Cyanobacterial Harmful Algal Blooms Forecasting Research
Learn More About CyAN
The cyanobacterial harmful algal bloom forecasts posted on this webpage will be paused until Spring 2025, as the probabilities of harmful algal blooms during colder months are generally lower. Weekly U.S. forecasts will resume on EPA’s HABs webpage.
Most species of algae are not harmful, but when certain types grow rapidly, they can cause harm to human and pet health, aquatic ecosystems, and local economies. Referred to as harmful algal blooms (HABs), this excessive growth is usually associated with algae that can produce toxins.
Even though they are classified as bacteria, cyanobacteria exhibit characteristics of algae and sometimes referred to as blue-green algae. Cyanobacterial HABs (cyanoHABs), can appear in water bodies across the country, are an indicator of poor water quality, and can potentially cause serious environmental concerns, including human and aquatic health effects. When harmful blooms occur in recreational waters or source waters used for drinking, the toxins released can cause respiratory or skin irritation and even illness in humans, domestic animals, and wildlife.
EPA researchers are working on ways to forecast when and where cyanoHABs may form. The data on this page are from a beta test of an experimental cyanoHAB forecasting model.
Interpreting the Data
In accordance with the goals of the Cyanobacteria Assessment Network (CyAN) project, the forecasts described here are from a beta test of an experimental cyanoHAB forecasting model for 2024. This forecasting model was trained and validated with the CyAN satellite data, but may contain errors and will be continuously improved over time. The focus of this forecast model is to provide weekly probabilities of cyanoHAB occurrences in 2,192 lakes in the contiguous United States that are resolvable by the Sentinel-3 satellite. For each lake, the provided probability is that the lake exceeds the World Health Organization (WHO) recreation Alert Level 1 threshold (≥12 micrograms per liter chlorophyll-a with cyanobacteria dominance). The weekly bloom probability forecasts are presented in the forecast table, produced following the workflow described in detail in Schaeffer et al. 2024.
What does a “weekly" forecast mean exactly?
The forecast model is structured to generate weekly cyanoHABs probabilities that are valid for a seven-day period extending Sunday through Saturday. For example, if forecasts are generated on a Sunday (5/5/2024), they will be valid (i.e., representing the probability of a bloom occurring) through Saturday (5/11/2024). However, if forecasts are generated a day later, on Monday (5/6/2024), they will still be valid only through Saturday (5/11/2024). Currently, beta test forecasts require about a day to generate, so it is likely they will be made available on Tuesday or Wednesday each week and will only be valid for the remainder of the current week through Saturday.
How do we interpret the bloom probability and what part of the lake does the probability apply to?
The weekly forecasted bloom probability for each lake is the probability that the median lake chlorophyll-a in the surface (typically upper two meters or less) is ≥12 micrograms per liter (ug/L) across the lake area represented by resolvable/unmasked satellite pixels. The Sentinel-3 satellite imagery is divided into a grid of 300 × 300 meter pixels, each of which has a chlorophyll-a (with cyanobacteria dominance) concentration* associated with it.
During preprocessing of the satellite imagery, pixels that included cloud, cloud shadow, glint, straylight, snow, ice, or land cover were removed. From the retained pixels, a median chlorophyll-a concentration is calculated for each lake and is compared to the WHO recreation Alert Level 1 threshold of 12 ug/L. Thus, when applying the forecast model, the weekly forecasted bloom probability for each lake is the probability that the median lake chlorophyll-a is ≥12 ug/L at the surface across the lake area represented by resolvable/unmasked pixels. The higher the probability, the higher the likelihood of a cyanoHAB.
No forecast is perfect, and most forecast models will result in false positives and negatives. This model should not replace regular sampling or observation methods. Currently, the model overpredicts positive events, where additional complementary field sampling or additional remote sensing may be useful. The model has low false omission and is therefore more conservative in protecting health.
*The chlorophyll-a concentration for each pixel is calculated by first converting a digital number (DN) to cyanobacteria index value (CIcyano) and then to a chlorophyll-a value as detailed by Seegers et al. 2021 and NASA CyAN data website using these two equations:
- CIcyano = 10(DN × 0.011714 − 4.1870866)
- chlorophyll-a = 6620 × CIcyano − 3.07
Why is my lake not included in this dataset?
Currently, the model can only generate cyanoHAB forecasts for the 2,192 lakes that are satellite- resolvable. Whether or not a lake is satellite-resolvable depends on the spatial resolution of the satellite imagery of the lake. The spatial resolution of the Sentinel-3 satellite images used to train the forecast model is 300 × 300 meters, so the forecast model can only make forecasts for larger lakes. The feasibility of using higher resolution satellite imagery to forecast for a broader set of lakes is being evaluated.
Understanding the Table
| Date | State | Lake Name | % Chance of CyanoHAB | Latitude of Centroid | Longitude of Centroid | COMID |
|---|---|---|---|---|---|---|
| The period of time that the forecast is being made for (in a short-date format). | State names in abbreviated format. |
The lake name associated with the CyAN lakes shapefile on EnviroAtlas. Please note: there are some inconsistencies in lake names and we are working to resolve these errors. |
Probability of a bloom derived from EPA's forecast model. For more information on how to use this percentage and its limitations, please see the interpretation section. |
Latitude (in decimal degrees) of the lake centroid (an estimated center point of the resolvable lake). | Longitude (in decimal degrees) of the lake centroid (an estimated center point of the resolvable lake). | A unique set of numbers, letters, and or symbols to identify each lake in EPA's model. |
Forecast Data
Weekly forecasts will resume in the Spring 2025 on EPA’s HABs webpage.