Globally, pond aquaculture is the most common method of cultivating finfish for human consumption. Growing populations and increased seafood demand has led to the intensification of pond aquaculture to the point where these systems are on the very limits of their carrying capacity. With this comes a heightened risk of a pond oxygen crash caused by the destratification of the water column or an algae die-off. These crashes are a leading cause of fish kills on pond farms due to the decline in dissolved oxygen (DO) that results. Current methods used on farms to monitor DO levels and prevent crashes are labor intensive, expensive, and reactionary rather than preventative. Being able to better monitor and predict DO levels would be hugely beneficial to the industry and its development.
The Hybrid Aerial/Underwater robotiCs System (HAUCS) is an Internet of Things development for fish farms with the aim of improving DO monitoring efficiency and effectiveness by using autonomous sampling platforms to collect data to support a machine learning based DO prediction algorithm. This study focuses on the relevant environmental data that can be used to develop a robust prediction algorithm. The use of digital imagery as a proxy to algal bloom concentration data was also investigated. Data was collected June-September 2021 using ponds at Aqua Blue Cichlids in Fellsmere, FL. DO and water temperature were monitored 24/7 in each pond. A weather station on farm was used to collect air temperature, atmospheric pressure, wind speed/direction, solar irradiance, and rainfall. Images of the water’s surface were taken in the mornings and afternoons using a digital camera. The color information from the images was used in correlation comparisons with the weather variables, algal concentration, and DO.
Significant correlations (P < 0.05) were seen between DO minimum and water temperature, algal concentration, sunlight irradiance, and rainfall (Table 1). For the image data, significant correlations were seen between DO minimum and some color components (e.g., RGB). Additionally, several correlations were seen between some image components and weather variables such as solar irradiance, windspeed, rainfall. Some correlations were seen between one image component and algal concentrations.