Abstract:
Anything built from a seismic recording inherits the imperfections of that recording, whether it is a routine earthquake location or a probabilistic hazard assessment. This study assesses the recording performance of four broadband stations of the Pakistani seismic network, at Islamabad, Jhelum, Lahore and Peshawar, using a common set of eight regional earthquakes recorded during 2017 with magnitudes of 4.0 to 5.2 and epicentral distances of 115 to 1314 km.
P-wave arrival times were picked by hand and compared with times predicted from a standard velocity model, which gives one residual per event-station pair. Those residuals were then summarised at each station across the eight events. The four stations do not perform alike. Islamabad has the smallest residuals, a mean of 13.79 s and a root-mean-square of 17.22 s, and it records the lowest value for five of the eight events. Peshawar has the largest, 27.53 s and 29.67 s, and the highest value for six of the eight. Its mean and median sit close together, so its residuals are uniformly large rather than distorted by a few outliers. Jhelum and Lahore fall between the two at 20.79 s and 21.04 s, close enough to be indistinguishable on a sample this size, though they arrive there by different routes: Lahore's figure rests on a single extreme value, while Jhelum's builds up evenly.
Correlation with epicentral distance separates the stations more sharply than the residual sizes do. Islamabad shows no dependence (r = +0.13, p = 0.75). Lahore and Jhelum both give strong Pearson correlations, r = +0.81 and r = +0.77, but both weaken substantially under rank correlation, which means a handful of distant events is driving them. Peshawar's correlation is moderate (r = +0.68), and here the Pearson and Spearman values agree. The increase with distance at Peshawar is therefore real, and it sits on top of an offset that is already present at the shortest distance.
Power spectral density analysis, referenced to the Peterson (1993) noise models, places all four stations between the low- and high-noise bounds, and nearer the high-noise bound at short and intermediate periods. Twenty-eight of the thirty-two records reach maximum power between 0.93 and 1.98 s. That the same band recurs across a tenfold range of epicentral distance points to shallow structure beneath each site rather than to the source or the propagation path.
The clearest result comes from setting the two analyses side by side. Peshawar records the strongest and most reproducible spectra of the four stations and still carries the largest residuals; Islamabad shows the reverse. Signal strength therefore does not account for timing performance. That points towards station timing, metadata or picking practice rather than towards siting or instrumentation. The study closes with station-specific recommendations, and the same approach can be applied to further stations using records the network already produces.
