: The importance of designing efficient systems for transmit information underwater has increased with the increase in underwater military and commercial activity. Wired (optical fiber) and wireless (radio, optical, and acoustic waves) communication systems can be used to transmit information underwater. Wired systems require a heavy implementation and maintenance cost, therefore wireless systems are a better option for underwater communication systems. Sound waves are better and more widely used than radio and optical waves because of their higher coverage range in underwater space. On the other hand, underwater acoustic data transmission has many challengies such as low bandwidth, carrier frequency, severe time and frequency changes in their channel. To address these challenges, innovative designs such as multiple-iutput multiple-output (MIMO) cooperative system and orthogonal frequency division modulation (OFDM) have been proposed. OFDM has recently received a lot of attention because of its good bandwidth efficiency in underwater acoustic (UWA) channel. Due to the ability of MIMO cooperative systems to increase capacity and coverage range, these systems are very useful in underwater communications. On the other hand, it is important to know the channel state information (CSI) for data detection, power optimization, capacity enhancement, error reduction and other network parameters improvement. Previous researchs on the underwater acoustic channel estimation for MIMO cooperative systems is very limited. On the other hand, an efficient underwater acoustic channel estimation is essential to achieve the benefits of MIMO-OFDM cooperative systems. In this study, sparse UWA doubly-selective channels are estimated using compressed sensing methods considering a MIMO-OFDM cooperative system based on amplify-and-forward (AF) relay. Discrete stochastic optimization (DSO) algorithm is used for optimal design of pilot signal’s location, and the basis expansion model (BEM) of channel’s coefficients are estimated by orthogonal matching pursuit (OMP), block simultaneous orthogonal matching pursuit (BSOMP), block joint stage determined matching pursuit (JSdMP) algorithms and a novel block joint stage determined matching pursuit algorithm (BJSdMP). The normalised mean square error (NMSE) and bit error rate (BER) results show that compressed sensing estimation methods perform better than LS and among them BJSdMP method performs best. Simulation results show that better estimation of the channel can be achieved by increasing the number of relay antennas as well as base station antennas. Keywords: MIMO Cooperative Communication, Underwater Acoustic Doubly-Selective Channel, Channel Estimation, Compressed Sensing