In addition to the commodity value of lumber, forests can also provide the function of conserving water and land resources, enriching biodiversity, and modulating micro-climate, etc. To secure the effective supply of forest resources, it is essential for governments to formulate subsidy programs for afforestation. Under most current afforestation programs in the world, the amount of afforestation payment covers only the total cost of afforestation and the opportunity cost of agricultural land, in which the social benefits of forests are not taken into consideration, and the period of payment is mainly determined by the forest rotation years of most common trees planted. However, based upon our model, it is found that the current afforestation program in Taiwan will not be able to achieve the socially optimal result; it will also cause higher government expenditure in subsidizing farmers. Different from the previous literature that did not consider the effect of the penalty for the farmers who harvest forests before the contract year, and only used the forest maintenance year to determine the duration of carbon payment without considering its proper justification, we develop a landowner's behavior model to determine the optimal payment structure over the whole period, and analyze the effects of the penalty payment. Based upon our theoretical model, a simulation is conducted by using cunnignhania lancelata as an example to determine the optimal number of years and the optimal amount of carbon payment each year. It is shown that our model can be applied to different kinds of trees over various kinds of situations.