I am currently working on the 4th edition of my textbook "MATLAB Recipes for Earth Sciences" with Springer. There will be an interactive ebook for tablet computers in addition to the regular PDF ebook and the printed book. Furthermore, I am expanding the existing chapters of the book including statistical and numerical methods that became popular during the last couple of years. Is there anything you think I should add to the book, a method which is widely used in your field of expertise? Thanks for your help, contents of the book attached!
I suggest to open a new chapter on: Inverse methods by Bayesian inference.
This is a new class of statistical algorithms mainly applied in atmospheric phyiscs by Satellite
I can suggest the includion of Correspondence Analysis, a useful method to multivariate analysis when data sets includes categorical variables, common in Earth Sciences.Discriminant Analysis is another useful technique widely used in Earth Sciences. The increasing use of geostatistical simulation techniques justifies the inclusion in such book.
I suggest to open a new chapter on: Inverse methods by Bayesian inference.
This is a new class of statistical algorithms mainly applied in atmospheric phyiscs by Satellite
Space geodesists use mostly non-linear Least Squares, a topic that is not even mentioned, but would probably require an entire chapter devoted solely to it. It would be worth it though if you could and widely used! Since you are in Potsdam, I suggest discussing this with our colleagues at GFZ, who could provide you with examples of applications in all areas.
I suggest to cover the spatial autocorrelation analysis topic. Additonally, very usefull is the GWR (Geographically Weighted Regrssion) approach.
good luck
CC
Congratulations by the new edition of this book. It is an interesting and actual thematic with many applications.
I would like to suggest a section with a non-parametric statistics because much of the geological records are qualitative data, that do not follow theoretical distributions. Some statistical applications such as run test, Wilcoxon-Mann-Whitney test, Kolmogorov-Smirnov test, measures of association (Spearman and Kendall), Kruskal-Wallis analysis, etc... can be useful.
In chapter 5 could be useful to have a brief introduction to Markov chains.
In Chapter 9 could be useful to briefly refer to the Correspondence Analysis
In Chapter 10 could be useful to show the correlation coefficients between linear-circular and circular-circular variables.
Once again congratulations.
Best regards
Joaquim Góis
Auxiliary Professor
Mining and Geo-Environmental Engineering Department
School of Engineering - University of Porto
Portugal
I would suggest you add:
* Intelligent methods such as SVM, ANN, GO, Fuzzy logic for classification and time series modeling
* Least squares harmonic estimation in order to model harmonic components of an uneven time series.
It would be nice to include discriminant analysis method and classification (e.g. centroids etc.). Its a nice way to group your samples helping geochemists, geoscientists doing applied research and others as well.
Thanks for your valuable comments and suggestions! Here's a list of new sections that I have so far:
2.10 Creating Graphical User Interfaces
3.6 [Now including a lot more about principles of hypothesis testing]
3.10 The Kolmogorov-Smirnov Test
3.11 The Mann-Whitney Test
3.12 The Ansari-Bradley Test
4.10 [The nonlinear and weighted regression has been updated]
6.4 [Now includes deconvolution]
7.6 Exporting 3D Graphs to Create Interactive Documents
8.4 Importing, Processing and Exporting LANDSAT Satellite Images
8.5 Processing and Georeferencing TERRA ASTER Satellite Images
8.6 Analyzing EO-1 Hyperion Satellite Images
8.8 Image Enhancement, Correction and Rectification
8.12 Shape-based Object Detection in Images
9.4 Factor Analysis
9.5 Correspondence Analysis
9.6 Discriminant Analysis
9.8 Multiple Linear Regression
I'm not a great fan of Bayesian techniques although one of my doctoral students is working with it.
Yes, that's the problem. It's a frequentist's book, Popper style, and adding sections on Bayesian methods would require a discussion on frequentist vs. Bayesian statistics. Saying this in the preface, however, is a good idea.
Dear Martin,
First of all, thank you for very useful and important work.
It seems to me (for a first view) that wavelet approach description should be extended since it becoming more and more popular (and effective).
Lev
1) Multiple collocation is now used by many for spatial variability and error budget analysis. I introduced it for satellite winds, but now it is used for satellite wave heights and soil moisture as well, among others. So, fundamentals:
Stoffelen, A. (1998), Toward the true near-surface wind speed: Error modeling and calibration using triple collocation, J. Geophys. Res., 103(C4), 7755–7766, doi:10.1029/97JC03180.
