Big Data Platform for Smart Aquaculture
SANISMART: Big Data Platform for shrimp aquaculture in Peru
Sustainable shrimp production is data-intensive, requiring coordination among diverse stakeholders. Smart data platforms can offer a cost-effective solution by supporting this coordination. These platforms are typically shaped by national and regional needs and capacities, and can be developed with a wide range of features and services. In this use case, the focus is on AI and machine learning applications to predict pond conditions.
SANISMART is Peru's national Big Data Platform for smart shrimp aquaculture, developed through the FAO Smart and Sustainable Aquaculture through Effective Biosecurity and Digital Technology (SAB) project and funded by the Ministry of Oceans and Fisheries of the Republic of Korea. It was built in partnership with APOYO Consultants and FAO technical support, with biosecurity data governance coordinated through SANIPES (Autoridad Nacional de Sanidad e Inocuidad en Pesca y Acuicultura, National Authority for Health and Safety in Fisheries and Aquaculture). The platform focuses on semi-intensive shrimp farming in Tumbes.
SANISMART focuses on tracking pond conditions using hand-reported and sensor data from shrimp farms and biosecurity information from SANIPES. It provides a standard suite of Big Data Platform services as described in the guide, including real-time sensor monitoring, automated alerts, regulatory reporting, and GIS-based map visualization. Although not yet integrated into the platform, the SAB project also developed geospatial data-driven pond activity monitoring, allowing authorities and farm managers to track pond status and changes in production activity across farming areas.
On the AI/ML side, the platform includes advanced probabilistic dissolved oxygen (DO) forecasting, which uses hybrid statistical and machine learning approaches by combining methods such as ARIMA and regression with models like Random Forest or LSTM to deliver cost-effective and accurate pond condition forecasts for individual farms. It also leverages geospatial AI services, specifically deep learning applied to Copernicus Sentinel-1 and Sentinel-2 satellite imagery for pond boundary detection and production activity classification.
The platform is designed to allow users with different roles to navigate quickly to their areas of interest through a dashboard.
Dashboard of SANISMART, featuring quick-access buttons to work areas. The system implements a secure role-based user-management approach where users can only access pre-defined areas and features of the platform.
Dashboard panel – Farm summary. This panel provides a quick situational overview, allowing users to filter production units, view current and historical values for key indicators, and access a risk forecast and alert panel.
Dashboard panel – Biosecurity and health screen. This screen summarizes current conditions and historical indicator values, showing risk levels for key diseases, including NHP, VpAHPND and WSSV. These indicators are generated in the platform’s backend using a combination of statistical analysis and AI.
Dashboard panel – Water quality screen. This screen shows current and historical sensor data for dissolved oxygen, pH, salinity and temperature.
To illustrate how the platform works in practice, the examples below show how it displays sensor time series, generates forecasts through statistical analysis and AI, and raises alerts when user-defined thresholds are exceeded.
Time-series of an environmental indicator of Oxygen level (top) and time-series with a computed forecast to anticipate DO Events (bottom) .
Based on time-series analysis and a statistical / AI forecasts, an alert can be raised, in this example of a DO risk.
Similarly to the DO risk assessment, SANISMART produces forecasts and risk assessments for pH, salinity and water temperature. In each case, the platform back end collects sensor data and processes it using a combination of statistical analysis and ML for short-term and long-term predictions. The model adapts to measurements from each sensor.
Prediction of pH with SANISMART.
The sensor values are used to compute compound indicators that can also be displayed. These include indicators for mineralisation, water quality and overall stress.
In the context of the SANISMART platform, an indicator is a derived variable that synthesises the information of one or more parameters in order to facilitate the interpretation of multiple signals. It is often expressed as a simplified value, usually between 0 and 1, that reflects the status of the underlying set of signals. A risk indicator is one example of a compound indicator.
Time series of water quality risk indicators based on the constituent parameters. A compound indicator for water quality (top image) can be constructed from the time-series of its constituent parameters (DO, pH, salinity, temperature).
Similar approaches can be developed for other compound risks, such as phytoplankton composition and density, which can cause large oscillations in DO values.
SANISMART supports also disease monitoring, more specifically the risks associated with three main diseases: Necrotizing Hepatopancreatitis (NHP), Vibrio parahaemolyticus - Acute Hepatopancreatic Necrosis Disease (VpAHPND) and White Spot Syndrome Virus (WSSV).
Dashboard panel – Prediction of disease risks.
In the platform, the risks are computed based on environmental conditions and an analytical model that expresses risk on a scale from 1 to 100. Depending on the site manager’s knowledge, the status of the production cycle, and recommendations from national authorities, a data-driven decision table can be developed for management purposes, for example, to recommend an action if the risk level exceeds 40%.
An example of screen with risk values of three diseases in SANISMART, showing risk percentage.
For more detailed information, contact FAO Peru at FAO-PE@fao.org.