• Brotzge, J. A., and Duchon C. E. , 2000: A field comparison among a domeless net radiometer, two four-component net radiometers, and a domed net radiometer. J. Atmos. Oceanic Technol., 17 , 15691582.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Field, R. T., Fritschen L. J. , Kanemasu E. T. , Smith E. A. , Stewart J. B. , Verma S. B. , and Kustas W. P. , 1992: Calibration, comparison, and correction of net radiation instruments used during FIFE. J. Geophys. Res., 97 , 1868118695.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fröhlich, C., 1991: History of solar radiometry and the world radiometric reference. Metrologia, 28 , 111115.

  • Halldin, S., 2004: Radiation measurements in integrated terrestrial experiments. Vegetation, Water, Humans and the Climate, P. Kabat et al., Eds., Springer-Verlag, 167–171.

    • Search Google Scholar
    • Export Citation
  • Halldin, S., and Lindroth A. , 1992: Errors in net radiometry: Comparison and evaluation of six radiometer designs. J. Atmos. Oceanic Technol., 9 , 762783.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kipp and Zonen, 2002: Instruction manual CNR1 net radiometer. Manual version 0706, 46 pp. [Available online at http://www.kippzonen.com/pages/706/3.].

  • Kohsiek, W., Liebethal C. , Foken T. , Vogt R. , Oncley S. P. , Bernhofer Ch , and DeBruin H. A. R. , 2007: The Energy Balance Experiment EBEX-2000. Part III: Behaviour and quality of the radiation measurements. Bound.-Layer Meteor., 123 , 5575.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Marty, C., and Coauthors, 2003: Downward longwave irradiance uncertainty under Arctic atmospheres: Measurements and modeling. J. Geophys. Res., 108 .4358, doi:10.1029/2002JD002937.

    • Search Google Scholar
    • Export Citation
  • Philipona, R., 2001: Sky-scanning radiometer for absolute measurements of atmospheric long-wave radiation. Appl. Opt., 40 , 23762383.

  • Philipona, R., 2002: Underestimation of solar global and diffuse radiation measured at Earth’s surface. J. Geophys. Res., 107 .4654, doi:10.1029/2002JD002396.

    • Search Google Scholar
    • Export Citation
  • Philipona, R., Fröhlich C. , and Betz C. , 1995: Characterization of pyrgeometers and the accuracy of atmospheric long-wave radiation measurements. Appl. Opt., 34 , 15981605.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Philipona, R., and Coauthors, 2001: Atmospheric longwave irradiance uncertainty: Pyrgeometers compared to an absolute sky-scanning radiometer, atmospheric emitted radiance interferometer, and radiative transfer model calculations. J. Geophys. Res., 106 , 2812928142.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Vogt, R., Bernhofer Ch , Gay L. W. , Jaeger L. , and Parlow E. , 1996: The available energy over a Scots Pine plantation: What’s up for partitioning? Theor. Appl. Climatol., 53 , 2331.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wilson, K., and Coauthors, 2002: Energy balance closure at FLUXNET sites. Agric. For. Meteor., 113 , 223243.

  • View in gallery

    Setup of CNR1 and reference radiometers at the MeteoSwiss radiation test field: CNR1 V in front on the right side, CNR1 without heating and ventilation on the left side, and the reference instruments mounted in the center of the tripod.

  • View in gallery

    Irradiances measured by the reference instruments and CNR1 deviations using original and field-calibrated sensitivity coefficients for four particular days during the four seasons: (a)–(d) the four component shortwave downward and upward and longwave downward and upward are shown. (top) Irradiances measured by the reference instruments; (middle) CNR1 original sensitivity data minus reference, CNR1-A (solid line) and CNR1-B (dashed line); and (bottom) CNR1 field-calibrated sensitivity data minus reference, CNR1-A (solid line) and CNR1-B (dashed line).

  • View in gallery

    Absolute and relative rms differences and bias from different time averages of downward radiation components, SDR and LDR, and TNR compared to the reference measurements. Rms values are calculated on hourly, daily, and monthly averages as well as on the annual mean, which results in the bias between the CNR1 measurements and the reference. (left) Uncertainties with original sensitivity values used on CNR1 instruments and (right) uncertainties when used with field-calibrated sensitivities.

  • View in gallery

    Night offset of SDR of the two CNR1 radiometers as a function of net longwave radiation. The night offset is a function of the temperature difference between the instrument body and the dome of the instrument, which is cooled by the cold night sky.

  • View in gallery

    Net irradiance measured by the reference instruments and CNR1 deviations using original and field-calibrated sensitivity coefficients for four particular days during the four seasons: (top) net irradiance measured by the reference instruments; (middle) CNR1 net irradiance measured with original sensitivity minus the reference value, CNR1-A (solid line) and CNR1-B (dashed line); and (bottom) CNR1 net irradiance measured with field-calibrated sensitivity minus the reference value, CNR1-A (solid line) and CNR1-B (dashed line).

