Ground water model calibration using pilot points and regularization

Use of nonlinear parameter estimation techniques is now commonplace in ground water model calibration. However, there is still ample room for further development of these techniques in order to enable them to extract more information from calibration datasets, to more thoroughly explore the uncertainty associated with model predictions, and to make them easier to implement in various modeling contexts. This paper describes the use of pilot points as a methodology for spatial hydraulic property characterization. When used in conjunction with nonlinear parameter estimation software that incorporates advanced regularization functionality (such as PEST), use of pilot points can add a great deal of flexibility to the calibration process at the same time as it makes this process easier to implement. Pilot points can be used either as a substitute for zones of piecewise parameter uniformity, or in conjunction with such zones. In either case, they allow the disposition of areas of high and low hydraulic property value to be inferred through the calibration process, without the need for the modeler to guess the geometry of such areas prior to estimating the parameters that pertain to them. Pilot points and regularization can also be used as an adjunct to geostatistically based stochastic parameterization methods. Using the techniques described herein, a series of hydraulic property fields can be generated, all of which recognize the stochastic characterization of an area at the same time that they satisfy the constraints imposed on hydraulic property values by the need to ensure that model outputs match field measurements. Model predictions can then be made using all of these fields as a mechanism for exploring predictive uncertainty.

An advanced regularization methodology for use in watershed model calibration

A calibration methodology based on an efficient and stable mathematical regularization scheme is described. This scheme is a variant of so-called Tikhonov regularization in which the parameter estimation process is formulated as a constrained minimization problem. Use of the methodology eliminates the need for a modeler to formulate a parsimonious inverse problem in which a handful of parameters are designated for estimation prior to initiating the calibration process. Instead, the level of parameter parsimony required to achieve a stable solution to the inverse problem is determined by the inversion algorithm itself. Where parameters, or combinations of parameters, cannot be uniquely estimated, they are provided with values, or assigned relationships with other parameters, that are decreed to be realistic by the modeler. Conversely, where the information content of a calibration dataset is sufficient to allow estimates to be made of the values of many parameters, the making of such estimates is not precluded by preemptive parsimonizing ahead of the calibration process. White Tikhonov schemes are very attractive and hence widely used, problems with numerical stability can sometimes arise because the strength with which regularization constraints are applied throughout the regularized inversion process cannot be guaranteed to exactly complement inadequacies in the information content of a given calibration dataset. A new technique overcomes this problem by allowing relative regularization weights to be estimated as parameters through the calibration process itself. The technique is applied to the simultaneous calibration of five subwatershed models, and it is demonstrated that the new scheme results in a more efficient inversion, and better enforcement of regularization constraints than traditional Tikhonov regularization methodologies. Moreover, it is argued that a joint calibration exercise of this type results in a more meaningful set of parameters than can be achieved by individual subwatershed model calibration. (c) 2005 Elsevier B.V. All rights reserved.

Efficient accommodation of local minima in watershed model calibration

The Gauss-Marquardt-Levenberg (GML) method of computer-based parameter estimation, in common with other gradient-based approaches, suffers from the drawback that it may become trapped in local objective function minima, and thus report optimized parameter values that are not, in fact, optimized at all. This can seriously degrade its utility in the calibration of watershed models where local optima abound. Nevertheless, the method also has advantages, chief among these being its model-run efficiency, and its ability to report useful information on parameter sensitivities and covariances as a by-product of its use. It is also easily adapted to maintain this efficiency in the face of potential numerical problems (that adversely affect all parameter estimation methodologies) caused by parameter insensitivity and/or parameter correlation. The present paper presents two algorithmic enhancements to the GML method that retain its strengths, but which overcome its weaknesses in the face of local optima. Using the first of these methods an intelligent search for better parameter sets is conducted in parameter subspaces of decreasing dimensionality when progress of the parameter estimation process is slowed either by numerical instability incurred through problem ill-posedness, or when a local objective function minimum is encountered. The second methodology minimizes the chance of successive GML parameter estimation runs finding the same objective function minimum by starting successive runs at points that are maximally removed from previous parameter trajectories. As well as enhancing the ability of a GML-based method to find the global objective function minimum, the latter technique can also be used to find the locations of many non-global optima (should they exist) in parameter space. This can provide a useful means of inquiring into the well-posedness of a parameter estimation problem, and for detecting the presence of bimodal parameter and predictive probability distributions. The new methodologies are demonstrated by calibrating a Hydrological Simulation Program-FORTRAN (HSPF) model against a time series of daily flows. Comparison with the SCE-UA method in this calibration context demonstrates a high level of comparative model run efficiency for the new method. (c) 2006 Elsevier B.V. All rights reserved.

