As the city faces the threat of community transmission, municipal authorities are bringing in advanced computing methods to crack the mystery of COVID-19 cases with no known sources, especially in Padarayanapura.
Results of the random tests conducted by the BBMP have been less than satisfactory — over 920 tests in Hongasandra, Padarayanapura and Mangammanapalya containment zones have yielded only eight positive cases. This is in addition to four cases found during the targeted random testing at Mangammanapalya, two Palike officials said.
“But eight is not a small number in this context. From a percentage accuracy point of view, it is high and something to worry about,” explained Himanshu Tyagi, assistant professor, Department of Electrical Communication Engineering Department (ECE), Indian Institute of Science (IISc).
For the past three weeks, Dr Tyagi has been working with Professor Bharadwaj Amrutur and assistant professor Aditya Gopalan (also of ECE) to help the BBMP to develop a new testing strategy, using mathematical modeling of the containment areas.
“The BBMP reached out for advice on two aspects: First, how to sample individuals effectively for Covid-19 testing across various containment zones. Second, how to decide when a containment zone could be denotified, rather than waiting for a fixed deadline of 28 days,” Dr Tyagi said.
The scientists started by modelling Padarayanapura, the choice of which is self-evident, said Hephsiba Korlapati, MD of the Smart Cities Initiative.
Since April 7, when the first case was reported in this neighborhood of 35,213 people crammed into an area of 0.31 square kilometers, a total of 62 additional cases have been registered here as of Sunday, including the eight mystery cases found through random testing.
“One third of the city's active cases are from (Padarayanapura) and the numbers are adding on. We are looking for ways to contain the numbers,” she said.
Hephsiba added that the simulator framework, dubbed as the “Adaptive Stratified Random Sampling Strategy,” has already produced insights.
“One is that males aged 20 to 40 are most susceptible to the disease because they are more mobile,” she said. “Another is that how buildings such as apartments are laid out among smaller buildings affects the spread pattern.”
Modelling human interactions
The IISc scientists explained that they began to build a fine-grained agent-based simulator for populations ranging from 10,000 to 30,000 people.
“Within the simulator, all individuals in the population have been given various demographic attributes like age and gender, with the added presence of comorbidities. We have also modelled how people mix with each other and how they meet each other, which helps with identifying how the spread happens,” said Dr Gopalan.
The model is said to be capable of highlighting which streets in the containment zone require more testing.
The simulator’s hardware requirements were described as basic. “At the moment, because the number of people being modeled is not that large, we run the simulator using IISc’s many servers. If we were to do city-scale modelling, we would need the institute’s supercomputers,” explained Dr Tyagi.