International Summer School on Turbulence Diffusion 2006 Multifractal Analysis in B&W Soil Images Ana M. Tarquis Dpto. de Matemática.

Slides:



Advertisements
Similar presentations
ROMA (Rank-Ordered Multifractal Analysis) for Intermittent Fluctuations with Global Crossover Behavior Sunny W. Y. Tam 1,2, Tom Chang 3, Paul M. Kintner.
Advertisements

Design of Experiments Lecture I
A.M. Alonso, C. García-Martos, J. Rodríguez, M. J. Sánchez Seasonal dynamic factor model and bootstrap inference: Application to electricity market forecasting.
CHAPTER 2 Building Empirical Model. Basic Statistical Concepts Consider this situation: The tension bond strength of portland cement mortar is an important.
Characterizing Non- Gaussianities or How to tell a Dog from an Elephant Jesús Pando DePaul University.
Snow Trends in Northern Spain. Analysis and Simulation with Statistical Downscaling Methods Thanks to: Daniel San Martín, Sixto.
Challenge the future Delft University of Technology Blade Load Estimations by a Load Database for an Implementation in SCADA Systems Master Thesis.
Maximum Covariance Analysis Canonical Correlation Analysis.
The Global Digital Elevation Model (GTOPO30) of Great Basin Location: latitude 38  15’ to 42  N, longitude 118  30’ to 115  30’ W Grid size: 925 m.
The General Linear Model. The Simple Linear Model Linear Regression.
STAT 497 APPLIED TIME SERIES ANALYSIS
Mining for High Complexity Regions Using Entropy and Box Counting Dimension Quad-Trees Rosanne Vetro, Wei Ding, Dan A. Simovici Computer Science Department.
The Simple Linear Regression Model: Specification and Estimation
Resampling techniques Why resampling? Jacknife Cross-validation Bootstrap Examples of application of bootstrap.
Alon Arad Alon Arad Hurst Exponent of Complex Networks.
Earthquake spatial distribution: the correlation dimension (AGU2006 Fall, NG43B-1158) Yan Y. Kagan Department of Earth and Space Sciences, University of.
Sampling Distributions
1 Simple Linear Regression Chapter Introduction In this chapter we examine the relationship among interval variables via a mathematical equation.
STOCHASTIC GEOMETRY AND RANDOM GRAPHS FOR THE ANALYSIS AND DESIGN OF WIRELESS NETWORKS Haenggi et al EE 360 : 19 th February 2014.
Lecture II-2: Probability Review
Modelling and Simulation 2008 A brief introduction to self-similar fractals.
Seiji Armstrong Huy Luong Huy Luong Alon Arad Alon Arad Kane Hill Kane Hill.
Hydrologic Statistics
Entropy and some applications in image processing Neucimar J. Leite Institute of Computing
Introduction --Classification Shape ContourRegion Structural Syntactic Graph Tree Model-driven Data-driven Perimeter Compactness Eccentricity.
Informational Network Traffic Model Based On Fractional Calculus and Constructive Analysis Vladimir Zaborovsky, Technical University, Robotics Institute,
Copyright © 2010, 2007, 2004 Pearson Education, Inc. All Rights Reserved Section 10-3 Regression.
Alternative Measures of Risk. The Optimal Risk Measure Desirable Properties for Risk Measure A risk measure maps the whole distribution of one dollar.
Southern Taiwan University Department of Electrical engineering
BPS - 3rd Ed. Chapter 211 Inference for Regression.
1 LES of Turbulent Flows: Lecture 1 Supplement (ME EN ) Prof. Rob Stoll Department of Mechanical Engineering University of Utah Fall 2014.
Section Copyright © 2014, 2012, 2010 Pearson Education, Inc. Lecture Slides Elementary Statistics Twelfth Edition and the Triola Statistics Series.
Agronomic Spatial Variability and Resolution What is it? How do we describe it? What does it imply for precision management?
The Examination of Residuals. Examination of Residuals The fitting of models to data is done using an iterative approach. The first step is to fit a simple.
