Abstract

Thermal offset is a significant source of uncertainty for solar radiation measurements. This study assesses the influence of mechanical ventilation on the daytime thermal offset of pyranometers. Toward this goal, an intensive unprecedented campaign of measurements was conducted in Badajoz, Spain, during four selected summer days under cloud-free conditions, covering a large range of solar zenith angle, irradiance, and temperature. Three leading manufacturers participated in the campaign, providing secondary standard pyranometers and compatible ventilation units. The thermal offset was experimentally measured following the capping methodology. A total of 372 capping events were conducted, the largest number ever reported in the literature. Each pyranometer was tested under different operational conditions (with/without ventilation and measuring global/diffuse irradiance). Results show that mechanical ventilation generally reduces the thermal offset. The magnitude of this reduction is different for each pyranometer model and depends on whether the instrument is shadowed (for measuring diffuse irradiance) or not (for measuring global irradiance). Mechanical ventilation tends to homogenize the temperature around the pyranometer and therefore reduces the impact of environmental conditions on the thermal offset. CMP11 and SPP pyranometers show notable tendencies in the thermal offset even when mechanical ventilation is applied. The Dutton et al. model, which aimed to correct the daytime thermal offset, is evaluated. Results show this model performs well for the SPP pyranometer but underestimates the absolute value of thermal offset for the CMP11 and SR20 pyranometers.

1. Introduction

The incoming solar irradiance and its partitioning into its direct and diffuse components governs many climate processes, such as the hydrological cycle, plant photosynthesis, and changes in surface temperature due to Earth’s energy budget (Roderick et al. 2001; Gu et al. 2003; Wild 2009). Thus, changes in clouds and aerosols can modify the solar irradiance amount and its direct/diffuse partitioning, with notable consequences on local, regional, and global climates. However, the role of clouds and aerosols, as they affect the variability and distribution of solar radiation at Earth’s surface, remains unclear. Thus, while Long et al. (2009) claimed that the solar brightening over the United States after 1990 was caused by a general decrease in cloud cover, Wild (2012) attributed it to a decrease in aerosols. Therefore, highly accurate measurements of solar irradiance are required in order to detect small variations in the incoming solar radiation and its response to a range of factors.

A number of studies have examined the operation of pyranometers in order to accurately measure global and diffuse irradiance (Michalsky et al. 2005; Long and Shi 2008; Gueymard and Myers 2009). One of the main sources of uncertainty is the thermal offset caused by a difference in temperature between the detector and the inner dome of pyranometers. In fact, the overarching environmental cause is the difference between the instrument body temperature (thermopile cold junction) and the sky blackbody brightness temperature. This temperature difference causes an energy flow through the thermopile (generating a negative voltage signal) out the top of the detector to the inner dome, then from the inner dome to the outer dome, and then radiated to the sky. This thermal imbalance generates an exchange of infrared flux that is emitted by the detector and is superimposed on the output signal. Neglecting the thermal offset could cause an underestimation of up to 40% in measurements of diffuse irradiance (Reda et al. 2003).

The temperatures of the detector and the inner dome are highly dependent on the physical characteristics of the pyranometer (Cess et al. 2000; Haeffelin et al. 2001; Dutton et al. 2001; Philipona 2002; Bush et al. 2000). The design and the material of each pyranometer component determine the temperature and the thermal coupling between them and therefore the thermal offset. Thus, different designs produced by manufacturers exhibit different thermal offset values. For example, while the thermal offset for diffuse irradiance measurements under clear skies can reach −20 W m−2 for an unventilated Eppley PSP pyranometer (Haeffelin et al. 2001), it is around −17 W m−2 for an unventilated CM11 pyranometer (Sanchez et al. 2015).

Because of its relevance, some authors have proposed empirical models to correct the daytime thermal offset with the aim to improve the quality of recorded data. Some of these models rely on nighttime relationships between the offset and the instrument’s net IR (Dutton et al. 2001; Younkin and Long 2004). Another approach is to develop statistical relationships using daytime radiative and environmental magnitudes recorded simultaneously to the pyranometer measurements (Vignola et al. 2007, 2008, 2009; Serrano et al. 2015).

Several authors (Bush et al. 2000; Haeffelin et al. 2001; Vignola et al. 2007, 2008, 2009) have studied the influence of environmental conditions on the thermal offset, with wind speed being an important factor. Based on these results, some authors have suggested that thermal offsets could be minimized with adequate ventilation systems, which would force ambient airflow around the body and the domes (Dutton et al. 2001; Philipona 2002; Michalsky et al. 2003). Thus, Ji (2007) reported that mechanical ventilation with a standard 115-Vac fan reduced by 13 W m−2 the thermal offset of a PSP pyranometer while measured at noon on a clear day.

Until now, two main approaches have been followed to study the effect of ventilation on the thermal offset: 1) comparing ventilated to unventilated pyranometers (Dutton et al. 2001; Michalsky et al. 2003) and 2) turning on and off in an alternating pattern the ventilation unit of a pyranometer (Philipona 2002; Ji 2007). However, these methodologies exhibit important drawbacks. Studies comparing different pyranometers with and without ventilation ignore properties of the pyranometers that could contribute to differences in the signals, such as their time responses, cosine error, and the pyranometer design. On the other hand, turning off the ventilation unit does not reproduce the actual conditions of common unventilated pyranometers, since the ventilation unit modifies the environmental conditions around the pyranometer.

Within this framework, our study aims to evaluate the effect of ventilation units on the thermal offset of different pyranometers and their impact on instrument performance. Toward this goal the main manufacturers were invited to send their pyranometers along with their corresponding ventilation units to participate in an unprecedented intercomparison campaign to be held in Badajoz, Spain, during June and July 2015. Among the pyranometers analyzed, the CMP11 manufactured by Kipp & Zonen is studied. This instrument is used worldwide at global and national radiometric networks, including Baseline Surface Radiation Network (BSRN), Global Energy Balance Archives (GEBA), Southern African Universities Radiometric Network (SAURAN), and the automatic weather station network of Servicio Meteorológico Nacional (SMN-AWS) in Mexico. Two other new models recently commercialized participated in this campaign: the SPP pyranometer manufactured by Eppley and the SR20 pyranometer manufactured by Hukseflux. It is worth mentioning that the SPP pyranometer has been designed by Eppley as the substitute for the previous model PSP, which has been one of the main instruments used by the BSRN.

This intensive campaign was conducted during four selected cloud-free summer days when very high thermal offset values were expected. A total of 124 capping events were performed for each pyranometer participating in the campaign. The thermal offset was obtained for the pyranometers measuring under four different operational conditions: measuring global irradiance while mechanical ventilation is being applied, measuring global irradiance while no mechanical ventilation is being applied, measuring diffuse irradiance while mechanical ventilation is being applied, and measuring diffuse irradiance while no mechanical ventilation is being applied. The present work focuses on studying the thermal offset regarding its magnitude, its diurnal evolution, and its relationship with environmental conditions. To our knowledge, this is the first study comparing the thermal offset of different pyranometers under such a number of different configurations in actual field conditions.

Additionally, this study evaluates the performance of the Dutton et al. (2001) model to correct the daytime thermal offset error. This model proposes using a nighttime-fitted linear relationship between the instrument net IR irradiance and the thermal offset to estimate the daytime thermal offset. The validity of this model is controversial. It has been tested mainly for diffuse irradiance measurements, showing differences between ventilated and unventilated pyranometers (Dutton et al. 2001; Michalsky et al. 2003, 2005; Carlund 2013). Our study evaluates the performance of this model when applied to each participating unventilated pyranometer while measuring global or diffuse irradiance.

2. Data

Measurements were made on the rooftop of the Physics Department building at the University of Extremadura in Badajoz (38.98°N, 7.018°W; 199 m MSL). This location guarantees an open horizon in the nonindustrialized outskirts of the city. The region is characterized by a generally mild Mediterranean climate with high temperature and cloud-free conditions during summer, when high solar irradiance values are reached.

