A validation of satellite derived cyanobacteria detections with state reported events and recreation advisories across U.S. lakes

Abstract

Cyanobacteria harmful algal blooms (cyanoHABs) negatively affect ecological, human, and animal health. Traditional methods of validating satellite algorithms with data from water samples are often inhibited by the expense of quantifying cyanobacteria indicators in the field and the lack of public data. However, state recreation advisories and other recorded events of cyanoHAB occurrence reported by local authorities can serve as an independent and publicly available dataset for validation. State recreation advisories were defined as a period delimited by a start and end date where a warning was issued due to detections of cyanoHABs over a state's risk threshold. State reported events were defined as any event that was documented with a single date related to cyanoHABs. This study examined the presence-absence agreement between 160 state reported cyanoHAB advisories and 1,343 events and cyanobacteria biomass estimated by a satellite algorithm called the Cyanobacteria Index (CI_cyano). The true positive rate of agreement with state recreation advisories was 69% and 60% with state reported events. CI_cyano detected a reduction or absence in cyanobacteria after 76% of the recreation advisories ended. CI_cyano was used to quantify the magnitude, spatial extent, and temporal frequency of cyanoHABs; each of these three metrics were greater (r > 0.2) during state recreation advisories compared to non-advisory times with effect sizes ranging from small to large. This is the first study to quantitatively evaluate satellite algorithm performance for detecting cyanoHABs with state reported events and advisories and supports informed management decisions with satellite technologies that complement traditional field observations.

Keywords: Cyanobacteria; Human health; Recreation advisory; Satellite; Validation; Water quality.

Fig. 1.

A seven-day maximum composite of CI_cyano output derived from satellite imagery captured between July 5 and July 11, 2020. The brown pixels are land and black pixels are flagged by QA masks. The colored pixels represent CI_cyano estimates of cyanobacteria from low (purple) to high (red). Gray pixels are a valid observation of the lake surface with no detection of cyanobacteria. Land pixels were removed from the close-up of Falls Lake, North Carolina at the bottom of the Fig. to highlight the pixels that correspond to the resolvable lake surface. These were the pixels used to establish CI_cyano bloom presence and absence in this study.

Fig. 2.

A representation of the approach used to examine agreement between state reported events and state recreation advisories and CI_cyano bloom presence and absence.

Fig. 3.

A conceptual confusion matrix that outlines the four scenarios used to evaluate agreement between field monitoring and CI_cyano bloom presence and absence.

Fig. 4.

A graduated symbol map of the 1,343 state reported events coincident with CI_cyano. Each symbol represents a unique lake, and the size of the symbol represents the number of events that occurred in that lake. Circles and rectangles were solely used to help distinguish between the different number of events and have no bearing on any other attribute of the events.

Fig. 5.

A cumulative time series of CI_cyano bloom presence and absence within the 210 unique lakes with state reported events, calculated on a weekly basis during the MERIS and OLCI acquisition periods. The white areas in each panel represent the number of lakes with no satellite retrievals. These CI_cyano derived phenological patterns follow previously established ecological patterns of all satellite resolvable lakes across CONUS.

Fig. 6.

The four general observation categories reported with the records of state reported events along with the percentage of those records with a CI_cyano bloom presence or a CI_cyano bloom absence. The records of state reported events included in this plot are those that had a coincident satellite observation from MERIS or OLCI with sufficient data for evaluation (n = 1,125).

Fig. 7.

A confusion matrix with measures of agreement between state reported events and CI_cyano bloom presence and absence. A comparison of absence measures was not possible because the state reported event records only contained observations of cyanoHAB presence. There was no documentation for observations of state reported event absence. There were 218 matched state reported events with insufficient satellite data due to quality flagging that were discarded. The true positive rate of CI_cyano when compared to state reported events was 60%.

Fig. 8.

A graduated symbol map of 160 state recreation advisories coincident with CI_cyano. Each symbol represents a unique lake and the size represents the number of advisories that occurred in that lake. Circles and rectangles were solely used to help distinguish between the different number of events and have no bearing on any other attribute of the events.

Fig. 9.

The number of state recreation advisories by duration in days with a bar width of 20 days (a) and the number of state recreation advisories by the month that they started and ended (b).

Fig. 10.

A confusion matrix with measures of agreement between state recreation advisories and CI_cyano bloom presence and absence. Although not included in the Fig., the computed F-1 score was 0.73.

Fig. 11.

A demonstration of the potential spatial incongruency between recreation sites and satellite pixels using Falls Lake, North Carolina as an example. The colored pixels represent CI_cyano estimates of cyanobacteria from low (purple) to high (red). Grey pixels are a valid observation of the lake surface with no detection of cyanobacteria and black pixels were flagged by QA masks. The easternmost recreation site was not close to any satellite water pixels due to the narrow width of the system. While cyanobacteria were detected near one of southern recreation sites, the pixels are still at a distance from shore and may not capture the areas that can be easily sampled in the field. This is highlighted in the inset using a true color image from Sentinel-2 as reference.