The study presents a conceptual model of an aggregator who selectively pays farmers for altering farming practices in exchange for carbon offsets that the change in practices generates. Under the assumption that the offsets are stochastic and that the aggregator maximizes the sum of the offsets from the purchase that he/she can rightfully claim with a specified level of confidence subject to a budget constraint, we investigate the optimal discounting of expected carbon offsets. We use the model to estimate empirically the optimal discounting levels and costs for a hypothetical carbon purchasing project in the Upper Iowa River Basin.