Thermal offset was measured by means of capping events conducted on 26 and 27 June 2015, and 21 and 22 July 2015 (days of year 176, 177, 202, and 203, respectively). A total of 372 capping events, 124 for each pyranometer, were conducted during these 4 days, being the highest number of capping events ever reported in the literature.

Since the most interesting cases are those with a large thermal offset, the campaign of measurements was performed during summer days. The four selected days showed high irradiance values and therefore a large thermal offset was expected. Additionally, these days were selected because of their high stability and similar environmental and radiative conditions. The intense campaign focused on these four selected days allows also to study the diurnal evolution of the thermal offset, which has never been addressed using the capping methodology. The solar zenith angle sampled ranged from 15.5° to 82.0°.

Table 1 gives a fairly complete description of conditions during the actual data collection period for each day, including the ranges of air temperature (with the time when those extreme temperatures occurred), sky brightness temperature, relative humidity, wind speed, global and diffuse irradiances, and the net IR irradiance of a collocated pyrgeometer (instrument net IR). During the 4 days of the campaign, global and diffuse irradiance presented very similar diurnal evolution with maximum values at noon of around 1000 and 100 W m−2, respectively. As expected for clear days in this location, the ambient temperature quickly rose during the early morning and reached its maximum around 1600 UTC, remaining close to this maximum value until sunset. Air temperature ranged between 18.0° and 37.4°C during the campaign. The relative humidity and the instrument net IR irradiance also showed similar magnitude and evolution during the four campaign days. During these 4 days, the relative humidity ranged between 22% and 80%, and the instrument net IR irradiance ranged between −138.0 and −71.0 W m−2. On days 176 and 177, the wind was significantly different from days 202 and 203, remaining below 3.2 m s−1 during days 176 and 177 but reaching values of up to 6.0 m s−1 during days 202 and 203. To compare days with similar wind conditions, days 176 and 177 were used to study the effect of ventilation on the thermal offset in global irradiance measurements, and days 202 and 203 to study the effect of mechanical ventilation on the thermal offset in diffuse irradiance measurements. Thus, each pair of days guarantees similar environmental and radiative conditions. Since moderate winds up to 4.9 and 5.8 m s−1 were found on days 202 and 203, respectively, the thermal offset in diffuse irradiance is compared between mechanically ventilated and naturally ventilated by significant winds.

Table 1.

Period of data collection and interval of values sampled for each atmospheric condition during each day of the campaign.

Period of data collection and interval of values sampled for each atmospheric condition during each day of the campaign.
Period of data collection and interval of values sampled for each atmospheric condition during each day of the campaign.

It must be acknowledged that while the study involves a very large number of capping events and thus some study of the diurnal cycle is possible, they are limited to the narrow range of the ambient conditions sampled during the 4-day campaign. Thus, the intensive observation period by definition does not include a range of cloud cover and cloud type or low temperatures. In general, these conditions would not represent high thermal offset (Bush et al. 2000; Sanchez et al. 2015; Serrano et al. 2015), but occasionally it could with partial cloud cover and low wind speeds.

3. Instruments

a. Pyranometers

Three manufacturers lent secondary standard pyranometers with their compatible ventilation units for the duration of the campaign: 1) a SR20 pyranometer (3727) with its VU01 ventilation unit provided by Hukseflux (the Netherlands), 2) a CMP11 pyranometer (080493) and its CVF4 ventilation unit provided by Kipp & Zonen (the Netherlands), and 3) a SPP pyranometer (38172) with its VEN ventilation unit provided by Eppley Laboratories (United States). These models are widely used in meteorological stations worldwide.

All the pyranometers used in this study comply with the ISO 9060 (1990) criteria for an International Organization for Standardization (ISO) secondary standard pyranometer, being classified as ‘‘high quality’’ according to the WMO nomenclature (WMO 2008). They had been calibrated following standard WMO procedures according to ISO 9847 (1992). It is to note that this is the first time the thermal offset of these instruments is simultaneously measured and compared.

The pyranometers studied consist of 1) a black-coated thermopile acting as a sensor; 2) two concentric hemispherical domes to protect the sensor; 3) the body or case, which holds the key parts of a pyranometer; and 4) a sunscreen to protect the body and to reduce solar heating. Table 2 shows their main characteristics as described in the manufacturers’ instruction manuals and websites (Kipp & Zonen 2013; Hukseflux 2014; http://www.eppleylab.com/).

Table 2.

Main characteristics of the participating pyranometers provided by the manufacturers in their corresponding manuals and websites. Thermal offset in this table is commonly denominated as “offset A” in pyranometer manuals.

Main characteristics of the participating pyranometers provided by the manufacturers in their corresponding manuals and websites. Thermal offset in this table is commonly denominated as “offset A” in pyranometer manuals.
Main characteristics of the participating pyranometers provided by the manufacturers in their corresponding manuals and websites. Thermal offset in this table is commonly denominated as “offset A” in pyranometer manuals.

The pyranometer sensors are based on the thermoelectric principle. Although the detector construction differs between models, the same fundamental working principle is applicable to all of them. The sensing elements are made up of a large number of thermocouple junction pairs connected in series. While the reference junction keeps at a fixed temperature, the active junction is exposed to the solar radiation. Bonded on the active junctions to the thermocouples, there is a black-painted ceramic disk that helps to absorb the incident solar radiation. A blackened thermopile detector is preferred over others due to its flat spectral response, fast time response, and good angular behavior. The temperature of the active junctions increases when the incoming radiation passes through the two glass domes and is absorbed by the black ceramic disk. The temperature difference between the active and the reference junctions produces a small voltage that is a function of the absorbed irradiance.

The sensor is easily affected by environmental factors and the delicate black coating must be protected; therefore, the sensor is covered by domes. The use of two domes instead of one results in a better thermal equilibrium between the sensor and the inner dome, therefore reducing the thermal offset. The pyranometers participating in this study have glass domes with different sizes: the diameter of the outer dome is 60 mm in the SPP, 50 mm in the CMP11, and 40 mm in the SR20.

The body of a pyranometer is designed to provide high mechanical and thermal stability to the instrument. The pyranometers participating in the campaign showed differences both in the material and the design. For example, SR20 and CMP11 have light aluminum bodies, while SPP has a heavy body made of bronze.

There is a screen attached to the body, which is designed to protect it from the rainfall and to reduce solar heating. The material and design of this screen could play an important role regarding the thermal offset. It is in contact with the body, therefore affecting the thermal stability, and size and shape modify the airflow around the pyranometer. Thus, while SR20 and CMP11 have a screen made of plastic with a diameter of 150 mm, SPP has a flatter screen made of aluminum with a notably longer diameter of 190 mm.

During the campaign the pyranometers were installed on a Kipp & Zonen SOLYS2 sun tracker (Fig. 1a) equipped with shading balls. These balls were installed only on days 202 and 203 in order to measure diffuse irradiance. Similarly, ventilation units were installed only on days 176 and 203 in order to obtain measurements under mechanically ventilated conditions. The output signals of the pyranometers were recorded by a Campbell CR1000 datalogger every second.

Fig. 1.

Instruments participating in the campaign: (a) pyranometers (left to right) CMP11, SPP, and SR20 and (b) pyranometers inside their corresponding ventilation units (left to right) CVF4, VEN, and VU01.

Fig. 1.

Instruments participating in the campaign: (a) pyranometers (left to right) CMP11, SPP, and SR20 and (b) pyranometers inside their corresponding ventilation units (left to right) CVF4, VEN, and VU01.

b. Ventilation units

Ventilation units are primarily aimed at preventing dew and frost deposition on the domes, which could disturb the measurements. As a beneficial side effect, ventilation favors the thermal equilibrium between different parts of a pyranometer, thereby reducing the thermal offset (Kipp & Zonen 2013; Hukseflux 2014).

Three ventilation units participated in our campaign. Their main characteristics are summarized in Table 3. These ventilation units consist of a base holding a cover and a fan designed to work at 12-Vdc. Their size, shape, and material determine how the air is blown around the body and domes of the pyranometer, therefore affecting the thermal offset.