  • View in gallery

    Regression of deviation CNR1 minus reference to reference total net radiation for hourly averages and daily averages: values of (a) CNR1-A and (b) CNR1-B. The upper row of each plot corresponds to the data scaled with the original sensitivity; the lower row corresponds to the data scaled with field-calibrated sensitivities.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 271 217 26
PDF Downloads 131 108 6

Performance and Uncertainty of CNR1 Net Radiometers during a One-Year Field Comparison

View More View Less
  • 1 Institute of Meteorology, Climatology and Remote Sensing, Department of Environmental Sciences, University of Basel, Basel, Switzerland
  • | 2 Physikalisch-Meteorologisches Observatorium Davos, World Radiation Center, Davos Dorf, Switzerland
  • | 3 Institute for Atmospheric and Climate Science, ETH Zurich, Zurich, Switzerland
  • | 4 Institute of Meteorology, Climatology and Remote Sensing, Department of Environmental Sciences, University of Basel, Basel, Switzerland
  • | 5 MeteoSwiss, Payerne, Switzerland
© Get Permissions
Full access

Abstract

Net radiation flux in correlation with surface energy budget, snowmelt, glacier ice balance, and forest or agricultural flux exchange investigations is measured in numerous field experiments. Instrument costs and energy consumption versus performance and uncertainty of net radiation instruments has been widely discussed. Here the authors analyze and show performance and uncertainty of two Kipp and Zonen CNR1 net radiometers, which were compared to high standard reference radiation instruments measuring individual shortwave and longwave downward and upward flux components. The intercomparison was aimed at investigating the performance of the radiometers under different climatological conditions and was made over one year at the midlatitude Baseline Surface Radiation Network (BSRN) station in Payerne, Switzerland (490 MSL). Of the two CNR1 radiometers tested, one was installed in a ventilation and heating system, whereas the other was mounted without ventilation and heating. Uncertainties of the different flux components were found to be larger for shortwave than longwave radiation and larger for downward than upward components. Using the single sensitivity coefficient provided by the manufacturer, which for CNR1 radiometers conditions using all four sensors, rather large root-mean-square differences between 2 and 14 W m−2 were measured for the individual components for hourly averages and between 2 and 12 W m−2 for daily averages. The authors then performed a field calibration, comparing each individual sensor to the reference instrument for one particular day. With the individual field calibration the uncertainty of hourly averages was reduced significantly for all components of the ventilated and heated instrument. For the unventilated CNR1 uncertainties could not be reduced significantly for all sensors. The total net radiation uncertainty of both CNR1 is rather large with up to 26% on daily averages (∼10 W m−2) for the original sensitivity coefficients and without field calibration. Only with the field calibration and for the ventilated and heated CNR1 net radiometer is an uncertainty of 10% of the daily totals of total net radiation reached, as claimed by the manufacturer.

Corresponding author address: Dominik Michel, Institute of Meteorology, Climatology and Remote Sensing, Department of Environmental Sciences, University of Basel, Klingelbergstrasse 27, Basel 4056, Switzerland. Email: dominik.michel@unibas.ch

Abstract

Net radiation flux in correlation with surface energy budget, snowmelt, glacier ice balance, and forest or agricultural flux exchange investigations is measured in numerous field experiments. Instrument costs and energy consumption versus performance and uncertainty of net radiation instruments has been widely discussed. Here the authors analyze and show performance and uncertainty of two Kipp and Zonen CNR1 net radiometers, which were compared to high standard reference radiation instruments measuring individual shortwave and longwave downward and upward flux components. The intercomparison was aimed at investigating the performance of the radiometers under different climatological conditions and was made over one year at the midlatitude Baseline Surface Radiation Network (BSRN) station in Payerne, Switzerland (490 MSL). Of the two CNR1 radiometers tested, one was installed in a ventilation and heating system, whereas the other was mounted without ventilation and heating. Uncertainties of the different flux components were found to be larger for shortwave than longwave radiation and larger for downward than upward components. Using the single sensitivity coefficient provided by the manufacturer, which for CNR1 radiometers conditions using all four sensors, rather large root-mean-square differences between 2 and 14 W m−2 were measured for the individual components for hourly averages and between 2 and 12 W m−2 for daily averages. The authors then performed a field calibration, comparing each individual sensor to the reference instrument for one particular day. With the individual field calibration the uncertainty of hourly averages was reduced significantly for all components of the ventilated and heated instrument. For the unventilated CNR1 uncertainties could not be reduced significantly for all sensors. The total net radiation uncertainty of both CNR1 is rather large with up to 26% on daily averages (∼10 W m−2) for the original sensitivity coefficients and without field calibration. Only with the field calibration and for the ventilated and heated CNR1 net radiometer is an uncertainty of 10% of the daily totals of total net radiation reached, as claimed by the manufacturer.

