Paper
Stability of solar radiation sensor calibration in the NZ Climate Network
Published Jan 1, 2018 · Liley
1
Citations
1
Influential Citations
Abstract
The NIWA Climate Database holds solar radiant energy data for ~150 stations around New Zealand and the southwest Pacific, largely from Licor photodiode sensors but with Eppley PSPs at some sites. There are good multi-year data for about 70 sites, including all major population centres. There are hourly global, diffuse, and direct radiation data from Eppley sensors at Kaitaia, Paraparaumu, and Invercargill since 1988, but the best NZ solar radiation data are from the Baseline Surface Radiation Network (BSRN) station at Lauder, in Central Otago. Meteorological Service of New Zealand Limited (MetService) has regularly calibrated their own and other solar radiation sensors, at Paraparaumu, for the New Zealand Climate Network but since November 2013 NIWA’s instruments have been calibrated at Lauder. Instruments that remain in statistical control are redeployed to the next vacant site, but the practice of using the new calibration directly has led to dubious steps in the data. Records of calibrations and deployments are incomplete, and somewhat error-prone, but they show that dividing out applied calibrations actually reduces the variance in (near-to-) clear-sky values. For more reliable time series for analysis of patterns or trends, regular calibration should be combined with statistical control theory, so that instrument response that can be regarded as constant or slowly trending does not become an additional source of error. Submitted: 12 October 2018 Accepted: 28 November 2018 29 Weather & Climate Volume 38 Issue 1 Meteorological Service and subsequent MetService at Kaitaia, Paraparaumu, and Invercargill from the late 1980s with tracking pyrheliometers and a pyranometer with shade band, logged at hourly intervals. From 2000 the instruments were transferred to NIWA data loggers, with both hourly and 10-minute recording. In 2011, trackers were upgraded to EKO systems using a shade ball for diffuse measurements. The New Zealand National Climate Database (CLIDB), managed by NIWA and accessible at https://cliflo.niwa. co.nz/, holds global irradiance data for over 200 climate stations around New Zealand. Of those stations, 181 are open, and 100 have at least 10 years of data. Most of the stations are operated by NIWA, 48 belong to MetService, and some are owned by other CRIs, Airways Corporation, regional and district councils, universities, and others. The New Zealand global irradiance data find use in calculating solar heating, evaporation rates, plant growth, and the like. Within NIWA, they have been used to develop maps of solar flux (Tait and Liley 2009), to estimate UV radiation anywhere in the country (Bodeker and McKenzie 1996; Bodeker et al., 2002; Bodeker et al., 2006), and to calculate available solar energy on a panel of arbitrary tilt and bearing allowing for horizon shading (http:/solarview.niwa.co.nz). All of these NIWA products use the ratio of measured global irradiance to clear sky values, and that calculation raises a question about the stability of instrument calibration. MetService operates a calibration facility at Paraparaumu, and all pyranometers in the New Zealand climate network were previously calibrated there. Since November 2013, NIWA sensors have been calibrated at Lauder. Only 17 of the NIWA pyranometers are thermopile radiometers (all Eppley) that measure the 300 nm to 2800 nm integral (e.g., https://s.campbellsci.com/documents/au/manuals/ psp.pdf). The manufacturer suggests figures of 2% uncertainty in hourly and 1% in daily integrals (unstated, but presumably both 2σ) for these instruments, while the calibration certificates from the MetService’s Calibration Laboratory give a figure, including reference sensor uncertainty of ±2.8% (2σ), of ±6% (2σ) in hourly data. Most of the other sensors are Licor silicon photodiodes, for which the manufacturer suggests a calibration uncertainty of ±3% within 60° angle of incidence (https:// www.licor.com/env/products/light/pyranometer.html). MetService calibration certificates give a figure of ±8% (2σ) for hourly values in field data. The response of silicon sensors drops sharply from 1000 to 1100 nm and beyond, so the calibration against reference thermopile pyranometers is applicable only to the extent that the overall spectrum in field measurement is similar to that at calibration. For most meteorological and climate parameters, such as temperature, pressure, humidity, or wind speed, the reliability of time series depends on periodic calibration of sensor response against reference sensors or collocated instruments. Drifting instrumental response may be suspected, and detectable in the calibration record, but it cannot usually be inferred just from field data with a single instrument. With solar radiation data, a further test arises naturally for checking both stability and absolute response when the aerosol optical depth is low. As demonstrated below, fitting the diurnal variation with a suitable model to detect days with clear sky (minimal cloud) provides a ‘field calibration’ or verification method that readily detects any major errors in response or alignment. Here I compare the results of such analysis with the formal calibration record.
Regular calibration and statistical control theory are crucial for reliable solar radiation data in New Zealand, as dividing applied calibrations reduces variance in clear-sky values.
Full text analysis coming soon...