And subsequently Google "triple collocation".
2) What about spatial structure function or spectral analyses?
I suggest to insert methodologies for signals analysis as Hilbert Huang Tranforrm (HHT), to analyze non stationary and non linerar signals.
Don't have anything specific to add towards the book, but would like to complement you for using this crowd sourcing method to meet researchers needs while compiling book. Plus kudos to the responders providing input!
I would love to see some chapters related to advanced methods of modeling.
Multigrid methods, SPH, and also FEM. Texts are there but applications in Earth Science of these would be entertained a lot. Or you can include different methods for data interpolation which will not only help in statistical analysis but also it will help in construction of meshfree shape functions.
Looking forward to next addition of your book.
Regards,
Pankaj K Mishra
I would suggest to add a new section related to point pattern analysis. I mean the application of Ripley- and Besag- functions to process clustering of points such as volcanic vent distributed in an area.
Related to this section you can add an application of kernel tecniques to give a probabistic distribution for future vent opening.
Finally, an other section can process data (such as earthquakes hypocenter or any type of locations) to give a reliable and suitable probability density function.
Dear all,
Thanks for your contribution! There are some really great ideas, while others are a bit too special for a 350 page book with a little bit about everything. The chapters on time series analysis and signal processing are already very long since this is what I have done a lot myself. The image processing and analysis chapter will be much longer since I got involved in a remote sensing project. Chapters 9 (multivariate stats) and Chapter 10 (directional data, I really don’t like this chapter very much and it wasn’t there in the first edition of the book) are still a bit light.
Here are some comments on your suggestions:
Ch 3 Wilcoxon-Mann-Whitney test
suggested by Joaquim Góis. Section has been completed already
Ch 3 Kolmogorov-Smirnov test
suggested by Joaquim Góis. Section has been completed already
Ch 3 Kruskal-Wallis analysis
suggested by Joaquim Góis. Maybe not because it is similar to
the Mann-Whitney test, isn’t it? I got the Mann-Whitney, KS
test and Ansari-Bradley test, which I used in a 2009 paper. Is
the KW analysis as important as the others I already have?
Ch 3 Machine learning, decision trees etc.
suggested by John Kern,
Ch 4 Spearman and Kendall
suggested by Joaquim Góis. Could be interesting, I’ll think
about it!
Ch 4 Nonlinear Least Squares
suggested by Erricos Pavlis. This is already in the book.
Ch 5 Markov chains
suggested by Joaquim Góis. Great idea, I’ve used it one in a
time-series simulation, maybe a section on it would be good to
have
Ch 5 Least squares harmonic estimation in order to model harmonic
components of an uneven time series.
suggested by Amir Souri. I got the Lomb-Scargle periodogram
in Ch 5 already, I think that’s it?
Ch 5 Extended discussion on wavelets
suggested by Lev Eppelbaum. Yes, maybe, but isn’t 6 pages
enough for the beginner? I refer to the excellent Wavelet
package and manual by Torrence and Compo for those who need
more.
Ch 7 Spatial autocorrelation analysis, Multiple-point geostats,
Geographically Weighted Regression etc.
suggested by Bart Rogiers, Christos Chalkias and others. That’s
a good comment. Ch 7, after having introduced the most popular
DEMs, is on gridding, contouring, and there is a great section
on Kriging by my colleague Robin Gebbers. I know that I should
expand that chapter to include more spatial analysis in addition
to the sections I have.
Ch 7 Finite element method
as suggested by Pankaj Mishra. VERY GOOD POINT, and this has
been said already in the only very negative review I got on
the very first edition of a book by a guy who isn’t easy to
google as his name is Tom Jones. I have never ever used it,
in contrast to most other methods explained in the book, so
it would be a tough job to add a chapter on it. I will keep
this in my mind, as I do so since the bad review.
Ch 9 Corresponding analysis
suggested by Joaquim Góis. Indeed, I have planned a section
on it.
Ch 9 Discriminant analysis
suggested by I Bazoitis. Indeed, I have planned a section
on it.
Ch 10 Correlation coeff between lin-circ and circ-circ variables
suggested by Joaquim Góis. This is a valuable suggestion. I’m
not at all an expert in circular data. Therefore I rely on
your help, what are the methods used by structural geologists
or people analyzing paleocurrent directions?
