The probability of a species to be present on a certain site is a quantity of interest for monitoring programs. Data for the occupancy of a species is recorded along with habitat covariates that are suspected to relate with its status (presence/absence). The objective of the analysis is to estimate the proportion of sites in which the species is present and the effects of the habitat covariates on the status of the species. Nevertheless, it is possible to have some errors in the observations. The most popular approach to estimate occupancy while accounting for the detection error is that of multiple surveys, for which every site is visited several times. The effectiveness of this approach relies on two main assumptions: 1) during the time of the study the population is closed and 2) the replicate visits at every site are independent from each other. In my thesis I evaluated the multiple surveys approach under two perspectives: the statistical properties and the sensibility of its assumptions. For the former, I found that the estimates are unstable when the number of visited sites and the number of surveys are small. To overcome these flaws in the estimation procedure, I developed an alternative estimation method, based on penalized likelihood, that provides better estimates for small number of sites and surveys. The analysis of the sensibility of the assumptions revealed that the violation of the assumptions could lead to biased estimates. The single survey model does not require the closed population assumption, but the popular belief for the non-identifiability of the parameters sanctioned its use. I tested the reliability of the estimates for the probability of occupancy and detection from information collected from a single survey and found that, contrary to popular belief, the parameters are identifiable under certain conditions. Finally, I developed a model (the cluster sampling model) to include the dependence between sites. This model allows estimation of the site occupancy using information collected on a single survey from sites that are correlated.