Corresponding author address: Dominik Michel, Institute of Meteorology, Climatology and Remote Sensing, Department of Environmental Sciences, University of Basel, Klingelbergstrasse 27, Basel 4056, Switzerland. Email: dominik.michel@unibas.ch

1. Introduction

Accurate net radiation flux measurements are indispensable for the determination of the surface energy budget. Net radiation, being an important part of the surface energy balance, forms together with the storage flux the available energy that is partitioned into the turbulent heat fluxes. The fact that, under ideal conditions, the available energy should equal the sum of turbulent heat fluxes—in general addressed as energy balance closure—makes the measured net radiation a plausible check for the measurement of turbulent heat fluxes; the latter is usually done applying an eddy covariance technique (Halldin 2004). This kind of quality control consideration has also been extended to flux measurements of carbon dioxide (Wilson et al. 2002). Regarding the important role of net radiation, the literature dealing with the uncertainty of net radiation instruments is surprisingly scarce. Most of the few studies focus on pyrradiometers (Field et al. 1992; Halldin and Lindroth 1992; Vogt et al. 1996)—a type of net radiometer that has disadvantages (different shortwave and longwave transmissivities of the plastic domes, which can easily be damaged) and that is no longer the first choice. Nowadays, in surface energy balance studies, net radiation is often measured using an instrument that combines two pyranometers and two pyrgeometers like the Kipp & Zonen CNR1 net radiometer. Apart from the manufacturer’s specifications (e.g., uncertainty of 10% on daily totals) there is little information regarding the performance of this instrument (Brotzge and Duchon 2000; Kohsiek et al. 2007).

Here we present the results of an intercomparison between two CNR1 net radiometers and a set of reference radiometers measuring the four components of the radiation balance:
i1520-0426-25-3-442-e1
where TNR denotes total net radiation, SDR and SUR denote shortwave downward radiation and shortwave upward radiation, and LDR and LUR denote longwave downward radiation and longwave upward radiation. The performance and uncertainty of the CNR1 radiometers is analyzed with respect to the individual flux measurements as well as the total net radiation. The intercomparison was performed at the MeteoSwiss radiation test field, where the Baseline Surface Radiation Network (BSRN) station Payerne is located (46°49′N, 06°57′E at 490 MSL), during fall 2004 to fall 2005.

2. Instrumentation

a. CNR1 net radiometers

The Kipp & Zonen CNR1 is a widely used compact net radiometer, which consists of four individual low standard [according to WMO standards] radiometers. The uncertainty of the CNR1 is 10% for daily total net radiation (Kipp and Zonen 2002). Solar radiation is measured by two CM3 pyranometers, one measuring shortwave downward radiation from the sky and the other facing downward measuring the reflected solar radiation. These two measurements allow one to determine the albedo, the ratio of reflected and downward shortwave radiation. Thermal infrared radiation is measured by two CG3 pyrgeometers, one for measuring longwave downward radiation from the atmosphere and the other for longwave upward radiation from the soil surface. The instrument signals are adjusted with shunt resistors (one for each instrument), which trim the sensitivities of each individual sensor to a common instrument sensitivity. Hence, the manufacturer provides only one sensitivity coefficient for the four sensors. The CNR1 can be used in two ways: measuring the four components separately [the Four Separate Components Mode (4SCM)] or measuring only Net Radiation [the Net Radiation Mode (NRM)]. In the NRM one single voltage output is created by connecting the downward-looking instruments in antiseries to the upward-looking instruments to substract upward radiation from downward radiation. In this experiment the 4SCM was chosen to investigate the behavior of each sensor. In the 4SCM the component outputs are not serially connected but are read individually. Hence, for each component the output voltage is only modified by the internal instrument resistance plus the shunt resistance. Recalibration of one component output is independent of the others. Case temperature is measured with a Pt-100 resistance thermometer located in the CNR1 body. Thus, it does not measure the exact temperature of the pyrgeometer thermopile cold junction, but rather gives a good approximation of the overall pyrgeometer temperature. The CNR1 includes an optional 12 V dc heater to eliminate frost and dew deposition. Kipp & Zonen points out that by heating measurement errors can be introduced and suggests that the heater should only be operated at night. In this experiment the built-in heater has not been used in either of the instruments.

