Application of Artificial Neural Networks to Complex Groundwater Management Problems

As water quantity and quality problems become increasingly severe, accurate prediction and effective management of scarcer water resources will become critical. In this paper, the successful application of artificial neural network (ANN) technology is described for three types of groundwater prediction and management problems. In the first example, an ANN was trained with simulation data from a physically based numerical model to predict head (groundwater elevation) at locations of interest under variable pumping and climate conditions. The ANN achieved a high degree of predictive accuracy, and its derived state-transition equations were embedded into a multiobjective optimization formulation and solved to generate a trade-off curve depicting water supply in relation to contamination risk. In the second and third examples, ANNs were developed with real-world hydrologic and climate data for different hydrogeologic environments. For the second problem, an ANN was developed using data collected for a 5-year, 8-month period to predict heads in a multilayered surficial and limestone aquifer system under variable pumping, state, and climate conditions. Using weekly stress periods, the ANN substantially outperformed a well-calibrated numerical flow model for the 71-day validation period, and provided insights into the effects of climate and pumping on water levels. For the third problem, an ANN was developed with data collected automatically over a 6-week period to predict hourly heads in 11 high-capacity public supply wells tapping a semiconfined bedrock aquifer and subject to large well-interference effects. Using hourly stress periods, the ANN accurately predicted heads for 24-hour periods in all public supply wells. These test cases demonstrate that the ANN technology can solve a variety of complex groundwater management problems and overcome many of the problems and limitations associated with traditional physically based flow models.     

Kinematic reworking and exhumation within the convergent Alpine Orogen

Kinematic data from the internal zones of the Western Alps indicate both top-to-SE and top-to-NW shearing during synkinematic greenschist facies recrystallisation. Rb/Sr data from white micas from different kinematic domains record a range of ages that does not represent closure through a single thermal event but reflects the variable timing of synkinematic mica recrystallisation at temperatures between 300 and 450 °C. The data indicate an initial phase of accretion and foreland-directed thrusting at ca. 60 Ma followed by almost complete reworking of thrust-related deformation by SE-directed shearing. This deformation is localised within oceanic units of the Combin Zone and the base of the overlying Austroalpine basement, and forms a regional scale shear zone that can be traced for almost 50 km perpendicular to strike. The timing of deformation in this shear zone spans 9 Ma from 45 to 36 Ma. The SE-directed shear leads to local structures that cut upwards in the transport direction with respect to tectonic stratigraphy, and such structures have been interpreted in the past as backthrusts in response to ongoing Alpine convergence. However, on a regional scale, the top-to-SE deformation is related to crustal extension, not shortening, and is coincident with exhumation of eclogites in its footwall. During this extension phase, deformation within the shear zone migrated both spatially and temporally giving rise to domains of older shear zone fabrics intercalated with zones of localised reworking. Top-NW kinematics preserved within the Combin Zone show a range of ages. The oldest (48 Ma) may reflect the final stages of emplacement of Austroalpine Units above Piemonte oceanic rocks prior to the onset of extension. However, much of the top-to-NW deformation took place over the period of extension and may reflect either continuing or episodic convergence or tectonic thinning of the shear zone.⁴⁰Ar/³⁹Ar data from the region are complicated due to the widespread occurrence of excess ⁴⁰Ar in eclogite facies micas and partial Ar loss during Alpine heating. Reliable ages from both eclogite and greenschist facies micas indicate cooling ages in different tectonic units of between 32 and 40 Ma. These ages are slightly younger than Rb/Sr deformation ages and suggest that cooling below ca. 350 °C occurred after juxtaposition of the units by SE-directed extensional deformation.Our data indicate a complex kinematic history involving both crustal shortening and extension within the internal zones of the Alpine Orogen. To constrain the palaeogeographic and geodynamic evolution of the Alps requires that these data be integrated with data from the more external zones of the orogen. Complexity such as that described is unlikely to be restricted to the Western Alps and spatially and temporally variable kinematic data are probably the norm in convergent orogens. Recognising such features is fundamental to the correct tectonic interpretation of both modern and ancient orogens.     

