Wireless Sensor Network (WSN) is highly budgeted by energy since sensor nodes are mostly battery-powered and deployed in hard-to-reach area for prolonged duration. Moreover radio communication is very expensive for WSN. At the same time, a substantial portion WSN applications require periodic data collection. In this paper we investigate this issue in depth and present a solution architecture: 2PDA, that eliminates repeated transmission. The solution is founded upon temporal linear correlation among sensor data. Instead of sending each data packet we model them using method of least square that exploits temporal correlation among sensor data. 2PDA observes sensor data and performs operation parameterized by application accuracy. After successful computation only the parameters of the model are sent over the radio to the application-end or sink. We have evaluated 2PDA on real-time sensor data indicating substantial energy-savings in overall transmission. We have implemented 2PDA in TinyOS for telosb sensor node that reflects its deployment plausibility in terms of memory requirement. Finally we have proposed a simplified multi-path routing strategy to enhance the reliability of data which is of prime importance in all WSN applications.