Development and evaluation of neural network freeway incident detection models using field data

This paper discusses a multi-layer feedforward (MLF) neural network incident detection model that was developed and evaluated using field data. In contrast to published neural network incident detection models which relied on simulated or limited field data for model development and testing, the model described in this paper was trained and tested on a real-world data set of 100 incidents. The model uses speed, flow and occupancy data measured at dual stations, averaged across all lanes and only from time interval t. The off-line performance of the model is reported under both incident and non-incident conditions. The incident detection performance of the model is reported based on a validation-test data set of 40 incidents that were independent of the 60 incidents used for training. The false alarm rates of the model are evaluated based on non-incident data that were collected from a freeway section which was video-taped for a period of 33 days. A comparative evaluation between the neural network model and the incident detection model in operation on Melbourne's freeways is also presented. The results of the comparative performance evaluation clearly demonstrate the substantial improvement in incident detection performance obtained by the neural network model. The paper also presents additional results that demonstrate how improvements in model performance can be achieved using variable decision thresholds. Finally, the model's fault-tolerance under conditions of corrupt or missing data is investigated and the impact of loop detector failure/malfunction on the performance of the trained model is evaluated and discussed. The results presented in this paper provide a comprehensive evaluation of the developed model and confirm that neural network models can provide fast and reliable incident detection on freeways. (C) 1997 Elsevier Science Ltd. All rights reserved.

Evaluation of multifrequency bioelectrical impedance data for predicting lean tissue mass in beef cattle

Multifrequency bioimpedance analysis has the potential to provide a non-invasive technique for determining body composition in live cattle. A bioimpedance meter developed for use in clinical medicine was adapted and evaluated in 2 experiments using a total of 31 cattle. Prediction equations were obtained for total body water, extracellular body water, intracellular body water, carcass water and carcass protein. There were strong correlations between the results obtained through chemical markers and bioimpedance analysis when determined in cattle that had a wide range of liveweights and conditions. The r(2) values obtained were 0.87 and 0.91 for total body water and extracellular body water respectively. Bioimpedance also correlated with carcass water, measured by chemical analysis (r(2) = 0.72), but less well with carcass protein (r(2) = 0.46). These correlations were improved by inclusion of liveweight and sex as variables in multiple regression analysis. However, the resultant equations were poor predictors of protein and water content in the carcasses of a group of small underfed beef cattle, that had a narrow range of liveweights. In this case, although there was no statistical difference between the predicted and measured values overall, bioimpedance analysis did not detect the differences in carcass protein between the 2 groups that were apparent following chemical analysis. Further work is required to determine the sensitivity of the technique in small underfed cattle, and its potential use in heavier well fed cattle close to slaughter weight.

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.

Prediction of the chemical composition of lamb carcasses from multi-frequency impedance data

Multi-frequency bioimpedance analysis (MFBIA) was used to determine the impedance, reactance and resistance of 103 lamb carcasses (17.1-34.2 kg) immediately after slaughter and evisceration. Carcasses were halved, frozen and one half subsequently homogenized and analysed for water, crude protein and fat content. Three measures of carcass length were obtained. Diagonal length between the electrodes (right side biceps femoris to left side of neck) explained a greater proportion of the variance in water mass than did estimates of spinal length and was selected for use in the index L-2/Z to predict the mass of chemical components in the carcass. Use of impedance (Z) measured at the characteristic frequency (Z(c)) instead of 50 kHz (Z(50)) did not improve the power of the model to predict the mass of water, protein or fat in the carcass. While L-2/Z(50) explained a significant proportion of variation in the masses of body water (r(2) 0.64), protein (r(2) 0.34) and fat (r(2) 0.35), its inclusion in multi-variate indices offered small or no increases in predictive capacity when hot carcass weight (HCW) and a measure of rib fat-depth (GR) were present in the model. Optimized equations were able to account for 65-90 % of the variance observed in the weight of chemical components in the carcass. It is concluded that single frequency impedance data do not provide better prediction of carcass composition than can be obtained from measures of HCW and GR. Indices of intracellular water mass derived from impedance at zero frequency and the characteristic frequency explained a similar proportion of the variance in carcass protein mass as did the index L-2/Z(50).

