The ensemble Kalman filter (EnKF) is a Monte Carlo approximation of the Kalman filter for high dimensional linear Gaussian state space models. EnKF methods have also been developed for parameter inference of static Bayesian models with a Gaussian likelihood, in a way that is analogous to likelihood tempering sequential Monte Carlo (SMC). These methods are commonly referred to as ensemble Kalman inversion (EKI). Unlike SMC, the inference from EKI is asymptotically biased if the likelihood is non-linear and/or non-Gaussian and if the priors are non-Gaussian. However, it is significantly faster to run. Currently, a large limitation of EKI methods is that the covariance of the measurement error is assumed to be fully known. We develop a new method, which we call component-wise iterative EKI (CW-IEKI), that allows elements of the covariance matrix to be inferred alongside the model parameters at negligible extra cost. This novel method is compared to SMC on a linear Gaussian example as well as four examples with non-linear dynamics (i.e. non-linear function of the model parameters). The non-linear examples include a set of population models applied to synthetic data, a model of nitrogen mineralisation in soil that is based on the Agricultural Production Systems Simulator, a model predicting seagrass decline due to stress from water temperature and light, and a model predicting coral calcification rates. On our examples, we find that CW-IEKI has relatively similar predictive performance to SMC, albeit with greater uncertainty, and it has a significantly faster run time.

Factor Fixing (FF) is a common method for reducing the number of model parameters to lower computational cost. FF typically starts with distinguishing the insensitive parameters from the sensitive and pursues uncertainty quantification (UQ) on the resulting reduced‐order model, fixing each insensitive parameter at a fixed value. There is a need, however, to expand such a common approach to consider the effects of decision choices in the FF‐UQ procedure on metrics of interest. Therefore, to guide the use of FF and increase confidence in the resulting dimension‐reduced model, we propose a new adaptive framework consisting of four principles: (a) re‐parameterize the model first to reduce obvious non‐identifiable parameter combinations, (b) focus on decision relevance especially with respect to errors in quantities of interest (QoI), (c) conduct adaptive evaluation and robustness assessment of errors in the QoI across FF choices as sample size increases, and (d) reconsider whether fixing is warranted. The framework is demonstrated on a spatially‐distributed water quality model. The error in estimates of QoI caused by FF can be estimated using a Polynomial Chaos Expansion (PCE) surrogate model. Built with 70 model runs, the surrogate is computationally inexpensive to evaluate and can provide global sensitivity indices for free. For the selected catchment, just two factors may provide an acceptably accurate estimate of model uncertainty in the average annual load of Total Suspended Solids (TSS), suggesting that reducing the uncertainty in these two parameters is a priority for future work before undertaking further formal uncertainty quantification.

The eWater Source modelling framework has been modified to support the Great Barrier Reef (GBR) Dynamic SedNet catchment modelling concept, which is used to simulate fine sediment and particulate nutrient generation, loss, and transport processes across GBR catchments. Catchment scale monitored data sets are used to calibrate and evaluate models. Model performance is assessed qualitatively and quantitatively. Modelling predicts that approximately half of generated sediment is delivered to the GBR lagoon; the remainder is deposited on floodplains, trapped in reservoirs or lost through other minor processes (e.g. irrigation extractions). Gullies are the major source of sediment, with comparable contributions from hillslopes and streambanks. Hillslope sources are considered the major source of particulate nutrients across the GBR catchments. We demonstrate that using locally developed, customised models coupled with a complementary monitoring program can produce credible modelled estimates of pollutant loads and provide a platform for testing catchment scale assumptions and scenarios.

