An experiment combining the use of two ecosystem models was conducted to search for effective protection strategies for the Gulf of Finland (Baltic Sea). Reference and scenario simulations were first run with a one-dimensional (1D) model for seven main basins of the entire Baltic Sea until steady state was achieved. The obtained basinwise distributions of inorganic nitrogen (N) and phosphorus (P), as well as sediment labile P, were then used to initiate 5-y simulations with a three-dimensional (3D) ecosystem model. The results suggest that relatively small local load reductions (the "Finland" scenario) would improve only the state of adjacent coastal waters significantly. This would be the case, even for runs covering several decades, which clearly exceed the residence times of nutrients in the Gulf of Finland. A significant decrease from a substantial loading source to the Gulf (the "St. Petersburg" scenario) would decrease cyanobacterial biomasses in the entire Gulf of Finland and also immediately outside it. A reduction in the current Polish nutrient loads would improve the situation in the whole Baltic Proper and cause an extensive decline in cyanobacterial biomasses in the Gulf of Finland, as well. However, it would take several decades until the improvement caused by reducing loads in the "Poland" scenario is seen, while in the "St. Petersburg" scenario the corresponding time lag would only be a few years. Our results suggest that the common water protection policy in the Baltic Sea region should have the largest nutrient sources as its primary target, regardless of their location and country.