Ch XX Baysian inference
as suggested by Rodolfo Guzzi. As Rodolfo has pointed out
correctly it would require an entirely new chapter. I know
Baysian inference, although not new, has become quite popular
in some fields while it is ignored in others. Personally I
am not a great fan of it, as a Popper-style frequentist, but
I acknowledge people using it, including one of my doctoral
students.
Thanks for your encouraging comments, in particular I would like to thank Joaquim, Lev and Kevin! When I wrote the first edition in 2004/05, I did not expect that it will be so successful. I think I’ve tried to keep a balance between the various methods used in the different fields. And that’s why this discussion here is quite helpful.
Kind regards, Martin
I would suggest to add Canonical correlation analysis CCA to multivariate statistics.
Canonical correlation analysis (CCA) is an important multivariate statistical tool for reducing the dimensionality of an original data set. CCA is most commonly used in the context where there are two sets of random multidimensional and correlated variables, x = {x1,x2, . . .,xn} and y = {y1,y2, . . .,ym}. For instance, x could be a set of n mineral proportions, while y could represent a set of bulk-rock chemistry m different oxides/trace elements values. CCA enables one to identify the dominant linear modes of covariability between the sets x and y. In other words, CCA identifies two new groups of artificial" composite" variables (canonical variables).
Yes, Khalid, I got Section 9.5 Correspondence Analysis already in the new edition.
Dear Azar, I remember a Monte Carlo example from my own studies in the 80s when we had to estimate the volume of an irregular body. I have use a bit of MC modeling to simulate bioturbation in the deep sea (published in C&G in early 2013) but this is a very special application. Do you know a good example from earth sciences? There is a lot of Monto Carlo type modeling in my book but there isn't a specific section on it. Kind regards, Martin
Dear Edoardo,
Thanks for this very kind and generous comment. It's my first Question to ResearchGate and it seems that it works very well!
Kind regards, Martin
what about adding Singular spectrum Analysis in the chapter of Multivariate Statistics?
SSA is similar to PCA but it might be interesting to add a section to introduce this technique.
Thanks for updated. I am going to teach something like "data analysis in petroleum exploration and production". I am looking for an appropriate tool for my students to do excercises and think Matlab would be a good choice. Please comments if my point is in appropraite.
Martin, when the book released please let me know
Dear Minh,
Thanks for your post. I think MATLAB is the best choice for teaching students in data analysis. There is a student version for 99 USD. Of course there are free alternatives (R, Python ...), some of which are even compatible with MATLAB (Octave). Students of course would prefer those but professionals mostly prefer the commercial product as there is an excellent support with a very short response time. In the industries MATLAB is a lot more popular than the free alternatives, for the same reason. My students love it. Some tried Octave first, but then switched to MATLAB after a while.
You can use the 3rd edition of my textbook which is available from Springer since 2010, or any online bookstore such as Amazon. See my Profile page for the contents of the 3rd edition. The example files and MATLAB recipes are available online at Springer Extras. There are two other great books on slightly different topics: E. Holzbecher, Environmental Modeling Using MATLAB (Springer) and A. Quateroni et al., Scientific Computing with MATLAB and Octave (Springer).
Let me know (via email) if you need anything that helps you designing your course! The supplementary electronic material available at Springer Online also includes presentation slides with all graphs and figures from the book. I am teaching courses internationally at several universities worldwide based on the book, including an English and German course here in Potsdam in March.
Kind regards, Martin
Shortcourses on MATLAB Recipes for Earth Sciences
I teach shortcourses on MATLAB Recipes for Earth Sciences (MRES) at U Kiel, U Bremen, U Bratislava, U Ghent, UA Barcelona, BGR Hannover, UC London, U Munich, BGI Bayreuth, U Nairobi, U Köln, U Stockholm, U Amsterdam, NHM Vienna, GNS Science Wellington, Brown U Providence, Aberystwyth University and U Potsdam. From 2013 I also teach shortcourses on the new book MATLAB and Design Recipes for Earth Sciences (MDRES) together with the designer Elisabeth Sillmann. Please send me an email if you interested to organize a course at your university.
Next shortcourses in Potsdam: 10-14 March 2014 (in English) and 17-21 March 2014 (in German).