The cost of uniqueness in groundwater model calibration

Calibration of a groundwater model requires that hydraulic properties be estimated throughout a model domain. This generally constitutes an underdetermined inverse problem, for which a Solution can only be found when some kind of regularization device is included in the inversion process. Inclusion of regularization in the calibration process can be implicit, for example through the use of zones of constant parameter value, or explicit, for example through solution of a constrained minimization problem in which parameters are made to respect preferred values, or preferred relationships, to the degree necessary for a unique solution to be obtained. The cost of uniqueness is this: no matter which regularization methodology is employed, the inevitable consequence of its use is a loss of detail in the calibrated field. This, ill turn, can lead to erroneous predictions made by a model that is ostensibly well calibrated. Information made available as a by-product of the regularized inversion process allows the reasons for this loss of detail to be better understood. In particular, it is easily demonstrated that the estimated value for an hydraulic property at any point within a model domain is, in fact, a weighted average of the true hydraulic property over a much larger area. This averaging process causes loss of resolution in the estimated field. Where hydraulic conductivity is the hydraulic property being estimated, high averaging weights exist in areas that are strategically disposed with respect to measurement wells, while other areas may contribute very little to the estimated hydraulic conductivity at any point within the model domain, this possibly making the detection of hydraulic conductivity anomalies in these latter areas almost impossible. A study of the post-calibration parameter field covariance matrix allows further insights into the loss of system detail incurred through the calibration process to be gained. A comparison of pre- and post-calibration parameter covariance matrices shows that the latter often possess a much smaller spectral bandwidth than the former. It is also demonstrated that, as all inevitable consequence of the fact that a calibrated model cannot replicate every detail of the true system, model-to-measurement residuals can show a high degree of spatial correlation, a fact which must be taken into account when assessing these residuals either qualitatively, or quantitatively in the exploration of model predictive uncertainty. These principles are demonstrated using a synthetic case in which spatial parameter definition is based oil pilot points, and calibration is Implemented using both zones of piecewise constancy and constrained minimization regularization. (C) 2005 Elsevier Ltd. All rights reserved.

Methodologies for calibration and predictive analysis of a watershed model

The use of a fitted parameter watershed model to address water quantity and quality management issues requires that it be calibrated under a wide range of hydrologic conditions. However, rarely does model calibration result in a unique parameter set. Parameter nonuniqueness can lead to predictive nonuniqueness. The extent of model predictive uncertainty should be investigated if management decisions are to be based on model projections. Using models built for four neighboring watersheds in the Neuse River Basin of North Carolina, the application of the automated parameter optimization software PEST in conjunction with the Hydrologic Simulation Program Fortran (HSPF) is demonstrated. Parameter nonuniqueness is illustrated, and a method is presented for calculating many different sets of parameters, all of which acceptably calibrate a watershed model. A regularization methodology is discussed in which models for similar watersheds can be calibrated simultaneously. Using this method, parameter differences between watershed models can be minimized while maintaining fit between model outputs and field observations. In recognition of the fact that parameter nonuniqueness and predictive uncertainty are inherent to the modeling process, PEST's nonlinear predictive analysis functionality is then used to explore the extent of model predictive uncertainty.

A simulation model of cereal-legume intercropping systems for semi-arid regions I. Model development

Cereal-legume intercropping plays an important role in subsistence food production in developing countries, especially in situations of limited water resources. Crop simulation can be used to assess risk for intercrop productivity over time and space. In this study, a simple model for intercropping was developed for cereal and legume growth and yield, under semi-arid conditions. The model is based on radiation interception and use, and incorporates a water stress factor. Total dry matter and yield are functions of photosynthetically active radiation (PAR), the fraction of radiation intercepted and radiation use efficiency (RUE). One of two PAR sub-models was used to estimate PAR from solar radiation; either PAR is 50% of solar radiation or the ratio of PAR to solar radiation (PAR/SR) is a function of the clearness index (K-T). The fraction of radiation intercepted was calculated either based on Beer's Law with crop extinction coefficients (K) from field experiments or from previous reports. RUE was calculated as a function of available soil water to a depth of 900 mm (ASW). Either the soil water balance method or the decay curve approach was used to determine ASW. Thus, two alternatives for each of three factors, i.e., PAR/SR, K and ASW, were considered, giving eight possible models (2 methods x 3 factors). The model calibration and validation were carried out with maize-bean intercropping systems using data collected in a semi-arid region (Bloemfontein, Free State, South Africa) during seven growing seasons (1996/1997-2002/2003). The combination of PAR estimated from the clearness index, a crop extinction coefficient from the field experiment and the decay curve model gave the most reasonable and acceptable result. The intercrop model developed in this study is simple, so this modelling approach can be employed to develop other cereal-legume intercrop models for semi-arid regions. (c) 2004 Elsevier B.V. All rights reserved.