The net radiation intercomparison experiment was carried out using two CNR1 instruments. One of them (CNR1-A) was equipped with a white coated housing and a ventilation and heating system (CNR1 V, as provided by Kipp & Zonen). The other one (CNR1-B) was mounted in its original configuration. The CNR1 V is an optional device and its use is not explicitly recommended by Kipp & Zonen. Comparing two CNR1, one of which is equipped with the CNR1 V, allows one to investigate advantages and disadvantages of ventilating and heating CNR1 instruments. The CNR1 V housing protects the instrument body from heating by solar radiation absorption, resulting in a more stable instrument temperature, as will be shown. This may result in a more quiescent pyrgeometer reading since the CG3 pyrgeometer signal is referred to the body temperature of the instrument. The housing is designed such that the air intake is followed by a filter and a heating coil that produces a constant warm airflow around the radiometer body. The warm air finally leaves the housing through openings provided for the pyranometer glass dome and the pyrgeometer silicon window. The airflow is primarily intended to prevent dew and ice formation on the glass dome and the silicon window. Ideally, it should also keep the instrument dome and window in thermal equilibrium with the body of the instrument. However, this is not the case, as will be shown; on the contrary it produces even larger dome − body temperature differences. The CNR1 V uses a nonadjustable 12-V power source.

b. Reference instruments and uncertainty

High standard (according to WMO standards) radiometers were used as reference instruments for the intercomparison. Shortwave radiation was measured by two Kipp & Zonen CM22 pyranometers and longwave radiation by two Kipp & Zonen CG4 pyrgeometers. These instruments are widely used for radiation measurements where minimal uncertainty is required. According to Kipp & Zonen, the uncertainty for daily totals is on the order of 1% for the CM22 and 3% for the CG4. The sensitivity coefficients of the four reference instruments were determined at the World Radiation Center at Davos (WRC), which holds the primary standards for pyranometer and pyrgeometer calibrations. At WRC pyranometer calibration is traced to the World Radiometric Reference (WRR), which consists of seven absolute pyrheliometers (Fröhlich 1991) for direct solar radiation measurements. For diffuse solar radiation measurements a reference standard CM22 pyranometer is used (Philipona 2002). Pyrgeometer calibration is traced to the absolute sky scanning radiometer (Philipona 2001) and a group of four reference standard pyrgeometers that have been internationally compared (Philipona et al. 2001; Marty et al. 2003). WRC calibrations warrant for the present intercomparison an uncertainty of within ±1% for daily totals for both the reference pyranometers and pyrgeometers.

In this experiment all four reference sensors were equipped with the same PMOD ventilation and heating system (PMOD-VHS; Philipona 2002). Owing to differences in design the effects of the reference and the CNR1 heating systems differ. In the PMOD-VHS the air is drawn in from below and flows vertically through a filter and around the instrument to a circular heating coil where the air is heated prior to flow over the instrument dome. Thus, the radiometer body is not heated, and the warm airflow only helps to keep the difference between the temperature of the instrument dome, which is exposed to the cold ambient sky, and the instrument body temperature rather small.

c. Instrument mounting and data acquisition

The reference instruments (Fig. 1, center) as well as the two CNR1 instruments (Fig. 1, front left and right) were mounted together on a massive tripod and horizontally leveled 2 m AGL. The upward-looking instruments had a hemispheric view with only minor obstacles within the field of view. The surface underneath the tripod was short grass, which was cut regularly, and the downward-looking instruments had a similar field of view with hardly any obstruction other than the legs of the tripod.

All instruments were logged on two Campbell Scientific CR23X dataloggers, one for the reference instruments and one for the two CNR1 radiometers. Thermopile voltage as well as case temperature measurements were sampled every 2 s, and 2-min averages were stored in the datalogger memory. Every 4 h, the data were downloaded from the datalogger to a personal computer.

3. Data quality control and evaluation

For the whole data series false data due to artifacts caused during the maintenance or other effects have been taken out. Dewfall should normally be taken out since its coverage of the sensor windows leads to an error. In this experiment it is of interest to observe whether the CNR1 V is able to keep the domes free of water during dewfall events and to know what impact on the measurements this would have, so these events were kept as recorded. Nighttime shortwave radiation values were set to zero (except in Figs. 2, 4 and 5). The effect of the CNR1 V on shortwave radiation data during night (night offset) will be shown in section 4.

In a first step, the four individual radiation fluxes are evaluated with the original sensitivity coefficients of the CNR1 radiometers and compared to the readings of the reference instruments. Since shunt resistors are used to compensate for the individual thermopile outputs, each of the four sensors of the CNR1-A and CNR1-B, respectively, was scaled with the same original sensitivity coefficient provided by the manufacturer. In a second step, the individual sensors of each CNR1 were field calibrated with respect to the reading of the corresponding reference instrument. Hence, each sensor received an individual sensitivity coefficient. The Pt-100 sensor has not been recalibrated. Since the original sensitivity coefficient of the CNR1-A was evaluated by Kipp & Zonen without heating and ventilation, one may expect that, when the CNR1 V is used, rescaling with a field-calibrated sensitivity coefficient improves the performance because the physical properties of the housing itself and the heated airflow to the domes affect the instrument behavior. As shown in section 4, not only the CNR1-A but also the CNR1-B performance improves after individual recalibration of the sensors.