Thick lava flows of Karisimbi Volcano, Rwanda: insights from SIR-C interferometric topography

 We use a digital elevation model (DEM) derived from interferometrically processed SIR-C radar data to estimate the thickness of massive trachyte lava flows on the east flank of Karisimbi Volcano, Rwanda. The flows are as long as 12 km and average 40–60 m (up to >140 m) in thickness. By calculating and subtracting a reference surface from the DEM, we derived a map of flow thickness, which we used to calculate the volume (up to 1 km³ for an individual flow, and 1.8 km³ for all the identified flows) and yield strength of several flows (23–124 kPa). Using the DEM we estimated apparent viscosity based on the spacing of large folds (1.2×10¹² to 5.5×10¹² Pa s for surface viscosity, and 7.5×10¹⁰ to 5.2×10¹¹ Pa s for interior viscosity, for a strain interval of 24 h). We use shaded-relief images of the DEM to map basic flow structures such as channels, shear zones, and surface folds, as well as flow boundaries. The flow thickness map also proves invaluable in mapping flows where flow boundaries are indistinct and poorly expressed in the radar backscatter and shaded-relief images. Received: 6 September 1997 / Accepted: 15 May 1998     

The role of carbon and oxygen in cosmic gases: some applications to the chemistry and mineralogy of enstatite chondrites

A new condensation sequence appears if the $\text{C}\text{O}$ ratio in a gas of otherwise solar composition is increased by less than a factor of two. As the ratio increases from the solar value of 0.6 to ? 1 the gas becomes extremely reduced, the condensation temperatures of silicates and oxides are depressed markedly ～ 400 K and a new suite of refractory minerals appears: AIN, CaS, MgS, SiC, TiN, graphite, Si₂N₂O and probably metastable (Fe,Ni)³C. Many of these minerals are unique to enstatite chondrites and may be analogues of the refractory silicates and oxides found in more oxidized meteorites such as Allende.The change in chemistry is related to the stability of CO, the most stable C or O compound at high T. Since the elements occur in a 1:1 ratio in CO, only the element which is in excess is free to form other compounds. But as T decreases CO reacts with H₂ to form graphite, CH₄ or other hydrocarbons thereby freeing O to form H₂O. If equilibrium is maintained oxides and silicates form at about 1000 K ($\text{C}\text{O}\text{> 1}$, P_τ = 10^?4atm) as products of reactions among the carbides, nitrides, sulfides and the gas. The possibility that equilibrium was not maintained among the C-bearing species was also investigated. If either graphite or CH₄ does not form as predicted the stability fields of the reduced minerals expands to lower temperatures. If neither graphite nor CH₄ form as predicted, CO remains stable and the nebular gas is highly reduced at all temperatures.Enstatite chondrites appear to have originated in a region of the nebula where the $\text{C}\text{O}$ ratio was somewhat higher than the solar value. Various fractionation mechanisms are considered. An interesting possibility is that graphite, which is quite refractory under a wide range of conditions, survived the collapse of the solar nebula.     

Dislocation substructure of mantle-derived olivine as revealed by selective chemical etching and transmission electron microscopy

Cleaved and mechanically polished surfaces of olivine from peridotite xenoliths from San Carlos, Arizona, were chemically etched using the techniques of Wegner and Christie (1974). Dislocation etch pits are produced on all surface orientations and they tend to be preferentially aligned along the traces of subgrain boundaries, which are approximately parallel to (100), (010), and (001). Shallow channels were also produced on (010) surfaces and represent dislocations near the surface that are etched out along their lengths. The dislocation etch channel loops are often concentric, and emanate from (100) subgrain boundaries, which suggests that dislocation sources are in the boundaries. Data on subgrain misorientation and dislocation line orientation and arguments based on subgrain boundary energy minimization are used to characterize the dislocation structures of the subgrain boundaries. (010) subgrain boundaries are of the twist type, composed of networks of [100] and [001] screw dislocations. Both (100) and (001) subgrain boundaries are tilt walls composed of arrays of edge dislocation with Burgers vectors b=[100] and [001], respectively. The inferred slip systems are {001} 〈100〉, {100} 〈001〉, and {010} 〈100〉 in order of diminishing importance. Exploratory transmission electron microscopy is in accord with these identifications. The flow stresses associated with the development of the subgrain structure are estimated from the densities of free dislocations and from the subgrain dimensions. Inferred stresses range from 35 to 75 bars using the free dislocation densities and 20 to 100 bars using the subgrain sizes.     

