Final Report: Early Detection and Prediction of Harmful Algal Blooms Using Low Cost, Networked IOT Sensors and Machine Learning

EPA Contract Number: 68HERC20C0027
Title: Early Detection and Prediction of Harmful Algal Blooms Using Low Cost, Networked IOT Sensors and Machine Learning
Investigators: Lee, Christopher
Small Business: AquaRealTime, Inc.
EPA Contact: Richards, April
Phase: I
Project Period: March 1, 2020 through August 31, 2020
Project Amount: $100,000
RFA: Small Business Innovation Research (SBIR) - Phase I (2020) RFA Text |  Recipients Lists
Research Category: Small Business Innovation Research (SBIR) , SBIR - Clean and Safe Water


Every day, 2 million tons of municipal, industrial and agricultural waste are discharged into the world’s water. AquaRealTime’s mission is to improve the health of our reservoirs, lakes, rivers, and oceans globally, through real time detection of contamination.

Harmful Algal Blooms, stemming from increased water pollution and temperatures, cause $17B of damages annually. AquaRealTime provides an early detection system and decision making tool, enabling industrial, municipal, and recreational stakeholders to decrease their contamination control expenses, and protect the public from toxic water.

We leverage patent pending IoT sensor buoys and predictive analytics with a SaaS model to provide drinking water utilities, recreational areas, desalination plants,and private stakeholders with a 24/7 decision making tool, resulting in large cost savings and decreased downtime.

Summary/Accomplishments (Outputs/Outcomes):

  •  Pilots: AquaRealTime today has 11 AlgaeTrackers deployed in CA, CO, WI, IN, MA, and OH. We have orders for 6 more units as soon as we can manufacture them, and another 7 shortly after those. We have been receiving more requests beyond these but have not had the manufacturing capacity yet.
  •  Sales projections: from talking to dozens of stakeholders in the US, South America, and Australia, our soft commit pipeline for 2021 is about 300 units. We have remaining engineering and commercialization work to do to reach this capacity.

“The AlgaeTracker buoy has provided some great insight into pond management right from our desktop in an easy-to-use interface. We've been able to monitor pond conditions, place pre-emptive phone calls to clients, and treat before algae becomes a major issue. Turbidity measurements also allow us to predict and monitor effects of runoff and make more timely management decisions. This is a great product that should have far reaching implementation possibilities for private pond owners, public beaches, and municipalities. We can't wait to get more units in the water.” Jeff Stelzer from Lake and Pond Solutions.

AquaTechNex in Long Beach, CA

AquaTechNex in Long Beach, CA

Technical progress:

  • We were able to complete the first manufacturable version of the optical head PCB that measures phycocyanin, chlorophyll-a, and turbidity. Specifically we were able to address the challenges around ambient light rejection and extracting the tiny fluorescence signal from the large amount of noise and disturbances. We ordered the first 12 articles of this PCB. (Due to costs and small modifications, it does not make sense to order more first articles of PCBs).
  • This optical head and the related firmware signal processing has reached very high accuracy levels with linearity values of R^2 = 0.999 for both chlorophyll-a and phycocyanin in lab tests. The measurements are impervious to solar light.
  • We completed 3 revisions of the communications PCB which has enabled consistent communications from the buoys to the cloud dashboard. We ordered 12 articles of this PCB.
  • We updated the buoy CAD design to improve the brushing system. We more recently have begun a needed improvement to the brushing system to reduce the impact of floating debris and plant life on the brushing.
  • We finalized the molds required for several components of the Tracker and had 25x first articles ordered and built. All machined mechanical and optical parts were ordered and built in qty 25, along with all other parts required to build 25 AlgaeeTrackers, except for the above listed PCBs.
  •  We built 10 complete AlgaeTrackers and deployed them to customer sites. We also built one ‘LabTracker’ that is a lab size version of the Tracker, and deployed it to our academic partner, Miami University.
  • We spent hundreds of hours testing in the lab, at our Lagerman reservoir test site, at Coot Lake in Boulder, and at customer sites. We have many sheets of data stored in our Google Drive folders detailing results from voltages on sensors in single lab experiments up to long term, high resolution tracking of whole buoy performance at customer sites.
  •  Team members Kara Wolley and Christopher Lee developed firmware for power, solar, communications, and external sensor components, as well as the low level microcontroller firmware for optical head measures of turbidity, phycocyanin, and chlorophyll-a.
  • Yuanming Chen and Kara Wolley have made large improvements to the software based on customer inputs. However we are still early stage here and the potential for service improvement is huge, as we plan to add more and more predictive ability to our current early detection ability
  • Many iterations on calibrations were coded and tested. The calibrations are looking good for chlorophyll-a and phycocyanin, though conversion of the calibrated Relative Fluorescence Units into biological densities is an active area of development now. Turbidity calibration has vastly improved and we added the capability to calibrate the device on site from handheld devices.
  •  We integrated Adaptive Battery Usage that throttles measurement frequency based on state of charge. Ongoing work here will address situations where charge can be lost prematurely.


