Fluxes Over Snow Surfaces, Phase II (FLOSSII)

Looking east toward Medicine Bow mountains from FLOSS site.

Measurement challenges due to rime.


FLOSSII is the second phase of the FLOSS project, studying the surface meteorology of snow-covered rangeland in the North Park region of Colorado, near Walden.

If you reached this page from a search engine, click here to see the full report, with frames.

The first phase of that project is described in a separate document.

This document is a standard product of NCAR/ATD/RTF which gives an overview of the measurements taken using the Integrated Surface Flux Facility (ISFF) and conditions during the FLOSSII field experiment.

FLOSSII will operate in conjunction with the NASA Cold Land Processes Field Experiment (CLPX).


Links to other FLOSS Partipants


Near Walden, Colorado, in the Arapahoe Wildlife Refuge near the former Case Ranch. We thank the Refuge for their generous use of this location.

The Refuge also has supplied us with an aerial photo of the area around our site from a few years ago, which I've anotated with approximate locations of the FLOSS and FLOSSII sites.

We've produced a map; using Topo of the site, with tower locations for FLOSS01 and FLOSSII indicated with red dots. Note that the labels are to the NE of the corresponding dot -- not necessarily the closest.

I have written a set of driving directions for the convenience of the many visitors we expect to see! Also, see my contact list for our phone numbers, etc.


Between the periods of staffed observations, data will be taken in a minimal maintaince mode, with short service visits and remote data monitoring.

Measurement Sites

NCAR ISFF sensors were installed at three sites during FLOSSII.

34 meter walkup tower (stations 0 and 1)

Station numbering

In the data files, and on the plots, sensors on the tower or the nearby radiation stands are designated as either station #0 or station #1. Sensors which are unique to the tower are regarded as being at station 0. Sensors which correspond to similar sensors at the lake and sage sites, like the T/RH, sonic and fast hygrometer at 2 meters, and the radiation and soil sensors at the grass site, are designated as station 1.

sagebrush site (station 2)

lake site (station 3)

Site Locations

Site Latitude (N) Longitude (W) Elevation [approx, from map](m)
34m tower 40° 39' 32" 106° 19' 26" 2476.5
sagebrush 40° 40' 04" 106° 19' 08" 2475.5
lake 40° 39' 50" 106° 19' 40" 2474.5
Base trailer 40° 39' 39" 106° 19' 18" -


Here is the plan for the sensor array.

Sonic Anemometers

CSAT3 sonic anemometers were used all wind sensing for FLOSSII. Overall, their performance was excellent, with occasional spikes probably due to precipitation in the path or accumulating on the path. Sage and Lake sonics sampled at 20 samp/s, main tower sonics were at 60 samp/s. During the pre-operations phase, the (sage and lake) sonics were in "grab sample" mode. They later were changed to average 3 samples at 60 samp/s.

Specific notes:

Sonic tilt angles have been evaluated using all the data.

FLOSSII Sonic Tilt angles

Sensor w-bias (m/s) Lean (deg) LeanAz (deg) Plot
1m -0.028 0.8 122.7 plot
2m -0.018 0.4 -154.1 plot
5m -0.024 0.4 114.6 plot
10m: Before 18 Dec -0.014 1.2 132.2 plot
10m: 18 Dec-6 Jan -0.034 2.1 -155.2 plot
10m: After 6 Jan -0.034 1.2 142.2 plot
15m: Before 18 Feb -0.004 1.4 58.1 plot
15m: After 18 Feb -0.025 2.7 23.8 plot
20m -0.001 1.5 155.2 plot
30m 0.035 1.3 122.5 plot
sage 0.014 0.3 95.8 plot
lake 0.009 1.6 -85.5 plot

We attempted to level all of these sensors during installation by using an electronic level to adjust (boom) guy wire tension and then attaching the anemometer. This should have been accurate to about 0.2 deg. Thus, we assume that these above "tilts" actually indicate real air motion due to local topography. Thus, none of these tilts have been removed from the data. If the user is interested in applying our rotation procedure, see our tilt correction page.

Krypton Hygrometers

These were our only sensor for measuring water vapor fluxes, but they did not perform well. The Lake and Sage sensors were connected (using the PMB-version serializers) on 10/10. On 11/18 at sage, 11/19 at lake, and all the time (after about 11/20), all kryptons (but one) were set to a pathlength of 2.6cm, following guidance from Arian van Dyke's paper. Previously, all ISFF kryptons had pathlengths of about 1.2cm.

Here is a list of serial numbers:

2m	1389 (2.6cm)
10m	1391 (2.6cm)
20m	1392 (2.6cm) changed to 1390 (2.0cm) on 6 Jan
30m	1394 (2.6cm)
sage	1133 (1.2cm) changed to 1395 (2.6cm) on 18 Nov
lake	1393 (1.242cm) changed to 1397 (2.6cm) on 18 Nov

All kryptons had extensive periods of bad data, especially on cold (below -15C) nights. We now suspect that the cause of this is fog. To test the theory that this was due to frost build-up, we attached a hot-air blower to main.2m on 12/16 (not operational until 12/18), which helped a bit on 12/22-24, but not as much as expected. The pump on this blower seized up the next day. On the morning of 1/7, this problem recurred and was observed to be related to the presence of ice fog. Whether fog was always the reason is still unknown. (The fog on 1/7 was accompanied by extensive "feather" rime which was visible on the Axis camera images, but the camera didn't show this rime during other bad krypton events.) Tests of 3 kryptons (both head and electronics) in the lab showed normal operation down to -40 C, so this problem is not simply a failure from low temperatures.

