登录后查看更多精彩内容~
您需要 登录 才可以下载或查看，没有帐号？立即注册
x
本帖最后由 hillside 于 2014-1-10 12:26 编辑
克里格法是传统的地统计学方法，已在GIS领域得到普遍的应用。下面简介一篇论文《地统计学新方法——贝叶斯最大熵地统计学方法》的摘要，GIS爱好者们可以看看。
张贝 李卫东 杨勇 汪善勤 蔡崇法
【摘要】：贝叶斯最大熵(Bayesian Maximum Entropy,BME)地统计学方法是近年来出现的一种时空地统计学新方法。相对于传统的克里金方法,该法具有坚实的认识论框架和方法学基础。它不需要作线性估值、空间匀质和正态分布的假设,能够融入先验知识和软数据,并且不会损失其中蕴含的有用信息,提高了分析精度。本文首先介绍了BME的基本理论及其估值方法,随后简单描述了该方法的理论发展过程及其在土壤和环境科学上的应用情况,最后对该方法的应用做了总结与展望。经过国外研究者多年的开发和实践,BME方法已经被证明是一个理论上较为成熟,能够应用到实际研究中的优秀地统计学方法,在资源环境评估上有着广泛的应用前景。
【作者单位】： 华中农业大学资源与环境学院;华中农业大学农业部亚热带农业资源与环境重点开放实验室;
【关键词】： 贝叶斯最大熵 地统计学 土壤学 环境科学
【基金】：国家自然科学基金项目(40971269,40801082)资助
BMElib BMElib description BMElib is a powerful MATLAB numerical toolbox of Modern Spatiotemporal Geostatistics implementing the Bayesian Maximum Entropy (BME) theory. BMElib provides a one stop library for 1. both spatial and space/time Geostatistics, dealing with 2. both continuous and categorical variables, implementing 3. the best unbiased non linear estimator (BME), which 4. rigorously processes hard and soft data, but also leads to 5. the (simple, ordinary, universal) kriging methods as linear limiting cases.
Space/Time Geostatistics: Unlike other Geostatistical package, BMElib has been implemented from the start for space/time Geostatistics. Hence BMElib implements the concepts of space/time metrics, space/time neighborhood search, space/time covariance analysis, and space/time estimation. As a result BMElib provides better Geostatistical functions for space/time analysis than classical Geostatistics software (where time is included as merely another spatial dimension), while providing excellent functions in the purely spatial case.
Non-Linear and Non Gaussian Geostatistics: BMElib implements the powerful BME method of Modern Spatiotemporal Geostatistics. BME is a rigorous mathematical framework that allows processing a wide variety of knowledge bases, and leads to the best unbiased non linear estimator. Hard data, soft data, non-Gaussian distributions, etc are automatically integrated in BMElib.
Classical kriging : When restricting the estimation to linear estimation of continuous variables, BME becomes the best unbiased linear estimators of classical Geostatistics. Hence simple kriging, ordinary kriging, universal kriging, etc. are all part of the BMElib toolbox.
Download BMElib is a publicly available software for BMElib users. To get on the BMElib users list, send an email to patrick.bogaert@uclouvain.be (European contact) ormarc_serre@unc.edu (USA contact) with your contact information (name, position, work address) stating your interest to be added to the list. Once added to the list, you will receive an email with a user name and password to download the latest version of BMElib. The latest version of BMElib is the following: Windows users:
MATLAB versions 6.1 to 7.3: Download BMELIB2.0b.zip (restricted access), which is a zipped file containing the source code for BMElib version 2.0b. Just download this file and unzip to a directory anywhere on your hard drive (e.g. C:). This will create your BMElib directory (e.g. C:\BMElib2.0b) with several subdirectories (bmecatlib, …, graphlib, …, mvnlib, etc.).
MATLAB version 7.4 to version 7.7: Do the step described above to download and install BMELIB2.0b.zip, then download executablesForMvnlibMATLAB7.3.zip , unzip its content of five *.mexw32 files and place them in the mvnlib subdirectory. Then delete the corresponding five *.dll files with the same names.
MATLAB version 7.8 (release 2009a) and later: Do the same as above using the five *.mexw32 files and five *.mexw64 files from executablesForMvnlibMATLAB2009a.zip . Note that this provides both the 32-bit and 64-bit files, and MATLAB will choose the correct one depending on whether you are using window 7 with a 32-bit or 64-bit operating system.
Linux users: BMELIB2.0bLinux.tar.gz (restricted access) is a tarred + zipped file containing the source code and executables for BMElib version 2.0b (for Intel Linux users; compiled with GCC v.3.2.2).
Mac users: BMELIB2.0bMacOSX.tar.gz (restricted access) is a tarred + zipped file containing the source code and executables for BMElib version 2.0b (for Macintosh users; compiled with GCC v.3.4.3).
Getting started with BMELIB version 2.0b
Once you have installed BMElib version 2.0b, start MATLAB, move your working directory to the BMElib directory and type startup.m to add the BMElib directory to MATLAB’s search path. Alternatively you can create a shortcut in Windows that starts MATLAB directly in the BMElib directory (in which case you do not need to type startup.m).
