Calibration and evaluation of individual-based models using Approximate Bayesian Computation

This paper investigates the feasibility of using approximate Bayesian computation (ABC) to calibrate and evaluate complex individual-based models (IBMs). As ABC evolves, various versions are emerging, but here we only explore the most accessible version, rejection-ABC. Rejection-ABC involves running models a large number of times, with parameters drawn randomly from their prior distributions, and then retaining the simulations closest to the observations. Although well-established in some fields, whether ABC will work with ecological IBMs is still uncertain. Rejection-ABC was applied to an existing 14-parameter earthworm energy budget IBM for which the available data consist of body mass growth and cocoon production in four experiments. ABC was able to narrow the posterior distributions of seven parameters, estimating credible intervals for each. ABC’s accepted values produced slightly better fits than literature values do. The accuracy of the analysis was assessed using cross-validation and coverage, currently the best available tests. Of the seven unnarrowed parameters, ABC revealed that three were correlated with other parameters, while the remaining four were found to be not estimable given the data available. It is often desirable to compare models to see whether all component modules are necessary. Here we used ABC model selection to compare the full model with a simplified version which removed the earthworm’s movement and much of the energy budget. We are able to show that inclusion of the energy budget is necessary for a good fit to the data. We show how our methodology can inform future modelling cycles, and briefly discuss how more advanced versions of ABC may be applicable to IBMs. We conclude that ABC has the potential to represent uncertainty in model structure, parameters and predictions, and to embed the often complex process of optimizing an IBM’s structure and parameters within an established statistical framework, thereby making the process more transparent and objective.

Efficient incorporation of channel cross-section geometry uncertainty into regional and global scale flood inundation models

This paper investigates the challenge of representing structural differences in river channel cross-section geometry for regional to global scale river hydraulic models and the effect this can have on simulations of wave dynamics. Classically, channel geometry is defined using data, yet at larger scales the necessary information and model structures do not exist to take this approach. We therefore propose a fundamentally different approach where the structural uncertainty in channel geometry is represented using a simple parameterization, which could then be estimated through calibration or data assimilation. This paper first outlines the development of a computationally efficient numerical scheme to represent generalised channel shapes using a single parameter, which is then validated using a simple straight channel test case and shown to predict wetted perimeter to within 2% for the channels tested. An application to the River Severn, UK is also presented, along with an analysis of model sensitivity to channel shape, depth and friction. The channel shape parameter was shown to improve model simulations of river level, particularly for more physically plausible channel roughness and depth parameter ranges. Calibrating channel Manning’s coefficient in a rectangular channel provided similar water level simulation accuracy in terms of Nash-Sutcliffe efficiency to a model where friction and shape or depth were calibrated. However, the calibrated Manning coefficient in the rectangular channel model was ~2/3 greater than the likely physically realistic value for this reach and this erroneously slowed wave propagation times through the reach by several hours. Therefore, for large scale models applied in data sparse areas, calibrating channel depth and/or shape may be preferable to assuming a rectangular geometry and calibrating friction alone.

Optimizing mixing during the sponge cake manufacturing process

Sponge cakes have traditionally been manufactured using multistage mixing methods to enhance potential foam formation by the eggs. Today, use of all-in (single-stage) mixing methods is superseding multistage methods for large-scale batter preparation to reduce costs and production time. In this study, multistage and all-in mixing procedures and three final high-speed mixing times (3, 5, and 15 min) for sponge cake production were tested to optimize a mixing method for pilot-scale research. Mixing for 3 min produced batters with higher relative density values than did longer mixing times. These batters generated well-aerated cakes with high volume and low hardness. In contrast, after 5 and 15 min of high-speed mixing, batters with lower relative density and higher viscosity values were produced. Although higher bubble incorporation and retention were observed, longer mixing times produced better developed gluten networks, which stiffened the batters and inhibited bubble expansion during mixing. As a result, these batters did not expand properly and produced cakes with low volume, dense crumb, and high hardness values. Results for all-in mixing were similar to those for the multistage mixing procedure in terms of the physical properties of batters and cakes (i.e., relative density, elastic moduli, volume, total cell area, hardness, etc.). These results suggest the all-in mixing procedure with a final high-speed mixing time of 3 min is an appropriate mixing method for pilot-scale sponge cake production. The advantages of this method are reduced energy costs and production time.