All models are imperfect, so it is important to consider uncertainty in their predictions.When calibrating models to measured data, uncertainty in the model can be simultaneously estimated, although there are multiple methods available for doing this.In this study, we applied ensemble smoothers and Bayesian inference to calibrate a model of nitrogen mineralization in soils.We obtained mineralization measurements from a previously published study that measured changes in inorganic nitrogen over long-term laboratory incubations in a soil located in the Mackay Whitsundays region (North Queensland).Simulations were performed using the Agricultural Production Systems Simulator (APSIM).We inferred two parameters that characterize the size of the simulated soil organic carbon pools (fbiom and finert) because it is difficult to estimate these parameters from measurements only.For the calibration, we considered two different sources of uncertainty: measurement noise and noise of unknown origin, the latter of which includes all nonmeasurement related errors.We found that ignoring noise of unknown origin can result in an overly optimistic representation of the model error (Figure 1a,c).On the contrary, incorporating noise of unknown origin can lead to a more accurate representation of the uncertainty in the predictions, with model predictions providing adequate coverage of measurements (Figure 1b,d).We show that parameterizing fbiom and finert is difficult because these parameters are correlated, hence different combinations of parameters can equally well simulate the measured data.We suggest that future work needs to provide a means of parameterizing at least one of these fractions independently to facilitate parameter identifiability.Figure 1.Measured (grey dots) and simulated (orange lines) nitrogen mineralization (mg N kg -1 ) for the ensemble smoother with multiple data assimilation (ES-MDA) (a), flexible iterative ES-MDA (b), Bayesian inference with measurement noise only (c), Bayesian inference with measurement noise and additional noise (d).The error bars in the measured data represent the error of the laboratory method.Shaded areas represent the predicted 95% credible intervals.

Measuring stream pollutant loads across the Great Barrier Reef (GBR) catchment area (GBRCA) is challenging due to the spatial extent, climate variability, changing land use and evolving land management practices, and cost. Thus, models are used to estimate baseline pollutant loads. The eWater Source modelling framework is coupled with agricultural paddock scale models and the GBR Dynamic SedNet plugin to simulate dissolved inorganic nitrogen (DIN) generation and transport processes across the GBRCA. Catchment scale monitoring of flow and loads are used to calibrate the models, and performance is assessed qualitatively and quantitatively. Modelling indicates almost half (47%) of the total modelled DIN load exported to the GBR lagoon is from the Wet Tropics, and almost half of the total modelled DIN load is from sugarcane areas. We demonstrate that using locally developed, customised models coupled with a complementary monitoring program can produce reliable estimates of pollutant loads.

With the underlying aim of increasing efficiency of computational modelling pertinent for managing&protecting the Great Barrier Reef, we perform a preliminary investigation on the use of deep neural networks for opportunistic model emulation of APSIM models by repurposing an existing large dataset containing outputs of APSIM model runs. The dataset has not been specifically tailored for the model emulation task. We employ two neural network architectures for the emulation task: densely connected feed-forward neural network (FFNN), and gated recurrent unit feeding into FFNN (GRU-FFNN), a type of a recurrent neural network. Various configurations of the architectures are trialled. A minimum correlation statistic is used to identify clusters of APSIM scenarios that can be aggregated to form training sets for model emulation. We focus on emulating 4 important outputs of the APSIM model: runoff, soil_loss, DINrunoff, Nleached. The GRU-FFNN architecture with three hidden layers and 128 units per layer provides good emulation of runoff and DINrunoff. However, soil_loss and Nleached were emulated relatively poorly under a wide range of the considered architectures; the emulators failed to capture variability at higher values of these two outputs. While the opportunistic data available from past modelling activities provides a large and useful dataset for exploring APSIM emulation, it may not be sufficiently rich enough for successful deep learning of more complex model dynamics. Design of Computer Experiments may be required to generate more informative data to emulate all output variables of interest. We also suggest the use of synthetic meteorology settings to allow the model to be fed a wide range of inputs. These need not all be representative of normal conditions, but can provide a denser, more informative dataset from which complex relationships between input and outputs can be learned.