Hi Martin
your book looks great...some suggestions:
a) chapter 3
may be something more of no-parametric statistics like
wilcoxon test by two independent samples, wilcoxon test by two dependent samples
in the t test be more specific in dependent or independent sample sand in this last one include with homecelastic sindependent samples (equal variances) and no-homocelastic independendent samples .
may be you can add a Q-Q test by normal distribution test
b) Chapter 4
add Spearman correlation coeficient...
may be if you see the linear regression like a special case of one taylor polinomium you can see multi regression analysis.
c) chapter 7 may be add a mach-up test
d) chapter 9 in the principal component analysis (PCA) you can separated Empirical orthogonal function (EOF), Standart EOF, Scores in the PCA numerical resolution and include the eigenvector plot (Factor Analysis) in the graphical PCA resolution.
About the mach up analysis , Pearson and Spearman correlation coefficients you can check
Santamaría-del-Angel E., R. Millán-Núñez, A. González-Silvera, R. Cajal-Medrano (2011) A comparison of Chl a concentrations estimated in situ and Chl a concentrations determined via remote sensing: A statistical examination of the match-up approach. 241-260 Chapter 17 in Handbook of Satellite Remote Sensing Image Interpretation: Applications for Marine Living Resources Conservation and Management (2011), EU PRESPO and IOCCG Edited by: Jesus Morales, Venetia Stuart, Trevor Platt and Shubha Sathyendranath J.
and by PCA things (EOF ZEOF, Scores numerical resolution )
Santamaría-del-Ángel E, A. González-Silvera, R. Millán-Núñez, M. E. Callejas-Jiménez, R. Cajal-Medrano (2011) Determining Dynamic Biogeographic Regions using Remote Sensing Data 273-293. Chapter 19 in Handbook of Satellite Remote Sensing Image Interpretation: Applications for Marine Living Resources Conservation and Management (2011), EU PRESPO and IOCCG Edited by: Jesus Morales, Venetia Stuart, Trevor Platt and Shubha Sathyendranath J.
and by factor analysis
Santamaría-del-Ángel, E., Millán-Núñez, R., González-Silvera, A., Callejas-Jiménez, M., Cajal-Medrano, R., and Galindo-Bect, M. S. (2011). The response of shrimp fisheries to climate variability off Baja California, México. – ICES Journal of Marine Science, 68: 766–772. doi: 10.1093/icesjms/fsq186.
this 3 "papers" are in my contribution or
the chapters you can download by http://www.ioccg.org/handbook.html
and the paper
http://icesjms.oxfordjournals.org/content/68/4/766.full
i hope these suggestion can help
Saludos
Hi Eduardo, many thanks for your suggestions! Although some of them have been made before, it's quite helpful if people have similar suggestions! Kind regards, Martin
one more time
Hi Martin
Thanks by you replay
about Kolmogorov-Smirnov test always obtain that the sample have a normal distribution ..even if you take a no-normal artificial sample.....i believed that this is because the Dmax calculated is only one different (of n different possible)..and the critical valuesis calculated using a number of total observation
Any way may be we need use the new generation of normality test like the Q-Q test...
Really is amazing the academic potential that have the Researgate..you way to focus you book is very nice
Hi Eduardo,
I use the Chi-2 test instead of the K-S test, but use the K-S sometimes as an additional test if the Chi-2 test result isn't very obvious. That's why the K-S wasn't in the previous editions of the book and the K-S is now added to the list of tests, in addition to the parameter free Mann-Whitney and Ansari-Bradley tests.
Thanks for the compliments, more than 30 replies, RG seems to work fine! I sold thousands of copies of the book in the last couple of years. Many more are reading the ebook version of it. I am thinking of establishing a ResearchGate Project where readers, students and lectures can meet and discuss these things, e.g. what are the methods an undergrad student should learn about data analysis.
Springer was planning to establish "living books" with an online environment, discussion forum etc., addition material for instance. The new book will be a lot more dynamic in the way it is kept up to date, correcting mistakes etc., even adding more chapters. If it is an interactive ebook (the experiments with interactive graphs look great) you can easily update and correct it online, and the readers get a free download of the new version. That's the ideas so far.
Does anyone use/run Projects here?
Kind regards, Martin
Great book for MATLAB users. Rodolfo earlier suggested to add a chapter on Bayesian inversion based methods. This Bayesian based method is also becoming popular in Seismic exploration in petroleum industry.