Table 3.

Main characteristics of the ventilation units provided by the manufacturers in their corresponding manuals and websites.

Main characteristics of the ventilation units provided by the manufacturers in their corresponding manuals and websites.
Main characteristics of the ventilation units provided by the manufacturers in their corresponding manuals and websites.

The base of the ventilation unit holds the fan and has available space to install the pyranometer. The ventilation unit bases of our study are made of aluminum, but their sizes and shapes differ. While VEN and VU01 have a circular base with a diameter of 23.0 and 16.5 cm, respectively, the CVF4 has a circular base of 23.0-cm diameter and an additional branch (Fig. 1b). The shape of the base is determined by the relative position of the fan with respect to the pyranometer. In VEN and VU01, the fan is placed under the pyranometer, continuously blowing air upward over the case and the dome. In contrast, the CVF4 ventilation unit uses a “squirrel cage” design with the fan located aside the pyranometer.

The VEN ventilation unit manufactured by Eppley uses a bladed fan that generates an airflow of 85 m3 h−1. The Huskeflux V01 ventilation unit uses a dc centrifugal fan. The intake of air is in the axial direction and the exhaust is centrifugal. In that sense, the working principle most closely resembles the squirrel cage design. This is called a “radial” fan to differentiate it from “axial” fans, which have both an intake and exhaust in the axial direction. The fan uses a backward-curved impeller. The fan airflow is 84 m3 h−1 at a nominal voltage of 12 V. The manufacturer has estimated an effective airflow of 50 m3 h−1 at the VU01 cover outlet (Hukseflux 2016). The VCF4 ventilation unit manufactured by Kipp & Zonen follows a squirrel cage design. It generates an air-free airflow of 28 m3 h−1.

The annular area available between the outer dome and the ventilation cover outlet limits the air blown by the fan that arrives at the dome. The size of this annual area is determined by the difference in diameter between the outer dome (Table 2) and the ventilation cover outlet (Table 3). This area is 14.2 mm wide for SR20 + V01, 11.5 mm for CMP11 + CVF4, and nearly zero for SPP + VEN. These different sizes and the fan airflow rates help to explain the feeling that the airflow is stronger for VU01, followed by CVF4, and finally by VEN.

c. Additional measurements

The radiometric station where the campaign was conducted is equipped with a nonventilated CG1 pyrgeometer. This pyrgeometer is designed to provide highly reliable and accurate measurements of infrared irradiance with an error lower than 20 W m−2 under changing temperature conditions up to 5 K h−1. It guarantees stability in time, with the change in sensitivity lower than 1% yr–1 (Kipp & Zonen 2003). The calibration factors provided by the manufacturer Kipp & Zonen for the pyrgeometer CG1 were applied. This calibration used a CG1 pyrgeometer as a reference, resulting in a sensitivity error of 5% under conditions of a temperature of 20°C and an irradiance of 140 W m−2 (according to the Kipp & Zonen calibration sheet for our pyrgeometer). Additionally, in order to confirm the stability of Kipp & Zonen’s calibration factors, the pyrgeometer was compared with a CG4 pyrgeometer located at a nearby (400 m from our station) meteorological station managed by the Spanish Meteorological Agency [Agencia Estatal de Meteorología (AEMET)]. The AEMET pyrgeometer is periodically calibrated every 2 years by intercomparison with the AEMET reference, which in turn is calibrated every 2 years by the World Radiation Center in Davos, Switzerland (WRC-Davos). The AEMET radiometric network follows WMO standard protocols and procedures for calibration and measurement, as certified according to ISO 9001 (2008). The comparison between our pyrgeometer and the AEMET pyrgeometer extended over the period August–December 2015 and showed a very good agreement, with a mean relative difference of 2%.

Figure 2 shows the temporal evolution of the ambient air temperature (Tair), the pyrgeometer case temperature (Tcase), the sky brightness temperature (Tsky), and the difference between Tcase and Tsky throughout the campaign. This last temperature difference is a descriptor of the infrared imbalance between the pyrgeometer and the environment, which provides interesting information to understand the thermal offset of the pyranometers. A similar diurnal pattern in the different temperatures analyzed is found between days 176 and 177, and slightly less similar between days 202 and 203. All days show Tcase values slightly higher than Tair and an increase in the difference between Tcase and Tsky along the day.

Fig. 2.

Temporal evolution of the ambient air temperature Tair, pyrgeometer case temperature Tcase, the sky brightness temperature Tsky, and the difference between Tcase and Tsky throughout the campaign.

Fig. 2.

Temporal evolution of the ambient air temperature Tair, pyrgeometer case temperature Tcase, the sky brightness temperature Tsky, and the difference between Tcase and Tsky throughout the campaign.

The direct irradiance is measured by a CHP1 pyrheliometer installed on a sun tracker SOLYS2, all of them manufactured by Kipp & Zonen. This pyrheliometer was calibrated in 2013 at the AEMET Radiometric Laboratory in Madrid, Spain. For this calibration, the Kipp & Zonen CH1 pyrheliometer 050408, which is directly traced to WRC-Davos, was used as reference. The calibration factor obtained in this campaign differs by less than 0.1% with respect to that provided by the manufacturer.

The station is also equipped with two Kipp & Zonen CM11 pyranometers, one for measuring global irradiance and the other one, mounted on a CM121 shadow ring, for measuring diffuse irradiance. These two pyranometers comply with the specifications of the first-class World Meteorological Organization classification (WMO 2008) for this instrument. Both pyranometers participated in the intercomparison campaign carried out in June 2013 at the “El Arenosillo” Atmospheric Sounding Station of the National Institute for Aerospace Technology (Huelva, Spain). In this campaign our two pyranometers were compared to the ventilated Kipp & Zonen CM21 pyranometer 041219, which had been recently calibrated. The calibration factors obtained in this campaign and the one provided by the manufacturer notably agree, with relative differences lower than 0.5%, proving the high stability of the response of both pyranometers.

The diffuse irradiance measurements have been corrected from the shadow ring error according to the methodology developed by Sanchez et al. (2012).

In this station, radiometric data are recorded by a Campbell CR1000 datalogger usually every minute. However, this study required higher temporal frequency; therefore, data were recorded every second during the campaign. These measurements give valuable information about the radiation field when capping events were being conducted.

Additionally, measurements of ambient temperature, relative humidity, and wind speed were registered every 10 min at a nearby meteorological station managed by AEMET. This station is only 400 m away from our radiometric station and follows WMO standard protocols and procedures for calibration and measurement. Wind speed is measured by a THIES Anemometer Compact, while ambient temperature and relative humidity are jointly measured by a THIES Compact probe. Temperature and relative humidity sensors operate inside a Stevenson wooden shelter to protect against precipitation and direct radiation. The accuracy is ±2% for relative humidity measurements and ±0.15K for temperature measurements. These measurements describe the environmental conditions while capping events were being performed.

4. Methodology

a. Capping events

In this study thermal offset measurements have been obtained following the capping methodology. This technique consists of recording the pyranometer output signal while the incoming solar radiation is blocked by a cap that covers the dome. When no solar radiation arrives at the sensor, the output signal is a measure of the exchange of infrared flux between the dome and the sensor, that is, the thermal offset. The capping event should last until the detector has responded to the blocking of shortwave radiation, but it should be short enough to prevent significant changes in the temperature of the dome.

It must be noted that this study has performed the highest number of capping events ever reported in the literature. In fact, the total number of performed capping events is greater than the measurements reported in the previous studies that have used the same methodology, such as in Haeffelin et al. (2001), Michalsky et al. (2003, 2005), and Bush et al. (2000). The low number of capping events conducted in previous studies is probably due to the great effort required to apply this methodology. Other alternative methodologies can easily provide more estimations of the thermal offset, but they suffer from other additional drawbacks. Thus, installing thermistors or barometers inside the pyranometer are intrusive methods that can affect the measurements, mainly if a thermometer is attached to the dome. On the other hand, using a negligible thermal offset pyranometer as reference generally underestimates the thermal offset, since the thermal offset of the reference is not strictly zero (Ji and Tsay 2010; Dutton et al. 2001). Moreover, other differences between the reference pyranometer and the pyranometer of study (such as their specific cosine error, spectral response, time response, and temperature dependence) are ignored.

