OPS 3-4
How NEON’s network of standardized sensor observations enable consistency testing for related data sets

Tuesday, August 11, 2015
Exhibit Hall, Baltimore Convention Center
Derek E. Smith, National Ecological Observatory Network (NEON, Inc.), Boulder, CO
Joshua A. Roberti, National Ecological Observatory Network (NEON, Inc.), Boulder, CO
Jeffrey R. Taylor, National Ecological Observatory Network (NEON, Inc.), Boulder, CO

The National Ecological Observatory Network (NEON) is set to begin full operations around 2017. The Observatory will collect and disseminate an array of ecological data across the United States for a period of 30 years. NEON has embarked upon the standardization of ecological measurements and data products on a scale that has not been seen before. With raw sensor data exceeding 100 TB/yr, many of the traditional “eyes on” approaches are no longer feasible. Thus, scientists must develop new and innovative data reduction methods for quality control purposes. One way NEON is confronting this challenge is through the use of consistency analyses to assess data quality of related data sets. This is possible though the standardization of measurement techniques and the traceable quantification of measurement uncertainties. Here we present a subset of currently developed consistency analyses for NEON’s terrestrial-based sensors.


Consistency analyses can vary widely in their complexity. A simplistic method is to verify whether the mean and variance between co-located sensors at a site are comparable to one another. An added benefit of simplistic approaches is that one method can be created and then replicated for varying use cases; however careful thought must be taken to ensure that the methods are modified to account for differences in measurement behavior. For example, if one is evaluating the rate of change between measurement observations, this will differ vastly for “slow” moving variables (e.g., pressure) than “fast” moving variables (e.g., wind speed). Another key aspect that must be considered is the homogeneity of the surrounding vegetation cover and/or other environmental variables, which may induce measurement differences among sensor locations. Simplistic analyses quickly expand in complexity when accounting for surrounding environmental variables that have both spatial and temporal dependencies. Along with traceable uncertainty estimates, consistency analyses can also be used in part to better inform sensor noise over time. For example, NEON can use nighttime measurements of photosynthetically active radiation (PAR) as a baseline to determine if laboratory based estimates of measurement uncertainty are realistic in the field. As the NEON network matures, consistency analyses and their thresholds will continue to be refined. This will enable more sophisticated approaches to be developed.