Evaluation of measurement accuracy and comparison of two new and three traditional net radiometers

Net radiation (R-n) is the sum of the radiant energy at the Earth's surface and is a major component of the surface energy balance. However, R. is difficult to measure accurately, and multiple instruments are available to measure it. Two new instruments (Hukseflux Thermal Sensors B.V., model NR01; Kipp & Zonen B.V., model CNR 2) have been released within the past two years. We compared these models, two less-expensive older models (Kipp & Zonen B.V., model NR-Lite; Radiation and Energy Balance Systems, Inc., model Q*7.1), and a more expensive older model (Kipp & Zonen B.V., model CNR 1) over a uniform turfgrass surface for 33 days in mid-summer. Three replicates of each radiometer were included in the study (except for the CNR 1). The instruments that independently measure the four components of R. (models CNR I and NR01) were typically the most accurate. Incoming shortwave measurements from the four component instruments were compared to a reference pyranometer, and outgoing longwave measurements were compared to infrared measurements of surface temperature. The differences from the reference pyranometer and surface temperature measurements were typically 2% or less. There was a difference of approximately 5% in incoming longwave measurements between these two radiometer models. This is likely due to differences in calibration approaches, which are discussed. This emphasizes the need for standardization of longwave calibration methods and establishment of a world reference for longwave radiation. The instruments that do not separate shortwave and longwave radiation into component measurements (net all-wave radiometers, models NR-Lite and (Q*7.1) were generally the least accurate, and had offsetting day and night differences that reduced daily total R-n differences relative to the reference. The CNR 2 measures net shortwave and net longwave, and is an intermediate between a four component instrument and a net all-wave instrument. The R,, measurement accuracy of the CNR 2 typically fell between that of the two groups. Differences among radiometers tended to be larger at night than during the day, indicating higher variability in longwave measurements. An inversion (flip) test in the field showed the NR-Lites and Q*7.1 s had well matched detectors, however two of the three replicate CNR 2s had mismatch errors greater than 5%. This becomes important for measurements over non-vegetated surfaces. The data presented here should be helpful in selecting the most cost effective instrument for a given application. (C) 2009 Elsevier B.V. All rights reserved.