Therefore, in the present study the capping methodology was preferred despite being highly demanding, since it has the advantage of providing realistic values of the thermal offset independently of other reference instruments and not requiring installation of thermometers nor barometers inside the pyranometers. This methodology has been extensively used for determining the thermal offset (Bush et al. 2000; Dutton et al. 2001; Haeffelin et al. 2001; Michalsky et al. 2005; Carlund 2013; Sanchez et al. 2015).

Three caps were specially manufactured for this experiment. The caps were made of polystyrene and coated outside and inside with laminated aluminum (Fig. 3). These materials were used in order to minimize affecting the dome and detector temperature when the pyranometer was capped. Thus, the polystyrene was used to insulate the dome and detector from the environment. Moreover, its external side was coated with laminated aluminum to reflect the radiation incoming from outside. The inside of the cap was also coated with aluminum because of its low IR emissivity (0.04), so as to reduce its emission. During the experiments, the caps were kept shadowed outdoors at the same ambient temperature as the pyranometers. Additionally, in order to study the possible radiative effect of the caps, their emission was measured. Toward this goal, the pyrgeometer was covered three times by each cap for 10 min on days 18 and 19 June 2015. The instrument net IR irradiance measured was similar for the three caps, with a mean value of 0.39 W m−2, which is negligible compared with the thermal offset values. The caps were designed to just cover the dome and part of the shield while not modifying the air circulation around the pyranometer.

Previous studies have shown that once the pyranometer is covered, the output signal decreases, reaches a minimum and then smoothly increases to approach a stable value (Bush et al. 2000; Sanchez et al. 2015). In the current study, the thermal offset was obtained as the minimum value reached by the signal during the capping event, as it was suggested by Bush et al. (2000). This criterion compares well with other estimations (Sanchez et al. 2015) while requiring no estimation of the time constant of the pyranometer, which can be difficult to determine.

To analyze the evolution of the output signal when the pyranometer is covered, preliminary long capping events were carried out. Six long capping events were conducted with each pyranometer, three of them with mechanical ventilation and three without mechanical ventilation. As a result of the information obtained from the preliminary measurements, the following capping events were decided to last 5 min. The capping events were performed approximately every 40 min to prevent memory effects from happening. The three pyranometers were covered simultaneously to guarantee the measurements were performed under the same environmental conditions.

b. Analysis

Previous studies have reported relationships between thermal offset and environmental variables (Bush et al. 2000; Haeffelin et al. 2001; Dutton et al. 2001; Sanchez et al. 2015). Among the main factors affecting the thermal offset, several authors pointed out the role played by the infrared downward irradiance at ground level. An increase of downward IR irradiance heats the dome and, consequently, reduces the difference in temperature between the dome and the detector. Several authors have quantified this influence using the instrument net IR as measured by a pyrgeometer and/or the difference of temperature between the pyrgeometer case and the sky brightness temperature (Dutton et al. 2001; Haeffelin et al. 2001; Younkin and Long 2004; Michalsky et al. 2005; Gueymard and Myers 2009).

Additionally, the thermal offset is affected by Tair, relative humidity RH, and wind speed W (Bush et al. 2000; Dutton et al. 2001; Haeffelin et al. 2001; Philipona 2002; Michalsky et al. 2003; Sanchez et al. 2015). The impact of these factors on the thermal offset will mainly depend on the pyranometer design. For example, the wind affects both the dome and the body, but while the dome is completely exposed, the body is partially protected by the pyranometer’s sunscreen.

The dependence of the daytime thermal offset on these environmental variables and the influence of mechanical ventilation have been evaluated for the three pyranometers participating in the campaign.

Additionally, the performance of the model proposed by Dutton et al. (2001) to correct the thermal offset was evaluated. This model proposes a relationship between the thermal offset of a pyranometer and the instrument net infrared irradiance measured by a collocated pyrgeometer (NetIR) as follows:

 
formula

According to the Dutton et al. proposal, the coefficient b is estimated using nighttime measurements and the obtained model is applied to correct the daytime thermal offset. The Dutton et al. method is intended to use the pyrgeometer as a tracer of the pyranometer. Therefore, in order to most accurately mimic the conditions that the pyranometer is experiencing, both must operate under the same ventilation conditions. Since the pyrgeometer used in this study was unventilated, the analysis is applied only to measurements registered by unventilated pyranometers. Nighttime data were selected as those measurements registered when the solar altitude was below −10° during nights 177 and 202 (no ventilation was applied).

Ordinary linear least squares fittings have been obtained for each pyranometer and the coefficient of determination R2 was calculated to quantify the goodness of fit. This statistic measures how well the observed outcomes are replicated by the model. Additionally, fittings with the intercept forced through (0, 0) have also been analyzed. This is a common practice, since a zero offset is expected when NetIR equals zero, that is, when the dome and the sensor of the pyrgeometer have the same temperature. Afterward, the nighttime-fitted model was employed to estimate the diurnal thermal offset. The differences between modeled and measured thermal offset values will be analyzed for unventilated pyranometers while measuring global or diffuse irradiance. The same analysis has been performed from the nigthtime thermal offset average of each pyranometer.

5. Results and discussion

a. Preliminary capping events

Figure 4a shows an example of the output signal measured by each pyranometer during a preliminary capping event conducted without mechanical ventilation. Vertical lines are plotted to highlight the time that each pyranometer takes to reach its thermal offset.

Fig. 4.

Output signal for each pyranometer during a preliminary capping event for (a) unventilated and (b) ventilated operational modes.

Fig. 4.

Output signal for each pyranometer during a preliminary capping event for (a) unventilated and (b) ventilated operational modes.

In all these six preliminary long capping events, the unventilated CMP11 pyranometer shows a sharp minimum with a well-defined thermal offset, taking a mean time of 10.1 s (standard deviation is 1.5 s) to reach the minimum. Meanwhile, the unventilated SPP and SR20 pyranometers requiring a mean time of 33.2 s (standard deviation is 5.9 s) and 63.8 s (standard deviation is 10.0 s) to reach their respective minimum values, taking several seconds to start increasing again.

Figure 4b shows an example of the evolution of the output signal measured by each pyranometer during a preliminary capping event while applying ventilation. The time needed by the CMP11 and SPP pyranometers to reach their minimum signal was similar to the values corresponding to the unventilated case. On the contrary, the time needed by the SR20 pyranometer notably increases when it is mechanically ventilated. The average time taken by this instrument to reach its minimum value while being ventilated was 150.5 s (with a standard deviation of 18.8 s). This long period of time could significantly affect the estimation of the thermal offset: the temperature of the dome could decrease, causing an underestimation of the thermal offset prior to the capping. In spite of this behavior, the ventilated SR20 has been included in this study; however, its results should be taken with caution.

b. Thermal offset measurements

1) Thermal offset for unventilated pyranometers while measuring global irradiance (UG)

Figure 5a shows the thermal offset values measured on day 177 by unventilated pyranometers measuring global irradiance. All pyranometers show a clear diurnal cycle, being more pronounced for the CMP11. This diurnal pattern shows three stages: 1) at early morning hours, the absolute thermal offset clearly increases (becomes more negative); 2) between midmorning and around 1700 UTC, the thermal offset oscillates around a constant value; and 3) afterward, the absolute thermal offset decreases (becomes less negative). A similar pattern has been previously reported by Sanchez et al. (2015) for two CM11 pyranometers. This behavior during cloudless summer days could be related to the diurnal cycle of case temperature to sky brightness temperature difference, as the overarching driver of the IR loss phenomenon.

Fig. 5.

