New analysis of cycling counters in Prague: Traffic increase since COVID is 82%

Městem na Kole has been monitoring cycling traffic intensities in Prague through automatic counters since 2010. In the years 2018-2020, there was a major overhaul of the installed counters, and the method of data provision changed. Our last analysis was conducted at the end of the „COVID“ year 2020, where we recorded a year-on-year increase of 32%. However, this analysis did not include all counters and was unable to detect certain errors, so we opted for the most conservative estimate at the time.

So, what has been the revised trend since 2019? In 2020, we recorded a 63% increase. The decline observed in 2021 was mainly due to extensive closures on cycling route A2, which significantly reduced traffic at heavily used counters, particularly on U Českých loděnic Street and Podolské nábřeží. The year 2022 saw a return to the peak values of the most active „COVID“ year, and 2023 surpassed the previous year by more than ten percent.

The development since 2010 shows fluctuations in traffic intensity. The increase in cycling traffic in 2011 was accelerated by the May public transport strike, while the drop in 2013 was caused by the closure of bike paths due to flooding. Several years of stagnation that followed the opening of the Blanka Tunnel may reflect a decline in infrastructure development, which slowed in 2012. The COVID lockdowns and reduced willingness to use public transport provided a boost to cycling traffic intensity in 2019. What further triggers will significantly affect cycling traffic intensity remains to be seen.

Individual counters

The cleaned data from the counters also allows us to examine selected individual counters.

Methodological Appendix

Data from the counters have never been particularly reliable. We were aware of counters with weak batteries that stopped measuring early in the morning, loops placed outside the cycling path for months due to changes in parking, and counters that provided no data for half a year. The new analysis confirmed that processing the data would remain a challenge even after 2019. To give you an idea, we’re sharing a few graphs to illustrate the kind of issues we encountered in some of the time series we worked with.

The PNK [Praha na kole – Prague on a bike] 2010 Index is calculated based on the year-over-year differences in average daily intensities at individual counters. Specifically, for counters that recorded data in two consecutive years, the average intensity during the summer season is determined. These intensities are then summed, and the year-over-year change in the index is the ratio of the sums for the two consecutive years. In the subsequent year, different counters may be used, as long as they have measured for at least two consecutive years. The advantage of this method is that counters can be replaced over time, and even significant outages at any of them won’t affect the index calculation, as they simply aren’t included. However, to calculate the index, we need a reasonably reliable daily average.

The original method dealt with deficiencies in daily data by excluding days with zero counts, manually removing excessively high totals, and calculating the daily average for the summer season. In the new calculation, this process has been largely automated. The average intensity, previously determined by the arithmetic mean, has been replaced with the median, which eliminates the impact of any remaining outliers, whether high or low. As a result, the final calculation is even more robust and reliable than the previous method, although the quality of the data being processed has not significantly improved over time—in fact, it may have deteriorated.