Docstoc

Alert SEARCH NIP calibrations

Document Sample
Alert SEARCH NIP calibrations Powered By Docstoc
					Alert SEARCH NIP calibrations at Bratt’s Lake
Observatory – January-February 2007
Prepared by David Halliwell, Environment Canada, March 2007


Introduction

        This document describes the procedures and results of the pyrheliometer comparisons
carried out between the Eppley NIPs from Alert and the pyrheliometers used at the BSRN station
at the Bratt’s Lake Observatory, in Saskatchewan, Canada. The NIPs were removed from service
in the fall of 2006 and shipped to the Bratt’s Lake Observatory, where they were operated from
January 15 to February 13, 2007. They were then re-installed at Alert in late February 2007.

        Two comparisons are presented here: the first using HF 20406 as a reference instrument,
and the second using CH1 940069 as a reference instrument. The comparison with HF 20406 is
the primary source of calibration, while the comparison with the CH1 is presented to increase the
data available for temperature sensitivity characterization.


Methods

        All instruments were installed and measured using the Bratt’s Lake Observatory’s BSRN
data acquisition system. This system uses an HP3458A multimeter, with data collected according
to the BSRN specifications: readings once per second, with statistics (mean, standard deviation,
maximum, and minimum) stored every minute. When weather is suitable, observatory staff
installed HF 20406, operating in a mode where the instrument is run through a calibration cycle
every 30 minutes. The calibration cycle takes 6 minutes, using the following pattern:

       - minute 0            shade on, heater off (stabilization)
       - minute 1            shade off, heater off (dark count)
       - minute 2,3          shade on, heater on (stabilization)
       - minute 4            shade on, heater on (calibration)
       - minute 5            shade off, heater off (stabilization)
       - minutes 6-29        taking irradiance readings

       The WRR reduction factor used for HF 20406 was the result from the NREL
Pyrheliometer Comparison 2006, carried out in Golden, CO, in September 2006. The operating
mode at the Bratt’s Lake Observatory provides 24 one-minute means per half-hour. The data
were then screened according to the following criteria:

       - the reference instrument irradiance reading was >= 700 W/m2.
       - the reference instrument’s standard deviation for the minute was <2.5 W/m2.
       - the standard deviation for each pyrheliometer was <2.5 W/m2.
         In the one-month period that the NIPs were operating, only six days were suitable for HF
operations, and only portions of those days provided suitable data. In total, 426 minutes of data
were used in the analysis. This may seem like a limited data set, but each value represents a
series of measurements taken over that minute, so 426 is an underestimate of the appropriate n-
value to use. The one-second sampling interval is faster than the response time of any of the
pyrheliometers, though, so 426*60 is too large an n-value. If individual readings were taken at
20-second intervals (as at the NPC), 426 minutes would be equivalent to over 1200 NPC points,
so the amount of data collected in this analysis is deemed sufficient for transfer of the HF
calibration to the NIPs. The dates and number of minutes of data from each date are given in
table 1:

Table 1: Times with acceptable HF data
Date                                              Number of minutes of data

January 15                                        137
January 16                                        17
January 18                                        134
January 19                                        58
February 7                                        31
February 12                                       49


Analysis

        For each minute of data, HF values were calculated and compared to the readings for
each pyrheliometer. In addition to the two NIPs, three Kipp and Zonen CH1s were included in
the comparison. The CH1s are the standard pyrheliometers used for BSRN data at the Bratt’s
Lake Observatory. For each pyrheliometer, a relative responsivity was calculated – the ratio
between the pyrheliometer’s irradiance reading and that of the reference instrument. The Alert
NIPs used the manufacturer’s responsivity (engraved on the side of the instrument) for data
reduction. These responsivities were 7.79 for NIP s/n 33822, and 8.03 for s/n 33856. The CH1
data were reduced using the responsivities determined through their history of calibration during
their service for BSRN readings. Zero offsets were not accounted for in any of the CH1s or NIPs.
Examination of the nighttime data shows that all instruments typically give readings in the range
+/-1 W/m2. The largest offsets appear to occur in NIP 33822, at times when temperature
fluctuations are large.

