Negative Offsets in Shortwave Radiometry (for V2.2)

This page addresses our adjustment to the SIROS and BSRN diffuse fluxes. Although this phenomenon is nothing new, we've noticed that nighttime fluxes reported by the upward-looking pyranometers are usually negative. Others have noticed that there are occasions when, during the daytime, the measured diffuse flux is lower than that calculated by a radiative transfer model which assumes nothing more than a Rayleigh atmosphere. (No clouds, no aerosol, etc...)

Eppley pyranometers, the kind used in the ARM program, output voltages which are dependent upon the "hot" and "cold" junctures of their thermopiles. The sensor is the hot juncture and the body of the instrument determines the temperature of the cold juncture. The difference in temperatures between the two junctures determines the voltage output, hence, the flux. At night, the domes of the instruments cool, and the "hot" juncture becomes colder than the "cold" juncture. So the voltage becomes negative. This longwave effect exists during the day also, but it is difficult to determine how it affects the reported unshaded pyranometer fluxes.

One of the ways you can actually correct for this during the day is to monitor the net IR signal (NETIR) produced at the pyrgeometer (LW instrument), and determine a relationship between this signal and the diffuse flux. It is safe to determine the relationship for the shaded pyranometer only, since the sun offers additional problems for the unshaded pyranometer. It is also safe to apply this relationship during the day, since we believe that the magnitude of the offset "fix" is smaller than it should be, because we are applying the effects of a longwave dome to that of a shortwave instrument.

NETIR is different from the net surface longwave flux found by subtracting the upwelling LW flux from the downwelling LW flux, as measured by the two pyrgeometers. Figure 1 illustrates this for 5 SIROS and 1 "BSRN" station, showing a net SFC LW that is higher than NETIR. The net SFC LW includes effects that are not present in NETIR. Because NETIR represents what is happening at the level of the instruments, it is the parameter of choice.

The net IR signal

The ARM method of getting a downwelling flux out of an up-looking pyrgeometer is

Flux = V * C1 + (sigma * Tc**4) + C2*sigma * (Tc**4 - Td**4)

which can be rewritten as

V*C1 = Flux - (sigma * Tc**4)  - C2*sigma * (Tc**4 - Td**4)

V*C1 is the net IR signal (NETIR), V being the thermopile voltage, and C1 the instrument sensitivity. Flux is the irradiance; C2 is the dome correction coefficient (4.0 for all ARM pyrgeometers); sigma, the Stefan-Boltzmann constant (5.67E-8); and Tc and Td, the body (case) and dome temperatures respectively.

The SIROS data stream available to the public provides the temperatures and flux, from which NETIR can be derived. The SIRS data stream (for CAGEX V3) provides the sensitivity as well as the thermopile voltage instead of the temperatures, so NETIR can still be found with the data available to the general public.

NETIR vs Reported Flux, Nighttime

The next three figures are as follows:

1. Figure 2: Nighttime pyrheliometer measurements vs NETIR for 5 SIROS and 1 "BSRN" station;
2. Figure 3: Nighttime pyranometer measurements vs NETIR for the same stations. These pyranometer measurements are from the pyranometer which is shaded during the day. The up-looking unshaded pyranometers show similar characteristics;
3. Figure 4: Nighttime pyranometer measurements from the downward-looking pyranometers for 5 SIROS stations.

In all three plots, the presence of rain, as reported by the SMOS instruments, is denoted by blue dots, while the rest of the scatter plot is red..

Figure 2 shows the relationship between negative fluxes reported by the pyrheliometer and the NETIR. It is evident that the negative fluxes are not a problem with the pyrheliometers, although there is a slight positive slope indicating that as NETIR increases, the reported flux increases.

Figure 3, the pyranometer vs NETIR, shows a relationship between the two parameters. The large blue plus sign indicates the origin (0,0). Ideally, we would like to see the slope line intersect with the origin. Obviously, this is not the case, except with the BSRN instrument. Our adjustment includes forcing the slope through the origin, by rotating it about the x and y centroids of the data. This ensures that there will be no adjustment when the NETIR signal is zero. However, it may cause a greater-than-needed adjustment for instances where NETIR is strongly negative. These plots are the important ones, from which we apply our adjustments.

Figure 4 shows the fluxes measured by the down-looking pyranometer. These are markedly different from those of the up-lookers. This is because the instrument is looking at the earth's surface, and the radiative cooling is not taking place. (The "BSRN" system does not have a down-looker). In these plots, the reported flux is zero when the NETIR signal is zero.

