In this thesis, a new dir=ltr In this thesis, the advantages of this decoding scheme are studied and certain characteristics of LDPCC code ensembles which affect the performance of these codes with the windowed decoder are identified. The performance of these ensembles and codes are considered over erasure channels with and without memory. The structure of LDPCC code ensembles is suitable to obtain a performance close to the theoretical limits over the memoryless erasure channel, both for the BP decoder and windowed decoding. However, the same structure imposes limitations on the performance over erasure channels with memory. Although the channels considered here are erasure channels, it should be noted that the WD scheme can be used over any channel. When this scheme is used to decode terminated LDPCC codes, an efficient way is provided to abtion a trade off decoding performance between the reduced latency and the termination operate. In order to have codes achieving a good decoding threshold, several design rules are suggested. Let us denote the minimum number of consecutive columns of a protograph that contain a protograph stopping set by . A good design rule for protographs must ensure the existence of large stopping set spans over erasure channels with memory. Some bounds for the maximal are obtained and some protographs are designed that achieve minimal spans close to this optimal values. Construction of LDPC block codes for bursty erasure channels are studied. The performance metric of a code over a bursty erasure channel is related to the maximum resolvable erasure burst length (MBL) denoted , that can be decoded by a BP decoder. Methods of optimizing codes for such channels focus on permuting columns of parity-check matrices to maximize . Instead of permuting columns of the parity-check matrix, designing ensembles are considered to maximize , while maintain the convolutional structure of the code.