In this paper two data assimilation methods for the two-dimensional ad
vection-diffusion equation are developed for the estimation and predic
tion of air pollution. First, a time-invariant Kalman filter is develo
ped by using Chandrasekhar-type filter algorithm. Then, as an alternat
ive, a simpler data assimilation method based on Kriging is also propo
sed. The two data assimilation methods are compared in a number of exp
eriments. (C) 1997 by Elsevier Science Inc.