Depending upon how much you want to expand your book on tools popular in seismic industry - you can add an example on simple seismic wave propagation, convolution/deconvolution, inverse theory, rock physics based analysis methods. There are already many published books or scripts in Matlab that you can refer along with your own examples.
Thanks, Dhananjay
Hello Martin
I have just two suggestions:
1) as you know Matlab is extensively used in Geophysics both for basic teaching and for research, and either at academic and industry levels. Perhaps it may be good to add some examples referring to applications to disciplines such as reflection seismic and potential methods. Many of your chapters (e.g. 5,6,8) seem appropriate to host such examples.
2) stochastic (global) optimization methods are being more and more used to solve multiminima inversion problems in geophysics and in many other Earth Sciences disciplines. Matlab has a specific toolbox (Genetic Algprithm, Simulated Annealing,..). It may be an useful methodological addition.
Best wishes for your book. I'm looking forward to see it published.
Dear Dhananjay and Alfredo,
Thanks for your comments, very helpful. I have written it mostly for people in the paleoclimate business and was surprised that many geophysicists are using it. No problem, as I am have a geophysicist, too, the other half is a geologist. My maths background comes from geophysics.
As I've said already, Bayesian methods won't fond their way into the book. Convolution is already in, a very short deconvolution section has been written already. I would like to add more about it as my own doctoral project was on a time-variable deconvolution technique in the frequency (Fourier) domain. That's almost 20 years ago, and I have made four papers out of my doctoral thesis but the most interesting part (the deconvolution) has never been published. I went through my MATLAB scripts a couple of weeks ago and had some difficulties to understand what's going on there. I was surprised that I even wrote a GUI for such a deconvolution. We'll see ...
A couple of other methods suggested here I have to google first, as I have never seen them being used in my field (paleoclimate dynamics). If they are used in geophysics by many people I my think about it.
Again, whoever is interested in the book, there is the 3rd edition out, okie? We are discussing the contents of the 4th edition here.
Kind regards, Martin
This is a great book, I have used it as textbook many times for graduate-level courses. I am exited to know that you are working on the next version.
The main thing missing in the book, from my perpective, are exercises.
Dear Jonás,
very good point! I didn't include the exercises in the book but I am sending exercises with solutions to instructors on request. Let's check whether I can use the "Projects" feature of ResearchGate for this.
Please everyone let me know using the email feature of ResearchGate if you wish to be invited to the Project.
Kind regards, Martin
I am neither an earth science person, nor a statistician. But here are a few off-hand suggestions based on my limited experience; please feel free to ignore them!
1. How about checking out the contents of the Venebles and Ripley book Modern Applierd Statitics with S-Plus (MASS, as it is popularly known) and adapting any interesting/relevant methods in the earth science context?
2. There's a large body of literature from many fields about the abuse of hypothesis testing and p-values in particular. How about helping the reader be at least aware about how to use and not use these "commodities"? Is multiple testing relevant to common earth science data analysis contexts?
3. Nonparametric statistics is a powerful paradigm that is not restricted to hypothesis testing alone. For example, nonparametric curve estimation methods (kernel smoothers, kernel density estimators, orthogonal series estimators, etc.) assume less, and hence tend to be more data-driven. As such, they can be used to "validate" parametric fits which assume more.
4. Similarly, the (nonparametric) bootstrap is a powerful tool to possess. I think it deserves a somewhat more detailed exposition. Some insight into how to deal with small-sample issues might be useful if relevant.
Best wishes, Mihir
Dear Mihir,
Thanks for your post! Yes, the abuse of hypothesis testing, it would indeed require a separate section. I have to admin, however, that my introductionary section on hypothesis testing is very light in the previous editions, although I am then introducing three tests (T, F, Chi2). Now there is such a section, but without mentioning the pitfalls of it, before I also introduce the KS, MW and AB tests. I am collecting all these suggestions for the 5th edition then, because making it an interactive book is a lot of work too!
Kind regards, Martin
Hi Martin,
Thanks for your comments. I will have a look at your suggested sites for more information. When my course content shaped out, I may come back to ask for your comments. Regards
Hi Martin,
may I suggest to add a chapter on the evergreen strain analysis techniques in section 8 (image processing)? Great job, regards
Dear Ahmad,
I'll do! But don't expect too much, I won't be able to include all the suggestions being made here in the forum. I will post a tentative contents in a couple of days again but right now I am busy with a manuscript of a doctoral student on fuzzy logic identification of annual layers, very interesting but difficult for a non-expert on fuzzy logic methods like me.
