## Beginning with a Water-Quality Failure: Why “More Equipment” Does Not Mean “A Smarter System” The incidents that most perplex aquaculture teams are often not those in which equipment alarms are triggered, but those in which **every device indicates normal operation while the fish nevertheless experience adverse effects**. Following feeding, the aerators continue operating, circulation pumps remain active, the automatic feeders report no faults, and every indicator on the monitoring platform is green. Yet several hours later, the fish begin to exhibit signs of stress and, in severe cases, mortality occurs. The problem may not lie with any individual piece of equipment. More commonly, each device has completed its assigned task, but the devices have failed to operate in a coordinated manner, ultimately resulting in **system-level mismatch**. Feeding simultaneously alters fish metabolism, oxygen consumption in the water, and the loading of nitrogenous waste. A study of channel catfish and blue catfish found that both oxygen consumption and ammonia excretion rates increased significantly within **2 hours** after feeding. Ammonia excretion in channel catfish peaked **6 hours** after feeding, and this temporal pattern was observed at both **25°C and 32°C**. Peak oxygen consumption rates reached **1.8 and 2.0 times** the corresponding fasting levels, respectively (Ott et al., 2025). Feeding, therefore, is never merely a production operation. It initiates a cascade of processes: feed intake, metabolism, oxygen consumption, ammonia excretion, and subsequent microbial transformation. Changes in feeding quantity, fish size, water temperature, or biomass will correspondingly alter the intensity and duration of this loading pulse. If the system continues to activate aeration according to a fixed schedule, or adjusts circulation pumps solely on the basis of a single Dissolved Oxygen (DO) probe, localized risks can readily be underestimated. In a study conducted at a commercial Atlantic salmon farm, real-time sensors were installed at the centers of **19 net pens** and at **4 reference locations**. During the same tidal phase, Dissolved Oxygen (DO) was **8.24 ± 0.29 mg/L** at one end of the farm but only **5.38 ± 0.34 mg/L** at the other, representing an average difference of approximately **2.86 mg/L** (Burke et al., 2021). A reading that appears to fall within the “normal range” may therefore conceal oxygen-depletion gradients near the pond bottom, return-water inlet, fish aggregation zones, or hydraulic dead zones. In field operations, I am particularly cautious of this form of “safety by average.” Fish do not inhabit average conditions. A localized decline in Dissolved Oxygen (DO) may not only affect cultured animals directly but also inhibit nitrification, allowing $NH_3$ and $NO_2^-$ to accumulate within a short period. This distinction is fundamental: possessing automated equipment and having system-level risk-control capability represent two different levels of technological maturity. The former addresses whether equipment has been activated and whether data have been uploaded. The latter must determine whether the biological load is moving beyond the safe operating range, how long this deviation is likely to persist, which system components will be affected as the risk propagates, and which intervention will entail the lowest overall cost. Technological development in aquaculture has progressed from manual pond inspections, experience-based feeding, and high rates of water exchange to single-point mechanization, connected equipment, and data visualization. The next step is not to add another dashboard, but to advance toward model-driven closed-loop control: identifying the current state, predicting near-term risks, coordinating multiple actuators, and subsequently verifying whether the intervention has been effective. The precision aquaculture framework summarizes this process as **Observe, Interpret, Decide, and Act**, with the objective of improving the accuracy, precision, and repeatability of operations while establishing more reliable decision support (Føre et al., 2018). New production models will not, of course, emerge in an orderly sequence and replace older models one by one. Open ponds, industrial flow-through systems, Recirculating Aquaculture Systems (RAS), biofloc systems, and integrated multi-trophic aquaculture are likely to coexist over the long term. These systems face different capital constraints, ecological boundaries, and operational capabilities; consequently, their required sensor densities, treatment processes, and levels of control precision will naturally differ. Open ponds depend more heavily on weather forecasting, zoned aeration, and sediment management, whereas high-density RAS place greater emphasis on continuous water-quality regulation, equipment redundancy, and fault-recovery mechanisms.
