Konstantin Krivoruchko is one of the world’s leading authorities on the subject of applied geostatistics. He has been at ESRI for more than ten years, devoted to building and promoting tools for the statistical analysis of geographic data. I recently spent two hours with Konstantin over sushi, talking about how he came to ESRI, his new book, what he sees as the biggest challenge in the field of geostatistics, and what might be on the horizon.
Konstantin, can you share a little about your background?
I was a professor at the International Sakharov Environmental University in Minsk, in the former Soviet Union, now Belarus. I was director of the GIS Laboratory there, and also a government scientific adviser on the Chernobyl accident. We developed a small GIS package, a large component of which was geostatistics. It was a Windows-based program we called MapStudio [note: no relation to the ArcWeb Services application called MapStudio]. We had several contracts for doing analysis of radioecological and epidemiological data collected after the Chernobyl accident, and I traveled to conferences to deliver presentations.
Screen shot from the original MapStudio software.
So how did you become connected with ESRI?
At one of these conferences, in Stockholm, Sweden, I believe it was in 1996, I was presenting MapStudio, and I met with representatives from ESRI. What followed were some correspondence, and then a visit to Redlands in 1997 where I spent one full week non-stop presenting the software. Jack Dangermond, ESRI president, proposed we move to Redlands, and the whole GIS Lab team of six people moved here in June 1998 with prototype software in hand.
What was the process for moving the geostatistics code from MapStudio to ArcGIS?
Basically, it was learning everything we could about ArcGIS. ESRI also hired the top spatial statisticians, including Professor Noel Cressie of Ohio State University, as consultants to advise us on the most recent advances in geostatistical theory. The whole process of building the Geostatistical Analyst extension to ArcGIS took about two years.
Another screen shot from the original MapStudio software.
Were you satisfied with the result?
I think that when it was released in 2001 it was the best geostatistical software application in the world. It has developed the largest following (estimated at about 20,000 users) of any geostatistical software. Geostatistical Analyst has not changed dramatically from that first release and today several additions are required to make sure that we are still the best. The number one addition we need to make is the inclusion of Bayesian models.
What have been the biggest challenges in all of this?
What I didn’t realize in the beginning was that building the software would not be the hardest part, that the biggest challenge is education. I was very surprised with a finding that many users simply use the default model, ignoring the largest and the most interesting part of the software. Geostatistical Analyst is the most complicated part of ArcGIS, and unfortunately, ESRI regional offices could not demonstrate its features to the local users. Most software documentation explains how to use the software, but with advanced geostatistics, we also had to try to explain in the software manual why and when to use a particular kriging model. Unfortunately this was not enough. Large efforts are required to help both the beginner and intermediate users to use Geostatistical Analyst properly.
How do you plan to address the education issue?
One way is through the book I’m writing. The book is called Introduction to Spatial Statistical Data Analysis for GIS Users and is an overview of the existing approaches to spatial data analysis. It’s more about scientific data analysis than it is just about geostatistics. It shows how various real-world problems can be solved using modern statistical theory in the context of GIS. The book includes a large number of case studies in the applied usage of spatial statistics across such data as radioecological, air pollution, agriculture, forestry, econometrics, epidemiological, crime, meteorological, and more.
Cover of Konstantin’s upcoming book.
How has it been progressing?
The original idea behind the book was to have simple explanations of geostatistical concepts with several detailed case studies, and the project kept growing and growing…it started out at about 100 pages, and now it is close to 1,000 pages. It is very visual, with more than 1,000 color illustrations. The writing is done; it’s now being edited and designed. ESRI Press plans to publish the book around October of 2009.
This illustration from Konstantin’s forthcoming book shows the construction of probability values for a study of cadmium concentrations in Austria.
Who is the target audience for the book?
Several types of people will benefit from reading the book:
• ArcGIS Geostatistical Analyst and ArcGIS Spatial Analyst users who are interested in case studies and detailed descriptions of statistical models.
• Students and teachers who are looking for largely qualitative explanations of the advantages and disadvantages (I believe, for practitioners, the models critique is as important as the models promotion) of various statistical models. In fact, this book is a textbook with a large number of exercises. A draft version of the book has been used in the UNIGIS online education network for two years.
• Researchers who need a practical introduction to statistical models of certain categories of spatial data.
• Readers who want to learn more about applications of spatial statistics.
In the rest of this interview, Konstantin discusses what he brought back from the recent GEOSTAT2008 conference in Santiago, Chile, and what we might expect to see in future releases of Geostatistical Analyst.