Rank-ordered multifractal analysis (ROMA) of magnetic intermittent fluctuations in the solar wind and in the magnetospheric cusps: evidence for global.
PCB 3043L - General Ecology Data Analysis. OUTLINE Organizing an ecological study Basic sampling terminology Statistical analysis of data –Why use statistics?
Actuarial Applications of Multifractal Modeling
Internal Tide Generation Over a Continental Shelf Summer 2008 internship Gaёlle Faivre Flavien Gouillon, Alexandra Bozec Pr. Eric P. Chassignet.
Modern Navigation Thomas Herring MW 11:00-12:30 Room
Percolation Percolation is a purely geometric problem which exhibits a phase transition consider a 2 dimensional lattice where the sites are occupied with.
FAT TAILS REFERENCES CONCLUSIONS SHANNON ENTROPY AND ADJUSTMENT OF PARAMETERS AN ADAPTIVE STOCHASTIC MODEL FOR RETURNS An adaptive stochastic model is.
Spatial Statistics in Ecology: Point Pattern Analysis Lecture Two.
Slide 1 NATO UNCLASSIFIEDMeeting title – Location - Date Satellite Inter-calibration of MODIS and VIIRS sensors Preliminary results A. Alvarez, G. Pennucci,
Z bigniew Leonowicz, Wroclaw University of Technology Z bigniew Leonowicz, Wroclaw University of Technology, Poland XXIX  IC-SPETO.
Local Predictability of the Performance of an Ensemble Forecast System Liz Satterfield and Istvan Szunyogh Texas A&M University, College Station, TX Third.
Chapter 8: Simple Linear Regression Yang Zhenlin.
STOCHASTIC HYDROLOGY Stochastic Simulation of Bivariate Distributions Professor Ke-Sheng Cheng Department of Bioenvironmental Systems Engineering National.
Review of Statistical Terms Population Sample Parameter Statistic.
Stochastic Hydrology Random Field Simulation Professor Ke-Sheng Cheng Department of Bioenvironmental Systems Engineering National Taiwan University.
Week 21 Order Statistics The order statistics of a set of random variables X 1, X 2,…, X n are the same random variables arranged in increasing order.
BPS - 5th Ed. Chapter 231 Inference for Regression.
Week 21 Statistical Model A statistical model for some data is a set of distributions, one of which corresponds to the true unknown distribution that produced.
WIS/COLLNET’2016 Nancy, France
Pianificare la città frattale
Linear Regression.
Dimension Review Many of the geometric structures generated by chaotic map or differential dynamic systems are extremely complex. Fractal : hard to define.
PCB 3043L - General Ecology Data Analysis.
Unsupervised Riemannian Clustering of Probability Density Functions
Chapter 5 STATISTICS (PART 4).
Digital Processing Techniques for Transmission Electron Microscope Images of Combustion-generated Soot Bing Hu and Jiangang Lu Department of Civil and.
Statistical Methods For Engineers
Basic Statistical Terms
Stochastic Hydrology Random Field Simulation
Volume 129, Issue 2, Pages (April 2007)
Berrilli F., Del Moro D., Giordano S., Consolini G., Kosovichev, A.
Nat. Rev. Cardiol. doi: /nrcardio
Testing for Multifractality and Multiplicativity using Surrogates
Statistics for Managers Using Microsoft® Excel 5th Edition
Volume 5, Issue 4, Pages e4 (October 2017)
Chap 7: Seasonal ARIMA Models
Presentation transcript:

International Summer School on Turbulence Diffusion 2006 Multifractal Analysis in B&W Soil Images Ana M. Tarquis Dpto. de Matemática Aplicada E.T.S.I. Agrónomos Universidad Politécnica de Madrid

International Summer School on Turbulence Diffusion 2006 INDEX Problem: motivation and start point. Fractals and multifractals concepts. Porosity images: resolved? Configuration Entropy Griding Methods

International Summer School on Turbulence Diffusion 2006 CONSERVATION OF NATURAL RESOURCES Agriculture : soil degradation and water contamination. Sustainable agriculture Quantification of soil quality index?

