Understanding the dynamics of a microclimate environment is difficult because of the number of factors that lead to changes in the environment. Minimizing the number of sensors needed to accurately characterize the environment results in low deployment and maintenance cost, while maximizing the utility of the data. Optimal sensor placement is difficult because it is dependent upon the properties of the environment, the types of obstacles in the environment, as well as the sensing phenomenon. We have deployed a sensor network in a botanical garden consisting of both static and portable nodes. Each node is equipped with temperature and humidity sensors, and readings were taken once per minute for over a month. Using the data from this deployment we present and evaluate two different approaches to sensor placement. The basis for our deployment approach is spatial-temporal statistical analysis that combines splinebased modeling, principal component analysis, and data partitioning. We prove that sensor network deployment is an NP-complete problem using a transformation from the dominating set problem. We also develop an integer linear programming (ILP) formulation that calculates the provably optimal solution to the network deployment problem.