GEOS 597e: Spatiotemporal Data Analysis Workshop

Homework 9: Independent projects

Last updated 10/27/06. Oral presentations to be completed by 12pm, Weds., Nov. 29th.
Goals
 Introduction
 Products
 Schedule
 Evaluation

Goals of this assignment
  1. Apply the empirical analysis tools you have built during our study of the GOSTA SST anomaly dataset to a different climate dataset of your choice.
  2. Analyze, interpret, and critically evaluate the results.
  3. Provide constructive peer review of scientific results; effectively incorporate constructive criticism on your work via peer evaluation.
  4. Improve your oral communication skills by presenting your results in short oral format.
Introduction  

At this point in the semester, we have developed tools for performing empirical orthogonal function analysis of  three dimensional datasets, and for singular spectrum analysis of time series, using a high-level scripting programming language.  We have also developed tools and intuition for the estimation of error and uncertainty in the products of these analyses. We have used these tools to (hopefully) wisely filter and interpret the resulting patterns, and we have honed those analytical skills by examining examples in the scientific literature of similar analyses.

In this assignment I ask you to now use the tools and skills you have developed to analyze and interpret the large-scale features in a three dimensional meteorological or climatic dataset of your choice.  Use the structure of previous Prework, Preview, and Homework assignments to develop, execute, and illustrate the results of your analysis.  I expect to see your project to cover at least the following topics; however, you are encouraged to probe more deeply into the data to produce a nuanced, thoughtful spatiotemporal analysis. 
Products

You will report your findings as a 15 minute oral presentation.  Some suggestions for developing an effective oral presentation:
  1. Start with the larger goal of the work.  Why is it important?  Fundamental climatological science question?  Societal relevance?  Other?  What are the implications of discovering the answer?
  2. Give a brief, but clear and concise statement of the specific research question which motivated your choice of dataset to analyze.  Sometimes this is best expressed in question form.  What did you expect to see result from this analysis, based on prior knowledge of the climate system or of the underlying physical climatology?
  3. Lay out the essential components of your experimental strategy so we can see where you're going.  In a short presentation, good organization will help you make even a subtle argument clear in a short amount of time. 
  4. Take some time to really dissect the figures you show us -- they are usually central to a scientific argument. Choose your figures carefully, and use them to illustrate your discussion points.  Explain your figures to the class as if you are explaining things to a non-expert. What are the axes and units? What is plotted? What points are made by presenting the figure?  Guide us through your interpretation of the figure.
  5. Take some time to discuss the major assumptions or uncertainties in the approach you have described.  You might anticipate, and answer, expected questions about your approach.
  6. Finish with a slide describing the major outcomes.  How can the results be applied to resolving the scientific question which motivated the work in the first place?
  7. Put enough text and description on your slides to remind yourself to hit the major points and to note important details, but not so much that you are tempted to read from your slides.  Instead, tell us the story of what you found from your analyses.  Note that giving a talk doesn't necessarily follow the same structure of a written report.
  8. Practice your presentation to yourself or to a classmate/friend.  Just running through it once out loud will help you work out timing and bugs, to fill in logical gaps, and make clearer slides.  A second rehearsal will often shock you by how much more smoothly it goes.
  9. All slides containing results or ideas taken from the literature should include citations, and your presentation should include a final slide of Cited References (your presentation of your project need not include reading us the reference list, of course).  For instance, the source of your analyzed data set should be given (e.g. Woodruff et al., 1998) and the Cited References slide should include the reference
Woodruff, S.D., Diaz, H.F., Elms, J.D., Worley, S.J., 1998. COADS Release 2 data and Metadata Enhancements for Improvements of Marine Surface Flux Fields. Phys. Chem. Earth 23, 517-526, accessed via Internet, 1 May 2003: http://www.cdc.noaa.gov/coads/egs_paper.html.

Schedule

You are free to make best use of your time over the next month to complete this assignment.  However, if you'd like some guidelines to structure your time, here is a roadmap for this assignment. 