4. Results

To visualize the differences between the CNR1 and the reference measurements under different conditions throughout the year, four days out of the four seasons have been selected. Figures 2 and 5 are a plot of the absolute values measured by the reference instruments as well as the differences between the CNR1 measurements and the references, first with the original calibration factors and second with the field calibration factors, for each component and the total net radiation. Note that nighttime shortwave values are kept as recorded in the figures, but for the statistical evaluations they have been set to zero since the night offset is regarded as an error. The root-mean-square differences of mean hourly, mean daily, and mean monthly values as well as the bias (difference between annual means) between CNR1 and the reference are shown in Table 1. Values of relative rms and bias differences are shown (Fig. 3) in percent and are also mentioned in the text.

a. Shortwave radiation

The shortwave downward radiation, in Fig. 2a shows differences up to 40 W m−2 with a strong diurnal cycle between the original scaled CNR1-A (with ventilation and heating system) and the reference for the September, March, and June days. The CNR1-B (without ventilation and heating system) measurements show smaller differences but can also reach large short-time differences. The shortwave upward radiation in Fig. 2b shows smaller differences for the ventilated CNR1-A than the unventilated CNR1-B. However, Figs. 2a and 2b show that, if the CNR1-A and CNR1-B are used with the field-calibrated sensitivities, they are much closer to the reference than with the original sensitivity coefficients. This is also obvious comparing the rms differences and the bias for shortwave downward and upward radiation given in Table 1, which are clearly reduced for both CNR1 with the field-calibrated sensitivity due to the fact that the overall difference is smaller and the deviations are more equally distributed between positive and negative values. The CNR1-A SDR hourly average difference of 13.8 W m−2 is large but is reduced to 2.2 W m−2 with the field calibration. The relative bias error for the SDR of CNR1-A and CNR1-B is reduced by the field calibration from 8.8% to 3.1%, and from 6.7% to less than 1%, respectively (Fig. 3). The annual mean SDR is 148.3 W m−2. For SUR the difference of CNR1-A cannot be reduced significantly (from 6.9% to 6.6%). The relative bias error of SUR CNR1-B is reduced from 9.9% to 4.7%. The annual mean SUR is 34.3 W m−2. As for the relative rms of daily averages of shortwave radiation, which correspond to the relative rms of daily totals, only CNR1-A SDR and CNR1-B SUR lie above the uncertainty range of 10% with 11.4% and 14.7%, respectively, when using the original sensitivities. With the field calibration the values reach 1.8% and 9.5%, respectively. All other shortwave daily average data lie well within the uncertainty range when using the original sensitivities (between 1.0% and 7.3%) and between 1.0% and 5.0% when using the field-calibrated sensitivities.

Comparing CNR1-A and CNR1-B, the effect of the CNR1 V ventilation and heating system is clearly visible. Although the deviations of the CNR1-A measurements with the original sensitivity coefficients are larger, they are more stable. This is likely due to the constant airflow, which prevents dew and ice formation and also stabilizes the instrument temperature. However, the ventilated and heated CNR1-A also shows more negative values during nighttime than the unventilated instrument. Figure 4 shows that this night offset is proportional to the longwave net radiation. For the ventilated and heated CNR1-A, the night offsets are twice as large as for the unventilated CNR1-B. In fact, since the ventilation and heating system on the CNR1-A primarily heats the body but not the dome of the pyranometers, the larger temperature difference between the instrument body and dome enhances the emission of thermal radiation by the sensor surface to the dome and hence produces the negative signal.

b. Longwave radiation

The longwave downward radiation scaled with the original sensitivity coefficient of CNR1-A (Fig. 2c) is in rather good agreement with the reference. There are, however, large deviations during nighttime in the CNR1-B longwave downward data series without recalibration, which appear primarily during dewfall. Besides dewfall, rainfall can also be a problem for the CG3 pyrgeometer because this device has a flat window rather than a dome so that water cannot easily run off. Dew and water droplets strongly absorb longwave radiation. During dewfall or after rainfall the thermopile is measuring the temperature of the water droplets on the window rather than the temperature of the sky. During daytime both CNR1 also show a diurnal deviation, which closely follows the signal of the shortwave radiation. These differences, with values up to 15 W m−2, are due to thermal effects of the pyrgeometer windows that are heated by solar radiation. This window or dome effect is due to temperature differences between the pyrgeometer window (dome) and the pyrgeometer body (Philipona et al. 1995). The error can be corrected by reducing the scaled longwave value of both CNR1 by 1.5% of the concurrent shortwave irradiance. The determination of the new sensitivity coefficient for recalibration has been conducted after the correction of longwave data by shortwave data, as described before. Hence, the recalibrated data displayed in Figs. 2c,d and 5 have been corrected by shortwave data and the field-calibrated sensitivity. The rms of hourly averaged longwave downward radiation of CNR1-A is 4.1 W m−2, whereas for the CNR1-B it is 15.6 W m−2 for the data without recalibration (Table 1). With field calibration CNR1-A reaches a difference of 1.8 W m−2; the situation for CNR1-B does not improve very much (14.7 W m−2). Nonetheless, the bias of LDR CNR1-B has been reduced from 7.2 to 4.8 W m−2. The longwave upward radiation data scaled with the original sensitivity coefficients of both CNR1 fit the reference rather well (Fig. 2d) except during dewfall, even on the CNR1-B LUR radiometer. Dew formation on a downward-looking radiometer window is very unusual and remarkable. The rms for the hourly average using original sensitivities is only 1.7 and 1.8 W m−2, respectively. This correlation cannot be improved significantly by the field calibration. The impact of dewfall is clearly visible in Fig. 2d. The bias relative errors of all pyrgeometers (downward and upward looking) are smaller than 1% (Fig. 3, for downwelling radiation). The annual mean LDR is 313.3 W m−2 and the annual mean LUR is 364.9 W m−2. The relative errors of daily average longwave radiation data (of all components) lie around 1%, except for the CNR1-B LDR which gives an error of 3.4%. This corresponds to the large effect of water droplets on the sensor.

