This work presents optimal precoder designs for the linear decentralized estimation of an unknown temporally correlated parameter in an orthogonal frequency division multiplexing (OFDM) based wireless sensor network (WSN). The precoder design is initially developed for a scenario with analog sensor measurement transmission with an objective to minimize the sum mean square error (sum-MSE) of estimation at the fusion center, incorporating constraints for distortionless reception as well as a total transmit power budget. Next, in order to facilitate digital transmission of the sensor measurements with limited bandwidth, a rate distortion theory based optimal bit allocation scheme is proposed to minimize the sum-distortion for a given bit budget. Subsequently, the optimal precoders are also designed for this scenario with quantized measurements to minimize the sum-MSE at the fusion center. The MSE performance of the centralized MMSE estimator is also derived, which serves as a benchmark to characterize the performance of the proposed decentralized schemes. Simulation results are presented to illustrate the performance of the proposed schemes and also compare them with analytical benchmarks.