To check that the startup command executed properly, move to a new working directory (anywhere on your hard drive), and type “help” at the MATLAB command prompt. MATLAB should list all the BMElib directories (including iolib, graphlib, …, tutorlib, exlib, testslib, etc.). Then type “help testslib”. This should list all the command in the testslib directory (including IOLIBtest, …, MVNLIBtest, etc.)
To test that BMElib is properly functioning, use the test functions in the testslib directory. In particular use “MVNLIBtest“ to test that the five FORTRAN compiled functions of mvnlib are executing properly.
To learn how to use the BMElib functions, use the tutorial functions in the tutorlib directory. Type “help tutorlib” at the command prompt to see a list of the tutorials, and run the tutorials in the order they are listed, i.e. starting with IOLIBtutorial, then GRAPHLIBtutorial, etc.
Finally, the book below contains a detailed explanation of BMElib functions, as well as examples that can be re-created using the functions in the exlib tutorial Christakos, G., P. Bogaert, and M.L. Serre, Temporal GIS: Advanced Functions for Field-Based Applications, Springer-Verlag, New York, N.Y., 217 p., CD ROM included, 2002
Description of earlier versions of BMELIB
BMELIB 1.0: This is the version in the Temporal GIS book, it works with MATLAB 5.3
BMELIB 1.0b: This version fixes some minor bugs of version 1.0, and it works with MATLAB 5.3 to 6.1
BMELIB 1.0c: This version fixes some minor bugs of previous versions, and it works with MATLAB 5.3 to 6.1
BMELIB 2.0b: This version provides a substantial upgrade for many of the functions of version 1.0c. In addition it introduces beta versions for an extension for s/t mapping, an extension for projection, as well as the bmecatlib directory for categorical data. This version works with MATLAB versions 6.1 or later. Obtaining earlier versions of BMElib:
Installation notes
Older files and versions
executablesForMvnlib/ - directory with executables for the mvnlib directory
BMELIB.zip (restricted access) is a zipped file containing the source code for BMElib version 1.0b (Windows users)
BMELIB.tar.gz (restricted access) is a tarred/zipped file containing the source code for BMElib version 1.0b (linux/unix users)
BMELIB1.0c.zip (restricted access) is a zipped file containing the source code for BMElib version 1.0c (Windows users). Use the same installation instructions as for version 1.0b, but using BMELIB1.0c.zip instead of BMELIB.zip.
Consulting services in the application of BME in space time statistics: · to find out about the increasing spread and popularity of spatiotemporal analysis as means to research many natural attributes and processes · to get informed about professional consulting services that are available for space-time geostatistics / statistics / analytical GIS tools · to explore more examples where Knowledge Synthesis / BME is applied
BMEGUI BMEGUI provides a Graphical Users Interface (GUI) to the Bayesian Maximum Entropy Library (BMElib) used to conduct Space/Time geostatistical analysis. Versions of BMEGUI
附：一篇国外最新文献
Improving Bayesian maximum entropy and ordinary Kriging methods for estimating precipitations in a large watershed: a new cluster-based approachBardia Bayat,a Mohsen Nasseri,a Gholamreza Naserb aSchool of Civil Engineering, University of Tehran, Tehran, Iran. bOkanagan School of Engineering, The University of British Columbia, 2329 West Mall, Vancouver, BC V6T 1Z4, Canada.
Corresponding author: Bardia Bayat (e-mail: brdbayat@ut.ac.ir).
Paper handled by associate editor Christine Rivard
Published on the web 28 October 2013. Received April 13, 2013. Accepted October 13, 2013.
Canadian Journal of Earth Sciences, 2014, 51(1): 43-55, 10.1139/cjes-2013-0062 ABSTRACTThe main purpose of this research is to investigate spatial variations of mean annual precipitation in a watershed. As a case study, the research focused on the Namak Lake watershed in Iran. Literature provides various techniques for studying spatial patterns of precipitation in a watershed. These techniques often require a large dataset. On the other hand, nonuniform data distribution in a watershed can reduce the accuracy and reliability of the predictions. To overcome these problems, this research applied the cluster method coupled with ordinary Kriging and Bayesian maximum entropy techniques. An estimated point was modified based on the distance from the point to the cluster center. The research considered elevation variations as a secondary variable. A cross-validation technique was used for evaluating the results of mean annual precipitations. The research compared the results of ordinary Kriging and Bayesian maximum entropy methods with and without the application of the clustering method. The research concluded that the cluster-based method can estimate the dynamics of long-term mean annual precipitation more reliably and accurately. The research also revealed more informative results for the cluster-based method.
| 
窥视卡
雷达卡
发表于 2014-1-1 21:21:31
提升卡
置顶卡
沉默卡
喧嚣卡
变色卡
抢沙发
千斤顶
显身卡
成长值: 0
发表于 2014-1-2 10:16:18
{:eb502:}