Human activities have a profound impact on marine ecology in Greenland, especially on narwhals, one of the endemic species of whales in Greenland. I will focus on movement data that are collected at irregular time points in a constrained spatial domain. I will present different stochastic models to analyse these data, especially stochastic Langevin diffusions or stochastic differential equations in a constraint domain. I will discuss more in details the case of multidimensional, hypoelliptic, penalized and partially observed diffusions. New estimation methods involve numerical schemes called splitting schemes, that prove to be theoretically and numerically efficient.