International Summer School on Turbulence Diffusion 2006 Soil structure Water, solutes and gas transport Soil resistance Roots morphology Microorganism populations PORE AND SOIL MATRIX GEOMETRY

International Summer School on Turbulence Diffusion 2006 Fractal structure: structured distribution of pore (and/or soil) in the space such that at any resolution the set is the union of similar subset to the whole.whole

International Summer School on Turbulence Diffusion 2006 Measure techniques The number-size relation is used normally to measure the fractal dimension of the defined measure (number of white or black pixels), or counting objects: Or covering the object with regular geometric elements of variable size:

International Summer School on Turbulence Diffusion 2006 “ Box-Counting” -m = fractal dimension, D 11 nn Black, white or interface interface

International Summer School on Turbulence Diffusion 2006 Multifractal analysis consider the number of black pixels in each box (pore density=m).Multifractal

International Summer School on Turbulence Diffusion 2006 Multifractal: density has an structured distribution in the space such that at any resolution the set is the union of similar subsets to the whole. But the scale factor at different parts of the set is not the same. More than one dimension is needed => the measure consider (M) is characterized by the union of fractal sets, each one with a fractal dimension.

International Summer School on Turbulence Diffusion 2006 11 nn DqDq q

Numerical Analysis of Multifractal Spectrum on 2-D Black and White Images

International Summer School on Turbulence Diffusion 2006 RANDOM AND MULTIFRACTAL IMAGES In this way a hierarchical probability tree was built generating an image of 1024x1024 pixels (ten subdivisions), as the soil images are normally analyzed. Probabilistic parameters are: { p 1, p 2, p 3, p 4 } Random images : p 1 = p 2 = p 3 = p 4 = 25% Multifractal images: p 1 = 50%, p 2 = 5%, p 3 = 25% and p 4 = 20% (by random arrangements or not).

International Summer School on Turbulence Diffusion 2006 Random multifractal

International Summer School on Turbulence Diffusion 2006 Generalized dimensions (Dq) obtained for two different distributions based on Stanley and Meakin (1988) formulas with their respective -  (q) curves.

International Summer School on Turbulence Diffusion 2006 Most common parameters calculated D 0 q=0  box counting dimension D 1 q=1  entropy dimension D 2 q=2  correlation dimension

International Summer School on Turbulence Diffusion 2006 Singularities of the measure (  ) For a given  there is a fractal dimension f(  ) of the set that support the singularity. At each area the relation number-size is applied: f(  ) 

International Summer School on Turbulence Diffusion 2006 f(  )  Multifractal Spectrum wfwf ww

International Summer School on Turbulence Diffusion 2006

INTERDENNY ABOKMUNCHONG 1500x1000 pixels

International Summer School on Turbulence Diffusion 2006 ¿How many points?

International Summer School on Turbulence Diffusion 2006 ADS BUSO EHV1

International Summer School on Turbulence Diffusion 2006 We have to compare

International Summer School on Turbulence Diffusion 2006

Obtaining D q Ehv1, porosity 46,7%

International Summer School on Turbulence Diffusion 2006 Calculating D q ADS, porosity 5,7%

International Summer School on Turbulence Diffusion 2006

Continuos line = random structure Dashed line = mfract structure Filled Square = values from image soils

International Summer School on Turbulence Diffusion 2006 Considerations on D q calculations Several authors have shown that the exact value of the generalized dimension is not an easy calculation to do. Vicsek proposed practical methods to compute the generalized dimension The main difficulty in using the multifractal formalism lies in the fact that the ideal limit cannot be reached in practice

