculated using conservative meteorology.The reference plant design is a 1000MWt pool-type design with metallic fuel. Because an integrated analysis tool comparable to MELCOR does not exist for SFR accident scenario analysis, it was necessary to write a computer code that would assess release of radionuclides from the fuel and transport within the reactor primary system and to link those analyses with results from existing computer codes that assess the dynamic response of the reactor, containment thermal–hydraulics, and radionuclide transport processes within the containment. The analyses indicate that the offsite source terms for SFR severe accident scenarios tend to be small because of the low melting temperature of the fuel, likelihood of significant retention of fission products within the sodium pool, augmentation of containment deposition processes by interaction with sodium oxide aerosols, and small driving force for release from the containment to the environment. A number of major sources of modeling uncertainty are identified as requiring further development effort. An integrated modeling capability, similar to the MELCOR code, is required to direct further research and ultimately to support the licensing of an SFR.

An investigation of mixing in a three-stream turbulent jet

Mixing models are required to account for the effects of molecular diffusion in probability density function (PDF) methods for reacting flows, and these models entail one of the greatest sources of modeling uncertainty in PDF calculations. In this work, we perform PDF calculations of a turbulent coaxial jet using both Reynolds-Averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES) based methods. The configuration of the coaxial jet comes from a recently developed experiment, in which the conditional diffusion and scalar dissipation rate of two conserved scalars are measured. These measurements present the unique opportunity to compare directly the conditional diffusion implied by PDF mixing models to the measured conditional diffusion. In the RANS-PDF calculations, three classic mixing models are evaluated: Interaction by Exchange with the Mean (IEM), Modified Curl (MC), and Euclidean Minimum Spanning Tree (EMST). Though calculations with all three mixing models yield time-averaged mean and RMS statistics of the two scalars in good agreement with the experimental measurements, the joint PDFs of the two scalars show a wide variability from the measurements. The EMST mixing model is unique in that it produces compositions which lie very close to the slow manifold identified in the experimental work. Next, LES calculations of the same flow are performed, and the dissipation rate and conditional diffusion are calculated. The resolved scalar dissipation rate is found to be in good agreement with the experimental data, but depends strongly on the resolution; the total dissipation rate from the RANS-PDF and LES calculations indicates significantly larger scalar dissipation rates than those measured experimentally. Finally, LES-PDF calculations of the same flow yield joint PDFs in very good agreement with the experimental data. The attenuation of variance production model is introduced and tested in this flow: this new model is found to yield calculations of improved accuracy in LES-PDF calculations on coarse grids.

The magnitude and causes of uncertainty in global model simulations of cloud condensation nuclei

Abstract. Aerosol–cloud interaction effects are a major source of uncertainty in climate models so it is important to quantify the sources of uncertainty and thereby direct research efforts. However, the computational expense of global aerosol models has prevented a full statistical analysis of their outputs. Here we perform a variance-based analysis of a global 3-D aerosol microphysics model to quantify the magnitude and leading causes of parametric uncertainty in model-estimated present-day concentrations of cloud condensation nuclei (CCN). Twenty-eight model parameters covering essentially all important aerosol processes, emissions and representation of aerosol size distributions were defined based on expert elicitation. An uncertainty analysis was then performed based on a Monte Carlo-type sampling of an emulator built for each model grid cell. The standard deviation around the mean CCN varies globally between about ±30% over some marine regions to ±40–100% over most land areas and high latitudes, implying that aerosol processes and emissions are likely to be a significant source of uncertainty in model simulations of aerosol–cloud effects on climate. Among the most important contributors to CCN uncertainty are the sizes of emitted primary particles, including carbonaceous combustion particles from wildfires, biomass burning and fossil fuel use, as well as sulfate particles formed on sub-grid scales. Emissions of carbonaceous combustion particles affect CCN uncertainty more than sulfur emissions. Aerosol emission-related parameters dominate the uncertainty close to sources, while uncertainty in aerosol microphysical processes becomes increasingly important in remote regions, being dominated by deposition and aerosol sulfate formation during cloud-processing. The results lead to several recommendations for research that would result in improved modelling of cloud–active aerosol on a global scale.

Sources of uncertainty in modeling the glaciohydrological response of a Karakoram watershed to climate change

[1] In the headwater catchments of the main Asian rivers, glaciohydrological models are a useful tool to anticipate impacts of climatic changes. However, the reliability of their projections strongly depends on the quality and quantity of data that are available for parameter estimation, model calibration and validation, as well as on the accuracy of climate change projections. In this study the physically oriented, glaciohydrological model TOPKAPI-ETH is used to simulate future changes in snow, glacier, and runoff from the Hunza River Basin in northern Pakistan. Three key sources of model uncertainty in future runoff projections are compared: model parameters, climate projections, and natural climate variability. A novel approach, applicable also to ungauged catchments, is used to determine which model parameters and model components significantly affect the overall model uncertainty. We show that the model is capable of reproducing streamflow and glacier mass balances, but that all analyzed sources of uncertainty significantly affect the reliability of future projections, and that their effect is variable in time and in space. The effect of parametric uncertainty often exceeds the impact of climate uncertainty and natural climate variability, especially in heavily glacierized subcatchments. The results of the uncertainty analysis allow detailed recommendations on network design and the timing and location of field measurements, which could efficiently help to reduce model uncertainty in the future.