Catchment water quality models are an important tool for understanding the impacts of land management practice on the water quality of the receiving waters of the Great Barrier Reef lagoon.As part of the Paddock to Reef program the Great Barrier Reef Catchment Loads Modelling Program estimates average annual loads of key pollutants (sediment, nutrients and pesticides) for each of the 35 catchments draining to the Great Barrier Reef.Since catchment models assume that constituent generation and transport within the catchment is largely controlled by rainfall and runoff, it is imperative that the hydrology calibration approach underpinning the catchment model is rigorous and achieves the best possible results.Because catchment models are conceptual representations of very complex landscape systems any forecasts and predictions they produce will be subject to uncertainty and quantifying uncertainty is an important aspect of analysing model performance.Various methods derived from a range of statistical frameworks have been applied to study uncertainty of rainfall-runoff models.Perhaps the most intuitive way of approaching uncertainty analysis is via the formalism of Bayes' theorem where some prior understanding of the model parameters is updated once exposed to relevant data.As elegant as Bayesian uncertainty analysis may be, there are practical limitations to implementing it.The equations defining Bayes' theorem often have no analytic solution, or at least one that is tractable, one must resort to numerical methods to complete the process.In practice, this usually involves a campaign of stochastically sampling from the Bayesian posterior distribution to construct a statistical facsimile.This can be a computationally exhausting process, particularly when an expensive model is used, the prior is significantly divergent from the posterior and a large number of parameters is involved.Ensemble methods such as the iterative ensemble smoother (IES) have been developed to alleviate much of the computational overhead demanded by the uncertainty quantification of environmental models, particularly those that involve high dimensional parameter spaces.On the face of it, the IES would seem to fit very well with the problem presented by catchment water quality models but to date, there is very little evidence of this.In this study, we apply a Gauss-Levenberg-Marquardt form of the IES to the calibration and uncertainty analysis of a rainfall-runoff model.The IES is found to be an efficient and powerful method for conditioning model parameters and providing robust uncertainty estimates adhering to the spirit of Bayesian statistics.

Catchment water quality models are notoriously over-parametrised. Given this condition, it is useful to be able to identify which parameters have the greatest influence on the model results. In theory, this could be accomplished through a detailed first principles interrogation of the mathematical structure of the model in an abstract manner. This however, is impractical in most instances owing to the complexity of the models and posteriori methods of parameter sensitivity analysis are more conventional. As with most aspects of large-scale modelling endeavours, a major consideration in choosing a technique for sensitivity analysis is efficiency and a compromise between computational effort and numerical accuracy is usually negotiated. ANOVA based sensitivity analysis methods are very popular as they offer a holistic survey of the parameter sensitivity by not only accounting for the response of the model output surface due to the activity of single parameters acting independently, but also due to the interaction between parameters. These global sensitivity indices are usually calculated by Monte Carlo simulation and may be too computationally demanding to be routinely applied in water quality modelling scenarios. We demonstrate the application of the group method of data handling (GMDH) inductive, self-organising modelling method to the sensitivity analysis of constituent generation parameters of an integrated hydrological and water quality model. By using a modestly sized sample input-output dataset, a GMDH neural network is used to synthesise a sparse, random-sampling high dimensional model representation (RS-HDMR) that can be used to calculate first and second order Sobol sensitivity indices. This algorithm potentially leads to reductions in computational cost of 2-3 orders of magnitude over Monto Carlo simulation. Although several other adaptive methods for efficiently constructing a sparse RS-HDMR have been reported in the literature, such as polynomial chaos expansions, the parameter selection and noise filtering characteristics of the GMDH network may result in more optimal HDMR expansion.

