Predictive error dependencies when using pilot points and singular value decomposition in groundwater model calibration

A significant practical problem with the pilot point method is to choose the location of the pilot points. We present a method that is intended to relieve the modeler from much of this responsibility. The basic idea is that a very large number of pilot points are distributed more or less uniformly over the model area. Singular value decomposition (SVD) of the (possibly weighted) sensitivity matrix of the pilot point based model produces eigenvectors of which we pick a small number corresponding to significant eigenvalues. Super parameters are defined as factors through which parameter combinations corresponding to the chosen eigenvectors are multiplied to obtain the pilot point values. The model can thus be transformed from having many-pilot-point parameters to having a few super parameters that can be estimated by nonlinear regression on the basis of the available observations. (This technique can be used for any highly parameterized groundwater model, not only for models parameterized by the pilot point method.)     

Water and salt balance modelling to predict the effects of land-use changes in forested catchments. 1. Small catchment water balance model

A long-term water balance model has been developed to predict the hydrological effects of land-use change (especially forest clearing) in small experimental catchments in the south-west of Western Australia. This small catchment model has been used as the building block for the development of a large catchment-scale model, and has also formed the basis for a coupled water and salt balance model, developed to predict the changes in stream salinity resulting from land-use and climate change. The application of the coupled salt and water balance model to predict stream salinities in two small experimental catchments, and the application of the large catchment-scale model to predict changes in water yield in a medium-sized catchment that is being mined for bauxite, are presented in Parts 2 and 3, respectively, of this series of papers. The small catchment model has been designed as a simple, robust, conceptually based model of the basic daily water balance fluxes in forested catchments. The responses of the catchment to rainfall and pan evaporation are conceptualized in terms of three interdependent subsurface stores A, B and F. Store A depicts a near-stream perched aquifer system; B represents a deeper, permanent groundwater system; and F is an intermediate, unsaturated infiltration store. The responses of these stores are characterized by a set of constitutive relations which involves a number of conceptual parameters. These parameters are estimated by calibration by comparing observed and predicted runoff. The model has performed very well in simulations carried out on Salmon and Wights, two small experimental catchments in the Collie River basin in south-west Western Australia. The results from the application of the model to these small catchments are presented in this paper.     

Studies on twelve common bivalve larvae,with notes on bivalve spawning seasons in New Zealand

Twelve common bivalve larvae occurring in the plankton from the Bay of Islands (35°15'S, 174°10'E), Wellington Harbour (41°16'S, 174°51'E), and off Raumati Beach (40°56'S, 174°58'E), New Zealand, during 1970–72 are described and, wherever possible, provisionally identified. The seasonal occurrences of these larvae in the plankton are also described. Information on the spawning cycles of some New Zealand adult bivalves is reviewed; although some species have a short (4 months or less) spawning season, for most it is much longer, possibly with ‘trickle’ spawning through several months of the year.     

Direct induction of compact rule-based classifiers for resource mapping

Abstract

Rule-based classifiers are used regularly with geographical information systems to map categorical attributes on the basis of a set of numeric or unordered categorical attributes. Although a variety of methods exist for inducing rule-based classifiers from training data, these tend to produce large numbers of rules when the data has noise. This paper describes a method for inducing compact rule-sets whose classification accuracy can, at least in some domains, compare favourably with that achieved by larger less succinct rule-sets produced by alternative methods. One rule is induced for each output class. The condition list for this rule represents a box in n-dimensional attribute space, formed by intersecting conditions which exclude other classes. Despite this simplicity, the classifier performed well in the test application prediction of soil classes in the Port Hills, New Zealand, on the basis of regolith type and topographic attributes obtained from a digital terrain model.     

Simulation of the effect of doubled atmospheric CO2 on the climate of northern and central California

A Local Climate Model (LCM) is described that can provide a high-resolution (10 km) simulation of climate resulting from a doubling of atmospheric CO₂ concentrations. A canonicalregression function is used to compute the monthly temperature (mean of daily-maximum-temperature) and precipitation for any point, given a set of predictor variables. Predictor variables represent the influence of terrain, sea-surface temperature (SST), windfields, CO₂ concentration, and solar radiation on climate. The canonical-regression function is calibrated and validated using empirical windfield, SST, and climate data from stations in the western U.S. To illustrate an application of the LCM, the climate of northern and central California is simulated for a doubled CO₂ (600 ppmv) and a control scenario (300 ppmv CO₂). Windfields and SSTs used to compute predictor variables are taken from general circulation model simulations for these two scenarios. LCM solutions indicate that doubling CO₂ will result in a 3 C° increase in January temperature, a 2 C° increase in July temperature, a 16 mm (37%) increase in January precipitation, and a 3 mm (46%) increase in July precipitation.     