c. Net radiation

Since CNR1 radiometers are primarily used to measure total net radiation, it is important to know how well these instruments compare with the total net radiation measured by the reference instruments. Figure 5 shows that by using the original sensitivity coefficients the diurnal maximum deviation measured by the CNR1-A is up to 50 W m−2 during the selected monitoring days in September, March, and June. The impact of dewfall on the CNR1-B instrument is clearly visible. The rms of CNR1-A is 16.5 W m−2 for the hourly average, and for CNR1-B it is 13.9 W m−2 (Table 1). In the case of CNR1-A, the large values (for an annual mean TNR of 62.4 W m−2) are mainly accounted for by the very large diurnal deviation of SDR (section 4a) and, in the case for CNR1-B, mainly by the effect of dew (section 4b). Using field calibration sensitivities the differences for CNR1-A on the four representive days are always smaller than 15 W m−2. This results in a hourly average of 6.6 W m−2. The bias reaches −0.8 W m−2, which is in good agreement. CNR1-B shows large deviations using either of the sensitivity coefficients due to the rain or dewfall errors. The uncertainty cannot be improved for hourly averages of the field-calibrated data. The bias of the CNR1-B TNR data is only slightly smaller for the field-calibrated data. The bias relative error of CNR1-A is reduced from 22.6% (original sensitivity) to 4.2% for the field-calibrated data. CNR1-B improves only from 16.4% to 15.4% (Fig. 3). The comparison of the daily averages for net radiation shows that the rms for both CNR1-A and CNR1-B is clearly larger than 10%. For daily total net radiation the CNR1-A measurements give a relative rms of 26.9%, whereas the CNR1-B measurements give 25% using the original sensitivity coefficient and not applying the calibration with shortwave data. For the field-calibrated data the relative rms of the CNR1-A net radiation is 10.1%. The large deviation of CNR1-B accounts for rain and dewfall situations and cannot be reduced significantly. The larger relative errors of monthly averages account for the very small monthly averages of net radiation from November to January (less than 5 W m−2). Nevertheless, the absolute rms difference is smaller than the rms of daily averages. Figure 6 shows the regressions of both CNR1 TNR to the reference TNR. It is clearly visible how the deviation for all time step averages converges to zero for CNR1-A when the data are calibrated. The scattering of points decreases with larger time average. The scattering of CNR1-B values stays much the same for the field-calibrated data, only there are more negative values, which decrease the bias error. The large positive hourly mean deviations up to 20 W m−2 at negative TNR reference values (Fig. 6a) and the even larger deviations up to more than 60 W m−2 (Fig. 6b) for data scaled with the original sensitivities are accounted for by a larger LDR in comparison to the reference. In Fig. 6a, a steeper rise of the CNR1-A LDR curve at sunrise, when shortwave components are still near zero and only longwave net radiation is composing the TNR, is responsible for the strong scattering. The steeper rise of the CNR1-A curve might be caused by higher Pt-100 measurements due to warming of the CNR1 body and a high body heat capacity. This is not corrected. In Fig. 6b, the large LDR measurements are affected by nighttime dew events. Dewfall on the sensor domes is registered very often in the data series. Absolute LDR deviations up to 60 W m−2 are not unusual. The impact of dew on the data quality is remarkably large when both figures are compared. The hourly average data scatterplot for CNR1-A shows outlying deviation values of more than 20 W m−2 (original sensitivity) and more than 10 W m−2 at a reference TNR between 0 and 100 W m−2. These deviations occurred when the sensors were not equally affected by shadow effects (e.g., clouds). Comparing the scatterplots of both CNR1 shows that the CNR1-B sensors are calibrated very well with the original sensitivity; the deviation cluster is distributed around zero in the first place. This is the reason why the differences of each individual sensor cannot be reduced significantly, or at all. It is only the effect of dew that leads to the very large error of total net radiation.

