Aquaculture is a broad term that incorporates many different fish species across multiple geographical regions. Within GAIN, we considered precision aquaculture implementations for, amongst others: salmon cage farms, seabass and seabream cages, land-based trout and carp farms, mussel and oyster cultivation sites, and shrimp pond systems. These farms are distributed across three continents giving huge variations in climate, environmental conditions, regulatory requirements, and maturity of industry and technological sophistication.

Precision aquaculture is founded on sensors and real-time data collected at farms to inform on current (and historic conditions). A huge variety of sensors exist which introduces complexity for standardisation and interoperability. Sensors can be categorised in several different manners:

•  locally connected sensors that return data to the farm operators via data loggers (manually) or an on-site computer (e.g. on the aquaculture barge)
•  online sensors communicating data sensor vendor’s proprietary cloud (the cloud system typically being provisioned by the sensor provider and designed to ingest those data).

Examples of the former include traditional data logger sensors where data is manually downloaded periodically while the latter include novel wireless sensors that communicate measurements to the cloud in real-time such as those provided by Realtime Aquaculture (https://rtaqua.com).

Precision aquaculture requires a system to ingest, parse, contextualise, and curate these different sensor datasets to allow data to be processed in an automated and standardised manner. Generally, data ingestion required a two-pronged approach whereby some sensor data is pushed to the IoT data from the sensor or farm level (requiring interaction at the sensor deployment stage or the farm management level), and some data is polled from external cloud storage services (via an Application Programming Interface (API) that exposes the sensor data).

Data collected at the farm are processed from standard data formats (CSV, Excel, ASCII) and communicated to a unified cloud (or on-premise) computer. Communication is generally via lightweight message-passing protocols such as MQTT (Message Queue Telemetry Transport) or AMQP (Advanced Message Queuing Protocol). MQTT and AMQP are standard messaging protocols for Internet of Thing services across all industries and have many advantages related to lightweight requirements, security, and standardisation across devices and systems.

The ocean consists of complex environmental conditions (tides, winds, water masses, ice), that impact farm operations, safety and health of the fish. Further, offshore farms (and indeed land-based farms) are highly exposed to weather conditions -- arguably less of a consideration for terrestrial farming where animals can at least be housed. Hence, information external to the cage are pertinent to operations and management. Satellite measured observations, weather data, and numerical models of the ocean all generate information impacting at the farm-scale. 

There are a wide variety of external data sources that are pertinent to aquaculture operations. Prominent examples are publicly available ocean sensor networks such as European Marine and Ocean Data Network (EMODNet) or the US Integrated Ocean Observing System; numerically generated ocean model reanalysis and forecasts such as Copernicus Marine Products; weather data in terms of observations and numerical products from sources such as ECMWF or IBM Environmental Intelligence Suite.

Precision Aquaculture implementations demand the ability to both integrate and ingest these datasets (and to contextualise or understand the data) and process that data towards meaningful insight