International Summer School on Turbulence Diffusion 2006 RESULTS AND DISCUSSION (1) For all of the soil images with different porosity we obtain convincing straight-line fits to the data having all of them r 2 higher than 0.98,

International Summer School on Turbulence Diffusion 2006 RESULTS AND DISCUSSION Finally, a comparison among the different images in each dimension is showed. In all of them, the points corresponding to porosities higher than 30% lie on the line representing the Dq calculated for the random generated images. Observing the difference between the fractal dimensions coming from multifractal and random images (discontinue line and continue line respectively) it is obvious that decreases when porosity increases in the images.

International Summer School on Turbulence Diffusion 2006 Configuration Entropy H(  ) The maximum value of j is  x  and the minimum value is 0 (Andraud et al., 1989)  11 ii n(  ) = boxes of size  from  = 1 to  = w /4 w N j = number of boxes with j black pixels inside

International Summer School on Turbulence Diffusion 2006 Configuration Entropy H(  ) The probability associated with a case of j black pixels in a box of size  (p j (  ))

International Summer School on Turbulence Diffusion 2006 Configuration Entropy H(  )  (pixels) H*(  ) w/4 H*(L) L

International Summer School on Turbulence Diffusion 2006 Methods: gliding, random walks, randomly Box size Jump step length Number of jumps

International Summer School on Turbulence Diffusion 2006 Thank you for your attention

International Summer School on Turbulence Diffusion 2006 Multifractal Analysis on a Matrix Ana M. Tarquis Dpto. de Matemática Aplicada E.T.S.I. Agrónomos Universidad Politécnica de Madrid

International Summer School on Turbulence Diffusion 2006 INDEX Field Percolation Soil Roughness Satellite images Time series

International Summer School on Turbulence Diffusion 2006 Z= 10 cmZ = 20 cmZ = 30 cm Z = 40 cmZ = 50 cmZ = 60 cm

International Summer School on Turbulence Diffusion 2006 % of blue vs. depth 50% 15 cm

International Summer School on Turbulence Diffusion 2006 Z = 25 cm blue staining 28,95%

International Summer School on Turbulence Diffusion 2006 Dye Tracer Distribution

International Summer School on Turbulence Diffusion 2006 Multifractal Analysis of the Dye Tracer Distribution B) Generalized dimensions A) f(  ) spectrum

International Summer School on Turbulence Diffusion 2006 Multispectral Satellite Images

International Summer School on Turbulence Diffusion 2006

Soil Rougness Roughness indices normally are based on transects data. One of the most used is the Random Roughness (RR). RR is the standard deviation of the soil heights readings from the transect. This implies that there is not an spatial component. Several authors have applied fractal dimensions to this type of data. Burrough (1989), Bertuzzi et al. (1990), Huang and Bradford (1992),

International Summer School on Turbulence Diffusion 2006 INTRODUCTION The aim of this work is to study soil height readings with multifractal analysis in the context of soil roughness. Several soils, with different textures, with different tillage methods have been analysed to compare their multifractal spectrum.

International Summer School on Turbulence Diffusion 2006 Soil measurements Three different soils with different textures. Three different treatments applying tillage: chisel, moldboard, seedbeds. Height measures of 2x2 m 2 plot area. Resolution of the measure each 2 cm

International Summer School on Turbulence Diffusion 2006

Soil texture

International Summer School on Turbulence Diffusion 2006

moldboard seedbeds chisel

International Summer School on Turbulence Diffusion 2006 moldboard seedbeds chisel

International Summer School on Turbulence Diffusion 2006 11 nn Box counting method Number of boxes depends on  i 22 33

International Summer School on Turbulence Diffusion 2006 MF analysis of Height Distribution (HD) Chhabra and Jenssen method

International Summer School on Turbulence Diffusion 2006 f(  )  Multifractal Spectrum wfwf ww