Automated forecasting of volcanic ash dispersion utilizing Virtual Globes

There are over 100 active volcanoes in the North Pacific (NOPAC) region, most of which are located in sparsely populated areas. Dispersion models play an important role in forecasting the movement of volcanic ash clouds by complementing both remote sensing data and visual observations from the ground and aircraft. Puff is a three-dimensional dispersion model, primarily designed for forecasting volcanic ash dispersion, used by the Alaska Volcano Observatory and other agencies. Since early 2007, the model is in an automated mode to predict the movement of airborne volcanic ash at multiple elevated alert status volcanoes worldwide to provide immediate information when an eruption occurs. Twelve of the predictions are within the NOPAC region, nine more within the southern section of the Pacific ring of fire and the others are in Europe and the Caribbean. Model forecasts are made for initial ash plumes ranging from 4 to 20 km altitude above sea level and for a 24-h forecast period. This information is made available via the Puff model website. Model results can be displayed in Virtual Globes for three-dimensional visualization. Here, we show operational Puff predictions in two and three-dimensions in Google Earth^®, both as iso-surfaces and particles, and study past eruptions to illustrate the capabilities that the Virtual Globes can provide. In addition, we show the opportunity that Google Maps^® provides in displaying Puff operational predictions via an application programming web interface and how radiosonde data (vertical soundings) and numerical weather prediction vertical profiles can be displayed in Virtual Globes for assisting in estimating ash cloud heights.     

Coal resources,production, and quality in the Eastern kentucky coal field: Perspectives on the future of steam coal production

The Eastern Kentucky coal field, along with adjacent portions of Virginia and southern West Virginia, is part of the greatest production concentration of high-heating-value, low-sulfur coal in the United States, accounting for over 27% of the 1993 U.S. production of coal of all ranks. Eastern Kentucky's production is spread among many coal beds but is particularly concentrated in a limited number of highquality coals, notably the Pond Creek coal bed and its correlatives, and the Fire Clay coal bed and its correlatives. Both coals are relatively low ash and low sulfur through the areas of the heaviest concentration of mining activity. We discuss production trends, resources, and the quality of in-place and clean coal for those and other major coals in the region.     

Chemical fractionations in meteorites—IX. C3 chondrites

Four C3V chondrites (Grosnaja, Kaba, Mokoia, Vigarano) and three C3O chondrites (Felix, Kainsaz, and Lancé) were analyzed by radiochemical neutron activation for 17 trace elements. Both classes show a typical chondritic step pattern, reflecting loss of volatiles during chondrule formation. Elements condensing above 1300 K (U, Re, Ir, Ni) are present in essentially Cl chondrite proportions, while moderately volatile elements condensing between 1300 K and 800 K (Ge, Rb, Ag) are depleted by a factor of 0.44. However, elements condensing below 700 K (S, Cs, Bi, Tl, Br, Se, Te, In, Cd) are depleted to a still greater degree, and more so in the Ornans subclass (factor of 0.24, except Cd 0.007) than in the Vigarano subclass (factor of 0.29). This additional depletion may be due to a slight (less than 3-fold) dust-gas fractionation, by settling of dust to the median plane of the solar nebula. Among other chondrite classes, ordinary chondrites show a similar depletion, but C2 chondrites do not. Possibly the undepleted meteorites formed in one of the convection zones of the nebula predicted by Cameron and Pine, whereas the depleted meteorites formed in a quiescent region.The condensation of chalocophile elements as a function of H₂S partial pressure is discussed, in an attempt to explain the drastic difference in Cd abundance between the two subclasses. It appears that the $\text{H}{}_2\text{S}\text{H}{}_2$ ratio is the key variable. C3O's seem to have condensed in a region where enough metallic Fe was present to buffer the H₂S pressure, while C3V's condensed in a more oxidized region, where H₂S was in excess. Accretion temperatures, for an assumed nebular pressure of 10^?5 atm, were between 415 and 430 K for C3O's and less than 440 K for C3V's.Two slightly volatile elements, Sb and Au, show variable depletion, presumably reflecting variable loss during chondrule formation. Indeed, their depletion correlates with the abundance of iron-poor olivine, a measure of the peak temperature and time during chondrule formation.