Week 1 (Nov 1st-Nov 7th):  Choose your target dataset and plan your experiment.  Please see Mike for a 10 minute conference during class this week, for help with and/or approval of your chosen project.  Use the answers to these questions to develop the introduction to your project report and presentation.
  1. Target dataset:
    1. What scientific question would you like to ask using the observed variable?
    2. What is the source of the dataset, including published references on its development?
    3. What are the major uncertainties in the raw data?
    4. How was this specific dataset produced?  What were the major processing steps? 
    5. What are the major potential uncertainties in the resulting product?
  1. Plan your experiment:
    1. How will you apply EOF and SSA analysis to this dataset?  Use the structure of the course and what you have learned from Preworks and Homeworks 1-8.
    2. What is your a priori expectation for the results, given your understanding of the underlying climate system?
    3. What tests will you apply to determine the robustness of your results?
  1. Logistics:
    1. Find a source for your data.  Most likely it will be from an online data repository of digitized data. 
    2. Be sure to document the source of your data, including both internet and peer-reviewed references.
    3. Be sure to download the data in a form which both you and Matlab can understand and read.  Some good sources of data are the LDEO/Ingrid Data Catalog; JISAO's Climate Data Archive; KNMI Climate Explorer.  Best standard formats for large datasets are binary random access, netcdf,dods.  I can help you directly if you get your data from Ingrid; if you have netcdf and/or dods sources, tools here and here are available to help you read these files directly into matlab (ask me or Eneida about getting these on clue machines asap).  Bottom line: take the easiest route to getting the data you can possibly find; this is a minor component of the project.
Week 2 (Nov. 8th-Nov. 14th):  Decompose and analyze your dataset.  Use your results to develop and illustrate your written project report and your oral presentation.   Digest your findings: this is easily the most important component of your project.  Be sure to allow adequate time to probe and really study the results that you find.
  1. Execute your experiment.
    1. Perform EOF decomposition of your spatiotemporal dataset.
    2. Analysis: Truncate and filter the data; test the results for error and stability using Monte Carlo techniques; if pertinent, perform rotation of selected EOF patterns and  examine  differences between rotated and unrotated EOFs, and improvements (or not) in interpretative value.
    3. Use the results of the EOF analysis to choose a representative index time series from the spatiotemporal dataset.
    4. Perform SSA decomposition of your time series dataset.
    5. Analysis: Truncate and filter the data; test the results for error and stability using Monte Carlo techniques.
    6. Analysis: perform any other calculations or analyses to support or illustrate your results.
  2.  Analyze the results.
    1. How many significant patterns are there?  How much of the variance do they represent?
    2. Are there plausible physical interpretations for any of the leading patterns?
    3. How would you use the results to filter the dataset?
    4. How robust are your results and interpretation, considering observational error, analysis uncertainity, and ability to resolve patterns?
  3. Digest the results.
    1. What are your primary findings (results and interpretation)?
    2. How do your results compare to results or interpretations in the scientific literature?
    3. How have you addressed the motivating scientific question you posed at the outset?
    4. Did you find anything particularly interesting or surprising given your a priori expectations?
Week 3 (Nov. 15th-21st):   Develop an oral presentation describing your analysis of the dataset.
  1. Starting with items from part (3) from Week 2, construct a clear, well-illustrated oral and visual presentation of your motivating scientific question and primary findings. See Products above for some suggestions on how to develop an effective oral scientific presentation.
Week 4 (Nov. 22nd - Nov 28th):  Have a nice Thanksgiving Break.

Week 5 (Nov. 29th - Dec 5th): Project presentations
  1. Oral presentations, to be made in class, Wednesday, Nov 29th; class to be held at Tree-Ring Lab West, Seminar Rm. 20, Mathematics East Building. 
    1. Important: please email your Powerpoint presentation to Mike no later than 12pm, Wednesday Nov 29th.  Name your presentation lastname.ppt.  I will load these all onto a classroom (windows) computer in time for class, and I will review them for grading.  If you cannot or do not wish to use Powerpoint, please talk to Mike well ahead of time about alternative arrangements; if you work outside of windows, you may wish to test your presentation on a windows box before class.  Each talk slot will be 15 minutes: 10 minutes to present, 4 minutes for questions, and 1 minute for evaluation and changeover to the next speaker.
    1. See the grading rubric for my expectations for this project.
    2. Submit completed peer reviews (see below).

Evaluation
  1. Be sure to review the grading rubric before beginning the assignment, so you understand my expectations for this assignment.  
  2. We will use the peer evaluation form to evaluate the presentations (I'll bring 8 x 8 copies to class).
  3. 50% of your grade will be determined by your classmates' evaluations of your presentation; 50% by MNE.  Together this will be 41% of your course grade.
  4. I will submit final grades to Anne Chase and distribute grades to each of you by email by end of week of Dec 11th .
  5. Please be sure to pick up any other graded materials of yours I may still have at my office in the W. Stadium, or give me a campus address to which I can send them.
Back to Schedule/Syllabus.