The aim of this paper is to address the Demodulation Reference Signal (DMRS) based channel estimation issue in railway tunnel scenarios. Thus, we propose a deterministic model to accurately generate the time-varying channel response. And to investigate the impact of these particular environments on a Cyclic-Prefix Orthogonal Frequency Division Multiplexing (CP-OFDM) system, we first explore the pilot patterns proposed in the 5G specification, then we apply a thresholding based channel estimation algorithm to extract the demodulated data symbols. By comparing the Bit Error Rate (BER) and the latency of seven slot configurations at different train speeds, the results allow us to make some recommendations depending on the intended railway application and the required Quality of Service (QoS) level.