Regional-scale, fully coupled modelling of stream-aquifer interaction in a tropical catchment

The planning and management of water resources in the Pioneer Valley, north-eastern Australia requires a tool for assessing the impact of groundwater and stream abstractions on water supply reliabilities and environmental flows in Sandy Creek (the main surface water system studied). Consequently, a fully coupled stream-aquifer model has been constructed using the code MODHMS, calibrated to near-stream observations of watertable behaviour and multiple components of gauged stream flow. This model has been tested using other methods of estimation, including stream depletion analysis and radon isotope tracer sampling. The coarseness of spatial discretisation, which is required for practical reasons of computational efficiency, limits the model's capacity to simulate small-scale processes (e.g., near-stream groundwater pumping, bank storage effects), and alternative approaches are required to complement the model's range of applicability. Model predictions of groundwater influx to Sandy Creek are compared with baseflow estimates from three different hydrograph separation techniques, which were found to be unable to reflect the dynamics of Sandy Creek stream-aquifer interactions. The model was also used to infer changes in the water balance of the system caused by historical land use change. This led to constraints on the recharge distribution which can be implemented to improve model calibration performance. (c) 2006 Elsevier B.V. All rights reserved.

New calibration technique for multiple-component stress wave force balances

Force measurement in hypervelocity expansion tubes is not possible using conventional techniques. The stress wave force balance technique can be applied in expansion tubes to measure forces despite the short test times involved. This paper presents a new calibration technique for multiple-component stress wave force balances where an impulse response created using a load distribution is required and no orthogonal surfaces on the model exist.. This new technique relies on the tensorial superposition of single-component impulse responses analogous to the vectorial superposition of the calibration loads. The example presented here is that of a scale model of the Mars Pathfinder, but the technique is applicable to any geometry and may be useful for cases where orthogonal loads cannot be applied.

Improving wheat simulation capabilities in Australia from a cropping systems perspective - III. The integrated wheat model (I_WHEAT)

Previous work has identified several short-comings in the ability of four spring wheat and one barley model to simulate crop processes and resource utilization. This can have important implications when such models are used within systems models where final soil water and nitrogen conditions of one crop define the starting conditions of the following crop. In an attempt to overcome these limitations and to reconcile a range of modelling approaches, existing model components that worked demonstrably well were combined with new components for aspects where existing capabilities were inadequate. This resulted in the Integrated Wheat Model (I_WHEAT), which was developed as a module of the cropping systems model APSIM. To increase predictive capability of the model, process detail was reduced, where possible, by replacing groups of processes with conservative, biologically meaningful parameters. I_WHEAT does not contain a soil water or soil nitrogen balance. These are present as other modules of APSIM. In I_WHEAT, yield is simulated using a linear increase in harvest index whereby nitrogen or water limitations can lead to early termination of grainfilling and hence cessation of harvest index increase. Dry matter increase is calculated either from the amount of intercepted radiation and radiation conversion efficiency or from the amount of water transpired and transpiration efficiency, depending on the most limiting resource. Leaf area and tiller formation are calculated from thermal time and a cultivar specific phyllochron interval. Nitrogen limitation first reduces leaf area and then affects radiation conversion efficiency as it becomes more severe. Water or nitrogen limitations result in reduced leaf expansion, accelerated leaf senescence or tiller death. This reduces the radiation load on the crop canopy (i.e. demand for water) and can make nitrogen available for translocation to other organs. Sensitive feedbacks between light interception and dry matter accumulation are avoided by having environmental effects acting directly on leaf area development, rather than via biomass production. This makes the model more stable across environments without losing the interactions between the different external influences. When comparing model output with models tested previously using data from a wide range of agro-climatic conditions, yield and biomass predictions were equal to the best of those models, but improvements could be demonstrated for simulating leaf area dynamics in response to water and nitrogen supply, kernel nitrogen content, and total water and nitrogen use. I_WHEAT does not require calibration for any of the environments tested. Further model improvement should concentrate on improving phenology simulations, a more thorough derivation of coefficients to describe leaf area development and a better quantification of some processes related to nitrogen dynamics. (C) 1998 Elsevier Science B.V.