        Temperatures were recorded for the three CH1s and for NIP 33822. In comparing
irradiance values to the HF reference instrument, no temperature compensation was performed.

         Figure 1 shows the relative responsivity (instrument irradiance / HF irradiance) for the
times with suitable HF values. The times are arranged chronologically, but not to scale. Dates are
indicated on the graph, but data within a single day may have time gaps that are not represented
(e.g., January 19, which clearly has at least two periods of data that are not continuous). For
comparison, results are included for the three BSRN CH1s, as well as the two Alert NIPs. Note
that all instruments show variation within a day, as well as from day-to-day. The primary cause
of this is temperature, which will be discussed later.

Figure 1:
                                        1.05

                                               Jan           Jan                   Jan                        Jan       Feb    Feb
                                                15            16                    18                         19        7     12
                                        1.04



                                        1.03



                                        1.02
  Relative responsivity (to HF 20406)




                                        1.01



                                          1



                                        0.99



                                        0.98



                                        0.97



                                        0.96



                                        0.95

                                                     CH1 940069    CH1 940074   NIP 33822   CH1 930017    NIP 33856



       New responsivities for each NIP were determined by calculating the mean of the
individual relative responsivities and using this to adjust the responsivity used in the data
reduction. These results are displayed in table 2:

Table 2: Responsivity (µV/W/m²) data
Instrument     Manufacturer’s Relative                                          Standard                 New            NOAA
               responsivity    Responsivity                                     Deviation                Responsivity   Responsivity
NIP 33822      7.79            0.9699                                           0.0037                   7.56           7.51
NIP 33856      8.03            0.9847                                           0.0039                   7.91           7.86

        For both NIPs, the new responsivity values are slightly higher (0.6-0.7%) than those
determined by NOAA prior to installation at Alert. These variations are within the 1% accuracy
of the instruments, however, and may be related to the range of temperatures used in the
calibrations. At Bratt’s Lake, the instrument temperatures ranged from -25°C to 0°C. Instrument
temperatures were recorded for all three CH1s, and for one of the NIPs, so the relative
responsivity for each instrument can be examined as a function of temperature. Figure 2 shows
this analysis. Note that the second NIP (33856) is graphed against the temperature of the first
NIP (33822), under the presumption that this is the most suitable temperature to use, of the
measurements available.

        Figure 2 shows that all five pyrheliometers show some temperature dependence that is
not fully compensated by the manufacturers’ internal methods. (This dependence is, however,
within manufacturers’ specifications.) The figure also shows that the NIPs are more subject to
changing temperatures than the CH1s, however: there is more variation about the trend in figure
2. Looking back at figure 1, this variation can also be seen as variations in relative responsivity
within a day – January 18 and 19 are particularly variable – and examination of the temperatures
on those days confirms that the responsivity variations are related to the rate of temperature
change.

Figure 2:
                                        1.05


                                        1.04


                                        1.03
  Relative responsivity (to HF 20406)




                                        1.02


                                        1.01


                                          1


                                        0.99


                                        0.98


                                        0.97


                                        0.96


                                        0.95
                                               -25   -20                   -15                     -10                    -5   0
                                                                           Instrument Temperature (°C)

                                                           CH1 940069   CH1 940074   NIP 33822   CH1 930017   NIP 33856



        In order to better define the temperature sensitivity of the NIPs, additional data were
analyzed using CH1 940069 as a reference instrument. This allowed inclusion of data when the
HF was not operational – either due to weather (high winds), or staff availability (e.g., sunny
weekends). The selection criteria were the same as for HF data (reference instrument >= 700
W/m2, standard deviation < 2.5 W/m2). The CH1 data also had additional temperature
compensation factors included – a linear adjustment based on 6-8 years of use and comparisons
to HF data as part of the BSRN operations. For this analysis, approximately 1500 minutes of data
are available.
        Figure 3 shows the relative responsivity of the other four pyrheliometers, compared to
CH1 940069, for the period January 20 to February 9. Dates where the HF was in use are not
included, so there is no duplication of results from figure 1. The other two CH1s show values
very close to unity, confirming that the linear calculation of temperature compensation represents
a significant improvement.

