Skip to main content
ARS Home » Midwest Area » St. Paul, Minnesota » Soil and Water Management Research » Research » Publications at this Location » Publication #308095

Title: Detecting drift bias and exposure errors in solar and photosynthetically active radiation data

Author
item WOOD, JEFFREY - University Of Minnesota
item GRIFFIS, TIMOTHY - University Of Minnesota
item Baker, John

Submitted to: Agricultural and Forest Meteorology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/21/2015
Publication Date: 6/15/2015
Publication URL: https://handle.nal.usda.gov/10113/61072
Citation: Wood, J.D., Griffis, T.J., Baker, J.M. 2015. Detecting drift bias and exposure errors in solar and photosynthetically active radiation data. Agricultural and Forest Meteorology. 206:33-44.

Interpretive Summary: Solar radiation is the primary driver for photosynthesis and transpiration and a key determinant of surface temperature. Furthermore, temporal trends in solar radiation may play a key tole in climate change. Consequently, it is important to measrue it accurately, and to have a process for detecting changes in calibration. We examined multiple years of solar radiation data recorded with a sensor considered to be a research grade instrument and noted a decline in sensitivity that was proportional to the cumulative amount of radiation received. We also noted a visual change in the sensing surface - a change from black to dark green. When we inspected multiple instruments we found that all exhibited this decay, and that it was proportional to the amount of radiation that each had received. During the course of this work we developed mathematical algorithms to analyze long-term radiation records from other sites, and were able to detect both gradual changes in sensitivity (drift) and sudden changes indicative of instrument problems such as shading or levelling errors. These results highlight the need for a regular program of calibration of radiation sensors, as well as analysis of past data to detect and correct systematic errors. These results will help to improve historical weather and climate data and allow better detection of climate trends.

Technical Abstract: All-black thermopile pyranometers are commonly used to measure solar radiation. Ensuring that the sensors are stable and free of drift is critical to accurately measure small variations in global solar irradiance (K'), which is a potential driver of changes in surface temperature. We demonstrate that the decreased responsivities of Eppley PSP pyranometers of -1.5% y-1, or -0.38% (GJ m-2)-1, were accompanied by a change in its spectral response owing to a discoloration of the sensing element. These observations motivated further work to develop routines to detect probable pyranometer drift in historical time-series. The temporal trends in the following ratios were used to detect pyranometer sensor drift: photosynthetically active radiation (PAR) to K', K' to KEX (extraterrestrial radiation at the top of the atmosphere) and PAR to KEX. Data from 8 AmeriFlux sites spanning latitudes from ~32–54°N were examined in this analysis. Probable drift in either a pyranometer or PAR sensor was identified at 7 of the 8 sites. The magnitudes of the drift represented changes of 0.15–0.85% y-1, which is sufficient to obscure actual trends in K', although these should be considered conservative low end drift estimates given that we were not making comparisons to co-located higher grade instruments. Deployment exposure errors caused by sensor shading were also discovered by comparing the daily correlations between (i) K' and KEX and (ii) PAR and KEX. Sensors drifting at rates similar to our defective PSP over a 5 year period would contribute to an underestimation of available energy of ~70 W m-2, which is non-trivial in the context of assessing eddy covariance energy balance closure, employing Penman-Monteith or Bowen ratio methods or calculating albedo radiative forcings. Our findings highlight the need for enhanced network-level QA/QC that assesses long-term segments of data rather than on a yearly basis or during 7 to 10 day inter-comparisons to identify possible sensor drift or other exposure errors. Given that probable drift was identified at multiple AmeriFlux sites, we recommend enhancing network access to calibration services that are traceable to a high quality gold standard.