Modification of the in vivo four-point loading model for studying mechanically induced bone adaptation

We modified the noninvasive, in vivo technique for strain application in the tibiae of rats (Turner et al,, Bone 12:73-79, 1991), The original model applies four-point bending to right tibiae via an open-loop, stepper-motor-driven spring linkage, Depending on the magnitude of applied load, the model produces new bone formation at periosteal (Ps) or endocortical surfaces (Ec.S). Due to the spring linkage, however, the range of frequencies at which loads can be applied is limited. The modified system replaces this design with an electromagnetic vibrator. A load transducer in series with the loading points allows calibration, the loaders' position to be adjusted, and cyclic loading completed under load central as a closed servo-loop. Two experiments were conducted to validate the modified system: (1) a strain gauge was applied to the lateral surface of the right tibia of 5 adult female rats and strains measured at applied loads from 10 to 60 N; and (2) the bone formation response was determined in 28 adult female Sprague-Dawley rats. Loading was applied as a haversine wave with a frequency of 2 Hz for 18 sec, every second day for 10 days. Peak bending loads mere applied at 33, 40, 52, and 64 N, and a sham-loading group tr as included at 64 N, Strains in the tibiae were linear between 10 and 60 N, and the average peak strain at the Ps.S at 60 N was 2664 +/- 250 microstrain, consistent with the results of Turner's group. Lamellar bone formation was stimulated at the Ec.S by applied bending, but not by sham loading. Bending strains above a loading threshold of 40 N increased Ec Lamellar hone formation rate, bone forming surface, and mineral apposition rate with a dose response similar to that reported by Turner et al, (J Bone Miner Res 9:87-97, 1994). We conclude that the modified loading system offers precision for applied loads of between 0 and 70 N, versatility in the selection of loading rates up to 20 Hz, and a reproducible bone formation response in the rat tibia, Adjustment of the loader also enables study of mechanical usage in murine tibia, an advantage with respect to the increasing variety of transgenic strains available in bone and mineral research. (Bone 23:307-310; 1998) (C) 1998 by Elsevier Science Inc. All rights reserved.

Flush air data system calibration using numerical simulation

The use of computational fluid dynamics simulations for calibrating a flush air data system is described, In particular, the flush air data system of the HYFLEX hypersonic vehicle is used as a case study. The HYFLEX air data system consists of nine pressure ports located flush with the vehicle nose surface, connected to onboard pressure transducers, After appropriate processing, surface pressure measurements can he converted into useful air data parameters. The processing algorithm requires an accurate pressure model, which relates air data parameters to the measured pressures. In the past, such pressure models have been calibrated using combinations of flight data, ground-based experimental results, and numerical simulation. We perform a calibration of the HYFLEX flush air data system using computational fluid dynamics simulations exclusively, The simulations are used to build an empirical pressure model that accurately describes the HYFLEX nose pressure distribution ol cr a range of flight conditions. We believe that computational fluid dynamics provides a quick and inexpensive way to calibrate the air data system and is applicable to a broad range of flight conditions, When tested with HYFLEX flight data, the calibrated system is found to work well. It predicts vehicle angle of attack and angle of sideslip to accuracy levels that generally satisfy flight control requirements. Dynamic pressure is predicted to within the resolution of the onboard inertial measurement unit. We find that wind-tunnel experiments and flight data are not necessary to accurately calibrate the HYFLEX flush air data system for hypersonic flight.

Image calibration to like-values in mapping shallow water quality from multitemporal data

The applicability of image calibration to like-values in mapping water quality parameters from multitemporal images is explored, Six sets of water samples were collected at satellite overpasses over Moreton Bay, Brisbane, Australia. Analysis of these samples reveals that waters in this shallow bay are mostly TSS-dominated, even though they are occasionally dominated by chlorophyll as well. Three of the images were calibrated to a reference image based on invariant targets. Predictive models constructed from the reference image were applied to estimating total suspended sediment (TSS) and Secchi depth from another image at a discrepancy of around 35 percent. Application of the predictive model for TSS concentration to another image acquired at a time of different water types resulted in a discrepancy of 152 percent. Therefore, image calibration to like-values could be used to reliably map certain water quality parameters from multitemporal TM images so long as the water type under study remains unchanged. This method is limited in that the mapped results could be rather inaccurate if the water type under study has changed considerably. Thus, the approach needs to be refined in shallow water from multitemporal satellite imagery.