Incomplete knowledge of the variation in site quality to be encountered during migration may make it necessary for birds to base their departure decisions on behavioural rules of thumb. Here I analyse four such rules for time-minimizers and show how they perform in spatially heterogeneous, autocorrelated environments. The "rst rule projects the currently experienced fuelling rate onto all the future sites, the second rule is based on a "xed expected rate along the route, the third rule assumes that the expected rate at sites nearby is close to the value currently experienced and the fourth rule uses the experienced values to form an expectation during migration. Simulations demonstrate that a global update rule which uses the current experience and interprets it as the global value to be expected at all the future sites is outperforming the second rule, which is based on "xed expectations, in nearly all circumstances and that it is very robust with respect to spatial autocorrelation and the degree of variability in fuelling rates. This rule leads to the repeatedly observed pattern of a shallow relationship between departure fuel load and fuel deposition rate. A third rule shows its best performance under the assumption that the quality of sites only changes little with distance and therefore comes close to the global update rule. The fourth departure rule, a modi"ed version of the global update rule, which allows some adjustments of the expected value through all the experienced values so far, leads only to small increases in performance.