Publications

Kato, S; Kouyama, T; Nakamura, R; Matsunaga, T; Fukuhara, T (2018). Simultaneous retrieval of temperature and area according to sub-pixel hotspots from nighttime Landsat 8 OLI data. REMOTE SENSING OF ENVIRONMENT, 204, 276-286.

Abstract
The monitoring of thermal anomalies such as from wildfires, coal fires, and volcanoes has been a major application of satellite remote sensing. The Landsat 8 Operational Land Imager (OLI) has achieved sufficient enhanced radiometric performance to quantify visible and near infrared (VNIR) to shortwave infrared (SWIR) radiance emitted from thermal anomalies during nighttime. We propose a method to simultaneously retrieve the brightness temperature and corresponding area of a subpixel hotspot assuming a single isothermal heat source using the spectral radiances in VNIR and SWIR regions (bands 3-7) from nighttime Landsat 8 01,1 data. We conducted a prescribed fire experiment coincident with a Landsat 8 overpass on January 24, 2015. The brightness temperature retrieved from the OLI data agreed with the peak temperature of the time series thermocouple record within 60 K. The retrieved area of fire was 19% of the true fire area. The blackbody assumption (emissivity = 1) might cause the underestimated retrieved area. We applied the hotspot brightness temperature and area retrieval to 17 Landsat 8 scenes that cover Kanto, Japan to validate the proposed method. More than 79% of the detected results were confirmed as true positives based on the land use for each scene. Meanwhile, the obvious false positive was only up to 1% of the detected results. The remaining ambiguous results, mostly found in areas regarding agriculture, were possibly caused by occasional burnings because these locations were never repetitively detected. Time series comparison of the retrieved results in cement plants as a typical stable heat source revealed that the brightness temperature is accurately retrieved, though the retrieval of area is somewhat sensitive to noise on OLI imagery. The proposed method had founded more than twice of the hotspots detected using Visible Infrared Imaging Radiometer Suite (VIIRS) fire products in the studied area. More than 82% among the hotspots from VIIRS were detected by OLI. However our method failed to detect spatially broad but slightly cool hotspots. Higher spatial resolution at an appropriate spectral configuration enabled enhanced sensitivity to detect small heat sources, as well as quantification of the subpixel hotspots.

DOI:
10.1016/j.rse.2017.10.025

ISSN:
0034-4257