KInd regards, Martin
I just found another new book by Shahab Araghinejad, "Data-Driven Modeling: Using MATLAB® in Water Resources and Environmental Engineering" (Springer 2013). It is mostly on modeling, contains chapters on Neural Networks, Support Vector Machines, Fuzzy Models, and Hybrid Models.
Advanced Hyperspctral image classification techniques such as SVM may be included
Hi Kamalesh,
I got a 6-page section on Hyperion data in the new edition, not more than importing, processing, displaying the int16 data. I'm afraid I can't do much more as it is not a book on remote sensing. The chapter where hyperspectral images are introduced also includes sections on thresholds (8.11) and detecting objects using the Hough transform (8.12). I need to keep a balance between the various methods. Anyway, thanks for your comment!
Regards, Martin
8 Image Processing
8.1 Introduction
8.2 Data Storage
8.3 Importing, Processing and Exporting Images
8.4 Importing, Processing and Exporting LANDSAT Images
8.5 Importing and Georeferencing TERRA ASTER Images
8.6 Importing, Processing and Exporting EO-1 Hyperion Images
8.7 Digitizing from the Screen
8.8 Image Enhancement, Correction and Rectification
8.9 Color-Intensity Transects of Varved Sediments
8.10 Grain Size Analysis from Microscope Images
8.11 Quantifying Charcoal in Microscope Images
8.12 Shape-Based Object Detection in Images
Martin,
I would suggest that you add a new chapter on linear and non-linear inverse methods. You touch on this in Ch 4 using polyfit but your treatment doesn't utilize the full matrix algebra capability of MATLAB where you are solving the equation y=Ax. Using this construct, the method can be easily generalized to fitting eigenfunctions and can be adapted to itteritive non linear least squares methods.
Dear Martin,
I would suggest a chapter on neural networks and other machine learning approaches. For example, support vector machines (SVM) or Learning Vector Quantizers (LVQ) are powerful classifier methods whereas self-organizing maps (SOM) are highly appreciated for data visualization. So, these methods (and maybe other) should be touched as alternatives to traditional multivariate methods.
Further, de-mixing of spectral signatures in hyperspectral data analysis is an important issue of remote sensing image analysis.
I support Bart Rogiers' suggestion to include something on sensitivity analysis, e.g. GLUE, since this is an often-used technique by hydrologists.
Daer Dean, Thomas and Simon,
Thanks for your suggestions! I am still struggling a bit with Chapter 9, started yesterday to add a section on the discriminant analysis for classification. Maybe I will add another one on multiple linear regression. I not using these techniques (except for the PCA) in my research and therefore I rely on your comments here: which are the methods which are really widely used, not only in remote sensing and hydrology?
As an example, my colleague Robin Gebbers wrote a section on Kriging already for the first edition in 2006. While this method is very popular in hydrology, agriculture, ecology and others, it isn't used in my field of research. We use (all kinds of) splines a lot more than Kriging. No problem, Robin got excellent reviews for his section and it's good to have it in the book.
What about factor analysis? Is it still used although being critized so much in the recent past? Do I need factor analysis if there is the PCA and even ICA (written by Norbert Marwan)? Do I need SVMs and similar methods of classification or is the discriminant analysis enough? I know that Chapter 5 and 6 include some very specialized things such as adaptive filters but that's what my thesis was about 20 years ago :-)
Kind regards, Martin
9 Multivariate Statistics
9.1 Introduction
9.2 Principal Component Analysis
9.3 Independent Component Analysis (by N. Marwan)
9.4 Discriminant Analysis
9.5 Cluster Analysis
9.6 Multiple Linear Regression
Dear all,
again about remote sensing. I have completed the sections on LANDSAT, ASTER and HYPERION images and believe that help people to import these complex types of data into MATLAB, do a litlte bit of processing and create RGB images for print.
Each of these include a little more than just that: The LANDSAT section explains despeckling or correcting the images for unusually high/low values (since the method can be used for all kinds of other data as well). The ASTER section explains how to georeference such images including automated registration.
