Bhopal Disaster: Was the Human Damage containable?

On December 3rd, 1984 a deadly methyl isocyanate gas leak from a chemical plant killed thousands of people in Bhopal city and the surrounding areas. Official death toll put the estimate at around 15,000.  The plant owned by a subsidiary of Union Carbide Inc. of United States was manufacturing Sevin(chemical name Carbaryl) an insecticide.  Union Carbide discovered carbaryl and introduced it commercially in 1958.

Damage from man-made disasters can minimized using technology.
Time magazine cover of the Bhopal tragedy. Courtesy: Time Inc

The plant produced and in-situ compound called Methyl Iso Cyanate (MIC) in the process of producing Carbaryl.
MIC is an extremely toxic and unstable substance and even in very small quantities is fatal. Large quantities of the MIC were stored in a steel tank.Water, inadvertently entered the storage tank and caused an exothermic, runaway reaction to occur. A vapour cloud about 100 m wide was formed above the plant consisting of the toxic gases MIC, Phosgene, Hydrogen Cyanide and Methyl Amine. It was carried by a north-westerly wind over adjacent parts of the town to the south of the plant where most of the victims died from respiratory failure.

Disaster Response:

Systematic disaster response at that time was still in nascent stage in India. There were apparently no maps available that would help the emergency crews in reaching the  most affected parts of the population. Simulation of the contaminant transport by air along with reliable population distribution data might have been able to minimize the casualties. Both, unfortunately, were not available at the time of disaster either due to lack of awareness or limits of technology of the time.

It is not to say no proper data were available, data from Census of India was available. Census surveys in India are carried diligently and are a very reliable source for estimation of population distribution and other socio-economic insights. This data if plotted on a map will give an idea of distribution of population according to their residence. But there is limitation to that that we will discuss in following paragraphs.

Population Datasets:

Most censuses count people at their nighttime residences. All population counts, even the most sophisticated high-resolution official censuses of advanced nations like the United States, are stochastic estimates, meaning they are  intrinsically non-deterministic. Accuracy and precision are limited by the census takers’ access to homes and even to whole neighborhoods; by the census takers’ understandings of personal work and travel habits; and by the frequency with which censuses can be undertaken. In addition, many nations are reluctant to release detailed census counts, and some release only a national total. For most of the world, the best available official census data are at province level (i.e., one administrative division below national) and of varying age, sometimes decades old.

In response to above limitations of above mentioned official census counts  in disaster prevention and response a new kind population distribution model was developed by Oak ridge National Laboratory, Tennessee, USA. This data set is created by collecting best available census counts for each country, projects aggregate populations to a target year, calculates a probability coefficients for each cell, and applies the coefficients to the census counts which are employed as control totals for appropriate areas(usually provinces). The probability coefficient is based on slope, proximity to roads, landcover, night time lights, and an urban density factor. The resulting Landscan dataset represents an ambient population which integrates diurnal movements and collective travel habits into a single measure. Below is how probability coefficients are based on various factors:

Roads, weighted by distance from cells to roads
Slope, weighted by favourability
Landcover, weighted by type, excluding certain types
Night Time lights of the World
The resulting coefficients are independent of census data. this generic models remains same for all regions but the probability weights for each variable are customized due to economic, physical, and cultural factors. To take an example it would be erroneous to give generic weight to nighttime lights of energy-poor North Korea. Similarly well lit roads in desertic but oil-rich countries like Kuwait might give false impression of habitation.
Coming back to Bhopal tragedy, if the data on ambient population were available and accessible to the emergency worker they would been in a position to make more informed decisions. Overlaying gas plume movement over the ambient population map would have been of immense helping saving many precious lives.

Below are some examples where Landscan has been actively used in disaster response:

www.disasterscharter.org

www.ornl.gov