Method of Adjustment

These are the steps we take to adjust the diffuse flux to account for longwave effects:

1. Determine a slope for nighttime diffuse vs. NETIR. This is done for the entire data set for each station (each station has a different slope) (figure5)
2. Throw out all data greater than 2 standard deviations from the slope. (figure 6)
3. Calculate a new slope, without the thrown-out data. (figure 7)
4. Find the X and Y centroids (X and Y means).
5. Rotating about the centroids, force the slope through the origin. The new slope becomes Y-centroid/X-centroid. (figure 8)
6. Using the simple y=mx+b relationship (where b=0), the offset becomes m*x, where m is the slope and x is the net IR signal. This offset is then subtracted from the reported diffuse flux (both daytime and "nighttime" values. (figure 9)

Notes

This is a fairly simplistic procedure. A rigorous application of this adjustment should include other factors, such as wind speed and other meteorology. At this writing, we do not understand why the net IR signal does not equal zero when the diffuse equals zero. This may be a calibration issue, but we're not sure. Application of this procedure to the CARE (CERES/ARM Radiation Experiment) data set results in a slope that indeed comes very close to the origin.

There may be a better way to force the slope through the origin, but theoretically, we should not have to take that step. Our method of rotation may add too much flux to instances where there is significant longwave cooling.

There are a few things that could have made the adjustment better. One is to use pyrgeometer data from approximately 3-4 minutes before the pyranometer measurement. The lag is due to different time constants in the single-domed pyrgeometer as opposed to the double-domed pyranometer (Bruce Forgan, personal communication, Feb. 1999). We will look into this in following versions of CAGEX.

Hyett (1998) proposed a method in which ten days of data were analyzed at a time. For example for day X, she used days X-5 to X+5 to identify a regression (this is only one of the methods she used in her analysis). We used the entire 38-days of CAGEX, and lumped it into one regression. The advantage to limiting the analysis to a 10-day domain for each regression is that the possibility of a seasonal effect on the netIR signal is eliminated. For CAGEX Version 3 and beyond, we will look into using a sliding 10-day scale for analysis.

Results

Results of this adjustment are part of the SIROS and BSRN record for CAGEX V2.2. A whole host of plots can be found on the plot page (link below). Analysis of the results followed three paths - the first was examination of sub-Rayleigh behavior of the reported diffuse fluxes. CAGEX V2.1 shows instance where the measured diffuse flux is actually less than that calculated using the Fu-Liou code with no aerosol input. This sub-Rayleigh behavior disappeared entirely in V2.2.

We also compared the old and new diffuse fluxes with a diffuse flux derived from RAMS and the pyrheliometer. We subtracted the pyrheliometer direct horizontal flux from the RAMS total SW downwelling flux to come up with a "RAMS-NIP diffuse", for lack of a better term. Plots of these results are also available (check out the link below).

The third indication that this adjustment is a "good thing" (as Martha Stewart would say) is the improvement in the clear-sky bias between calculations and measurements. As you can see by comparing V2.1 and V2.2 results (SW aerosol sensitivity tables for V2.1 and V2.2) The clear-sky diffuse bias decreases from 30 W/m² to 14 W/m². We believe this adjustment is the primary reason for this improvement (there were many changes between V2.1 and V2.2).

Plots

The following types of plots are available. Use the above link for timeseries'. The scatter plots can be accessed by using the links below:

• Scatter plot, non-adjusted diffuse vs. adjusted diffuse.
• Scatter plot, clear-sky diffuse V2.1 and V2.2 - calculations vs measurements.
• Scatter plots - RAMS-NIP diffuse vs. shaded pyranometer diffuse. RAMS-NIP diffuse is derived by subtracting SIROS direct horizontal flux from RAMS total SW. Shaded pyranometer diffuse is SIROS. One plot uses adjusted diffuse, one plot uses non-adjusted diffuse.
• Time series page
• V2.1 and V2.2 surface SW diffuse fluxes and biases (calculations minus measurements)
• V2.1 and V2.2 surface SW insolation fluxes and biases (calc - meas)
• V2.1 and V2.2 diffuse fluxes and flux biases (calculations with no aerosols)
• V2.2 SIROS and RAMS/NIP diffuse - fluxes and differences for three sites.
• V2.2 SIROS and RAMS insolations - Pyranometer (PYR) as well as Pyrheliometer + shaded pyranometer (P+SP) fluxes and differences for three sites.
• RAMS/NIP and SIROS diffuse fluxes, all three sites on one plot.
• SIROS Pyrheliometer + shaded pyranometer (P+SP) insolation, all three sites on one plot.
• SIROS Pyranometer insolation (PYR), all three sites on one plot.
• RAMS insolation, all three sites on one plot.
• SIROS direct horizontal surface fluxes, all three sites on one plot.

Note - RAMS is available for a few days only. Insolations should be regarded as pyrheliometer + shaded pyranometer, unless otherwise specified (except in the case of RAMS, which is all-pyranometer)