High-resolution bulk density images, using calibrated X-ray radiography of impregnated soil slices

Bulk density of undisturbed soil samples can be measured using computed tomography (CT) techniques with a spatial resolution of about 1 mm. However, this technique may not be readily accessible. On the other hand, x-ray radiographs have only been considered as qualitative images to describe morphological features. A calibration procedure was set up to generate two-dimensional, high-resolution bulk density images from x-ray radiographs made with a conventional x-ray diffraction apparatus. Test bricks were made to assess the accuracy of the method. Slices of impregnated soil samples were made using hardsetting seedbeds that had been gamma scanned at 5-mm depth increments in a previous study. The calibration procedure involved three stages: (i) calibration of the image grey levels in terms of glass thickness using a staircase made from glass cover slips, (ii) measurement of ratio between the soil and resin mass attenuation coefficients and the glass mass attenuation coefficient, using compacted bricks of known thickness and bulk density, and (iii) image correction accounting for the heterogeneity of the irradiation field. The procedure was simple, rapid, and the equipment was easily accessible. The accuracy of the bulk density determination was good (mean relative error 0.015), The bulk density images showed a good spatial resolution, so that many structural details could be observed. The depth functions were consistent with both the global shrinkage and the gamma probe data previously obtained. The suggested method would be easily applied to the new fuzzy set approach of soil structure, which requires generation of bulk density images. Also, it would be an invaluable tool for studies requiring high-resolution bulk density measurement, such as studies on soil surface crusts.

Changes in estimated coronary risk in the 1980s: data from 38 populations in the WHO MONICA Project

The World Health Organization (WHO) MONICA Project is a 10-year study monitoring trends and determinants of cardiovascular disease in geographically defined populations. Data were collected from over 100 000 randomly selected participants in two risk factor surveys conducted approximately 5 years apart in 38 populations using standardized protocols. The net effects of changes in the risk factor levels were estimated using risk scores derived from longitudinal studies in the Nordic countries. The prevalence of cigarette smoking decreased among men in most populations, but the trends for women varied. The prevalence of hypertension declined in two-thirds of the populations. Changes in the prevalence of raised total cholesterol were small but highly correlated between the genders (r = 0.8). The prevalence of obesity increased in three-quarters of the populations for men and in more than half of the populations for women. In almost half of the populations there were statistically significant declines in the estimated coronary risk for both men and women, although for Beijing the risk score increased significantly for both genders. The net effect of the changes in the risk factor levels in the 1980s in most of the study populations of the WHO MONICA Project is that the rates of coronary disease are predicted to decline in the 1990s.

Data analysis in multiple-frequency bioelectrical impedance analysis

The performance of three analytical methods for multiple-frequency bioelectrical impedance analysis (MFBIA) data was assessed. The methods were the established method of Cole and Cole, the newly proposed method of Siconolfi and co-workers and a modification of this procedure. Method performance was assessed from the adequacy of the curve fitting techniques, as judged by the correlation coefficient and standard error of the estimate, and the accuracy of the different methods in determining the theoretical values of impedance parameters describing a set of model electrical circuits. The experimental data were well fitted by all curve-fitting procedures (r = 0.9 with SEE 0.3 to 3.5% or better for most circuit-procedure combinations). Cole-Cole modelling provided the most accurate estimates of circuit impedance values, generally within 1-2% of the theoretical values, followed by the Siconolfi procedure using a sixth-order polynomial regression (1-6% variation). None of the methods, however, accurately estimated circuit parameters when the measured impedances were low (<20 Omega) reflecting the electronic limits of the impedance meter used. These data suggest that Cole-Cole modelling remains the preferred method for the analysis of MFBIA data.

Data partitioning for parallel spatial join processing

The cost of spatial join processing can be very high because of the large sizes of spatial objects and the computation-intensive spatial operations. While parallel processing seems a natural solution to this problem, it is not clear how spatial data can be partitioned for this purpose. Various spatial data partitioning methods are examined in this paper. A framework combining the data-partitioning techniques used by most parallel join algorithms in relational databases and the filter-and-refine strategy for spatial operation processing is proposed for parallel spatial join processing. Object duplication caused by multi-assignment in spatial data partitioning can result in extra CPU cost as well as extra communication cost. We find that the key to overcome this problem is to preserve spatial locality in task decomposition. We show in this paper that a near-optimal speedup can be achieved for parallel spatial join processing using our new algorithms.

