Crop Residue Burning (CRB) is a major contributor to air pollution in the Indo-Gangetic Plain (IGP) region during winter months, particularly in October and November. Satellite-based active fire counts and burnt area are two widely used remote sensing metrics to monitor CRB. However, these methods have some limitations. Coarser spatial and temporal resolutions, as well as obstructions due to smoke and clouds, result in data loss. Moreover, these methods do not differentiate between partial burning and complete burning incidents. These limitations lead to an underestimation of CRB. An accurate estimation of CRB helps evaluate the effectiveness of its mitigation policies. Moreover, it helps in the creation of accurate emission estimates that are fed into air quality models such as Delhi’s Air Quality Early Warning System (AQEWS). Therefore, it is important to assess the effectiveness of the remote sensing methods in detecting CRB incidents and their ability to differentiate between partially and completely burnt fields.
The study analysed the effectiveness of two remote sensing-based methods – active fire counts and burnt area – in monitoring CRB. It analysed the effectiveness of satellite-borne sensors such as the VIIRS and MODIS in detecting CRB incidents. A burnt area algorithm combining two burnt area indices – Burnt Area Index for Sentinel-2 and Tasseled Cap Brightness Index (TBI) was developed to assess its ability to detect burnt fields. The study also explored the ability of this burnt area algorithm to differentiate between partially burnt and completely burnt fields.
Key Highlights
- Satellite-borne sensors are unable to detect all farm fires. This analysis of the very high-resolution satellite images shows that 169 fields were set on fire on two consecutive days over an area of 74 sq km in Sangrur in 2020. Authors could only find one instance where very high-resolution satellite imagery was available for two consecutive days on Google Earth. VIIRS detected only seven fires over this region, and MODIS detected none on either of the two days.
- A burnt area algorithm that uses a combination of BAIS2 and TBI detected burnt pixels with 75 per cent accuracy.
- Differentiating partially and completely burnt fields is challenging. The algorithm developed to separate partially and completely burnt fields could only achieve an accuracy of 60 per cent, with a high false alarm ratio of 40 per cent.
- Central and state government agencies should consider using very-high-resolution satellite images to monitor CRB.
- However, procuring such imagery from commercial providers is expensive. Therefore, India should also move towards having its own satellites with performance comparable to that of Sentinel that complement the existing constellation. Also, air quality models should not rely only on fire count-based emission inventories, but should use burnt area-based emission inventories.
