Drive tests are important means to evaluate critical properties of a wireless network in operation. The network coverage is vital, and therefore, the received power of pilot signals from the base stations are monitored to estimate the spatial distributions and variations of the power gain. With uniform time-sampling and a varying velocity, the typical temporal filter fails to extract the interesting information. In this paper we apply convolutional spatial filtering, both causal and non-causal, to resolve the problem. Relations to spatial data analysis methods are also commented upon. Simulations indicate significant improvements.