Forest statistica! data, agriculture and forestry census data from 1960--19, 95 and mathematical model F/N=a ?exp(p (N/S)), (where F is forest area, N is population, S is city area, a and p are parameters), are employed to predict the urban forest decrease in Kanagawa prefecture Japan. The results show that per capita forest area is decreased because of the increase of population. If the population is increased by 10%, the per capita forest area will be decreased from 1%to 92%. Most of validation coefficients in different cities are over0.9, that means calculation results are very accurate, lnorder to prevent the decrease of urban forest and reduce the preservation cost, based on the deforestation rates and financial status of different governments, we suggest that Japanese governments should purchase the forests, forest land, and monitor the forest land conversion in a long term. These preservation strategies are useful for developing countries, such as P.R. China.