Figure 3:
                                          1.05

                                          1.04

                                          1.03

                                          1.02

                                          1.01
  Relative responsivity (to CH1 940069)




                                            1

                                          0.99

                                          0.98

                                          0.97

                                          0.96

                                          0.95

                                          0.94

                                          0.93

                                          0.92

                                          0.91

                                                 CH1 940074   NIP 33822   CH1 930017   NIP 33856




         Figure 4 shows the relative responsivity as a function of temperature, using CH1 940069
as the reference instrument. Again, the values near unity with no significant slope confirm the
CH1 temperature compensation adjustment. The results for the two NIPs are consistent with the
results in figure 2, which was based on the HF data. This strongly suggests that the NIP data
could be improved by applying additional temperature compensation, and that the difference
between the earlier NOAA calibrations and the current EC calibrations may be the result of
temperature rather than a change in the NIPs.

       In using CH1 940069 as the reference instrument, the new responsivities for the two NIPs
would be 7.54 and 7.86 (s/n 33822 and 33856 respectively), which are very close to the values of
7.56 and 7.91 (from the comparison with HF 20406) and the NOAA values of 7.51 and 7.86.
These results give increased confidence that the instruments are stable.
Figure 4:
                                          1.05


                                          1.04


                                          1.03
  Relative responsivity (to CH1 940069)




                                          1.02


                                          1.01


                                            1


                                          0.99


                                          0.98


                                          0.97


                                          0.96


                                          0.95
                                                 -30   -25   -20         -15            -10                -5               0   5   10
                                                                                 Temperature (°C)

                                                                   CH1 940074   NIP 33822     CH1 930017        NIP 33856

        The temperatures sensitivity of the two NIPs, as seen in figures 2 and 4, does suggest that
there is a drop in responsivity at colder temperatures. This could be significant for operations at
Alert, where significant periods of radiation data are collected at these lower temperatures. An
examination of this sensitivity in greater detail is probably worthwhile. NIP 33822 also seems to
show a maximum value around -10°C, with the relative responsivity decreasing both above and
below this temperature. It is impossible to determine whether difference between the NOAA
calibration and the EC calibration are due to temperature sensitivity without knowing the range
of temperatures covered by the NOAA calibration.

        Figure 5 shows the relationship between instrument temperature and air temperature, for
the points from the analysis using HF data. The instruments tend to be about 1-2° warmer than
air temperature. If additional temperature compensation calculations were to be performed using
air temperature instead of instrument temperature, this difference could be accounted for, but its
effect would be small.
Figure 5:
                                 0




                                 -5
  Instrument Temperature (°C)




                                -10




                                -15




                                -20




                                -25
                                      -30   -25     -20                 -15                  -10       -5   0
                                                                Air Temperature (°C)

                                                  CH1 940069   CH1 940074     NIP 33822   CH1 930017




Summary

        The results of calibrations in early 2007, at the Bratt’s Lake Observatory, on NIPs 33822
and 33856 given no indication that any changes in sensitivity have occurred in the three seasons
that the instruments have been in use at Alert. Small variations occur, within manufacturers’
specifications for the temperature compensation of the instruments. Knowledge of the
temperature range used for calibrations at NOAA prior to the 2004 deployment may help clarify
the minor differences between the NOAA and EC calibrations. Further knowledge of the
temperature sensitivity of the instruments may also assist in refining the field measurements, by
allowing additional temperature compensation calculations.

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:5
posted:9/16/2012
language:English
pages:7