Using Error Propagation and Data-Driven Modeling to Routinely Quantify Sources of Uncertainty in Large and Diverse Datasets

The National Ecological Observatory Network (NEON) is a continental-scale research platform with a projected lifetime of 30 years. NEON's purpose is to provide high quality data products that will facilitate discovering and understanding the impacts of climate change, land-use change, and invasive species on ecology. To accomplish this, NEON will perform in-situ, sensor-based measurements of approximately 55,000 high quality data streams and generate uncertainty estimates. Only when uncertainty is sufficiently quantified can meaningful interpretations be made about mean quantities and their interrelations, thus allowing for applied computations such as constructing or constraining process-based models.

NEON data and corresponding documentation will be publicly available with the goal to ensure all sources of uncertainty are identified and, if possible, quantified in a traceable and transparent manner. To meet this goal, NEON follows three main objectives: i) laboratory calibrations are traceable to nationally recognized standards, ii) measurement uncertainty estimates follow ISO and JCGM protocols, and iii) quantifiable as well as unquantifiable (i.e., those that can only be identified at current date) uncertainties are described in publicly available documents.

Preliminary results show the repeatability of temperature measurements is < ± 0.002⁰C, with an overall quantifiable uncertainty < ± 0.05⁰C, both at the 95% confidence level. Current unquantifiable measurement uncertainties include sensor and data acquisition system drifts as well as measurement error associated with varying magnitudes of aspiration and incoming solar radiation. Over time we aim to gain a better understanding of sensor-specific previously unquantifiable uncertainties. This approach will enable improved inferences between environmental drivers and responses from NEON data, such as site-specific meteorological-ecological interactions.