We will be applying for phase 2 in the next 6 weeks for the October 30 due date. I
won’t go into the details but here are examples of some of the tacks we will pursue:

  •  Technical progress:
  •  We will make several improvements to the brushing system, calibrations, and battery usage.
  •  We will be branching into the predictive analytics aspect of cyanobacteria and HAB monitoring: using weather data, satellite data, watershed nutrient load data, all in an automated package with the goal of pushing out the HAB warning timeline as much as possible. This will likely involve partnering with other HAB modeling entities e.g. universities or agencies.
  • We will perform a small pilot with an auxiliary Dissolved Oxygen sensor which is key to expanding to various other segments like aquaculture.
  •  Commercialization:
  • We will be expanding our business development team and addressing the needs of our core market segments: drinking water utilities, recreational and wildlife areas, and then desalination and industrial plants.
  •  We will provide various service and feature levels at different price points, though using all the same hardware.
  • We will take our contract manufacturing to the next level, where we can satisfy the demand for ~500 units, then 1000 and 5000 units per year. In tandem, we will improve our support and maintenance strategy.

Two Trackers in forest fire season at Lagerman Reservoir, CO

Two Trackers in forest fire season at Lagerman Reservoir, CO

OH1 Conc Rho mg/L Time R G B TU PC CA microT counter Add ml PC ug/L CA ug/L  
Rho 0 12:00 PM 3259.00 123.00 149.00       26.80 2.00 0.00 0.00 0.00 CALIBRATION
3-Sep 0.09236   3378.00 1063.00 315.00       27.00 4.00 6.00 3.55 15.01 CALIBRATION
  0.18204   3194.00 1827.00 474.00       27.10 5.00 12.00 6.99 29.59 CALIBRATION
  0.00000   3396.00 128.00 145.00 nan 0.00 0.00 27.10 1.00 0.00 0.00 0.00  
  0.01559   3349.00 293.00 173.00 nan 6.23 25.32 27.00 2.00 1.00 0.60 2.53  
  0.03109   3320.00 464.00 203.00 nan 12.69 52.44 27.00 3.00 2.00 1.19 5.05  
  0.04653   3388.00 627.00 231.00 nan 18.85 77.76 27.30 4.00 3.00 1.79 7.56  
  0.06188   3237.00 772.00 261.00 nan 24.32 104.89 27.20 5.00 4.00 2.38 10.06  
  0.07716   3369.00 925.00 289.00 nan 30.10 130.21 27.30 6.00 5.00 2.96 12.54  
  0.09236   3354.00 1071.00 321.00 nan 35.64 159.27 27.50 7.00 6.00 3.55 15.01  
  0.10749   3230.00 1200.00 350.00 nan 41.44 185.86 27.50 8.00 7.00 4.13 17.47  
  0.12255   3399.00 1322.00 373.00 nan 46.94 206.95 27.50 9.00 8.00 4.71 19.92  
  0.13753   3252.00 1437.00 401.00 nan 52.11 232.63 27.50 10.00 9.00 5.28 22.36  
  0.15244   3337.00 1567.00 424.00 nan 57.97 253.72 27.50 11.00 10.00 5.85 24.78  
  0.16727   3221.00 1694.00 458.00 nan 63.69 284.90 27.30 12.00 11.00 6.42 27.19  
  0.22590   3249.00 2181.00 562.00 nan 85.61 380.26 27.10 13.00 15.00 8.67 36.72  

Serial dilution results

Serial dilution results


SBIR Phase II:

Early detection and prediction of harmful algal blooms using low cost, networked IOT sensors and machine learning