Improving VIIRS Fire Radiative Power Estimates for Smoke Emission Modeling Applications
Abstract
Active Fire products from moderate 750m and imagery 375m resolution bands of the Visible Infrared Imaging Spectroradiometer (VIIRS) on the Suomi National Polar-orbiting Partnership (NPP) satellite and the NOAA-20 satellite are implemented in NOAA's operational production system and near-real-time pre-operational environment. One of the key downstream applications of the product is the use of VIIRS fire radiative power (FRP) in the smoke emission monitoring and modeling applications, such as the NOAA/ESRL system based on High Resolution Rapid Refresh model with smoke component (RAP/HRRR-Smoke). In the RAP/HRRR-Smoke model, the FRP data are used in biomass burning emissions calculation and in plume rise parameterization, and accurate FRP estimation in VIIRS active fire product is important for the smoke forecast.
In the VIIRS active fire algorithm, the FRP is retrieved from moderate resolution (750m) 4 µm channel (M13) measurements by relating difference between fire pixel and background brightness temperatures to emitted radiant fire energy. Ideally, each detected fire pixel would have non-zero radiative power and contribute 100% of its FRP to the total FRP over the specified pixel footprint area. However, there are several scenarios where FRP retrieval could be inaccurate in the current algorithm, or portion of individual pixel contribution to total FRP could not be easily determined. The first scenario corresponds to the situation where there are not enough valid surrounding pixels to determine background brightness temperature; it may occur either over complex land/water surfaces with significant amount of clouds, or within large multi-pixel fires. Without valid background brightness temperatures the algorithm sets FRP to zero. The 375m algorithm uses co-located 750m pixels for FRP retrievals, and if the co-located pixel is unavailable for any reason FRP is set to zero as well. All this could result in a situation where fire is detected in a pixel with high confidence but the pixel will not contribute to the total FRP for specified area. The second scenario corresponds to the situation when the fire is detected close to or within on-ground bowtie deletion zones. The VIIRS scans overlap rather significantly, especially towards the edges of the swath. Sometimes a fire is detected in one scan and not detected in the overlapped neighboring scan. In this case it is important not to exclude the pixel from total FRP despite its location in the deletion zone. In other cases however, the same fire is reported twice with only slight shift in location. Total FRP estimation could be challenging in such a case; ideally it should take into account overlapping area between appropriate fire pixels. While these scenarios considered to be relatively rare they may impact the total FRP for specific fire or over certain area which in turn could modify initial conditions for the RAP/HRRR-Smoke or similar atmospheric models and result in inaccurate smoke and air quality forecast. The study investigates frequency of such scenarios for both 750m and 375m products, and suggests possible solutions to improve FRP estimation as an input to atmospheric models.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFMIN33F0914T
- Keywords:
-
- 0394 Instruments and techniques;
- ATMOSPHERIC COMPOSITION AND STRUCTUREDE: 3360 Remote sensing;
- ATMOSPHERIC PROCESSESDE: 1640 Remote sensing;
- GLOBAL CHANGEDE: 4275 Remote sensing and electromagnetic processes;
- OCEANOGRAPHY: GENERAL