Abstract: We made linear regression to get the emprical formula of Tm-Ts and Tm-Ts、es、Ps using Hong Kong Kings Park sounding station(number 45004) data from 2003 to 2009,in which Tm means atmospheric weighting mean temperature,Ts means surface temperature,es means surface water vapour pressure and Ps means surface atmospheric pressure.It concludes that there is no significant difference between one-factor and multi-factor results,but the precision of regression formula based on local sounding data is higher than Bevis formula;The precision of regression formula can't be significantly increased by annually adding sample number,so it is sufficient to forecast Tm by Ts using sounding data of last year and the root mean square of estimation is 4.913 K;By classifying sounding sample data according seasons,Tm-Ts model coefficients a and b value vary little in spring and autumn,whereas they vary obviously in summer and winter;Comparing the accuracy of regression formula using data by season and no season classification,it is almost quite the same,so there is no need to regress the Tm-Ts model using different season data.It can be found that a and b value relatively uniform in one same climatic region and are correlative with climatic type