Fine sediment generated from catchments of the Great Barrier Reef (GBR) is considered to be one of the main pollutants of serious concern affecting the quality of water entering the reef.Estimation of the amount of sediment that is generated from these catchments and the proportion that enters the reef lagoon for a given combination of land use and land management is being carried out under the Paddock to Reef (P2R) Program using the eWater Source modelling platform and the Dynamic SedNet plugin.Currently, water quality monitoring data from selected sites, including end of system sites, are being used to manually calibrate and validate the water quality model.However, the manual calibration attempted so far has mainly been based on trial-and-error.The main objectives of the simple one-at-a-time (OAT) local sensitivity analysis reported here are to: (1) identify parameters that the model reacts most sensitively to in order to simplify and accelerate the calibration of the model; (2) examine if the results of the sensitivity analysis depend on the spatial and temporal scales of investigation; and (3) compare the results to a more computationally intensive global sensitivity analysis reported in an associated global sensitivity analysis paper to be presented in this conference (Bennett and Fentie, 2017).The choice of parameters included in the sensitivity analysis was determined based on a preliminary investigation on a relatively small test sub-catchment in the Burnett catchment by the first author.The preliminary investigation showed that fine sediment load is most sensitive to changes in three parameters (i.e., streambank sediment bulk density (ρs), hillslope sediment delivery ratio (HSDR) and gully sediment delivery ratio (GSDR)), which represent the three sources of sediment (streambank, hillslope and gully erosion, respectively) in the Source/Dynamic SedNet water quality model.The other three parameters included in the analysis -floodplain deposition settling velocity (V p ), channel average terminal fall velocity for fine sediment remobilization (ω dep ), and channel average terminal fall velocity for fine sediment deposition (ω mob ) -represent the effect of settling velocity of particles on sediment deposition on the floodplain and in the stream, and channel remobilization, respectively.In order to explore the effect of catchment size and other characteristics on the sensitivity analysis, three locations with differing catchment area were chosen in this study.The location with the smallest catchment area drains the upper parts of the Burnett catchment while the location with the largest catchment area represents the Burnett catchment end of system (EoS).The third location is at the outlet of the Mary catchment.Temporal variability in results of the sensitivity analysis was investigated by conducting the analysis at an annual time step summarised from the daily time-step model outputs of the 28 years modelling period of the Great Barrier Reef catchment modelling for Report Card 2016.Results show that: (1) the modelled fine sediment loads were relatively highly sensitive to changes in streambank sediment bulk density (ρ s ), hillslope sediment delivery ratio (HSDR) and gully sediment delivery ratio (GSDR), but were only marginally sensitive to changes in the other three parameters; (2) the sensitivity of fine sediment load to changes in parameters varied across the three locations; and (3) the sensitivity of fine sediment load to changes in parameters varied annually.Sensitivities from this study have similar rankings of parameters to those from Bennet and Fentie (2017).Both approaches determined that modelled fine sediment load is relatively insensitive to changes in particle settling velocity parameters for both stream deposition and channel remobilization.Based on these findings it is concluded that (1) the simple one-at-a-time sensitivity analysis conducted in this study is adequate in determining the sensitivity of modelled sediment to changes in model parameters; and (2) since results of the local sensitivity analysis vary spatially from catchment to catchment and temporally from year to year, the analysis needs to be carried out for each specific catchment and period of interest.

The aim of this study was to develop a bio-economic model to estimate the feasibility and net profit (or net costs) of achieving set water quality targets (sediment, nitrogen, phosphorus and herbicide load reductions) in the Burnett-Mary region within the southern portion of the Great Barrier Reef (GBR), southern Queensland, Australia. Two sets of targets were evaluated, namely (1) Reef Plan Targets (RPTs) representing currently agreed targets, and (2) the more ambitious Ecologically Relevant Targets (ERTs) designed to halt the decline and improve the condition of the GBR. This paper describes the construction of a bio-economic optimisation framework linking field and catchment scale biophysical model results and farm economic analysis to solve for RPTs or ERTs assigned either regionally or within discrete basins. Key outcomes from the study were that RPTs could be achieved whereas ERTs required significant additional investment and were infeasible if individual basins must meet the targets.

This paper describes the construction of a bio-economic optimisation framework to evaluate the feasibility and net profit (or net costs) of achieving water quality targets in the Burnett-Mary region within the southern portion of the Great Barrier Reef (GBR), southern Queensland, Australia. Key outcomes from the study were that current sediment, nitrogen and phosphorus load reduction targets could be achieved whereas more ambitious ecologically relevant targets required significant additional investment and were not feasible based on the model if individual basins must meet the targets.

