Guidance for design of ISCAT ISFF array

Introduction

We have one primary goal for this experiment -- to determine the surface flux of NO as the product of eddy-correlation measurements of the temperature flux w't' and a Modified Bowen Ratio (MBR). The MBR is determined from the ratio of the difference in NO concentration to the difference in temperature measured at two identical heights. This method assumes:

Similarity of transport of temperature and NO. This assumption is not exactly valid for two reasons. Temperature contributes to buoyancy and thus is not a totally passive scalar quantity. Even more important, NO is photochemically reactive and is not conserved in the presence of sunlight (which always will exist in summer at the South Pole!) The affect of this reactivity has not been estimated for this document.
That turbulence is the dominant transport mechanism, i.e. that the measurements are taken in the Planetary Boundary Layer (PBL). We have been told that this layer is only 3-6m high in shelf regions of Antarctica, but the only numbers we've know of for the South Pole are 50 - 1000m.

Lowest level, z1

Historically, micrometerological measurements have avoided the "roughness sublayer", typically estimated at 100 zo, where zo is the roughness length. From photos that we've seen, the dominant roughness elements are the sastrugi, which appear to be ~30cm in height. Over land, zo would be estimated as 10% of this value or 3cm. However, we expect snow to be more aerodynamic. Thus, I estimate zo = 1cm. I note that Tom processed Dave Fitzjarrald's data with zo = 0.014cm.

Due to practical considerations (such as the ability to define what the measurement height is), we have chosen the lowest level as 50cm. This would be 50 zo, which might be within the roughness sublayer. However, it is not obvious to Don or me that we have to avoid this region strictly when using the MBR method.

Temperature profile

We are deploying a 6-level temperature profile for a secondary goal of determining the height of the PBL. We have chosen to cover the full range of the existing 22m tower with this profile in geometric steps to maximize our ability to determine this height. With a lowest level of 0.5m heights with a constant ratio are: 0.50, 1.07, 2.27, 4.84, 10.32, and 22.00m.

Scalar Gradient

The biggest concern we have is that the NO gradient is measurable to sufficient resolution to be able to resolve the surface flux. From Marty Buhr, we have the 1-minute measurement precision for NO of 3e-5 umole/m3. From Tom's analysis of Dave Fitzjarrald's data, we have a median value of the resistance R of 37 s/m for z1=0.50m and z2=4.84m. Thus, half of the time, we could resolve a flux of (3e-5 umole/m3)/(37 s/m) * (86400 s/day) = 0.07 umole/m2/day. A comparison plot with several pairs of heights shows that moving the upper measurement to 10.32 m would improve the flux resolution only to 0.06 umole/m2/day.

For reference, the flux estimate by Davis et al (1999) was 0.12-0.48 umole/m2/day and by Jones et al (2000) was 0.012 umoles/m2/day. Thus, the Davis estimates should be measurable, but the Jones estimates would not.

Upper MBR level, z2

This height (z2) will be a tradeoff between several factors. Due to the reactivity of NO and the possibly shallow PBL depths, z2 should be as low as possible. However, with a fixed NO analyzer resolution, z2 should be as high as possible to maximize the concentration difference between measurements at z2 and z1.

Sonic anemometer placement

We need one sonic anemometer to measure w't' for use in the MBR method. This measurement typically would be made at the geometric mean of the gradient heights, z1 and z2. However, it isn't obvious to us that this is a requirement. The MBR method will generate an NO flux estimate at the height of the sonic anemometer. Thus, this height should be chosen for optimal performance of the anemometer.

A sonic anemometer averages over an acoustic path which filters turbulent eddies on the order of the path length and smaller. Since these eddies scale with height, the recommended measurement height is at least 20 times the path length. For our ATI K-probes, this is 20 * 0.15m = 3.0m. For a shallow PBL, we might be willing to accept some attenuation of high frequencies in order to be assured that our measurements were continously within the PBL. It should be noted that, in the expected stable conditions at the South Pole, the frequencies of turbulence will be higher than conditions when the "20 times" rule is used.

We are deploying a second anemometer to allow use to (linearly) extrapolate the w't' flux measurements to the surface. Again, this extrapolation will only be reasonable if both measurements are made within the PBL. However, if the sensors are placed too close together, most of the difference in their measurements may be due to sampling errors. Having the second anemometer be twice the height of the first should be close to optimal.

We also are restricted in where we can practically mount the sonic anemometers on the tower with our mounting hardware. Tom lists our choices as 3' increments, starting at 2.5', i.e. 2.5', 5.5', 8.5', etc.

Conclusion

We recommend a geometric profile of temperature sensors from 0.50m to 22m. The lower NO inlet should be at this lowest height. The upper NO inlet initially should be placed at 4.84m. If the NO instrument resolution was sufficient, it could be lowered to 2.27m. If the PBL appears to be above 10m (by examining previous temperature profile data) and the NO instrument resolution was marginal, the upper inlet could be moved to 10.32m.

I suggest placing the sonic anemometers at the arithmetic means of the adjacent temperature profile heights, starting with a height near 3m. This places the lower anemometer at 11.5' = 3.51m and the upper one at 23.5' = 7.16m.