Interior Least Tern (Sternula antillarum) breeding distribution and ecology: implications for population‐level studies and the evaluation of alternative management strategies on large, regulated rivers

Interior Least Terns (Sternula antillarum) (ILT) are colonial, fish-eating birds that breed within active channels of large sand bed rivers of the Great Plains and in the Lower Mississippi Valley. Multipurpose dams, irrigation structures, and engineered navigation systems have been present on these rivers for many decades. Despite severe alteration of channels and flow regimes, regulation era floods have remained effective at maintaining bare sandbar nesting habitat on many river segments and ILT populations have been stable or expanding since they were listed as endangered in 1985. We used ILT breeding colony locations from 2002 to 2012 and dispersal information to identify 16 populations and 48 subpopulations. More than 90% of ILT and >83% of river km with suitable nesting habitat occur within the two largest populations. However, replicate populations remain throughout the entire historical, geophysical, and ecological range of ILT. Rapid colonization of anthropogenic habitats in areas that were not historically occupied suggests metapopulation dynamics. The highest likelihood of demographic connectivity among ILT populations occurs across the Southern Plains and the Lower Mississippi River, which may be demographically connected with Least Tern populations on the Gulf Coast. Paired ecological and bird population models are needed to test whether previously articulated threats limit ILT population growth and to determine if management intervention is necessary and where. Given current knowledge, the largest sources of model uncertainty will be: (1) uncertainty in relationships between high flow events and subsequent sandbar characteristics and (2) uncertainty regarding the frequency of dispersal among population subunits. We recommend research strategies to reduce these uncertainties.

Bias-corrected short-range Member-to-Member ensemble forecasts of reservoir inflow

• We develop a short-term ensemble reservoir inflow forecasting system. • The ensemble attempts to sample all sources of uncertainty in the modeling chain. • Increasing the diversity of the ensemble greatly improves ensemble quality. • Bias correction of ensemble members offers significant additional improvement. • For the flashy case study basin, only recent errors are important in bias correction. A Member-to-Member ensemble forecasting system is developed for inflows to hydroelectric reservoirs that incorporates multiple numerical weather prediction models and multiple distributed hydrological models linked by a variety of downscaling schemes. Each hydrological model uses multiple differently-optimized parameter sets and begins each daily forecast from several different initial conditions. The ensemble thereby attempts to sample all sources of error in the modeling chain. The importance of sampling all sources of error is illustrated by comparing this ensemble with an ensemble comprised of single ‘best’ parameterization for each hydrological model. Degree-of-mass-balance bias correction schemes trained using data windows of varying lengths are applied to the individual ensemble members. Based on examination of various verification metrics, we determine that a bias corrector that uses a linearly-weighted combination of past errors calculated over a three-day moving window is able to significantly improve forecast quality for the flashy case study watershed in southwestern British Columbia, Canada. Incorporation of all sources of modeling uncertainty is found to greatly improve ensemble resolution and discrimination. The full potential for these improvements using ensembles is only realized after removal of bias.

Future Arctic Ocean primary productivity from CMIP5 simulations: Uncertain outcome, but consistent mechanisms

Net Arctic Ocean primary production (PP) is expected to increase over this century, due to less perennial sea ice and more available light, but could decrease depending on changes in nitrate (NO3) supply. Here Coupled Model Intercomparison Project Phase 5 simulations performed with 11 Earth System Models are analyzed in terms of PP, surface NO3, and sea ice coverage over 1900–2100. Whereas the mean model simulates reasonably well Arctic‐integrated PP (511 TgC/yr, 1998–2005) and projects a mild 58 TgC/yr increase by 2080–2099 for the strongest climate change scenario, models do not agree on the sign of future PP change. However, similar mechanisms operate in all models. The perennial ice loss‐driven increase in PP is in most models NO3‐limited. The Arctic surface NO3 is decreasing over the 21st century (−2.3 ± 1 mmol/m3), associated with shoaling mixed layer and with decreasing NO3 in the nearby North Atlantic and Pacific waters. However, the intermodel spread in the degree of NO3 limitation is initially high, resulting from >1000 year spin‐up simulations. This initial NO3 spread, combined with the trend, causes a large variation in the timing of oligotrophy onset—which directly controls the sign of future PP change. Virtually all models agree in the open ocean zones on more spatially integrated PP and less PP per unit area. The source of model uncertainty is located in the sea ice zone, where a subtle balance between light and nutrient limitations determines the PP change. Hence, it is argued that reducing uncertainty on present Arctic NO3 in the sea ice zone would render Arctic PP projections much more consistent.