The long-term changes of reference crop evapotranspiration （ET0） were estimated over the period 1960-2004 at six meteorological stations in Shandong Province based on seven different models driven by routine daily meteorological date. Firstly, the variability of predicted ET0 over the space and time scale has been compared, then the other six models were optimized by means of Least Absolute Deviation method according to Penman-Monteith model recommended by FAO and comparison values of ET0 between pre- and post-optimization. The results showed that Makkink model was the best of all, while Mass transfer model underestimated much from June to August, and Net Radiation model overestimated throughout the year at all the stations. After optimization, models gave a better estimation for monthly ET0 with a mean relative deviation error （MRE） less than 85%, and less than 10% for yearly ET0. Therefore, the models perhaps can be used instead where and/or when meteorological data were absent.