In this paper, we investigate the cross-layer design of medium access control (MAC) protocols and distributed source coding (DSC). We closely model the interaction between the source compression and the medium access control layer for sensor network applications and determine the parameters that are most relevant to both the layers. Specifically, we exploit the spatio-temporal redundancy in the measurements at the sensors and adjust the MAC parameters accordingly. We use short code block lengths, which are easily realizable on resource-constrained sensor nodes. Experimental results with real sensor nodes validate our approach. Whereas this approach yields high energysavings in sensor networks applications, it also preserves the accuracy of the data recovered at the sink nodes.