Nitrogen dynamics under Lolium perenne after a single application of different sewage sludge types from the same treatment stream

Three sludge types from the same treatment stream (undigested liquid, anaerobically digested liquid and dewatered, anaerobically digested cake) were used in a field based tub study. Amendments (4, 8, and 16 Mg dry solid (ds)ha(-1)) were incorporated into the upper 15 cm of a sandy loam soil prior to sowing with rye-grass (Lolium perenne L.). Nitrogen transformations in the soil were determined for the 80 d period following incorporation. Nitrogen uptake and crop yield were measured in the cut sward 35 and 70 d after sowing. The study showed that application of sewage sludge at rates as low as 4 Mgha(-1) can have a nutritional benefit to rye-grass over the two harvests. Differences in N transformation, and hence crop nutritional benefit, between sludge types were evident throughout the experiment. In particular, the dewatering process changed the mineral N characteristics of the anaerobically digested sludge, which, when not dewatered, outperformed the other sludges in terms of yield and mineralisation rate at both harvests. The dewatered sludge produced the lowest yield of rye-grass. The undigested liquid sludge had the lowest foliar N and soil NO(3)-N concentrations, possibly immobilised as the large oxidisable C component of this sludge was metabolised by the microbial biomass. Correlation data support the concept of preferential uptake of NH(4)-N over NO(3)-N in Lolium perenne. Results are discussed in the context of managing sludge type and application for a plant nutrient source and NO(3)-N release.

Assessment and improvement of biotransfer models to cow’s milk and beef used in exposure assessment tools for organic pollutants

The aim of this study was to assess and improve the accuracy of biotransfer models for the organic pollutants (PCBs, PCDD/Fs, PBDEs, PFCAs, and pesticides) into cow’s milk and beef used in human exposure assessment. Metabolic rate in cattle is known as a key parameter for this biotransfer, however few experimental data and no simulation methods are currently available. In this research, metabolic rate was estimated using existing QSAR biodegradation models of microorganisms (BioWIN) and fish (EPI-HL and IFS-HL). This simulated metabolic rate was then incorporated into the mechanistic cattle biotransfer models (RAIDAR, ACC-HUMAN, OMEGA, and CKow). The goodness of fit tests showed that RAIDAR, ACC-HUMAN, OMEGA model performances were significantly improved using either of the QSARs when comparing the new model outputs to observed data. The CKow model is the only one that separates the processes in the gut and liver. This model showed the lowest residual error of all the models tested when the BioWIN model was used to represent the ruminant metabolic process in the gut and the two fish QSARs were used to represent the metabolic process in the liver. Our testing included EUSES and CalTOX which are KOW-regression models that are widely used in regulatory assessment. New regressions based on the simulated rate of the two metabolic processes are also proposed as an alternative to KOW-regression models for a screening risk assessment. The modified CKow model is more physiologically realistic, but has equivalent usability to existing KOW-regression models for estimating cattle biotransfer of organic pollutants.

Coming of age in L3 initial stages transfer models: deriving developmental predictions and looking towards the future

Aims: Over the past decade in particular, formal linguistic work within L3 acquisition has concentrated on hypothesizing and empirically determining the source of transfer from previous languages—L1, L2 or both—in L3 grammatical representations. In view of the progressive concern with more advanced stages, we aim to show that focusing on L3 initial stages should be one continued priority of the field, even—or especially—if the field is ready to shift towards modeling L3 development and ultimate attainment. Approach: We argue that L3 learnability is significantly impacted by initial stages transfer, as such forms the basis of the initial L3 interlanguage. To illustrate our point, the insights from studies using initial and intermediary stages L3 data are discussed in light of developmental predictions that derive from the initial stages models. Conclusions: Despite a shared desire to understand the process of L3 acquisition in whole, inclusive of offering developmental L3 theories, we argue that the field does not yet have—although is ever closer to—the data basis needed to effectively do so. Originality: This article seeks to convince the readership for the need of conservatism in L3 acquisition theory building, whereby offering a framework on how and why we can most effectively build on the accumulated knowledge of the L3 initial stages in order to make significant, steady progress. Significance: The arguments exposed here are meant to provide an epistemological base for a tenable framework of formal approaches to L3 interlanguage development and, eventually, ultimate attainment.

