4G networks were the most affected by Sulawesi earthquake

OpenSignal measured the impact on smartphone users’ experience before and in the aftermath of the 7.5 magnitude earthquake with epicentre on the Indonesian island of Sulawesi, on Friday September 28, 2018. The earthquake sparked a tsunami with waves up to 6m high, and together they caused wide destruction to the city of Palu and the nearby area. The impact on mobile technology was extensive but according to our data it did not cause a complete service blackout throughout the whole area.

4G services took almost two weeks to return to normal

OpenSignal observed a drastic decline in 4G availability, with the service requiring almost two weeks to return back to the previous 30 days’ average for our smartphone users.

However, 4G availability did not see its worst decline on the very first day after the earthquake but on the second — Sunday, September 30 — when LTE availability dropped by almost 60%. This delayed impact on mobile network experience suggests that direct damage to the cell towers from the main shock was not the only cause of the slow recovery of the network.

After the earthquake smartphone users were less likely to find LTE signal

Plotting the geographical distribution of our data, we noticed that readings before the earthquake were fairly spread across the whole city and its surrounding boroughs, while data collected in the days following the natural disaster was mostly located in the city center and around the airport, with sparse measurements throughout the other areas. Such a distribution suggests that smartphone users were more likely to find mobile signal in the city center — where mobile operators usually deploy a higher number of cell sites to cope with the larger population density.

Map tiles by Stamen Design, under CC BY 3.0. Data by OpenStreetMap, under ODbL.

However, the maps highlight a difference in the readings coming from different mobile technologies. Before the natural disaster, smartphone users could access the 4G network almost anywhere they had a mobile signal, whereas afterwards the area covered by LTE services shrank compared to the area where devices had mobile connectivity, providing an additional hint on the LTE availability decline.

4G devices on average connected to more distant LTE cell sites

OpenSignal measured the average distance of 4G devices from LTE cell sites for each day following the natural disaster, and compared it with the average distance in the same locations in the 30 days before the earthquake. We observed that devices in the aftermath of the natural disaster connected on average to more distant cell sites, and that it took 11 days to return to normal.

On September 30 — two days after the earthquake — devices were on average connected to LTE sites located more than 1,400 meters away, whereas in the 30 days before the average distance in the same locations — as we saw, mostly the city center — was less than 500 meters.

Such a large increase in average connection distance suggests that fewer LTE sites were operational in the days following the earthquake, with devices not able to connect to the 4G network via live sites located as near as before connecting instead to more distant cells. This also suggests that on average more devices connected to the same cell site, increasing the strain on those parts of the network that remained live.

When natural disasters occur, mobile operators face significant challenges to restore normal service and depending on the size of the calamity, that can take days — as we measured with hurricane Florence in the U.S. — or weeks, as with Sulawesi earthquake in Indonesia. In such situations, operators not only have to deal with restoring those parts of the network that were disconnected, but also face a degrading mobile experience because of more users connecting to a smaller number of active cells, further increasing the strain on the network.

Our Methodology

OpenSignal measures the real-world experience of consumers on mobile networks as they go about their daily lives. We collect 3 billion individual measurements every day from tens of millions of smartphones worldwide.

Our measurements are collected at all hours of the day, every day of the year, under conditions of normal usage, including inside buildings and outdoors, in cities and the countryside, and everywhere in between. By analyzing on-device measurements recorded in the places where subscribers actually live, work and travel, we report on mobile network service the way users truly experience it.

For this particular analysis, 8,854,926 measurements were collected from 1762 devices in the city of Palu – Sulawesi island, Indonesia –- during the period: Aug. 29 – Oct. 20, 2018.

We aggregated the data on a daily basis and compared smartphone users’ mobile experience during the days following the earthquake, versus the average experience of the previous 30 days. We have calculated statistical confidence intervals and plotted them on the graphs. When confidence intervals overlap for a certain period, the results are too close to declare a statistically significant difference.