Underwater wireless sensor networks (UWSNs) widely used for maritime object detection or for monitoring of oceanic parameters that plays vital role prediction of tsunami to life-cycle of marine species by deploying sensor nodes at random locations. However, the dynamic and unpredictable underwater environment poses significant challenges in communication, including interference, collisions, and energy inefficiency. In changing underwater environment to make routing possible among nodes or/and base station (BS) an adaptive receiver-initiated deep adaptive with power control and collision avoidance MAC (DAWPC-MAC) protocol is proposed to address the challenges of interference, collisions, and energy inefficiency. The proposed framework is based on Deep Q-Learning (DQN) to optimize network performance by enhancing collision avoidance in a varying sensor locations, conserving energy in changing path loss with respect to time and depth and reducing number of relaying nodes to make communication reliable and ensuring synchronization. The dynamic and unpredictable underwater environment, shaped by variations in environmental parameters such as temperature (T) with respect to latitude, longitude, and depth, is carefully considered in the design of the proposed MAC protocol. Sensor nodes are enabled to adaptively schedule wake-up times and efficiently control transmission power to communicate with other sensor nodes and/or courier node plays vital role in routing for data collection and forwarding. DAWPC-MAC ensures energy-efficient and reliable time-sensitive data transmission, improving the packet delivery rati (PDR) by 14%, throughput by over 70%, and utility by more than 60% compared to existing methods like TDTSPC-MAC, DC-MAC, and ALOHA MAC. These enhancements significantly contribute to network longevity and operational efficiency in time-critical underwater applications.