Daily evolution of the thermal offset measurements for (a) unventilated global (UG), (b) ventilated global (VG), (c) unventilated diffuse (UD), and (d) ventilated diffuse (VD), for each pyranometer model.

Fig. 5.

Daily evolution of the thermal offset measurements for (a) unventilated global (UG), (b) ventilated global (VG), (c) unventilated diffuse (UD), and (d) ventilated diffuse (VD), for each pyranometer model.

The mean value of the UG measurements registered during the central period of day 177 is around −6.85 W m−2 (standard deviation of 2.0 W m−2) for the SPP and SR20 pyranometers but reaches −15.0 W m−2 for the CMP11 pyranometer. These values fall in the range of the thermal offset derived by Bush et al. (2000) for an unventilated PSP measuring global irradiance under cloudless and different sky conditions. They estimated the thermal offset using measurements of the temperature of the dome and the body, obtaining values between −17.0 and −6.5 W m−2. It must be noted that although the unventilated CMP11 shows the highest absolute thermal offset (most negative), it is very stable during the central hours of the day, with the standard deviation being 1.0 W m−2. It should be noticed that as offsets are usually negative, a reference to increasing offset means more negative and vice versa.

2) Thermal offset for ventilated pyranometers while measuring global irradiance (VG)

Figure 5b shows the thermal offset measured on day 176 by mechanically ventilated pyranometers measuring global irradiance. As expected, the main difference with respect to the unventilated case is the removal of the diurnal cycle. Instead, the VG for SR20 oscillates around −4.5 W m−2 all day long. The absolute value of the VG for the CMP11 slightly rises from the early morning to around 1500 UTC and then vaguely decreases until the sunset. Meanwhile, the absolute values of the thermal offset for the ventilated SPP pyranometer slightly decreases during the day.

There is a significant reduction in the absolute value of thermal offset when the pyranometers are mechanically ventilated. The average absolute value of VG is 4.04 W m−2 for the SR20, 5.16 W m−2 for the SPP, and 5.77 W m−2 for the CMP11 during the central period of the day. These values mean a reduction of 61% for CMP11 and 25% for SPP. In the case of the SR20 pyranometer, the 41% reduction obtained could be overestimated, taking into account the long time needed by this instrument to reach the thermal offset when it is ventilated.

A notable reduction in the variability between consecutive measurements of the thermal offset has been also found for the three pyranometers. The standard deviation is reduced to 1.3 W m−2 for the SR20, to 0.81 W m−2 for the CMP11, and to 1.5 W m−2 for the SPP. In addition, the thermal offset values of the different pyranometers are very similar. This higher similarity becomes relevant for comparisons of values reported for different pyranometers. Therefore, in general, installation of ventilation units is warranted considering the reduction in absolute thermal offset.

3) Thermal offset for unventilated pyranometers while measuring diffuse irradiance (UD)

Figure 5c shows the thermal offset values measured on day 202 by unventilated pyranometers measuring diffuse irradiance. A diurnal cycle is not evident for the UD values of the SPP. Instead, UD values for the SPP pyranometer oscillate around a mean value of −3.5 W m−2 during the entire day with a standard deviation of 2.5 W m−2. Meanwhile values of UD for the CMP11 and the SR20 are around −7.0 W m−2 before 1000 UTC, reaching −11.5 W m−2 around noon and then becoming less negative toward sunset. The mean thermal offset value for the SR20 and CMP11 pyranometers is −8.7 W m−2 and the standard deviation is around 2.0 W m−2.

Previous studies are inconclusive regarding the effect of global versus diffuse irradiance in affecting thermal offsets. While Philipona (2002) found no difference in the thermal offset between global and diffuse measurements, Bush et al. (2000) reported significant differences. In this study, a notably different diurnal pattern is found between days 177 and 202. However, this comparison must be taken with caution, since both days showed different environmental conditions. The main difference between the thermal offset for the global and diffuse measurements is the effect of direct radiation on the temperatures of the dome and sensor. The fast change in the direct component in early morning and later evening could be the main factor determining the increase and decrease in UG. In the absence of direct irradiance in unventilated diffuse measurements, this tendency disappears in the morning and is smoother during the afternoon. This result about the effect of the direct/diffuse partitioning on the thermal offset tends to agree with several studies (Bush et al. 2000; Cess et al. 2000; Sanchez et al. 2015). However, opposite conclusions have been also reported (Younkin and Long 2004; Vignola et al. 2007, 2008, 2009; Philipona 2002) and the issue remains controversial. This diversity in results is probably related to the distinct methodology, instrumentation, and location used in the different studies. Thus, further analysis is needed to assess possible differences in the thermal offset of pyranometers depending on whether they are measuring global or diffuse irradiance.

Most UD values obtained for the three pyranometers are less negative than the value −20 W m−2 given by Haeffelin et al. (2001) for a PSP pyranometer or the value −11.0 W m−2 reported by Dutton et al. (2001) for two CM21 pyranometers.

4) Thermal offset for ventilated pyranometers while measuring diffuse irradiance (VD)

Figure 5d shows the thermal offset measured on day 203 by mechanically ventilated pyranometers measuring diffuse irradiance. Ventilation affects the thermal offset of diffuse irradiance measurements similarly than when it was observed for global irradiance measurements. One of its most noticeable effects is the reduction in the variability between consecutive measurements of the thermal offset for all the pyranometers.

The thermal offset of the SR20 pyranometer oscillates around a mean value of −6.17 W m−2 during the day, with a standard deviation of 1.1 W m−2. However, this value could be higher considering the time needed by this instrument to reach the minimum when it is covered. In the case of the CMP11 pyranometer, the absolute value of the thermal offset rises from the early morning to around 1400 UTC and then decreases toward sunset. This pyranometer shows the largest reduction in the absolute thermal offset with a mean value of 3.4 W m−2 and a standard deviation of 2.0 W m−2. A decrease of around 50% in the maximum VD has been observed for the CMP11 with respect to that obtained when it was measuring UD. At the same time, a clear tendency is detected in VD for the SPP pyranometer. The absolute thermal offset of this pyranometer increases smoothly during the day and reaches a maximum value of 9.0 W m−2 during sunset. Its mean value for that day is 4.4 W m−2 and the standard deviation 2.5 W m−2. A possible explanation for this increase could be the small outlet of the ventilation unit that hinders the airflow supplied by the fan reaches the dome. With this design, the air driven by the fan circulates mainly around the body, while the dome is also exposed to natural air circulation and environmental conditions. The effect of the artificial ventilation on the thermal offset dependence on the environmental conditions will be thoroughly analyzed in the next section. The CMP11 and SPP pyranometers showed similar VD patterns as those observed for VG. This issue will be analyzed in the next section.

Other studies have obtained experimental VD values on cloudless skies by means of capping events (Dutton et al. 2001; Haeffelin et al. 2001; Michalsky et al. 2003). Haeffelin et al. (2001) and Dutton et al. (2001) reported thermal offset values between −15 and −5 W m−2 for a PSP pyranometer. These values differ from the results obtained in this study for the CMP11, SPP, and SR20 pyranometers under cloud-free summer conditions. These differences between pyranometer models emphasize the need to individually characterize each instrument, its ventilation unit, and the environmental conditions surrounding each experiment.

c. Effect of mechanical ventilation on the influence of environmental conditions

Mechanical ventilation modifies the air circulation around the pyranometers, particularly around the body and the dome. This change could directly affect the thermal offset and its relationship with the natural environmental conditions. In this section, the effect of ventilation on the relationship between the thermal offset and the environmental variables is examined. To assess these relationships, regression models between the thermal offset of each pyranometer and each independent variable were fitted using the least squares approach.

Figure 6 depicts the behavior of the thermal offset values versus W, Tair, RH, instrument NetIR, and the difference between the pyrgeometer case temperature and the sky brightness temperature (Tcase minus Tsky) data when the pyranometers were measuring global irradiance with and without mechanical ventilation are shown. Additionally, Table 4 shows the coefficients of the linear regression obtained with each independent variable.

