BMElib (Bayesian Maximum Entropy library)

Last checked: Sept 13, 2001





Marc Serre, University of North Carolina, USA
Patrick Bogaert, Université Catholique de Louvain, BELGIQUE

George Christakos, San Diego State University, USA


MATLAB for Windows & UNIX


Space/time Geostatistics (any spatial dimension)

Main Functions


  • general statistics (histograms, probability plots) 
  • variogram and cross variogram calculations in space/time (any spatial dimension)  


  • BME using probabilistic and hard data
  • BME using interval and hard data
  • BME using hard data only (simple kriging)
  • BME using probabilistic and hard data with a trend model
  • BME using interval and hard data with a trend model
  • BME using hard data with a trend model (ordinary kriging and kriging with trend)
  • all estimation can use vector fields (e.g. cokriging) and be in the space/time domain
  • all estimation can be with transformation to handle non-Gaussian data
  • LU simulation
  • sequential gaussian simulation 


  • all these functions run in MATLAB, takes advance of MATLAB graphics  

Source codes

All source codes of BMElib is available in MATLAB 

Good to know 

BMElib implements space/time estimation using the Bayesian Maximum Entropy (BME) theory.  BME is a very general framework, which leads to the well known kriging algorithms (simple, ordinary, with mean trend model) as a special case under some limiting conditions (when unsing only hard data, etc.), but provides a more comprehensive model when using both hard AND soft data (as well as many other sorts of knowledge bases).


BMElib is free, but the associated book (Temporal GIS: Advanced Functions for Field-Based Applications, by Christakos, Bogaert and Serre, Springer-Verlag, 2001 ) is indispensable and includes a CDROM containing the source code for BMElib, with executables or compiling instructions for either PC or Unix.



Download is available from with the permission from the authors. Check the website for more information or contact Marc Serre.