2018MNRAS.475.1323L

We present a principal component analysis method that tracks and compensates for short-time-scale variability in pulsar profiles, with a goal of improving pulsar timing precision. We couple this with a fast likelihood technique for determining pulse time of arrival, marginalizing over the principal component amplitudes. This allows accurate estimation of timing errors in the presence of pulsar variability. We apply the algorithm to the slow pulsar PSR J2139+0040 using an archived set of untargeted raster-scan observations at arbitrary epochs across four years, obtaining an improved timing solution. The method permits accurate pulsar timing in data sets with short contiguous on-source observations, opening opportunities for commensality between pulsar timing and mapping surveys.