5. Discussion of results

The rms difference of deviations becomes smaller if averaged over longer time periods. The results show that shortwave components in general produce larger errors than longwave components. Errors are usually also larger on downward versus upward components even though relative errors are larger for SUR due to a low annual average of only 34 W m−2. By carrying out a field calibration for the individual sensors and scaling the output signal with the field-calibrated sensitivity coefficients, differences have been strongly reduced, particularly for the ventilated and heated instrument. Original and field-calibrated sensitivity coefficients for the two CNR1 instruments are given in Table 2. Longwave downward radiation measurements show large deviations, primarily from dew and water droplets on the domes during rain or dewfall and from solar thermal effects of the silicon window. Dew and water droplets strongly absorb longwave radiation. Hence, the thermopile is measuring the temperature of the water droplets on the window rather than the temperature of the sky. Again this is in better control if the instrument is used with a ventilation and heating system, which allows preventing dew and frost formation. The solar window heating effect on longwave radiation can be removed by reducing longwave data by 1.5% of the concurrent shortwave value. The net radiation or the sum of the four components consists of a rather small absolute value of 62 W m−2 for the annual average but, since all differences of the four components add up, it results in a large percentage error. An uncertainty on the order of 10% or less of daily totals, as claimed by the manufacturer, was only achieved with the ventilated and heated instrument and a field calibration of the individual sensors. Recalibrating shortwave radiation is easier than longwave radiation since longwave measurements also imply the measurement of temperature, which is affected by the thermal properties of the body. During a radiation comparison it is very important to verify the horizontal exposition of the radiometer since an inclination produces an error dependent on the azimuth angle of the sun. Furthermore, it must be verified that the sensors fit in the cutouts of the ventilation and heating housing properly so that no shadow effects occur.

6. Conclusions

The results of the CNR1 net radiometer intercomparison show that for the net radiation an uncertainty, as claimed by the manufacturer, was only achieved with a ventilation and heating system and a field calibration of the individual sensors. Among the individual components only the shortwave downward radiation data of one CNR1 show significantly larger relative errors than the uncertainty mentioned above. In general, shortwave data show larger errors and downward radiation data show larger errors than upward radiation data. The results of the CNR1 net radiometer intercomparison show that a ventilation and heating system produces additional differences on the one hand, particularly due to the instrument heating. An unequal heating of the instrument body and dome leads to an exchange of thermal radiation between thermopile sensor and the dome and results in negative offsets. On the other hand, the results show that ventilating and heating CNR1 radiometers improves the instrument performance and keeps the instrument domes and windows free from dew and frost. The use of shunt resistors is very convenient since only one sensitivity coefficient is needed to scale the net radiometer signals. However, the experiment shows that single components can produce large differences, which eventually add up to large errors in the total net radiation. Conducting a calibration of all sensors individually improves the performance of a CNR1 significantly. This implies, of course, that the CNR1 reads every component separately and that high standard reference radiometers for each component are available for field calibration.

Acknowledgments

We thank the Swiss Federal Office for Meteorology and Climatology (MeteoSwiss) for providing space and logistics in their radiation test field as well as supervision and regular maintenance of the radiometers during the intercomparison.

REFERENCES

  • Brotzge, J. A., and Duchon C. E. , 2000: A field comparison among a domeless net radiometer, two four-component net radiometers, and a domed net radiometer. J. Atmos. Oceanic Technol., 17 , 15691582.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Field, R. T., Fritschen L. J. , Kanemasu E. T. , Smith E. A. , Stewart J. B. , Verma S. B. , and Kustas W. P. , 1992: Calibration, comparison, and correction of net radiation instruments used during FIFE. J. Geophys. Res., 97 , 1868118695.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fröhlich, C., 1991: History of solar radiometry and the world radiometric reference. Metrologia, 28 , 111115.

  • Halldin, S., 2004: Radiation measurements in integrated terrestrial experiments. Vegetation, Water, Humans and the Climate, P. Kabat et al., Eds., Springer-Verlag, 167–171.

    • Search Google Scholar
    • Export Citation
  • Halldin, S., and Lindroth A. , 1992: Errors in net radiometry: Comparison and evaluation of six radiometer designs. J. Atmos. Oceanic Technol., 9 , 762783.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kipp and Zonen, 2002: Instruction manual CNR1 net radiometer. Manual version 0706, 46 pp. [Available online at http://www.kippzonen.com/pages/706/3.].

  • Kohsiek, W., Liebethal C. , Foken T. , Vogt R. , Oncley S. P. , Bernhofer Ch , and DeBruin H. A. R. , 2007: The Energy Balance Experiment EBEX-2000. Part III: Behaviour and quality of the radiation measurements. Bound.-Layer Meteor., 123 , 5575.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Marty, C., and Coauthors, 2003: Downward longwave irradiance uncertainty under Arctic atmospheres: Measurements and modeling. J. Geophys. Res., 108 .4358, doi:10.1029/2002JD002937.