International Summer School on Turbulence Diffusion 2006 Considerations on MF calculations Height readings have been corrected for slope and tillage tool marks. The linearity in the  function were found in all cases from  =1 to  =64 cm. The range of q values used were from – 5 to +5 with increments of 0.5. All the R 2 obtained were higher than 0.97

International Summer School on Turbulence Diffusion 2006 HD Multifractal Spectrum

International Summer School on Turbulence Diffusion 2006 HD Multifractal Spectrum

International Summer School on Turbulence Diffusion 2006 HD Multifractal Spectrum

International Summer School on Turbulence Diffusion 2006 Results from the multifractal analysis

International Summer School on Turbulence Diffusion 2006 AV= SD =14.90 RR

International Summer School on Turbulence Diffusion 2006 AV= SD =6.62

International Summer School on Turbulence Diffusion 2006 AV= SD =8.92

International Summer School on Turbulence Diffusion 2006

CONCLUSIONS Fractal dimensions estimated from MF analyses of HD are useful descriptors. Multifractal parameters seem to be correlated depending on soil texture properties. Comparison between data structure and a random structure can be used to get a complementary index to RR.

International Summer School on Turbulence Diffusion 2006 Further research More work on correlating parameters from multifractal analysis to soil properties: we need to understand what represent each parameter. More work on application of multifractal parameters to the prediction of processes related to soil erosion.

International Summer School on Turbulence Diffusion 2006 WIND FLUCTUATIONS The study of wind-speed (w) is aimed at greenhouse control (heating and ventilation), since wind velocity influences both types of control. Wind increases heat losses in winter nights, so it is of interest to regulate the heating as a function of wind-speed and its realistic simulation is an important task in modeling and system design. To study the multifractal nature of this series and to fully characterize the dynamical system that supports it is the first step before any simulation could be successfully achieved. Time series data from 2004 were used in this study. Every ten minutes, the station recorded mean values of the wind velocity in m/s. Thus we handle in each yearly analysis a series of data points, and in the monthly analysis a minimum of values (February) and a maximum of

International Summer School on Turbulence Diffusion 2006 Stochastic process: fBm The minimum and maximum lag values are normally chosen. If the series is self-similar then: Hurst exponent H = 0.5 => random structure H > 0.5 => persistant structure H anti-persistant structure

International Summer School on Turbulence Diffusion 2006 Multifractal Analysis (MF) Multiscaling analysis determines the dependence of the statistical moments (and not only the covariance) of the time series on the resolution with which the data are examined. Different moments different exponent in the increments (q). Structure Function (M q )

International Summer School on Turbulence Diffusion 2006 Generalized Hurst exponent H(q) The minimum and maximum lag values are normally chosen. If the series is self-similar or self-affine then: monotonically non-decreasing function of q

International Summer School on Turbulence Diffusion 2006 CASES stationary processes have scale- independent increments and show invariance under translation => H(q)=0 non-stationary and monofractal processes => constant H(q) non-stationary and multifractal => non constant H(q)

International Summer School on Turbulence Diffusion 2006 Wind velocity time series

International Summer School on Turbulence Diffusion 2006

Histograms of wind fluctuations

International Summer School on Turbulence Diffusion 2006 Structure Functions (M) for February of 2004.

International Summer School on Turbulence Diffusion 2006  (q) and the corresponding H(q) function

International Summer School on Turbulence Diffusion 2006 COMMENTS AND CONLUSION There are several steps as number of data and lag values range chosen that influence the numerical results. February shows a different behavior from the other months, however the q values used are much higher that it is normally found in the literature. July shows a clear multiscaling pattern with a non constant H(q). December shows an almost constant H(q) All of them, as the annual time series analysis, show an anti-persistent character. Structure Functions is a way to usefully characterizing this multiscale heterogeneity. Based on this modeling simulation of wind fluctuations can be done in easy way and being realistic.

International Summer School on Turbulence Diffusion 2006 Thank you for your attention