Land cover classification by an artificial neural network with ancillary information

Abstract

Remote sensing is an important source of land cover data required by many GIS users. Land cover data are typically derived from remotely–sensed data through the application of a conventional statistical classification. Such classification techniques are not, however, always appropriate, particularly as they may make untenable assumptions about the data and their output is hard, comprising only the code of the most likely class of membership. Whilst some deviation from the assumptions may be tolerated and a fuzzy output may be derived, making more information on class membership properties available, alternative classification procedures are sometimes required. Artificial neural networks are an attractive alternative to the statistical classifiers and here one is used to derive a fuzzy classification output from a remotely–sensed data set that may be post–processed with ancillary data available in a GIS to increase the accuracy with which land cover may be mapped. With the aid ancillary information on soil type and prior knowledge of class occurrence the accuracy of an artificial neural network classification was increased by 29–93 to 77–37 per cent. An artificial neural network can therefore be used generate a fuzzy classification output that may be used with other data sets in a GIS, which may not have been available to the producer of the classification, to increase the accuracy with which land cover may be classified.     

First Data on the Isotopic Age of Zircons from Rocks of the Roseta Ultramafic Massif,Southern Margin of the São Francisco Craton (Brazil)

Doklady Earth Sciences - The results of isotope U–Pb dating of zircons from lherzolite and vein olivine orthopyroxenite composing the Roseta ultramafic massif are presented. The zircons...     

Wide-field VLBI imaging

We discuss the technique of Wide-field imaging as it applies to Very Long Baseline Interferometry (VLBI). In the past VLBI data sets were usually averaged so severely that the field-of-view was typically restricted to regions extending a few hundred milliarcseconds from the phase centre of the field. Recent advances in data analysis techniques, together with increasing data storage capabilities, and enhanced computer processing power, now permit VLBI images to be made whose angular size represents a significant fraction of an individual antenna's primary beam. This technique has recently been successfully applied to several large separation gravitational lens systems, compact Supernova Remnants in the starburst galaxy M82, and two faint radio sources located within the same VLA FIRST field. It seems likely that other VLBI observing programmes might benefit from this wide-field approach to VLBI data analysis. With the raw sensitivity of global VLBI set to improve by a factor 4–5 over the coming few years, the number of sources that can be detected in a given field will rise considerably. In addition, a continued progression in VLBI's ability to image relatively faint and extended low brightness temperature features (such as hot-spots in large-scale astrophysical jets) is also to be expected. As VLBI sensitivity approaches the μJy level, a wide-field approach to data analysis becomes inevitable.     

A portable extruder for accurate and precise high-resolution subsampling of unconsolidated sediment cores

Journal of Paleolimnology - We developed a portable extruder for precise and accurate high-resolution subsampling of unconsolidated sediment cores. This extruder is capable of producing subsamples...     

XMM–Newton observations of bright ROSAT selected active galactic nuclei with low intrinsic absorption

We present a sample of 21 ROSAT bright active galactic nuclei (AGNs), representing a range of spectral classes, and selected for follow-up snapshot observations with XMM–Newton . The typical exposure was between 5 and 10 ks. The objects were primarily selected on the bases of X-ray brightness and not on hardness ratio; thus the sample cannot be strictly defined as a 'soft'sample. One of the main outcomes from the XMM–Newton observations was that all of the AGN, including 11 type 1.8–2 objects, required low levels of intrinsic absorption  ( N _H≲ 10²¹ cm⁻²)  . The low absorption in type 2 systems is a challenge to account for in the standard orientation-based unification model, and we discuss possible physical and geometrical models which could elucidate the problem. Moreover, there does not appear to be any relation between the strength and shape of the soft excess, and the spectral classification of the AGN in this sample. We further identify a number of AGN which deserve deeper observations or further analysis: for example, the low-ionization nuclear emission regions (LINERs) NGC 5005 and NGC 7331, where optically thin thermal and extended emission is detected, and the narrow-line Seyfert 1 II Zw 177, which shows a broad emission feature at ∼ 5.8 keV.     

The excitation mechanism of [Fe XIV] 5303 å line in the inner regions of solar corona

The line intensity of the green coronal line and the continuum intensity are derived from the filter and white light photographs of the solar corona obtained during the 1980 total solar eclipse. Ratio of the line to continuum intensity is plotted against the radial distancer(=R/R₀,R ₀ is the solar radius), in various position angles. A simple model assuming an electron density dependence of the line and continuum intensities suggests a dominant collisional mechanism for the excitation of the line in the innermost regions (～ 1.4R ₀). The measured line to continuum ratio tends to a constant value at different radial distances in different position angles. The constancy of the measured line to continuum ratio indicates significant radiative excitation beyond 1.4 R₀, in some of the position angles.     