Palynological evidence for early Holocene aridity in the southern Sierra Nevada, California

Sediments of Balsam Meadow have produced a 11,000-yr pollen record from the southern Sierra Nevada of California. The Balsam Meadow diagram is divided into three zones. (1) The Artemisia zone (11,000–7000 yr B.P.) is characterized by percentages of sagebrush (Artemisia) and other nonarboreal pollen higher than can be found in the modern local vegetation. Vegetation during this interval was probably similar to the modern vegetation on the east slope of the Sierra Nevada and the climate was drier than that of today. (2) Pinus pollen exceeded 80% from 7000 to 3000 yr B.P. in the Pinus zone. The climate was moister than during the Artemisia zone. (3) Fir (Abies, Cupressaceae, and oak (Quercus) percentages increased after 3000 yr B.P. in the Abies zone as the modern vegetation at the site developed and the present cool-moist climatic regime was established. Decreased fire frequency after 1200 yr B.P. is reflected in decreased abundance of macroscopic charcoal and increased concentration of Abies magnifica and Pinus murrayana needles.     

Mortar coupling and upscaling of pore-scale models

Pore-scale models are becoming increasingly useful as predictive tools for modeling flow and transport in porous media. These models can accurately represent the 3D pore-structure of real media. Currently first-principles modeling methods are being employed for obtaining qualitative and quantitative behavior. Generally, artificial, simple boundary conditions are imposed on a model that is used as a stand-alone tool for extracting macroscopic parameters. However, realistic boundary conditions, reflecting flow and transport in surrounding media, may be necessary for behavior that occurs over larger length scales or including pore-scale models in a multiscale setting. Here, pore-scale network models are coupled to adjacent media (additional pore-scale or continuum-scale models) using mortars. Mortars are 2D finite-element spaces employed to couple independent subdomains by enforcing continuity of pressure and flux at shared boundary interfaces. While mortars have been used in the past to couple subdomains of different models, physics, and meshes, they are extended here for the first time to pore-scale models. The approach is demonstrated by modeling single-phase flow in coupled pore-scale models, but the methodology can be utilized to model dynamic processes and perform multiscale modeling in 3D continuum simulators for flow and transport.     

Topographic analyses of K*lauea Volcano, Hawai'i, from interferometric airborne radar

 We analyze digital topographic data collected in September 1993 over a ∼500-km² portion of K*lauea Volcano, Hawai'i, by the C-band (5.6-cm wavelength) topographic synthetic aperture radar (TOPSAR) airborne interferometric radar. Field surveys covering an ∼1-km² area of the summit caldera and the distal end of an ∼8-m-thick 'a'* flow indicate that the 10-m spatial resolution TOPSAR data have a vertical accuracy of 1–2 m over a variety of volcanic surfaces. After conversion to a common datum, TOPSAR data agree favorably with a digital elevation model (DEM) produced by the U.S. Geological Survey (USGS), with the important exception of the region of the ongoing eruption (which postdates the USGS DEM). This DEM comparison gives us confidence that subtracting the USGS data from TOPSAR data will produce a reasonable estimate of the erupted volume as of September 1993. This subtraction produces dense rock equivalent (DRE) volumes of 392, 439, and 90×10⁶ m³ for the Pu'u '*'*, K*pa'ianah*, and episode 50–53 stages of the eruption, respectively. These are 124, 89, and 94% of the volumes calculated by staff of the Hawaiian Volcano Observatory (HVO) but do not include lava of K*pa'ianah* and episodes 50–53 that flowed into the ocean and are thus invisible to TOPSAR. Accounting for this lava increases the TOPSAR volumes to 124, 159, and 129% of the HVO volumes. Including the ±2-m uncertainty derived from the field surveys produces TOPSAR-derived volumes for the eruption as a whole that range between 81 and 125% of the USGS-derived values. The vesicularity- and ocean-corrected TOPSAR volumes yield volumetric eruption rates of 4.5, 4.5, and 2.7 m³/s for the three stages of the eruption, which compare with HVO-derived values of 3.6, 2.8, and 2.1 m³/s, respectively. Our analysis shows that care must be taken when vertically registering the TOPSAR and USGS DEMs to a common datum because C-band TOPSAR penetrates only partially into thick forest and therefore produces a DEM within the tree canopy, whereas the USGS DEM is adjusted for vegetation. Received: 28 April 1998 / Accepted: 1 February 1999