The ABR service suffers greatly from the long propagation delays typical of satellite networks. It has, in fact, been shown that a long delay, in combination with a high bandwidth, can deceive ABR sources and make them believe that a break has occurred somewhere along the path. ABR sources respond to breaks by reducing their cell rate, so resource utilization decreases drastically. The solution to this problem has been to reduce the cell rate only when feedback from the network has been missing for a period longer than the round trip delay. However, this delay in reducing the cell rate causes many additional cell losses if a break actually occurs. In this paper, a framework is introduced for ABR traffic control in satellite ATM networks. The framework is composed of an accurate analytical model for standard ABR traffic behavior in long propagation delay networks, and new procedures which avoid false decreases in the cell rate. The analytical model is used to forecast false decreases of the cell rate and trigger the new procedures. Overall, the proposed framework avoids a reduction in network utilization without causing additional losses.