Wireless sensor networks (WSN) have been largely used in various applications due to its ease of deployment and scalability. The throughput of such networks, however, suffers from high bit error rates mainly because of medium characteristics. Maximizing bandwidth utilization while maintaining low frame error rate has been an interesting problem. Frame fragmentation into small blocks with dedicated error detection codes per block can reduce the unnecessary retransmission of the correctly received blocks. The optimal block size, however, varies based on the wireless channel conditions. In addition, blocks within a frame can have different optimal sizes based on the variations on interference patterns. This thesis studies two dynamic partial packet recovery approaches experimentally over several interference intensities with various transmission-power levels. It also proposes a dynamic data link layer protocol: Hybrid Seda (H-Seda). H-Seda effectively addresses the challenges associated with dynamic partitioning of blocks while taking the observed error patterns into consideration. The design of H-Seda is discussed in details and compared to other previous approaches, namely Seda+ and Seda. The implementation of H-Seda shows substantial enhancements over fixed-size partial packet recovery protocols, achieving up to 2.5x improvement in throughput when the channel condition is noisy, while delay experienced decreases to only 14 % of the delay observed in Seda. On average, it shows 35% gain in goodput across all channel conditions used in our experiments. This significant improvement is due to the selective nature of H-Seda which minimizes retransmission overhead by selecting the appropriate number of blocks in each data frame. Additionally, H-Seda successfully reduces block overhead by 50% through removing block number field reaching to better performance when channel conditions are identical.