Heart rate, blood lactate and kinematic data of elite colts (under-19) rugby union players during competition

Physiological and kinematic data were collected from elite under-19 rugby union players to provide a greater understanding of the physical demands of rugby union. Heart rate, blood lactate and time-motion analysis data were collected from 24 players (mean +/- s((x) over bar): body mass 88.7 +/- 9.9 kg, height 185 +/- 7 cm, age 18.4 +/- 0.5 years) during six competitive premiership fixtures. Six players were chosen at random from each of four groups: props and locks, back row forwards, inside backs, outside backs. Heart rate records were classified based on percent time spent in four zones (>95%, 85-95%, 75-84%, <75% HRmax). Blood lactate concentration was measured periodically throughout each match, with movements being classified as standing, walking, jogging, cruising, sprinting, utility, rucking/mauling and scrummaging. The heart rate data indicated that props and locks (58.4%) and back row forwards (56.2%) spent significantly more time in high exertion (85-95% HRmax) than inside backs (40.5%) and outside backs (33.9%) (P < 0.001). Inside backs (36.5%) and outside backs (38.5%) spent significantly more time in moderate exertion (75-84% HRmax) than props and locks (22.6%) and back row forwards (19.8%) (P < 0.05). Outside backs (20.1%) spent significantly more time in low exertion (< 75% HRmax) than props and locks (5.8%) and back row forwards (5.6%) (P < 0.05). Mean blood lactate concentration did not differ significantly between groups (range: 4.67 mmol.l(-1) for outside backs to 7.22 mmol.l(-1) for back row forwards; P < 0.05). The motion analysis data indicated that outside backs (5750 m) covered a significantly greater total distance than either props and locks or back row forwards (4400 and 4080 m, respectively; P < 0.05). Inside backs and outside backs covered significantly greater distances walking (1740 and 1780 m, respectively; P < 0.001), in utility movements (417 and 475 m, respectively; P < 0.001) and sprinting (208 and 340 m, respectively; P < 0.001) than either props and locks or back row forwards (walking: 1000 and 991 m; utility movements: 106 and 154 m; sprinting: 72 and 94 m, respectively). Outside backs covered a significantly greater distance sprinting than inside backs (208 and 340 m, respectively; P < 0.001). Forwards maintained a higher level of exertion than backs, due to more constant motion and a large involvement in static high-intensity activities. A mean blood lactate concentration of 4.8-7.2 mmol.l(-1) indicated a need for 'lactate tolerance' training to improve hydrogen ion buffering and facilitate removal following high-intensity efforts. Furthermore, the large distances (4.2-5.6 km) covered during, and intermittent nature of, match-play indicated a need for sound aerobic conditioning in all groups (particularly backs) to minimize fatigue and facilitate recovery between high-intensity efforts.

A framework for selecting appropriate remotely sensed data dimensions for environmental monitoring and management

Two major factors are likely to impact the utilisation of remotely sensed data in the near future: (1)an increase in the number and availability of commercial and non-commercial image data sets with a range of spatial, spectral and temporal dimensions, and (2) increased access to image display and analysis software through GIS. A framework was developed to provide an objective approach to selecting remotely sensed data sets for specific environmental monitoring problems. Preliminary applications of the framework have provided successful approaches for monitoring disturbed and restored wetlands in southern California.

Exploring sequential data: Commentary on Bowers, Jentsch, Salas, and Braun (1998)

In an investigation intended to determine training needs of night crews, Bowers et al. (1998, this issue) report two studies showing that the patterning of communication is a better discriminator of good and poor crews than is the content of communication. Bowers et al. characterize their studies as intended to generate hypotheses for training needs and draw connections with Exploratory Sequential Data Analysis (ESDA). Although applauding the intentions of Bowers ct al., we point out some concerns with their characterization and implementation of ESDA. Our principal concern is that the Bowers et al. exploration of the data does not convincingly lead them back to a better fundamental understanding of the original phenomena they are investigating.

