Lake level fluctuations provide evidence about past variations in the global hydrological balance. The geostatistical approach is here used to more objectively identify global patterns using an ensemble of lake level databases by examining spatial autocorrelation between sites. The spatial structures of the lake level data are then modeled and grids produced for the last 12,000 years at 3000-year intervals using ordinary and indicator kriging techniques. The two gridding techniques produced almost identical estimated regional lake status patterns, thus suggesting a robust estimation. The resulting lake-status grids are in general agreement with previous paleoclimatic reconstructions using only site-by-site lake status point maps; however, the reduction of local fine-scale variability resulted in more coherent regional spatial patterns in areas of high local variability. The 6 ka lake-status grids were compared to simulations of four atmospheric general circulation models to illustrate their usefulness in validating broad-scale climate model outputs. ¿ 2001 American Geophysical Union