Fig. 6.

Dependence on the environmental conditions when the pyranometers are measuring global irradiance with ventilation (gray points) and without ventilation (black points).

Fig. 6.

Dependence on the environmental conditions when the pyranometers are measuring global irradiance with ventilation (gray points) and without ventilation (black points).

Table 4.

Fitting coefficients of the linear relationships between the thermal offset and W, Tair, RH, NetIR, and Tcase minus Tsky when the pyranometers were measuring global irradiance. With and without mechanical ventilation cases are shown.

Fitting coefficients of the linear relationships between the thermal offset and W, Tair, RH, NetIR, and Tcase minus Tsky when the pyranometers were measuring global irradiance. With and without mechanical ventilation cases are shown.
Fitting coefficients of the linear relationships between the thermal offset and W, Tair, RH, NetIR, and Tcase minus Tsky when the pyranometers were measuring global irradiance. With and without mechanical ventilation cases are shown.

The thermal offset in the unventilated CMP11 pyranometer measuring global irradiance shows a positive tendency with respect to the instrument NetIR (with R2 = 0.713) and relative humidity (with R2 = 0.475). It shows a negative tendency with respect to ambient air temperature (with R2 = 0.5) as expected, since relative humidity and air temperature are anticorrelated. This behavior agrees with the behavior observed by Serrano et al. (2015) for two unventilated CM11. Similarly, a negative tendency with respect to the difference between Tcase and Tsky is expected, since this variable is anticorrelated with NetIR. These tendencies are, however, less evident for the SPP (with R2 below 0.43) and imperceptible for the SR20 (with R2 values below 0.22).

When the pyranometers are measuring global irradiance while being ventilated (gray points in Fig. 6), the decrease in the thermal offset described in section 5b(1) is confirmed. While the SR20 shows no clear patterns (with R2 below 0.38), the CMP11 and SPP pyranometers show similar tendencies to those observed for unventilated measurements (UG) but notably smoother. The CMP11 pyranometer shows the largest attenuation of these tendencies due to ventilation. In general, it is observed that the air generated by the fan seems to homogenize the conditions around the pyranometer, favoring the reduction of the thermal offset and its dependence on the environmental conditions.

The persistence of the trends on the CMP11 and SR20 when these two pyranometers are ventilated could be related to the actual air volume rate out of their corresponding ventilation units. There are many factors involved in the actual amount of ventilation that finally flows around the dome, such as the fan volume and type (bladed/squirrel/radial) and its design. As it was described in section 3b, the SR20-VU01 has a fan with a very high volume rate, and a large annular area between the dome and the ventilation cover outlet. Conversely, CMP11-CVF4 and SPP-VEN have lower volume rate fans, and the annular areas available for the airflow to reach the dome are notably smaller.

Figure 7 shows the thermal offset versus W, ambient temperature T, RH, and instrument NetIR when the pyranometers were measuring diffuse irradiance with and without ventilation. Additionally, Table 5 shows the coefficients of the linear regression obtained with each independent variable. One of the most important results observed for UD is the lack of tendencies with respect to the environmental variables being analyzed, with R2 being below 0.22 for all pyranometers. A certain diurnal pattern is observed only for the CMP11 pyranometer.

Fig. 7.

Dependence on the environmental conditions when the pyranometers are measuring diffuse irradiance with ventilation (gray points) and without ventilation (black points).

Fig. 7.

Dependence on the environmental conditions when the pyranometers are measuring diffuse irradiance with ventilation (gray points) and without ventilation (black points).

Table 5.

Fitting coefficients of the linear relationships between the thermal offset values and W, Tair, RH, NetIR, and Tcase minus Tsky when the pyranometers were measuring diffuse irradiance. With and without mechanical ventilation cases are shown.

Fitting coefficients of the linear relationships between the thermal offset values and W, Tair, RH, NetIR, and Tcase minus Tsky when the pyranometers were measuring diffuse irradiance. With and without mechanical ventilation cases are shown.
Fitting coefficients of the linear relationships between the thermal offset values and W, Tair, RH, NetIR, and Tcase minus Tsky when the pyranometers were measuring diffuse irradiance. With and without mechanical ventilation cases are shown.

The SR20 pyranometer shows no clear tendency with respect to any environmental variable while measuring diffuse irradiance and being ventilated (R2 below 0.10). On the contrary, the CMP11 and SPP pyranometers show marked relationships with respect to several environmental variables. The thermal offset of diffuse irradiance measurements tends to increase with wind speed, ambient temperature, and Tcase minus Tsky, and to decrease with instrument NetIR and relative humidity when the pyranometers are ventilated.

d. Thermal offset correction obtained from nighttime data

1) Nighttime fittings

Nighttime thermal offset versus the instrument NetIR values and its corresponding linear least squares fit (solid gray line) are shown in Fig. 8 for each pyranometer. The fitting coefficients are summarized in Table 6 together with the root-mean-square error (RMSE) and R2.

Fig. 8.

Unventilated nighttime thermal offset vs instrument NetIR: nighttime average offset (dashed black line), fit forced through zero (solid black line), and nonforced fit (solid gray line).

Fig. 8.

Unventilated nighttime thermal offset vs instrument NetIR: nighttime average offset (dashed black line), fit forced through zero (solid black line), and nonforced fit (solid gray line).

Table 6.

Fitting coefficients, RMSE, and R2 for the linear fittings between the nighttime thermal offset and the instrument NetIR.

Fitting coefficients, RMSE, and R2 for the linear fittings between the nighttime thermal offset and the instrument NetIR.
Fitting coefficients, RMSE, and R2 for the linear fittings between the nighttime thermal offset and the instrument NetIR.

The SPP pyranometer shows a slope of 0.0337, which is similar to the values reported by Haeffelin et al. (2001) and Michalsky et al. (2005) for a ventilated PSP. However, R2 is below 0.16 in all cases, indicating that most of the variance is not explained by the linear model. Several authors have previously reported no trend between the nighttime offset and instrument NetIR (Michalsky et al. 2003, 2005; Gueymard and Myers 2009). In our particular case, it is difficult to conclude this, since the low significance of the fitting could be likely due to the lack of cases with small values of instrument NetIR during the analyzed nights. To obtain a more robust fit, fewer sparse data and a finer datalogger resolution (it is currently about 0.3 W m−2) would be needed.

In usual practice, the intercept is forced through (0,0), since a zero offset is expected when the instrument NetIR equals zero, that is, when the dome and the sensor of the pyrgeometer are at the same temperature. The no-intercept regression line for each pyranometer has been plotted as a solid gray line in Fig. 8.

The slopes obtained for the no-intercept fittings are around 0.01 for the SR20 pyranometer, and 0.04 for the CMP11 and SPP pyranometers. All regression coefficients were statistically significant at the 95% confidence level. Dropping the intercept results in a small change in the fitting for the SPP, since it was the one to show the smallest intercept in the nonforced-to-zero fitting. The CMP11 fitting gets closer to the SPP fitting when the intercept is forced to be zero.

Additionally, the nighttime values of each unventilated pyranometer have been averaged, obtaining −1.26, −3.31, and −3.50 for the SR20, CMP11, and SPP pyranometers, respectively.

2) Application to correct the daytime thermal offset

The regression models obtained in the previous section were applied to daytime instrument NetIR measurements, and the resulting daytime thermal offset was compared to the experimental values measured using capping events. The nighttime-averaged value has been also tested as an indicator for the daytime thermal offset. Figure 9 shows the absolute differences found between the modeled and measured daytime thermal offset values for days 177 (UG) and 202 (UD).

Fig. 9.

Absolute difference between daytime experimental thermal offset and values estimated with Dutton et al.’s methodology (fit forced through zero; black points), nonforced fit (gray triangles), and the nighttime average (crosses).

Fig. 9.

Absolute difference between daytime experimental thermal offset and values estimated with Dutton et al.’s methodology (fit forced through zero; black points), nonforced fit (gray triangles), and the nighttime average (crosses).

