The role of network architecture in surface-based in-situ climate observations

Since the establishment of a nationwide Cooperative Observer Program (COOP), many more atmospheric observational networks have been deployed in the US for a variety of purposes. These networks provide a wealth of information for initializing and validating atmospheric models critical to weather and hydrological issues, monitoring conditions in urban environments for health purposes, measuring critical variables for agriculture, detecting hazards for transportation, and fulfilling many other stakeholder needs. Even still, other networks have been developed for making science-quality measurements of the environment for purposes of understanding earth system processes and climate change. Many of these networks have little in common in regard to station placement, infrastructure, observation redundancy, quality assurance (QA), and even units of measurement. As multiple data sources become available, it is often up to the stakeholder or user to decide whether mixing various data sets from different network platforms will impact their overall findings. To begin to address these issues, this study is intended to analyze effects of network architecture on in-situ observations of variables important to climate monitoring.

The Earth Resources Observation and Science (EROS) research center in Sioux Falls, SD offers an opportunity to compare measurements from many observational networks in one location. This site hosts stations belonging to the US Climate Reference Network (USCRN), COOP Network, Automated Weather Station (AWS) Network, Soil Climate Analysis Network (SCAN), Surface Radiation (SURFRAD) Network, and Canadian Reference Climate Station (RCS) Network. Apart from near-surface temperature, initial findings indicate a higher degree of variation than expected among the measurements taken by the networks of precipitation, and soil moisture. These differences were particularly pronounced and could not be explained by microclimates alone. Precipitation tended to be sensitive to station infrastructure (existence and type of precipitation shield, heated gauges, and wetness sensors), observational redundancy, and QA algorithm. Soil moisture, on the other hand, tended to be influenced by observational techniques (approaches to sampling, and conversion from raw measurement units to soil moisture units), sensor installation, observational redundancy, and QA methods to detect faulty sensors. These subtle differences between network architectures can lead to heterogeneities in data quality across the suite of networks, which is very important to recognize as data from more observational networks become more easily accessible. Moreover, it is suggested here that potential users have a strong understanding of possible weakness and biases of the networks from which their data originated.