Automatic Determination of Stellar Atmospheric Parameters Using Neural Networks and Instance-Based Machine Learning

In this article we show how machine learning methods can beeffectively applied to the problem of automatically predictingstellar atmospheric parameters from spectral information, a veryimportant problem in stellar astronomy. We apply feedforwardneural networks, Kohonen's self-organizing maps andlocally-weighted regression to predict the stellar atmosphericparameters effective temperature, surface gravity and metallicityfrom spectral indices. Our experimental results show that thethree methods are capable of predicting the parameters with verygood accuracy. Locally weighted regression gives slightly betterresults than the other methods using the original dataset asinput, while self-organizing maps outperform the other methods when significant amounts of noise are added. We also implemented a heterogeneous ensemble of predictors, combining the results given by the three algorithms. This ensemble yields better results than any of the three algorithms alone, using both the original and the noisy data.     

NORTHERN AND SOUTHERN HEMISPHERIC POLAR LOWS——A COMPARATIVE STUDY

Northern Hemispheric“polar lows”,i.e.maritime,small,but fairly intense cyclonic systems have been studied formore than two decades.Recently researchers have turned their attention towards the occurrence of polar lows and othersmall-scale cyclonic disturbances on the Southern Hemisphere.In this paper a short account of our present knowledgeof Northern Hemispheric polar lows will be given,followed by some preliminary results of an investigation of the natureand occurrence of polar lows on the Southern Hemisphere.     

A MESO-β SCALE SIMULATION OF THE EFFECTS OF BOREAL FOREST ECOSYSTEM ON THE LOWER ATMOSPHERE

Based on the Intensive Field Campaign(IFC-1)data of Boreal Ecosystem-Atmosphere Study(BOREAS).a three-dimensional meso-β scale model is used to simulate the effect of boreal forests onthe lower atmosphere.A fine horizontal resolution of 2 km×2 km is used in order to distinguish thevegetative heterogeneity in the boreal region.A total of 20×25 grid points cover the entire sub-modeling area in BOREAS' South Study Area(SSA).The ecosystem types and their coverage ineach grid square are extracted from the North American Land Cover Characteristics Data Base(NALCCD)generated by the U.S.Geographical Survey(USGS)and the University of Nebraska-Lincoln(UNL).The topography of the study area is taken from the Digital Elevation Map(DEM)of USGS.The model outputs include the components of the energy balance budget within the canopyand at the ground.the turbulence parameters in the atmospheric boundary layer and the wind.temperature and humidity profiles extending up to a height of 1500 m.In addition to the fine timeand spatial step,the unique feature of the present model is the incorporation of both dynamic andbiological effects of the Boreal forest into the model parameterization scheme.The model resultscompare favorably with BOREAS' IFC-1 data in 1994 when the forest was in the luxuriant growingperiod.     

The first World Atlas of the artificial night sky brightness

We present the first World Atlas of the zenith artificial night sky brightness at sea level. Based on radiance-calibrated high-resolution DMSP satellite data and on accurate modelling of light propagation in the atmosphere, it provides a nearly global picture of how mankind is proceeding to envelop itself in a luminous fog. Comparing the Atlas with the United States Department of Energy (DOE) population density data base, we determined the fraction of population who are living under a sky of given brightness. About two-thirds of the World population and 99 per cent of the population in the United States (excluding Alaska and Hawaii) and European Union live in areas where the night sky is above the threshold set for polluted status. Assuming average eye functionality, about one-fifth of the World population, more than two-thirds of the United States population and more than one half of the European Union population have already lost naked eye visibility of the Milky Way. Finally, about one-tenth of the World population, more than 40 per cent of the United States population and one sixth of the European Union population no longer view the heavens with the eye adapted to night vision, because of the sky brightness.     

抗差估计的初值选择

对L1 范估计的平差值和LS估计的平差值作为抗差估计的初值进行了比较 ,指出在 ρ函数是严凸函数时 ,二者无明显的区别。因为LS估计简单且易于计算 ,因此建议用LS估计的平差值作为抗差估计的初值。     

Phasing of segmented mirrors: a new algorithm for piston detection

The point spread function of a segmented-mirror telescope is severely affected by segment misalignment, which can nullify the performance of adaptive optics systems. The piston and tilt of each segment must be precisely adjusted in relation to the other segments. Furthermore, the direct detection of the alignment error with natural stars would be desirable in order to monitor the errors during astronomical observation.
We have studied the lost information of the piston error caused by the presence of atmospheric turbulence in the measurements of curvature, and present a new algorithm for obtaining the local piston using the curvature sensor. A phase-wrapping effect is shown as responsible for the loss of curvature information and so the piston errors can no longer adequately be mapped; this happens not only in the presence of atmospheric turbulence, but also in its absence.
Good results are obtained using a new iterative method for obtaining the local piston error map. In the presence of atmospheric perturbation, the turbulent phase information obtained from a Shack–Hartmann sensor is introduced in our new iterative method. We propose a hybrid sensor composed of a curvature sensor and a Shack–Hartmann sensor, in order to complete all the information for the phasing. This design takes a short computation time and could be used in real time inside an adaptive optics system, where tilt and piston errors must be corrected.     

Particle-mesh simulations of the Lyα forest

We have analysed short-exposure high-resolution images obtained on the William Herschel Telescope (WHT), Big Telescope Azimuthal (BTA) and CanadaFranceHawaii Telescope (CFHT) in order to investigate mechanical vibrations that are capable of reducing the high angular resolution allowed by speckle-interferometric methods. After filtering the photon noise we have computed power spectra of the image centres of gravity. In these spectra we have found vibration features that differ from one telescope to another and that vary in particular with the zenith angle. In the case of the WHT we discuss these results in terms of possible causes and in terms of degradation of the transfer function. We present some means to improve the data.     

广西芒场锡多金属矿田稳定同位素组成对矿床成因探讨

笔者曾对广西芒场锡多金属矿田的稳定同位素进行研究。本文根据硫、铅、氢、氧、碳等稳定同位素组成和锶的初始值提供的信息.探讨了矿床成因。并结合矿田矿床地质特征、控矿条件及有关统计参数,参考前人对矿床认识的基础上,修正提出了该矿田混合热液成矿模式,可供类似矿床研究和找矿的参考。