The aim of this study was to develop a bio-economic model to estimate the feasibility and net profit (or net costs)s of achieving set water quality targets (sediment, nitrogen and phosphorus load reductions) in the Burnett Mary region of northern Queensland, Australia to with the aim of protecting the southern portion of the Great Barrier Reef (GBR). Two sets of targets were evaluated – Reef Plan Targets (RPTs) which are the currently formally agreed targets, and more ambitious Ecologically Relevant Targets (ERTs) which current science suggests might be needed to better protect the values of the GBR. This paper describes the construct of a bio-economic optimisation framework which has been used to underpin a Water Quality Improvement Plan (WQIP) for the Burnett Mary region. The bioeconomic model incorporates the available science developed from paddock and catchment scale biophysical model results and farm economic analysis. The model enabled transparent assessment and optimisation of net profits and costs associated with four categories of best management practices (cutting edge unproven technologies called ‘A’ practice, current best-management practices called ‘B’, common industry or ‘C’ practices, and below industry standards or ‘D’ practice) in the grazing and sugar cane industries. The bioeconomic model was able to solve for RPTs or ERTs assigned to either the entire region or within each of five discrete river basins. Key outcomes from the study were that RPTs could be achieved at an annual cost of $3M/year on a whole of region basis. In contrast ERTs could be achieved on a whole of region basis at as net cost of $16M/year. ERTs were not able to be feasibly met on a basin by basin basis. This is the first time such a comprehensive and integrated bio-economic model has been constructed for a region within GBR using environmental software that linked available biophysical and economic modelling.

BACKGROUNDField studies of diuron and its metabolites 3-(3,4-dichlorophenyl)-1-methylurea (DCPMU), 3,4-dichlorophenylurea (DCPU) and 3,4-dichloroaniline (DCA) were conducted in a farm soil and in stream sediments in coastal Queensland, Australia.RESULTSDuring a 38 week period after a 1.6 kg ha(-1) diuron application, 70-100% of detected compounds were within 0-15 cm of the farm soil, and 3-10% reached the 30-45 cm depth. First-order t(1/2) degradation averaged 49+/-0.9 days for the 0-15, 0-30 and 0-45 cm soil depths. Farm runoff was collected in the first 13-50 min of episodes lasting 55-90 min. Average concentrations of diuron, DCPU and DCPMU in runoff were 93, 30 and 83-825 microg L(-1) respectively. Their total loading in all runoff was >0.6% of applied diuron. Diuron and DCPMU concentrations in stream sediments were between 3-22 and 4-31 microg kg(-1) soil respectively. The DCPMU/diuron sediment ratio was >1.CONCLUSIONRetention of diuron and its metabolites in farm topsoil indicated their negligible potential for groundwater contamination. Minimal amounts of diuron and DCMPU escaped in farm runoff. This may entail a significant loading into the wider environment at annual amounts of application. The concentrations and ratio of diuron and DCPMU in stream sediments indicated that they had prolonged residence times and potential for accumulation in sediments. The higher ecotoxicity of DCPMU compared with diuron and the combined presence of both compounds in stream sediments suggest that together they would have a greater impact on sensitive aquatic species than as currently apportioned by assessments that are based upon diuron alone.

Collisions between 98-keV ${\mathrm{N}}^{7+}$ ions and a HCl target have been investigated experimentally. The kinetic-energy distribution of fragment ${\mathrm{H}}^{+}$ ions originating from multiple electron capture was detected at angles in the range 20\ifmmode^\circ\else\textdegree\fi{}--160\ifmmode^\circ\else\textdegree\fi{} with respect to the incident beam direction. Proton energies as large as 100 eV were observed, and calculations made in the simple Coulomb explosion model suggest that up to seven target electrons may be involved during the collision. Using the Landau-Zener model, we show that the ${\mathrm{N}}^{7+}$ projectile mainly captures outer-shell electrons from HCl. From the experimental data we derived multiple-capture cross sections which we compared with results from a model calculation made using the classical over-barrier model and also with a semiempirical scaling law. For the specific case of double capture, several structures appeared, which were assigned using ab initio calculations to states of ${\mathrm{HCl}}^{2+}$.

