There are three components of monitoring that need to be achieved — universality, identifiability and timeliness.
Consider universality first. We need to monitor all point sources, which would enable us to clearly differentiate between units that conform and those that do not. In a country such as India there are scores of units that are unregistered, and many that are registered may not be as polluting units, and yet others may be mis-categorised. In other words, we cannot depend upon government records to decide which units to monitor, which not to. The solution will need to be a technological one that can monitor a large expanse.
Next consider identifiability. Any identification exercise can suffer from errors of omission and commission. While we cannot afford errors of omission, errors of commission can be acceptable if there is a built-in method of confirming the initial identification of the polluter.
And finally, the time element. Simply installing pollutant-ameliorating equipment does not mean it is being used. Floor managers may be using the wrong practices, be lax, or the unit management may find it too expensive to operate, or the input materials may be contaminated — the list behind possible causes can be long. Constant monitoring therefore is an essential element of good monitoring.
However, there are hundreds of coal power plants, tens of thousands of brick kilns, and hundreds of thousands of construction sites. How can we possibly measure pollution levels at each of these point sources, do it in real time, capture divergence from the expected norm, and investigate further for corrective action?
Remote sensing technologies including those by satellites are a good solution to the problem. Satellites can monitor large expanses on a 24x7 basis. Whether it is industrial chimneys or coal power plants, all units that emit fumes can be imaged at different wavelengths. Algorithms can constantly analyse the data and red-flag potentially polluting entities for further investigation.
Not only can this method achieve universal coverage, it can also capture points such as unregistered units that are otherwise invisible to the government. It can red-flag potential polluters if they diverge any time of the day and night. Though these red flags may not constitute proof, it can constantly identify units for further investigation.
Consider such a watchdog mechanism that is constantly monitoring and putting the data in the public domain. When connected to off-the-shelf mapping software, it can make polluting points accessible not just to the government and regulator but also to the public. With public accessibility, independent researchers can build their own algorithms for evaluation of potential polluters. In other words, the risk of monitoring being “compromised” is also eliminated with public availability of pollution information.
The beauty of artificial intelligence is the low-cost scalability of such technologies. Economies of scale and scope are both very high. Depending on the bandwidths being captured, the same image can be used both to monitor different kinds of pollutants from varying sources. Take garbage dumps, for example. Many illegal garbage dumps exist, and these are not just repositories of the much-maligned plastics, but also emit significant levels of methane. Garbage dumps tend to be warmer than surrounding areas, and sometimes the temperature difference can be as much as 10 degrees centigrade or even higher, a temperature difference easily captured by remote cameras.
Riverbed sand mining leaves a large signature that can be captured by satellites. This is also true of stone mining. Illegal fishing is rampant, with limited ability to police “no entry” zones or fishing closures. Opening the data for individual experimenters in universities, research institutes or even amateurs can spread and accelerate the process of a cleaner environment, while accelerating skill formation in an emerging technology.
The writer is an environmental economist and heads Indicus Foundation