The HYPERION section explains how to convert the DN values into radiance BUT (and this is the question) doesn't explain how to convert radiance into reflectance values (is there a simple set of equations for doing this if you don't want to use FLAASH for instance?).
Kind regards, Martin
8 Image Processing
8.1 Introduction
8.2 Data Storage
8.3 Importing, Processing and Exporting Images
8.4 Importing, Processing and Exporting LANDSAT Images
8.5 Importing and Georeferencing TERRA ASTER Images
8.6 Importing, Processing and Exporting EO-1 Hyperion Images
8.7 Digitizing from the Screen
8.8 Image Enhancement, Correction and Rectification
8.9 Color-Intensity Transects of Varved Sediments
8.10 Grain Size Analysis from Microscope Images
8.11 Quantifying Charcoal in Microscope Images
8.12 Shape-Based Object Detection in Images
Dear all,
I have posted (and deleted) an extra question about animated 3D animated objects from MATLAB into iBooks Author and PDF files to create interactive eBooks. Comments on this are also welcome. Anybody working on this, maybe to created animated PDFs or Presentations?
Furthermore, I am seeking colleagues which are very experienced in the use of MATLAB in data analysis in research and teaching who would like to review the new sections, maybe only some in their fields of interest or expertise. There isn't an official reviewing process for textbooks but the earlier editions have taken great benefit from such colleagues.
You won't have to correct the English as I have an excellent proofreader for this. Read the text and try the MATLAB examples, that's all. Please send me a privat message and I will reply as soon as possible.
Kind regards, Martin
Dear all,
Here is an example of an interactive PDF with sound, movie and animated 3D graphics. What do you think about it? I also got a version created with Apple's iBook Authors for Macs and iPads which looks a lot prettier than this one.
Kind regards, Martin
Hi Martin,
A somewhat delayed response, but I would have to throw my support behind bootstrap and simple Monte Carlo techniques. These are two very powerful approaches that I regularly use in my work (paleomagnetism) and are invaluable when we have no parametric description of our data. They are big topics and entire books have already been written on them, but I think introduction of some basics would be a great idea.
Simple bootstrapping in MATLAB is fairly straightforward with inbuilt functions such as bootci() and bootstrp(), but can easily be adapted for more complex functions and needs. Perhaps this could form a part of a more general section/chapter on random sampling techniques?
Thanks for the good work on this extremely useful book!
Cheers
Greig
Hi Martín,
Is in your book any chapter on compositional data analysis? It would be good to have such an entry.
Best regards,
Vera
Hi Martín,
Is in your book any chapter on Markov Chain.
Best regards,
Dear Vera and Khalid,
There isn't a chapter on composition data but there is a discussion on closed data and Aitchison's log-ratio transformation of such data. Is there anything particular that I should include, a specific method particularly designed for such data? I also mention log-normal distributions as a good model for such data.
Markov Chains are on my shopping list. I have just published a paper on bioturbation modelling using a Markov-Chain-type approach but I haven't really thought about a nice example to use them. That's always difficult, finding a good example to demonstrate the advantage of a particular method. What are you using them for?
Kind regards, Martin
Dear Martin,
There is a lot of new results related to the log-ratio approach put forward by John Aitchison. Log-normal distributions are actually at most a good approximation, as they assume R^+ es sample space, while the sample space of compositional data is constraint. At least, I think you should include the isometric log-ratio transformation (Egozcue et al., 2003), and in particular balances (Egozcue and Pawlowsky-Glahn, 2005). You can find good tools for a basic approach in the free package CoDaPack (look in Google or in www.compositionaldata.com).
Best regards,
Vera
Dear Martin,
You can add geostatistical algorithms, such as variograms, co-variograms, kriging and cokriging. There are several open source MATLAB codes that you can obtain from Computer and Geosciences.
I have a good application for ground water in the paper titled,
Parra, J., and Emery, X., Geostatistics applied to cross-well reflection seismic for imaging carbonate aquifers,Journal of Applied Geophysics 92 (2013) 68–75.
This work was done using open source MATLAB codes,
Best regards,
Jorge
Dear Jorge,
Thanks very much, but Kriging has been included since the first edition of the book :-) I will post the update of the table of contents in the next post.