Interplanetary Nanodust Detection by the Solar Terrestrial Relations Observatory/WAVES Low Frequency Receiver

New measurements using radio and plasma-wave instruments in interplanetary space have shown that nanometer-scale dust, or nanodust, is a significant contributor to the total mass in interplanetary space. Better measurements of nanodust will allow us to determine where it comes from and the extent to which it interacts with the solar wind. When one of these nanodust grains impacts a spacecraft, it creates an expanding plasma cloud, which perturbs the photoelectron currents. This leads to a voltage pulse between the spacecraft body and the antenna. Nanodust has a high charge/mass ratio, and therefore can be accelerated by the interplanetary magnetic field to the speed of the solar wind: significantly faster than the Keplerian orbital speeds of heavier dust. The amplitude of the signal induced by a dust grain grows much more strongly with speed than with mass of the dust particle. As a result, nanodust can produce a strong signal despite its low mass. The WAVES instruments on the twin Solar TErrestrial RElations Observatory spacecraft have observed interplanetary nanodust particles since shortly after their launch in 2006. After describing a new and improved analysis of the last five years of STEREO/WAVES Low Frequency Receiver data, we present a statistical survey of the nanodust characteristics, namely the rise time of the pulse voltage and the flux of nanodust. We show that previous measurements and interplanetary dust models agree with this survey. The temporal variations of the nanodust flux are also discussed.     

Distinguishing lower and upper crustal processes in magmas erupted during the buildup to the 7.7 ka climactic eruption of Mount Mazama, Crater Lake, Oregon, using 238U–230Th disequilibria

Uranium-series isotope ratios determined for 35 volcanic rocks and 4 glass separates erupted from ~36 to 4.8 ka at Mt. Mazama, Crater Lake, Oregon, identify both ²³⁰Th-excess and ²³⁸U-excess components. U–Th isotope compositions cover a wide range, exceeding those previously measured for the Cascade arc. Age-corrected (²³⁰Th/²³²Th) and (²³⁸U/²³²Th) activity ratios range from 1.113 to 1.464 and from 0.878 to 1.572 (44.4 % ²³⁰Th-excess to 8.8 % ²³⁸U-excess), respectively. The most distinctive aspect of the data set is the contrast in U–Th isotope ratios between low and high Sr (LSr, HSr) components that have been previously identified in products of the 7.7 ka caldera-forming climactic eruption and preclimactic rhyodacite lavas. The LSr component exclusively contains ²³⁸U-excess, but the HSr component, as well as more primitive lavas, are marked by ²³⁰Th-excess. ²³⁰Th-excesses such as those recorded at Mt. Mazama are commonly observed in the Cascades. Melting models suggest that high ²³⁰Th-excesses observed in the more primitive lavas evolved through mixing of a mantle melt with a partial melt of a mafic lower crustal composition that contained garnet in the residuum that was produced through dehydration melting of amphibolite that was initially garnet free. Dehydration melting in the lower crust offers a solution to the “hot-slab paradox” of the Cascades, where low volatile contents are predicted due to high slab temperatures, yet higher water contents than expected have been documented in erupted lavas. The ²³⁸U-excess observed at Mt. Mazama is rare in Cascade lavas, but occurs in more than half of the samples analyzed in this study. Traditionally, ²³⁸U-excess in arc magmas is interpreted to reflect slab fluid fluxing. Indeed, ²³⁸U-excess in arcs is common and likely masks ²³⁰Th-excess resulting from lower crustal interaction. Isotopic and trace element data, however, suggest a relatively minor role for slab fluid fluxing in the Cascades. We propose that ²³⁸U-excess reflects melting and assimilation of young, hydrothermally altered upper crust. The processes related to generating ²³⁸U-excess are likely important features at Mt. Mazama that accompanied development of a large-scale silicic magma chamber that led to the caldera-forming eruption.     

Comparison of dust‐to‐gas ratios in luminous,ultraluminous, and hyperluminous infrared galaxies

The dust‐to‐gas ratios in three different samples of luminous, ultraluminous, and hyperluminous infrared galaxies are calculated by modelling their radio to soft X‐ray spectral energy distributions (SED) using composite models which account for the photoionizing radiation from H II regions, starbursts, or AGNs, and for shocks. The models are limited to a set which broadly reproduces the mid‐IR fine structure line ratios of local, IR bright, starburst galaxies. The results show that two types of clouds contribute to the IR emission. Those characterized by low shock velocities and low preshock densities explain the far‐IR dust emission, while those with higher velocities and densities contribute to the mid‐IR dust emission. Clouds with shock velocities of 500 km s^–1 prevail in hyperluminous infrared galaxies. An AGN is found in nearly all of the ultraluminous infrared galaxies and in half of the luminous infrared galaxies of the sample. High IR luminosities depend on dust‐to‐gas ratios as high as ∼0.1 by mass, however most hyperluminous IR galaxies show dustto‐gas ratios much lower than those calculated for the luminous and ultraluminous IR galaxies. (© 2007 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)