Research

Recent Updates

February 2026: New Publication & Dataset Release:

Our latest manuscript detailing the expansion of CyanNet is now in press. Key highlights include:

The successful zero-shot extension of CyanNet to the Great Lakes Region and
The estimation of CIcyano using MODIS data (2012–2015).
We have also generated a comprehensive 25-year continuous time series (2000–2024) for the U.S. Great Lakes.
More research highlights from this paper is here.
Paper with dataset in Environmental Research: Ecology, https://doi.org/10.1088/2752-664X/ae4631.

Current Research

CyanNet Skill Assessment
Hyperspectral remote semsing of HABs

Project Gallery

CYANNET: Bridging the Satellite Data Gap

CyanNet is an advanced deep learning framework that leverages MODIS-Terra satellite imagery to estimate the Cyanobacteria Index (CIcyano) in freshwater systems. This framework provides CIcyano estimates that are consistent with measurements from the MERIS and OLCI sensors. By utilizing 12 spectral bands, CyanNet effectively fills the temporal gap between the MERIS (2002–2012) and OLCI (2016–present) data. A brief but relevant discussion on cyanNet is also available in the Publications page under AI-ML Applications for CyanoHAB Detection and Multi-sensor Data Fusion section.

Core Function: CyanNet generates a consistent CIcyano from the Rayleigh-corrected surface reflectance obtained from MODIS-Terra. This allows for a long-term, continuous record (from 2000 to the present) of cyanobacterial harmful algal blooms (cyanoHAB) in freshwater lakes and water bodies.
Proven Accuracy: The framework demonstrates strong performance, featuring approximately a 17% median difference in bloom magnitude when validated in locations such as Lake Okeechobee, Lake Apopka, and Lake George.
High Scalability: CyanNet successfully generalizes across various locations, including Lake Erie and Lake Apopka, without the need for site-specific retraining.
Journal Article in Remote Sensing: Mishra et al. (2023)

Evaluation of cyanobacteria index (CI) bloom detection algorithm using field-measured microcystin data

Visual Summary

Fig. 1. Schematic Diagram showing the data processing workflow and algorithm evaluation strategy

Cyanobacteria Index (CI) is widely used to detect and quantify cyanobacteria density. However, only a few studies evaluated the algorithm for its detection accuracy. In this study, we validated CI algorithm (Fig. 1) for its effectiveness in identifying lakes with toxin-producing blooms using MERIS and OLCI data in 11 states across the United States over 11 years (2005–2011, 2016–2019) (Fig. 2).

Fig. 2. The location map of in situ MC samples (n=281) from the lakes located in 11 states across the United States. Bubble size is proportional to the microcystin sample count in each lake.

We used in situ Microcystins (MCs) data as a proxy of the presence of CyanoHABs in water bodies. Although we are aware that the satellite sensors cannot detect toxins, we used MCs as the indicator of cyanobacterial health risk and as a confirmation of cyanoHAB presence in a lake. Another reason for using MCs is that they are the most common laboratory measurement researchers or water quality managers make during CyanoHABs.

How accurate is the CI algorithm?

We evaluated algorithm performance by its ability to detect a CyanoHAB 'Presence' or 'Absence.' We found the overall accuracy of CyanoHAB detection to be 84% with same-day matchups. Precision - the percentage of algorithm-detected bloom events that are 'true' bloom events of the bloom detection was 87%, and recall - the portion of 'true' bloom events detected by the algorithm was 90%. Based on a bootstrapping simulation, the overall accuracy of the algorithm can vary between 77% and 87% (95% confidence) (Fig. 3).

Fig. 3. Distribution of overall accuracy calculated using boot-strapping method from10,000 runs with a random draw of 98 samples. PDF is the fitted probability density function

These findings demonstrate that the CI algorithm has utility for routine monitoring of toxic cyanobacteria blooms in lakes across the United States on a synoptic scale.

Citation

Mishra, S., Stumpf, R.P., Schaeffer, B.A., Schaeffer, B., Werdell, J., Loftin and Meredith, A. (2021). Evaluation of satellite-based cyanobacteria blooms detection algorithm using field-measured Microcystin data. Science of The Total Environment, p.145462.

Page updated

Google Sites

Report abuse