PDF Version

Investigating Drift Estimates and Corrections for NEON Data 

J. Csavina, G. Litt, C. Sturtevant, C. Florian, E. Ayres and J. Roberti

National Ecological Observatory Network (NEON), Boulder, CO 80301; 720-965-4901, E-mail: jcsavina@battelleecology.org

Drift is defined by the Joint Committee for Guides in Metrology (JCGM) as an estimate of incremental change in an instrument’s measurement over time or between calibration events. Two primary components of drift are sensor fouling and instrument degradation. To quantify sensor fouling, field techniques such as measurements of pre and post cleaning, are necessary, but regular cleaning can mitigate significant data quality impacts. Pre and post deployment calibrations can estimate instrument degradation, but in-field comparisons to co-located sensors or reference matter can provide more continuous monitoring of degradation. Here we show the National Ecological Observatory Network (NEON)’s effort to investigate these methods deployed on a range of instruments. We found that the assumption of a linear degradation utilizing pre and post deployment calibrations worked well for some instruments but was not a full observatory solution. Where sensor system design allowed, co-located sensors or reference material drift corrections could be deployed more rapidly but still had complications on automated quality thresholds. The overall goal of NEON is to provide high quality data with appropriate quality information which includes drift estimates and corrections. We want to ensure we are applying the best drift estimate approach for each sensor, which may mean unique applications for the given sensors while being transparent with our approaches. Drift corrections and uncertainty estimates are imperfect, and thus, maintaining frequent sensor calibration may be the best approach to ensuring collection of high quality data.  

Figure 1

Figure 1. Histogram of annual drift estimated for a population of 312 pyranometers (Hukseflux SR01) with averages for expanded uncertainty of calibration, drift, and correction algorithm of 2.3%, 2.6% and 0.18%, respectively.