Abstract: Covariance　function 　fittin g　for 　stochastic 　si gnal 　is 　a　key　problem　for 　collocation．In 　theprocess 　of 　covariance 　function　estimation we　often 　assume 　stochastic 　si gnals　with　the 　characteristic　of isotro pic but　anisotropic 　is　more　generall y．Combined　the 　expressions 　of 　error 　component 　in　different directionsa　method　of 　covariance 　function 　fittin g　based　on　anisotropic 　is 　given．Further adaptive collocation which　constructed　by　variance 　component 　estimations 　can　ad just 　the 　contribution 　to　model parameters 　by　observation　errors 　and　stochastic 　si gnals　and　weaken　the 　affect 　leadin g　by　their 　uncertaint yis 　proposed　and　applied 　in　missing　InSAR　data 　fittin g．A　practical 　example 　shows 　that 　collocation 　can　well　supply　the　missing　dataand　adaptive 　collocation　based　on　anisotropic 　covariance 　function 　can　give 　hi gher　accurac y　in　missing　data 　fittin g．