Implementation and Feasibility of On-Orbit Satellite Anomaly Detection, Diagnostics, and Resolution, using Artificial Intelligence

Generating information from data requires several predicates, chief among them is the quality of the data. Bad data yields bad information. The transference model must also score a level of fidelity that captures the story that the data is trying to tell. This relationship between the two is often tainted by molding the story to support goals, not to express facts as measured by quantifiable metrics. A credible transference model must therefore be driven by scientific tools, strict objective analysis, hence enabling unrestricted outcomes. The quality of the model is measured by the consistency of outcome information generated from the transference of the same source data. In other words, the exact same data, regardless of its quality and validity, must yield the exact same information. The quality and validity of the extracted information are strongly correlated to the quality of source data. This correlation can be linear, simple nonlinear, or complex nonlinear. Data may drive information about important indicators of changes in environment, behavior of a phenomenon, and success or lack thereof of tactical and strategic planning. The value of such information, however, is only meaningful if it flows into a decision-making process, economy-driven governance, and organizational leadership.

In this paper, we discuss the data to information (D2I) transfer part, the generic and NOAA-Centric transfer models, and NESDIS role in the process. We address the potential risks associated with D2I transference models. We do not include in this paper the critical part of how information flows into decision-making, and how such a process would bring about quantifiable results that would be expressed by specific metrics and indicators of change.

In our approach, we adopt a D2I transference modeling that propagates along multi-levels of abstraction. The top-down approach assigns the highest level of abstraction to the most generic transference model, labeled as the General Case. Raw Data in this General Case model flows by securely ingesting from multiple Acquisition Systems. The Raw Data then undergoes multiple processes to generate Higher Levels of Data Products. Through multiple processes, the data is transferred into Raw Information. These processes include Integration, Data Analytics, Ontological Classifications, AI/ML Interpretation, etc. The Raw Information then undergoes stringent Validation and Verification processes. These processes yield Refined Information. The level of refinement and quality of the resulting information are very dependent on the strength of the Validation and Verification processes.

To develop the NOAA D2I model, ND2I, we use the General Case as a foundation, then reduce the Abstraction to capture NOAA’s processes. In this case, the Acquisition Systems are represented by NESDIS’ portfolio assets, i.e., GEO, LEO, and SWO. Similarly, the resulting high-level products are represented by various products such as the Climate Data Records. The Refined Information products include various reports such as weather, climate, ocean, fishery, and environment reports. This ND2I is further de-abstracted to generate application-specific models. In this paper, we present the Requirement case, where the Information Products are evaluated against the needs of the end users. In this model, the Requirements that govern the development of the Acquisition Systems are refined. This refinement informs the future development of NESDIS’ portfolio assets. These refinements in Requirements would result in Refined Information products that bring higher value to the End User.

Lastly, we identify the likely areas for Risks to occur in the ND2I transference model. These include weak user engagement, low quality of needs identification, misrepresentation of the needs, and incomplete requirements.

In this paper, we explain the relationship between data and information, and we emphasize the importance of the quality of the data and the transference models. We present a General Case D2I transference model, then explain how to implement it on the specific domain of NOAA, hence ND2I. We explain the End User role in validating that the resulting Information Products meet their needs or would inform future developments of NESDIS’ portfolio assets. We also explain the risks associated with the transference models.

We conclude by summarizing how data transference to information is a critical element that flows into decision-making processes and governance. The main lesson is that all these elements work together toward data and information-based decision-making in defining NOAA’s ability to meet the data and information needs of End Users, to guide the Decision-Making process.