This thesis is a study on the issues involved in implementing asynchronous transfer mode (ATM) over satellite links and the ways in which ATM can be optimised to achieve better performance over such links. The main issue is the impact of errors introduced by the satellite links on the performance of ATM. Options to improve the performance of ATM over SATCOM have been identified and they are: (i) increase transmit power, (ii) FEC, (iii) adoption of an alternative ATM architecture, (iv) extensive to commercial-off-the-shelf (COTS) ATM equipments and (v) construction of ATM protocol enhancers. The fifth option which uses a combination of protocol conversions and error control techniques is preferred over the others as it allows maximum use of standard COTS ATM equipments and provides a framework for experimentation with multiple versions of ATM equipments (that are evolving rapidly). As part of the strategy of using protocol enhancement to improve the performance of ATM, two techniques have been proposed. The first technique is called cell header duplication, and it improves cell loss ratio by compressing the information in the cell headers and using the extra room in each header to carry the duplicate compressed header information of the previous cell. When a cell header is corrupted, the corrupted header information is replaced with the duplicate copy carried in the next cell, provided that too is not corrupted. This technique was compared with cell header interleaving and error tolerant addressing which are cell header protection techniques proposed by others. The second technique is called selective cell retransmission, which employs a partial retransmission (hybrid) ARO strategy to perform error recovery for only the individually errored non-realtime cells allowing it to achieve higher reliability of data transfer and more efficient utilisation of satellite bandwidth. The overhead in the proposed scheme varies according to the error conditions in the satellite channel making it an adaptive system. This is achieved by using a unique method of sending error erasure information (inserted into cells carrying negative acknowledgement messages) from the receiver to the transmitter which will use this information to pin-point the corrupted cells. The scheme is also compared with concatenated coding (using a Reed-Solomon code) and link protocols (using selective repeat (full retransmission) ARQ).