Comparison of aerosol optical properties above clouds between POLDER and AeroCom models over the South East Atlantic Ocean during the fire season

Aerosol properties above clouds have been retrieved over the South East Atlantic Ocean during the fire season 2006 using satellite observations from POLDER (Polarization and Directionality of Earth Reflectances). From June to October, POLDER has observed a mean Above-Cloud Aerosol Optical Thickness (ACAOT) of 0.28 and a mean Above-Clouds Single Scattering Albedo (ACSSA) of 0.87 at 550 nm. These results have been used to evaluate the simulation of aerosols above clouds in 5 AeroCom (Aerosol Comparisons between Observations and Models) models (GOCART, HadGEM3, ECHAM5-HAM2, OsloCTM2 and SPRINTARS). Most models do not reproduce the observed large aerosol load episodes. The comparison highlights the importance of the injection height and the vertical transport parameterizations to simulate the large ACAOT observed by POLDER. Furthermore, POLDER ACSSA is best reproduced by models with a high imaginary part of black carbon refractive index, in accordance with recent recommendations.

A process-based evaluation of dust-emitting winds in the CMIP5 simulation of HadGEM2-ES

Despite the importance of dust aerosol in the Earth system, state-of-the-art models show a large variety for North African dust emission. This study presents a systematic evaluation of dust emitting-winds in 30 years of the historical model simulation with the UK Met Office Earth-system model HadGEM2-ES for the Coupled Model Intercomparison Project Phase 5. Isolating the effect of winds on dust emission and using an automated detection for nocturnal low-level jets (NLLJs) allow an in-depth evaluation of the model performance for dust emission from a meteorological perspective. The findings highlight that NLLJs are a key driver for dust emission in HadGEM2-ES in terms of occurrence frequency and strength. The annually and spatially averaged occurrence frequency of NLLJs is similar in HadGEM2-ES and ERA-Interim from the European Centre for Medium-Range Weather Forecasts. Compared to ERA-Interim, a stronger pressure ridge over northern Africa in winter and the southward displaced heat low in summer result in differences in location and strength of NLLJs. Particularly the larger geostrophic winds associated with the stronger ridge have a strengthening effect on NLLJs over parts of West Africa in winter. Stronger NLLJs in summer may rather result from an artificially increased mixing coefficient under stable stratification that is weaker in HadGEM2-ES. NLLJs in the Bodélé Depression are affected by stronger synoptic-scale pressure gradients in HadGEM2-ES. Wintertime geostrophic winds can even be so strong that the associated vertical wind shear prevents the formation of NLLJs. These results call for further model improvements in the synoptic-scale dynamics and the physical parametrization of the nocturnal stable boundary layer to better represent dust-emitting processes in the atmospheric model. The new approach could be used for identifying systematic behavior in other models with respect to meteorological processes for dust emission. This would help to improve dust emission simulations and contribute to decreasing the currently large uncertainty in climate change projections with respect to dust aerosol.

What controls the vertical distribution of aerosol? Relationships between process sensitivity in HadGEM3–UKCA and inter-model variation from AeroCom Phase II

The vertical profile of aerosol is important for its radiative effects, but weakly constrained by observations on the global scale, and highly variable among different models. To investigate the controlling factors in one particular model, we investigate the effects of individual processes in HadGEM3–UKCA and compare the resulting diversity of aerosol vertical profiles with the inter-model diversity from the AeroCom Phase II control experiment. In this way we show that (in this model at least) the vertical profile is controlled by a relatively small number of processes, although these vary among aerosol components and particle sizes. We also show that sufficiently coarse variations in these processes can produce a similar diversity to that among different models in terms of the global-mean profile and, to a lesser extent, the zonal-mean vertical position. However, there are features of certain models' profiles that cannot be reproduced, suggesting the influence of further structural differences between models. In HadGEM3–UKCA, convective transport is found to be very important in controlling the vertical profile of all aerosol components by mass. In-cloud scavenging is very important for all except mineral dust. Growth by condensation is important for sulfate and carbonaceous aerosol (along with aqueous oxidation for the former and ageing by soluble material for the latter). The vertical extent of biomass-burning emissions into the free troposphere is also important for the profile of carbonaceous aerosol. Boundary-layer mixing plays a dominant role for sea salt and mineral dust, which are emitted only from the surface. Dry deposition and below-cloud scavenging are important for the profile of mineral dust only. In this model, the microphysical processes of nucleation, condensation and coagulation dominate the vertical profile of the smallest particles by number (e.g. total CN  >  3 nm), while the profiles of larger particles (e.g. CN  >  100 nm) are controlled by the same processes as the component mass profiles, plus the size distribution of primary emissions. We also show that the processes that affect the AOD-normalised radiative forcing in the model are predominantly those that affect the vertical mass distribution, in particular convective transport, in-cloud scavenging, aqueous oxidation, ageing and the vertical extent of biomass-burning emissions.