It can be observed that both zero-forced and nonforced fittings notably underestimate the absolute value of the daytime thermal offset for the SR20 and CMP11 pyranometers. This underestimation is larger when the CMP11 pyranometer measures global irradiance than when it is shadowed, and it shows a marked relationship with the solar zenith angle. A similar pattern was reported by Vignola et al. (2007) when measuring diffuse irradiance with a PSP pyranometer while being ventilated.

The regression models obtained with nighttime data perform better for the SPP pyranometer, with absolute differences between the modeled and measured values being lower than 5 W m−2 and almost uniformly scattered around the zero line. Thus, although our results apply to unventilated pyranometers, these results for the SPP pyranometer are in line with the widely application of the Dutton et al. model for the ventilated PSP pyranometer, manufactured also by Eppley.

For the cloud-free summer days analyzed in this study, using averaged nighttime values performs similarly to using nonforced fittings. This result is found for all the analyzed pyranometers regardless of whether they were measuring global or diffuse irradiance. This fact reveals that, although being statistically significant, the nighttime-fitted slopes are notably different from the daytime-fitted slope between the thermal offset and NetIR. These differences are larger for the unventilated pyranometers CMP11 and SPP measuring global irradiance (see Tables 4 and 6). Nevertheless, it should be noted that this result could be also related to the lack of cases with instrument NetIR close to zero and the small sample size available to derive the relationships.

The previous studies that apply the Dutton et al. model mainly focused on diffuse irradiance measurements (Dutton et al. 2001; Michalsky et al. 2005, 2007). They have reported differences in the daytime thermal offset between the model and capping-event measurements of around 1.5 W m−2 for ventilated pyranometers and 7 W m−2 for unventilated pyranometers. Thus, our results for the SPP and SR20 pyranometers fall within the expected range. On the other hand, the CMP11 pyranometer shows somewhat larger differences up to 11 W m−2, and notable differences between global and diffuse measurements are found as well.

6. Conclusions

An intensive campaign to assess the effect of artificial ventilation on the thermal offset of pyranometers using dc fans was carried out during selected cloud-free days in June and July 2015 in Badajoz, Spain. The SR20 (Hukseflux), CMP11 (Kipp & Zonen), and SPP (Eppley) secondary standard pyranometers with their corresponding VU01, CVF4, and VEN ventilation units participated in this campaign. The daytime thermal offset was experimentally measured by means of capping events for each pyranometer while measuring global or diffuse irradiance, and with and without ventilation. To our knowledge, this is the first study to measure and compare the thermal offset of the CMP11, SPP, and SR20 pyranometers on such configurations and in actual field conditions.

Preliminary measurements revealed limitations to apply the capping methodology to the SR20 pyranometer while being ventilated. In this configuration the output signal takes 150 s to reach the minimum once the pyranometer is capped. This long period of time could significantly affect the temperature of the dome, resulting in an underestimation of the thermal offset prior to its capping. As well, the effect on the dome temperature for the 33 and 10 s for the SPP and CMP11 responses, respectively, from the temperature at the time of initial capping is unknown but is presumed to be small.

The comparison between unventilated and ventilated thermal offset when the pyranometers are measuring global or diffuse irradiance shows interesting results. Generally, mechanical ventilation tends to homogenize the temperature between the sensor and the domes and therefore to reduce the thermal offset. Additionally, a notable decrease in the short-term variability of the thermal offset was detected when the pyranometers were ventilated.

Results obtained in this study confirm the differences in the thermal offset between different pyranometer models as reported by previous studies (Dutton et al. 2001; Michalsky et al. 2003) but also reveal notable disparities in their daytime behavior. These differences between pyranometer models are generally reduced when mechanical ventilation is applied. This homogenization of the thermal offset becomes relevant for the comparison of different pyranometers.

To correct the daytime thermal offset, several authors have suggested mathematical expressions based on environmental variables (Vignola et al. 2007, 2008, 2009; Dutton et al. 2001; Serrano et al. 2015). Along these lines, the present study shows important differences in the relationships with environmental magnitudes depending on whether the pyranometer is ventilated. The tendencies detected for the CMP11 and SPP pyranometers while ventilated could be related to the small size of the outlet of the ventilation unit, which hinders the airflow from reaching the dome.

Additionally, the correction model proposed by Dutton et al. (2001) was essayed in this study. The relationship between the nighttime thermal offset and the instrument NetIR was analyzed for each pyranometer under unventilated conditions. No significant relationship was found (R2 below 0.16 for all regression fittings), being in line with other authors (Michalsky et al. 2003; Gueymard and Myers 2009). However, the low significance could be also related to the lack of cases with instrument NetIR close to zero and the small sample size available to derive the relationships. Additionally, regression models forced through (0,0) were also fitted to nighttime data. Then, the fitted models were applied to estimate the daytime thermal offset. Differences between modeled and measured daytime values indicated that Dutton et al. model performs well for the unventilated SPP but seems to be unsuitable for the CMP11 and SR20. However, this result for the CMP11 and SR20 models could be nondefinitive due to the limited robustness of the nighttime fits and limited data.

The analysis performed in this study highly recommends the general use of dc fan ventilation units. However, it must be emphasized that each pyranometer model must be individually studied with its ventilation unit under actual working conditions.

This study aims to contribute to better knowledge of the thermal offset error in pyranometers, focusing on the effect of ventilation. The methodology for comparisons described in this paper can be applied at other locations, with other instruments and environmental conditions. The findings can be especially relevant for long-term historical radiation series recorded worldwide by networks such as BSRN, GEBA, SAURAN, SMN-AWS, etc. The reduction in the thermal offset achieved in this campaign when mechanical ventilation is applied has been quantified to be between 1.3 and 9 W m−2. This value is in the order of the general changes in global radiation that occurred during the 1960–90 dimming period (variation between −5.1 and −1.6 W m−2 decade−1) and the later brightening period (variation between +2.2 and +5.1 W m−2 decade−1) (Wild 2009). It is also similar to the trend (+3 W m−2 decade−1) in the diffuse irradiance in the United States for the period 1996–2007 reported by Long et al. (2009) and to the trend (−1.3W m−2 decade−1) in Girona (Spain) for the period 1994–2014 recently reported by Calbó et al. (2017). Thus, neglecting the thermal offset error could hinder the detection of these long trends, and the mechanical ventilation has been revealed in this study as an efficient method to reduce such error.

This study also suggests directions for future research concerning the development of models to correct the daytime thermal offset based on environmental and radiative magnitudes. It would be useful in further work to assess offset and ventilation characteristics with instruments of the same type. This might provide some general relationships typical of the pyranometer model.

Acknowledgments

This study was partially supported by the research projects CGL2011-29921-C02-01 and CGL2014-56255-C2-1-R granted by the Ministerio de Economía y Competitividad from Spain and the European Union through FEDER, and by Ayuda a Grupos GR15137 granted by Junta de Extremadura and Fondo Social Europeo (FEDER). The authors thank Eppley, Hukseflux, and Kipp & Zonen for providing their instruments to this study and the Spanish Agencia Estatal de Meteorología (AEMET) for the provision of meteorological data. Guadalupe Sanchez Hernandez thanks the Ministerio de Economía y Competitividad for the predoctoral FPI Grant BES-2012-054975. Thanks to the reviewers for their comments and suggestions, which notably improved this paper. The data analyzed in this study are available from the authors upon request (guadalupesh@unex.es).

