Multi-Sensor Harmful Algal Bloom Severity and Speciation Dataset

Study Areas Southern California- Certain species of Pseudo-nitzchia form HABs in Southern California. This marine diatom produces domoic acid, a neurotoxin known to cause neurological impacts on marine wildlife. Gulf of Mexico- Karenia is a marine dinoflagellate genus and a subset of these species produce brevetoxins, linked with neurotoxic shellfish poisoning. Time Periods Data from JPSS1 VIIRS, SNPP VIIRS, AQUA MODIS, Sentinel-3A, and Sentinel-3B is included in the time period of 06/01/2018 - 12/31/2019. For the tests using PACE, the time range of 2024/03/05 - 2025/03/31 was used. Direct use of reflectance was chosen here for two reasons: 1) likelihood of missed latent patterns, especially in complex waters, as mentioned above, when only using downstream ocean color parameters to proxy phytoplankton presence, and 2) inconsistent availability of various parameters across all of the instruments used. The Sentinel-5P TROPOMI-based red SIF (TROPOSIF) products have been generated based on the retrieval approach from Köhler et al., (2020), and the data is hosted on ftp://fluo.gps.caltech.edu/ or data.caltech.edu. Köhler et al., (2020) implemented a variant of an established far-red SIF retrieval scheme (Guanter et al., 2012, 2015; Joiner et al., 2013; Köhler et al., 2015, 2018) to estimate red SIF from TROPOMI measurements for aquatic science. The TROPOSIF data generated for the 2018-2019 time period has only been produced in a daily ungridded format, so this data was taken and gridded at its native 7km resolution. Köhler et al., (2020) and Luis et al., (2023) highlighted the potential that red SIF has for improving our understanding of global phytoplankton photosynthesis and HABs. Specifically, Luis et al., (2023) found that red SIF provided more than twice the data than nFLH, and thus provided a new monitoring capability to obtain critical information on HABs. The TROPOSIF data is not available for the 2024-2025 time period, so it was not used in conjunction with the PACE data here. The areas used for test cases were the Gulf of Mexico and coastal Southern California. Reflectance data was pulled from the Ocean Biology Distributed Active Archive Center (OBDAAC). The latest versions available of the 4km daily Level-3 Mapped (gridded) reflectance data were used (Version 2 for VIIR, Version 2022 for MODIS and Sentinel-3, and Version 3 for PACE). Geometric and radiometric calibration has been done by the science data processing pipelines of the various missions, and would not need to be done by anyone else looking to curate the same data. Further information for each instrument can be found in each of the publicly available Level-1 algorithm theoretical basis documents (ATBDs). Because the reflectance data can be pulled directly from the publicly accessible OBDAAC and the TROPOSIF from the CalTech-based FTP site, we only host the output severity and speciation maps here. The data is split into two subdirectories, one for Southern California and one for the Gulf of Mexico. Under each of these subdirectories, there is a directory named for each of the instruments used in this study (AQUA_MODIS, JPSS1_VIIRS, SNPP_VIIRS, S3A, S3B).Each individual file for daily products contains 'DAY' in the filename, the date in the format YYYYMMDD, and the type of severity map (karenia_brevis, total_phytoplankton, pseudo_nitzschia_delicatissima, pseudo_nitzschia_seriata, or alexandrium. Each file is either contains the actual discretized severity map, or the Data Quality Indicator (DQI). For the DQI files, 'DQI' is also appended to the filename. The montly files contain the 'Monthly' indicator string in their filename, along with the date in the format YYYYMM, the severity map string (described above), and 'DQI' if it is a DQI file.All data have been put in GeoTiff format. The GeoTiff data format natively contains geolocation metadata internally, and can be interfaced with via C/C++/Python GDAL packages, or other python packages that wrap GDAL, like rasterio and rioxarray . The documentation for SIT-FUSE , the package with which the labels were generated, also has examples on how to read and interface with various data formats, including GeoTiffs. Lastly, this data can be interfaced with using Geographic Information Systems (GIS), like the free and open-source QGIS.Model weights for this study can be found here - HuggingFace