Land acquisition and ownership is an important part of modern agriculture in North America. Given the unique nature of farmland as a good, this paper develops a multi-agent simulation of farmland auction markets in a Canadian context. The model is used to generate data on land transactions between farm agents to determine if a particular auction design or type is better suited to farmland transactions. The simulation uses three different sealed-bid auctions, as well as an English auction. The auctions are compared on the basis of efficiency, stability, and perceived surplus. We find that the form of agent learning about land markets affects both sale price and the variance of sale prices in all of the studied auctions. The second-price-sealed-bid auction generates the most perceived surplus, most equitable share of surplus, and also decreases uncertainty in the common-value element of prices. But on a macroscopic level, it appears that auction choice does not influence market structure or evolution over time.