SPADE (SPArse DEnse) is a family of advanced geospatial generative Machine Learning models by Mathclick, designed to enhance the quality and utility of sparse data streams, such as those coming from sensors on buoys or drones, or coming from periodic surveys. It operates similarly to a Generative Pretrained Transformer model, like GPT4 by OpenAI, using learned patterns to generate new content.
In practical applications, SPADE can be used in an industrial port setting to nowcast (provide present and past values up to 3 years back) and forecast (up to 5 days into the future) the water depth in any area of the port that undergoes periodic water depth surveys. To do that, SPADE uses the Copernicus Marine data stream and (where available) locally deployed buoys. This predictive maintenance capability can improve operations, reduce costs, and decrease the carbon footprint of port operations. It can lead to more efficient use of surveys, improved and less expensive dredging, enhanced decision-making abilities, and the transformation of survey costs into strategic investments.
In another application, SPADE can be used to increase the spatial or temporal resolution of sensors, by "filling the gaps".