Kind regards, Martin
UPDATE:
Dear all,
Here's an updated table of contents and commented list of suggestions of new sections/methods from this question/answer. Please check this first before adding more suggestions :-) Comments on my comments on the suggestions are also very welcome.
I completed writing most of the new stuff and won't be able to add much more to it right now. Some of the sections within the chapters have been expanded significantly, for instance Section 3.5 on Empirical Distributions.
The new edition should come out within less than a year and making it an interactive ebook will keep us very busy for months. SO ... thanks again to everybody helping to identify "missing statistical methods"!
Kind regards, Martin
Dear Martin,
What about cokrigng for processing surface seismic data and/ or cross well sesimic data?
Jorge
Dear Martin,
you and colleagues have done a big work. Another effort should be Monte Carlo Markov Chain and Kalman filter. Markov chain, Monte Carlo and Kalman should be better understood introducing the Bayesian inference. I too I am not a fan of Bayesian inference, however I recognize some problems can be better solved with Bayesian inference.
As a matter of comparison you can see the table of content of my book: Introduction of Inverse Methods with application to Geophysics and Remote Sensing (Springer), where there are some chapters on the topics I have mentioned. Unfortunately the book has been actually written in italian.
http://www.springer.com/earth+sciences+and+geography/book/978-88-470-2494-6
Thanks, Rodolfo, our library has the ebook version of your book that I will check in a minute! Kind regards, Martin
Dear Marin.
I'm waiting for a new edition. When would you like to finish?
Iwona
Dear Iwona,
Maybe by the end of the year or early next year. Thanks for your interest!
Kind regards, Martin
Hi Martin,
Thanks for your effort in writing this book-this is a great idea! I think that some examples from geochemical data analysis (e.g. PCA, cluster analysis, etc.) would be really useful. It is especially important for geochemical exploration methods (Journal of Geochemical Exploration provides numerous examples).
Best wishes,
Inna
Dear Inna,
Thanks for your nice comment and interest! Please let me clarify again, the book already exists in its 3rd edition ... and it already contains the PCA and cluster analysis :-)
Trauth, M.H. (2010) MATLAB Recipes for Earth Sciences – Third Edition. Springer, 336 p. and CD-ROM, Hardcover, ISBN: 978-3642127618.
Actually I am currently working on Chapter 9 on Multivariate Statistics, which now also includes Multiple Linear Regression and Discriminant Analysis (as an example for classification and support vector machines). The RG survey here really helped a lot to identify the really most important methods in multivariate statistics out of a great variety of methods.
We are a bit struggling to find a nice synthetic example that better explains/demonstrates the difference between the PCA and ICA (both included in the 3rd edition), in particular on the influence of normality on the quality of linear unmxing of both methods. Today I think we found something but it requires a lot more testing before being used in the book.
Thanks again!
Kind regards, Martin
Dear all,
Writing of the new edition has been completed. The proofreader is now working on it. As you can see from the contents (attached), some of your suggestions of "missing methods" actually made their way into the book. Numerous minor suggestions, such as alternatives to Pearson's correlation coefficient are now included within existing sections.
One of the major issues, however, was the new R2014b release of MATLAB. I got the beta just after having finish all the code within the book. Although I consider the new release as a good one it has required lots of modifications of the MATLAB code within the book. The new release will change the entire graphics of MATLAB, as well as it replaces old functions such as HIST by new ones, which are object oriented and much better handling graphics properties.
I have made most of the necessary changes and will now work on the graphics over the next couple of months. And then I will hopefully submit the new edition before the end of the year! Thanks again to everyone for his/her valuable contributions to the discussion!
Kind regards, Martin
A delayed response, but I think a section in the signal processing chapter on calculating the significance of spectral peaks would be helpful and interesting.
Dear Broxton,
Thanks for your response! In fact the book has been largely written and we are currently working on the layout, in particular on the interactive eBook version of it. Chapter 5 has a section (already in the previous edition) on Lomb-Scargle Periodograms which introduces two methods to calculate the significance on the spectral peaks.
Kind regards, Martin
Dear all,
The 427 page 4th edition of "MATLAB Recipes for Earth Sciences" has been submitted to the publisher and will be available in spring 2015. Thanks to everyone who contributed to this discussion!
Kind regards, Martin
http://www.springer.com/earth+sciences+and+geography/book/978-3-662-46243-0
Here's a little movie showing the interactive iPad version of the book that will be published soon.