Relative dispersions of intra-albumin transit times across rat and elasmobranch perfused livers, and implications for intra- and inter-species scaling of hepatic clearance using microsomal data

It is recognized that vascular dispersion in the liver is a determinant of high first-pass extraction of solutes by that organ. Such dispersion is also required for translation of in-vitro microsomal activity into in-vivo predictions of hepatic extraction for any solute. We therefore investigated the relative dispersion of albumin transit times (CV2) in the livers of adult and weanling rats and in elasmobranch livers. The mean and normalized variance of the hepatic transit time distribution of albumin was estimated using parametric non-linear regression (with a correction for catheter influence) after an impulse (bolus) input of labelled albumin into a single-pass liver perfusion. The mean +/- s.e. of CV2 for albumin determined in each of the liver groups were 0.85 +/- 0.20 (n = 12), 1.48 +/- 0.33 (n = 7) and 0.90 +/- 0.18 (n = 4) for the livers of adult and weanling rats and elasmobranch livers, respectively. These CV2 are comparable with that reported previously for the dog and suggest that the CV2 Of the liver is of a similar order of magnitude irrespective of the age and morphological development of the species. It might, therefore, be justified, in the absence of other information, to predict the hepatic clearances and availabilities of highly extracted solutes by scaling within and between species livers using hepatic elimination models such as the dispersion model with a CV2 of approximately unity.

Effects of data characteristics on the results of reserve selection algorithms

We tested the effects of four data characteristics on the results of reserve selection algorithms. The data characteristics were nestedness of features (land types in this case), rarity of features, size variation of sites (potential reserves) and size of data sets (numbers of sites and features). We manipulated data sets to produce three levels, with replication, of each of these data characteristics while holding the other three characteristics constant. We then used an optimizing algorithm and three heuristic algorithms to select sites to solve several reservation problems. We measured efficiency as the number or total area of selected sites, indicating the relative cost of a reserve system. Higher nestedness increased the efficiency of all algorithms (reduced the total cost of new reserves). Higher rarity reduced the efficiency of all algorithms (increased the total cost of new reserves). More variation in site size increased the efficiency of all algorithms expressed in terms of total area of selected sites. We measured the suboptimality of heuristic algorithms as the percentage increase of their results over optimal (minimum possible) results. Suboptimality is a measure of the reliability of heuristics as indicative costing analyses. Higher rarity reduced the suboptimality of heuristics (increased their reliability) and there is some evidence that more size variation did the same for the total area of selected sites. We discuss the implications of these results for the use of reserve selection algorithms as indicative and real-world planning tools.

V4 region of small subunit rDNA indicates polyphyly of the Fellodistomidae (Digenea) which is supported by morphology and life-cycle data

There is no morphological synapomorphy for the disparate digeneans, the Fellodistomidae Nicoll, 1909. Although all known life-cycles of the group include bivalves as first intermediate hosts, there is no convincing morphological synapomorphy that can be used to unite the group. Sequences from the V4 region of small subunit (18S) rRNA genes were used to infer phylogenetic relationships among 13 species of Fellodistomidae from four subfamilies and eight species from seven other digenean families: Bivesiculidae; Brachylaimidae; Bucephalidae; Gorgoderidae; Gymnophallidae; Opecoelidae; and Zoogonidae. Outgroup comparison was made initially with an aspidogastrean. Various species from the other digenean families were used as outgroups in subsequent analyses. Three methods of analysis indicated polyphyly of the Fellodistomidae and at least two independent radiations of the subfamilies, such that they were more closely associated with other digeneans than to each other. The Tandanicolinae was monophyletic (100% bootstrap support) and was weakly associated with the Gymnophallidae (< 50-55% bootstrap support). Monophyly of the Baccigerinae was supported with 78-87% bootstrap support, and monophyly of the Zoogonidae + Baccigerinae received 77-86% support. The remaining fellodistomid species, Fellodistomum fellis, F. agnotum and Coomera brayi (Fellodistominae) plus Proctoeces maculatus and Complexobursa sp. (Proctoecinae), formed a separate clade with 74-92% bootstrap support. On the basis of molecular, morphological and life-cycle evidence, the subfamilies Baccigerinae and Tandanicolinae are removed from the Fellodistomidae and promoted to familial status. The Baccigerinae is promoted under the senior synonym Faustulidae Poche, 1926, and the Echinobrevicecinae Dronen, Blend & McEachran, 1994 is synonymised with the Faustulidae. Consequently, species that were formerly in the Fellodistomidae are now distributed in three families: Fellodistomidae; Faustulidae (syn. Baccigerinae Yamaguti, 1954); and Tandanicolidae Johnston, 1927. We infer that the use of bivalves as intermediate hosts by this broad range of families indicates multiple host-switching events within the radiation of the Digenea.