REFERENCES

REFERENCES
Bush
,
B. C.
,
F. P. J.
Valero
, and
A. S.
Simpson
,
2000
:
Characterization of thermal effects in pyranometers: A data correction algorithm for improved measurement of surface insolation
.
J. Atmos. Oceanic Technol.
,
17
,
165
175
, doi:.
Calbó
,
J.
,
J. A.
González
, and
A.
Sanchez-Lorenzo
,
2017
:
Building global and diffuse solar radiation series and assessing decadal trends in Girona (NE Iberian Peninsula)
.
Theor. Appl. Climatol.
, doi:, in press.
Carlund
,
T.
,
2013
: Pyranometer comparison. Baltic region pyrheliometer comparison 2012, World Meteorological Organization Instruments and Observing Methods Rep. 112,
10
35
.
Cess
,
R. D.
,
Q.
Taotao
, and
S.
Moguo
,
2000
:
Consistency tests applied to the measurement of total, direct, and diffuse shortwave radiation at the surface
.
J. Geophys. Res.
,
105
,
24 881
24 887
, doi:.
Dutton
,
E. G.
,
J. J.
Michalsky
,
T.
Stoffel
,
B. W.
Forgan
,
J.
Hickey
,
D. W.
Nelson
,
T. L.
Alberta
, and
I.
Reda
,
2001
:
Measurement of broadband diffuse solar irradiance using current commercial instrumentation with a correction for thermal offset errors
.
J. Atmos. Oceanic Technol.
,
18
,
297
314
, doi:.
Gu
,
L. H.
,
D. D.
Baldocchi
,
S. C.
Wofsy
,
J. W.
Muger
,
J. J.
Michalsky
,
S. P.
Urbanski
, and
T. A.
Boden
,
2003
:
Response of a deciduous forest to the Mount Pinatubo eruption: Enhanced photosynthesis
.
Science
,
299
,
2035
2038
, doi:.
Gueymard
,
C. A.
, and
D. R.
Myers
,
2009
:
Evaluation of conventional and high-performance routine solar radiation measurements for improved solar resource, climatological trends, and radiative modeling
.
Sol. Energy
,
83
,
171
185
, doi:.
Haeffelin
,
M.
,
S.
Kato
,
A. M.
Smith
,
C. K.
Rutledge
,
T. P.
Charlock
, and
J. R.
Mahan
,
2001
:
Determination of the thermal offset of the Eppley precision spectral pyranometer
.
Appl. Opt.
,
40
,
472
484
, doi:.
Hukseflux
,
2014
: User manual SR20: Secondary standard pyranometer. Manual v1407, 47 pp.
Hukseflux
,
2016
: User manual VU01. Manual v1615, 25 pp. [Available online at www.hukseflux.com/sites/default/files/product_manual/VU01_manual_v1615.pdf.]
ISO 9001
,
2008
: Quality management systems—Requirements. International Organization for Standardization, 41 pp.
ISO 9060
,
1990
: Solar energy—Specification and classification of instruments for measuring hemispherical solar and direct solar radiation. International Organization for Standardization, 11 pp.
ISO 9847
,
1992
: Solar energy—Calibration of field pyranometers by comparison to a reference pyranometer. International Organization for Standardization, 24 pp.
Ji
,
Q.
,
2007
:
A method to correct the thermal dome effect of pyranometers in selected historical solar irradiance measurements
.
J. Atmos. Oceanic Technol.
,
24
,
529
536
, doi:.
Ji
,
Q.
, and
S.-C.
Tsay
,
2010
:
A novel nonintrusive method to resolve the thermal dome effect of pyranometers: Instrumentation and observational basis
.
J. Geophys. Res.
,
115
, D00K21, doi:.
Kipp & Zonen
,
2013
: Instruction manual: CMP series pyranometers and CMA series albedometers. 46 pp.
Long
,
C. N.
, and
Y.
Shi
,
2008
:
An automated quality assessment and control algorithm for surface radiation measurements
.
Open Atmos. Sci. J.
,
2
,
23
37
, doi:.
Long
,
C. N.
,
E. G.
Dutton
,
J. A.
Augustine
,
W.
Wiscombe
,
M.
Wild
,
S. A.
McFarlane
, and
C. J.
Flynn
,
2009
:
Significant decadal brightening of downwelling shortwave in the continental United States
.
J. Geophys. Res.
,
114
,
D00D06
, doi:.
Michalsky
,
J. J.
, and Coauthors
,
2003
:
Results from the first ARM diffuse horizontal shortwave irradiance comparison
.
J. Geophys. Res.
,
108
,
4108
, doi:.
Michalsky
,
J. J.
, and Coauthors
,
2005
:
Toward the development of a diffuse horizontal shortwave irradiance working standard
.
J. Geophys. Res.
,
110
,
D06107
, doi:.
Michalsky
,
J. J.
,
C.
Gueymard
,
P.
Kiedron
,
L. J. B.
McArthur
,
R.
Philipona
, and
T.
Stoffel
,
2007
:
A proposed working standard for the measurement of diffuse horizontal shortwave irradiance
.
J. Geophys. Res.
,
112
, D16112, doi:.
Philipona
,
R.
,
2002
:
Underestimation of solar global and diffuse radiation measured at Earth’s surface
.
J. Geophys. Res.
,
107
,
4654
, doi:.
Reda
,
I.
,
T.
Stoffel
, and
D.
Myers
,
2003
:
A method to calibrate a solar pyranometer for measuring reference diffuse irradiance
.
Sol. Energy
,
74
,
103
112
, doi:.
Roderick
,
M. L.
,
G. D.
Farquhar
,
S. L.
Berry
, and
I. R.
Noble
,
2001
:
On the direct effect of clouds and atmospheric particles on the productivity and structure of vegetation
.
Oecologia
,
129
,
21
30
, doi:.
Sanchez
,
G.
,
A.
Serrano
,
M. L.
Cancillo
, and
J. A.
Garcia
,
2012
:
Comparison of shadow-ring correction models for diffuse solar irradiance
.
J. Geophys. Res.
,
117
, D09206, doi:.
Sanchez
,
G.
,
A.
Serrano
,
M. L.
Cancillo
, and
J. A.
Garcia
,
2015
:
Pyranometer thermal offset: Measurement and analysis
.
J. Atmos. Oceanic Technol.
,
32
,
234
246
, doi:.
Serrano
,
A.
,
G.
Sanchez
, and
M. L.
Cancillo
,
2015
:
Correcting daytime thermal offset in unventilated pyranometers
.
J. Atmos. Oceanic Technol.
,
32
,
2088
2099
, doi:.
Vignola
,
F.
,
C. N.
Long
, and
I.
Reda
,
2007
: Evaluation of methods to correct for IR loss in Eppley PSP diffuse measurements. Optical Modeling and Measurements for Solar Energy Systems Proceedings, D. R. Meyers, Ed., International Society for Optical Engineering (SPIE Proceedings, Vol. 6652), 66520A, doi:.
Vignola
,
F.
,
C.
N. Long
, and
I.
Reda
,
2008
: Modeling IR radiative loss from Eppley PSP pyranometers. Optical Modeling and Measurements for Solar Energy Systems II, B. K. Tsai, Ed., International Society for Optical Engineering (SPIE Proceedings, Vol. 7046), 70460E, doi:.
Vignola
,
F.
,
C. N.
Long
, and
I.
Reda
,
2009
: Testing a model of IR radiative losses. Optical Modeling and Measurements for Solar Energy Systems III, B. K. Tsai, Ed., International Society for Optical Engineering (SPIE Proceedings, Vol. 7410), 741003, doi:.
Wild
,
M.
,
2009
:
Global dimming and brightening: A review
.
J. Geophys. Res.
,
114
, D00D16, doi:.
Wild
,
M.
,
2012
:
Enlightening global dimming and brightening
.
Bull. Amer. Meteor. Soc.
,
93
,
27
37
, doi:.
WMO
,
2008
: Guide to meteorological instruments and methods of observation. 7th ed., WMO 8, 716 pp. [Available online at http://library.wmo.int/pmb_ged/wmo_8_en-2012.pdf.]
Younkin
,
K.
, and
C. N.
Long
,
2004
: Improved correction of IR loss in diffuse shortwave measurements: An ARM value-added product. Tech. Rep. DOE/SC-ARM/TR-009, 50 pp. [Available online at https://www.arm.gov/publications/tech_reports/arm-tr-009.pdf.]

Footnotes

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).