Saving lives in flood-prone India may require getting the right warnings to the right people
Cyclone Amphan was South Asia's first major natural disaster of the Covid-19 era. As it bore down on coastal border regions of India and Bangladesh in late May, public officials struggled to prepare for floods and destruction while under a partial lockdown to fight the coronavirus.
This cyclone is an important portend for the policy responses needed in the coming monsoon months. Lockdowns and social distancing guidelines will reduce the efficacy of current emergency responses to monsoonal riverine floods. Specifically, our research in Bihar found that villagers rely on flood alerts delivered in-person via personal contacts. A reluctance to interact physically due to Covid-19 threatens to slow these vital information channels.
Can we swiftly leverage technology in the next few months to create new and effective information channels?
We are researchers working through the Yale Economic Growth Center, in collaboration with Google, on testing mechanisms to ensure villagers have robust access to a reliable early warning system (EWS) for flooding in the Ganges-Brahmaputra river basin. These floods led to the death of over 16,000 people and the displacement of over 200 million more from 2000-2010, and roughly $20 billion in economic damage. The increasing pace of climate change suggests that their severity is only going to increase.
Floods impose physical and mental health costs in addition to economic losses in lower-income, flood-prone countries like India and Bangladesh. An effective early warning system can lower such costs. However, getting the right warnings to the right people is often difficult due to underdeveloped dissemination infrastructure and systems. For instance, in India, flood alerts are issued centrally through the media and do not indicate specific areas where flooding will occur.
In 2019, Google launched a pilot EWS for floods covering 11,600 sq. kilometers of the Ganges-Brahmaputra river basin. Google's system uses artificial intelligence to reliably predict when and where flooding will occur. This information is then pushed as a notification to Android smartphones in specific affected areas which are highlighted by Google Maps links.
In summer 2019, the research team piloted outreach mechanisms during monsoons, followed by a detailed household survey to 810 households across 81 flood-affected villages in the Indian state of Bihar. The aim was to pilot outreach efforts and administer a household survey in order to gain information that could help design on-the-ground interventions to increase the reach and impact of flood alerting in 2020.
This research is still underway, but we have some early insights that will inform our work in the coming months as we develop EWS in a Covid-19 world.
1. Existing flood alerts are widespread and timely, and villagers recognize their value.
In every surveyed village, at least one household received an alert and, in most villages, over half of households received one. Among households that received a flood alert, roughly 70% got it before flood waters reached the village. In 9 out of 10 villages, at least one household received a flood alert before the water arrived, and in the average village, nearly one-third of households received a flood alert before the water.
People pay attention to, and act on early warnings: 65% of households that received a flood alert took some steps to avoid damage before the water reached the household, 45% moved food to safety, 40% took precautions to safeguard livestock, 35% moved their assets, and 20% moved household members to safer locations. So, early warnings are definitely worth the effort. Even in an area suffering from low literacy, limited education, and high poverty, a majority of citizens act on information they receive.
2. Villagers want more specific information provided by comfortable modes of communication. Can technology help?
We asked households about the type of information that would be useful to include in flood alerts. A majority responded with the time when flood waters will reach the village and the velocity of flood water. However, traditional EWS provide only basic information: only 20% of households report receiving information on arrival time and 30% on water velocity. Similarly, 40% of households would like to be informed about the depth of water when it reaches the village, but only 17% got that information. They want to know the length of time the village would be inundated by flood waters, and 2% received that information.
EWS can add immense value by tailoring forecasts in accordance with household preferences, prioritizing the provision of localized water depth and flood timing in warnings.
3. Poor, rural communities generally rely on village networks and trusted contacts. In this era of social distancing, how can we create effective interfaces with smartphones?
Bihari villagers trust their leaders, and this was reflected in our surveys: more than 45% of households said they would like to receive alerts from local leaders (that is, the head of the village or members of the village council). However, that was not how they generally got the information in the past: of the households who received an alert in 2019, very few (less than 5%) did receive it from such leaders. While 86% received a flood alert in person, for 80%, that was from friends, neighbors, or family members in the village.
Smartphone penetration here is low: only one in five of our survey respondents had one. So, the technology they preferred as the vehicles for warnings were loudspeakers (25%) and phone calls (24%). However, of the households that received an alert in 2019, less than 2% received it via loudspeakers and less than 6% via phone calls.
Importantly, while relatively few villagers own smartphones, over 95% of their local leaders do.
This presents not only a challenge but also an opportunity. Other research has shown that marginalized populations are more likely to evacuate based on messages from leaders. Even in developed countries like the United States, people are more likely to act on warnings from leaders – human beings who have provide trustworthy in the past.
So, the key may not be to reach all people with an early warning, but the right people. And we've shown this is possible. We conducted a pilot of a selective EWS in 2019 and successfully sent flood alerts to almost all local leaders in our sample via WhatsApp.
Given Covid-19, how can we plan early warning systems that complement smartphone alerts with on-the-ground interventions, that utilize local leaders but follow social distancing guidelines? The Indian government's focus on contact tracing apps will help emphasize the salience of smartphones as a public health tool among rural officials.
We intend to seize this moment. Over the next two years, we intend to educate local leaders about the opportunities offered by EWS, and establish robust protocols for local leaders to alert villagers via more traditional but socially distant modes of communication (e.g. loudspeakers or phone calls). We will describe this work in future blog posts.