Collisions between 98-keV N7+ ions and a HCl target have been investigated experimentally. The kinetic-energy distribution of fragment H+ ions originating from multiple electron capture was detected at angles in the range 20 degrees-160 degrees with respect to the incident beam direction. Proton energies as large as 100 eV were observed, and calculations made in the simple Coulomb explosion model suggest that up to seven target electrons may be involved during the collision. Using the Landau-Zener model, we show that the N7+ projectile mainly captures outer-shell electrons from HCl. From the experimental data we derived multiple-capture cross sections which we compared with results from a model calculation made using the classical over-barrier model and also with a semiempirical scaling law. For the specific case of double capture, several structures appeared, which were assigned using ab initio calculations to states of HCl2+.

The most widely employed industrial process for producing alumina ( Bayer process) involves the dissolution of available aluminium hydroxide minerals present in raw bauxite into high temperature sodium hydroxide solutions. On cooling of the solution, or liquor in the industrial vernacular, Al is precipitated from solution in the form of gibbsite (Al(OH)(3)). In order to optimise the process, a detailed knowledge of factors influencing gibbsite solubility is required, a problem that is confounded by the presence of liquor impurities. In this paper, the use of the Group Method of Data Handling (GMDH) polynomial neural network for developing a gibbsite equilibrium solubility model for Bayer process liquors is discussed. The resulting predictive model appears to correctly incorporate the effects of liquor impurities and is found to offer a level of performance comparable to the most sophisticated phenomenological model presented to date.

Photo-double ionization of hydrogen iodide has been investigated over the photon energy range 29.0–32.5 eV by threshold photoelectrons coincidence (TPEsCO) spectroscopy and by ab initio calculations that included spin–orbit interaction. Good agreement is found between experiment and theory for the adiabatic ionization potentials for the formation of the X3Σ−, a 1Δ and b1Σ+ states of HI2+ and for the vibrational separations and relative vibrational intensities within the states. The spin–orbit splitting of the X3Σ− state of HI2+ is calculated to be 0.17924 eV for ν2+=0. Because this splitting is close to the calculated vibrational separations within the spin–orbit components of this state [e.g., for ν2+=0–1, 0.22382 eV in 3Σ0− and 0.22163 eV in 3Σ1−], there is an overlapping of vibrational structure within the TPEsCO spectrum at the resolution used that is confirmed by the simulation spectrum.

Although attempts to measure spectra of NO2+ using ion beam spectroscopy have been unsuccessful to date we show that the Newcastle ion-beam/infrared laser beam spectrometer should be capable of recording the spectrum calculated here. We report ab initio calculations of infrared spectra of NO2+ between levels of the A2Π electronic state. The accuracy of our potential energy function is established by using it to synthesize a previously recorded TPEsCO spectrum of NO2+.

We have obtained hyperfine-resolved infrared spectra of a (P)Q(23)(N) branch line in the u =2-1 band of the X (3)Sigma(-) state of the molecular dication (DCl2+)-Cl-35. Analysis of the hyperfine structure allows us to estimate the magnitude of the Fermi contact interaction for the chlorine nucleus; b(F)(Cl) = 167 (25) MHz.

Vibrationally resolved spectra of HCl2+ appear to show five vibrational levels for the X3Σ− ground electronic state, whereas calculations of vibrational levels supported by ab initio potential energy curves have been able to locate only three vibrational levels below the barrier; this discrepancy is resolved by considering vibrational states that the potential function supports in the continuum above the barrier maximum. A low resolution spectrum produced from first principles is compared with a spectrum obtained with threshold photoelectrons in coincidence (TPEsCO) spectroscopy, with agreement sufficient to suggest that care must be taken in the inversion of vibrational spectroscopic data for molecular dications to avoid generating potential functions that are too strongly bound.