A mathematical framework for earth science data provenance tracing

Abstract

This paper identifies three distinct data production paradigms for Earth science data, each having its own versioning structure: Climate data record production, used when the data producer’s dominant concern is providing a homogeneous error structure for each data set version, particularly when the data record is expected to cover a long time period Operational data set production, used when the producer must ensure low latency and service continuity with less attention to error homogeneity across the entire record Exploratory production, used for validation or research in which the producer decides which processes to apply by interacting with the data. In this paradigm, there may not be a common versioning structure from one production episode to another