    • Search Google Scholar
    • Export Citation
  • Philipona, R., 2001: Sky-scanning radiometer for absolute measurements of atmospheric long-wave radiation. Appl. Opt., 40 , 23762383.

  • Philipona, R., 2002: Underestimation of solar global and diffuse radiation measured at Earth’s surface. J. Geophys. Res., 107 .4654, doi:10.1029/2002JD002396.

    • Search Google Scholar
    • Export Citation
  • Philipona, R., Fröhlich C. , and Betz C. , 1995: Characterization of pyrgeometers and the accuracy of atmospheric long-wave radiation measurements. Appl. Opt., 34 , 15981605.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Philipona, R., and Coauthors, 2001: Atmospheric longwave irradiance uncertainty: Pyrgeometers compared to an absolute sky-scanning radiometer, atmospheric emitted radiance interferometer, and radiative transfer model calculations. J. Geophys. Res., 106 , 2812928142.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Vogt, R., Bernhofer Ch , Gay L. W. , Jaeger L. , and Parlow E. , 1996: The available energy over a Scots Pine plantation: What’s up for partitioning? Theor. Appl. Climatol., 53 , 2331.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wilson, K., and Coauthors, 2002: Energy balance closure at FLUXNET sites. Agric. For. Meteor., 113 , 223243.

Fig. 1.
Fig. 1.

Setup of CNR1 and reference radiometers at the MeteoSwiss radiation test field: CNR1 V in front on the right side, CNR1 without heating and ventilation on the left side, and the reference instruments mounted in the center of the tripod.

Citation: Journal of Atmospheric and Oceanic Technology 25, 3; 10.1175/2007JTECHA973.1

Fig. 2.
Fig. 2.

Irradiances measured by the reference instruments and CNR1 deviations using original and field-calibrated sensitivity coefficients for four particular days during the four seasons: (a)–(d) the four component shortwave downward and upward and longwave downward and upward are shown. (top) Irradiances measured by the reference instruments; (middle) CNR1 original sensitivity data minus reference, CNR1-A (solid line) and CNR1-B (dashed line); and (bottom) CNR1 field-calibrated sensitivity data minus reference, CNR1-A (solid line) and CNR1-B (dashed line).

Citation: Journal of Atmospheric and Oceanic Technology 25, 3; 10.1175/2007JTECHA973.1

Fig. 3.
Fig. 3.

Absolute and relative rms differences and bias from different time averages of downward radiation components, SDR and LDR, and TNR compared to the reference measurements. Rms values are calculated on hourly, daily, and monthly averages as well as on the annual mean, which results in the bias between the CNR1 measurements and the reference. (left) Uncertainties with original sensitivity values used on CNR1 instruments and (right) uncertainties when used with field-calibrated sensitivities.

Citation: Journal of Atmospheric and Oceanic Technology 25, 3; 10.1175/2007JTECHA973.1

Fig. 4.
Fig. 4.

Night offset of SDR of the two CNR1 radiometers as a function of net longwave radiation. The night offset is a function of the temperature difference between the instrument body and the dome of the instrument, which is cooled by the cold night sky.

Citation: Journal of Atmospheric and Oceanic Technology 25, 3; 10.1175/2007JTECHA973.1

Fig. 5.
Fig. 5.

Net irradiance measured by the reference instruments and CNR1 deviations using original and field-calibrated sensitivity coefficients for four particular days during the four seasons: (top) net irradiance measured by the reference instruments; (middle) CNR1 net irradiance measured with original sensitivity minus the reference value, CNR1-A (solid line) and CNR1-B (dashed line); and (bottom) CNR1 net irradiance measured with field-calibrated sensitivity minus the reference value, CNR1-A (solid line) and CNR1-B (dashed line).

Citation: Journal of Atmospheric and Oceanic Technology 25, 3; 10.1175/2007JTECHA973.1

Fig. 6.
Fig. 6.

Regression of deviation CNR1 minus reference to reference total net radiation for hourly averages and daily averages: values of (a) CNR1-A and (b) CNR1-B. The upper row of each plot corresponds to the data scaled with the original sensitivity; the lower row corresponds to the data scaled with field-calibrated sensitivities.

Citation: Journal of Atmospheric and Oceanic Technology 25, 3; 10.1175/2007JTECHA973.1

Table 1.

Rms differences (W m−2) for different time averages and bias difference: A denotes CNR1-A and B denotes CNR1-B. The first value is the difference for the original sensitivity; the value within parentheses is the difference for the field-calibrated sensitivity.

Table 1.
Table 2.

Original and field-calibrated sensitivities [μV (W m−2)−1] of all radiation components. Asterisks denote a sensitivity that is equal to the value presented in the first row.

Table 2.
Save