Mapping Brick Kilns to Support Environmental Impact Studies around Delhi Using Sentinel-2
| Autor: | Ram Avtar, Prakhar Misra, Sachiko Hayashida, Wataru Takeuchi, Ardhi Adhary Arbain, Ryoichi Imasu |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: |
010504 meteorology & atmospheric sciences
Kiln soil-degradation Geography Planning and Development Air pollution lcsh:G1-922 SDG 010501 environmental sciences transfer learning medicine.disease_cause 01 natural sciences Training (civil) Normalized Difference Vegetation Index Earth and Planetary Sciences (miscellaneous) medicine land-use Computers in Earth Sciences Emission inventory 0105 earth and related environmental sciences environment impact Land use emission inventory Vegetation open-data urban growth Remote sensing (archaeology) Environmental science Physical geography lcsh:Geography (General) object-detection |
| Zdroj: | ISPRS International Journal of Geo-Information Volume 9 Issue 9 ISPRS International Journal of Geo-Information, Vol 9, Iss 544, p 544 (2020) |
| ISSN: | 2220-9964 |
| DOI: | 10.3390/ijgi9090544 |
| Popis: | Cities lying in the Indo-Gangetic plains of South Asia have the world&rsquo s worst anthropogenic air pollution, which is often attributed to urban growth. Brick kilns, facilities for producing fired clay-bricks for construction are often found at peri-urban region of South Asian cities. Although brick kilns are significant air pollutant emitters, their contribution in under-represented in air pollution emission inventories due to unavailability of their distribution. This research overcomes this gap by proposing publicly available remote sensing dataset based approach for mapping brick-kiln locations using object detection and pixel classification. As brick kiln locations are not permanent, an open-dataset based methodology is advantageous for periodically updating their locations. Brick kilns similar to Bull Trench Kilns were identified using the Sentinel-2 imagery around the state of Delhi in India. The unique geometric and spectral features of brick kilns distinguish them from other classes such as built-up, vegetation and fallow-land even in coarse resolution imagery. For object detection, transfer learning was used to overcome the requirement of huge training datasets, while for pixel-classification random forest algorithm was used. The method achieved a recall of 0.72, precision of 0.99 and F1 score of 0.83. Overall 1564 kilns were detected, which are substantially higher than what was reported in an earlier study over the same region. We find that brick kilns are located outside urban areas in proximity to outwardly expanding built-up areas and tall built structures. Duration of brick kiln operation was also estimated by analyzing the time-series of normalized difference vegetation index (NDVI) over the brick kiln locations. The brick kiln locations can be further used for updating land-use emission inventories to assess particulate matter and black carbon emissions. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|