Anthropogenic Specular Interference in the Operational GOES-R Fire Product

Dataset for the research paper "Anthropogenic Specular Interference in the Operational GOES-R Fire Product".

Large reflective structures like solar power plants and commercial greenhouses sometimes reflect sunlight directly into GOES-R sensors. These anthropogenic specular reflections, or "sparkles", cause commission errors in operational GOES-R ABI products like the Fire Detection and Characterization Algorithm (FDCA). Using the abi-sparkle library for Python (Dove-Robinson, 2023), we generated a dataset containing both detected anthropogenic specular reflection pixels and the coincident FDCA commission errors caused by them for the GOES-16 CONUS domain during the 2020 calendar year.

The dataset consists of two exported PostgreSQL tables: sparkle_pixels_g16_abi_conus_2020, which contains the detected anthropogenic specular reflection pixels at 500 m resolution, and fdca_commission_error_clusters_g16_abi_conus_2020, which contains clustered FDCA false alarm fire pixels caused by anthropogenic specular reflection at 2 km resolution. The FDCA pixels were only processed for fire mask codes 10-15 and 30-35; see Table 3.11 in Schmidt et al., 2013 for fire code definitions.

Each row in sparkle_pixels_g16_abi_conus_2020 is a detected specular reflection pixel in a GOES-16 CONUS image from the 2020 calendar year with a unique numeric ID sparkle_id and associated metadata from the detection algorithm abi-sparkle. The column sparkle_geom is a PostGIS geometry ST_Point object that can be used to plot the pixels on a map.

FDCA fire pixels at 2 km resolution were clustered based on their connectivity in a 3x3 pixel kernel and assigned a UUID fire_cluster_id in the table fdca_commission_error_clusters_g16_abi_conus_2020. Only the fire clusters that overlapped with sparkle pixels in time and space were retained in the table. In this way, each row of fdca_commission_error_clusters_g16_abi_conus_2020 is a unique cluster of errant FDCA fire pixels caused by anthropogenic specular reflection in every available scan start time for the GOES-16 CONUS domain in 2020. Every fire cluster centroid has a PostGIS geometry object fire_cluster_centroid_geom that can be used to plot the errant fire pixel clusters on a map.

The two tables relate with the column sparkle_ids in fdca_commission_error_clusters_g16_abi_conus_2020, which is an array of overlapping sparkle IDs from the sparkle_pixels_g16_abi_conus_2020 table. The combined dataset may therefore be generated with a simple SQL INNER JOIN:

SELECT * FROM fdca_commission_error_clusters_g16_abi_conus_2020 fcecgac
INNER JOIN sparkle_pixels_g16_abi_conus_2020 spgac ON spgac.sparkle_id = ANY(fcecgac.sparkle_ids);

The tables can be imported into a PostgreSQL database version 12 or newer with PostGIS extensions installed. For example, to import the tables into a database in a Linux environment, run the following commands:

gunzip -c sparkle_pixels_g16_abi_conus_2020.sql.gz | psql -d your_database_name
gunzip -c fdca_commission_error_clusters_g16_abi_conus_2020.sql.gz | psql -d your_database_name