We will have removed the majority of obviously bad data by using the criteria: |kh2o-H2O]>1.8 g/m3 or kh2oV<50mV. Some bad data will slip through this filter -- especially during the transitions between rimed and ice-free conditions.

Also, the 20m krypton showed a diurnal variation not seen by other kryptons. The onset appears to be related to incoming short-wave radiation, but this problem continues into darkness. We saw a hint of similar behavior in FLOSS01 that appeared to be caused by corrosion in the electronics box that was fixed during IHOP02. The 20m sensor was replaced on 1/6.

Carbon Dioxide

We deployed our LiCor 6251 (closed-path IRGA) with an inlet at 2m at the main tower to measure CO2 fluxes close to the surface. It was installed 12/16, and valve control (for the gas calibration system) started 12/18. Only a Lo-Cal cylinder was attached initially, since one gas regulator was bad. We installed a new regulator 1/6 and got Hi-Cal working. On 1/7, we swapped Lo-cal cylinders since the pressure was somewhat low. On 1/8, the calibration cycles were shortened to 30 seconds (down from 2 minutes) to extend the life of the calibration gases. Due to an apparent slight leak in the LoCal plumbing, it was necessary to change the LoCal cylinder 3 times during the experiment, whereas only one HiCal cylinder was used throughout (see plot).

For the first time, we monitored pressure in the line close to the sample cell. It read about 10% lower than ambient pressure, a value which probably is typical of most of our deployments -- certainly FLOSS01. However, on 1/6 the inlet flow rate was changed from 6 to 7 lpm, which caused this pressure sensor to go offscale. We lowered the flow rate back to 6 on the next visit. During the period when it ran offscale, Plicor appeared to be approximately 82% of P, so we will synthesize a corrected Plicor from P at stn1 (itself a derived quantity that implements a correction for sensor temperature -- see "Barometer" below). (The top and bottom magenta lines in this plot show the uncorrected and corrected Plicor values.)

We have investigated data from the calibration periods and find that most of the data are well fit by voltage linearly related to CO2 density, computed using the calibration gas mixing ratio and the sample cell temperature and pressure (see plot). The data from the first 3 weeks (before the Hi-Cal regulator was installed) appear to have an offset from the rest of the data. Also, the Lo-Cal data at the end of the program are more scattered -- even from earlier data using the same Lo-Cal cylinder. The cause of both of these effects is as yet unknown. Thus, we have applied the regression from the middle 2 months (7 Jan - 21 Mar) to the entire period, noting that only the sensor gain is important for the flux measurements desired for FLOSS.

To obtain a slightly reduced residual, we added an extra term to the fit: V ~ a1 + a2*m*rho + a3*P. With this fit, we compute CO2 mixing ratios during the calibration periods. Our plot of the mixing ratios thus computed versus those of the cylinders shows that there still is a disturbing amount of scatter. However, the barely-discernable horizontal lines on this plot indicate that the first and third quartiles are generally within a range of +/-2 ppmV of the calibration cylinder values (except for those from the 323.26 ppmV cylinder used at the beginning).

As a further check of these calibrations, we show data from "supplementary calibrations" done on 20 and 27 Feb. During these calibrations, gas from other CO2 calibration cylinders were introduced to the Li6251 at the inlet, rather than at the analyzer. Both of these calibrations were performed (on purpose?) at the time of automatic calibration cycles. A plot of these supplementary calibrations shows rather poor absolute values. (In this plot, the horizontal lines represent the value of the automatic calibration cylinders, the dots are the value of the supplementary calibration cylinder and indicate the supplementary calibration time, and the black line is the LiCor output calibrated using the fit described above.) The 20 Feb case shows that the above calibration produces values which are about 9ppmV lower than the supplementary cals but also about 5-7ppmV higher than the automatic cals. However, the 27 Feb case shows the supplementary cals now 12ppmV lower, with the automatic cals nearly matched. For such differences to exist between automatic and supplemental cals, there must be a response of our system to pressure or perhaps temperature variations that we have not characterized. (Even though the automatic cals themselves go through a wide range of pressures and temperatures.) Clearly, it will take more work to interpret the usefulness of these supplemental cals!

CO2 data have NOT yet been included in the final data.


These generally worked, though main tower profiles were often not strictly monotonic. Some of these differences may be real (since fetch changes with height), however at least one swapping test showed a sensor problem. Nevertheless, all sensors agreed with the laboratory reference to better than 0.1 C in post-calibrations, suggesting a sensor exposure problem. We ran a side-by-side intercomparison for 2 weeks on the FLAB roof immediately following the field phase to investigate these difference further (see table, below). As with the lab tests, these showed differences of at most 0.1 C, and more typically 0.03 C. (RH differences were 2.5% max, 1% typical.) All sensors were operated as they had been at the time of tear-down, except for the replacement of a bad fan on the 20m sensor. Many inlet bells also were noted as being broken (due to ice during operations). Thus, the obvious potential problems of poor fan aspiration and dirty Gortex filters wouldn't appear to have caused these differences, though the fans <could> have had less electrical power on the tower. The laboratory-determined differences listed in the below table have been removed from the data.

Note also that these sensors ingested ice crystals during riming events (and possibly snow) -- see logbook, e.g., comments 169 and 173 and photos. No attempt will be made to eliminate these data.

There were several cases that fans were noted as not working (see below). During the days before these times, temperatures could be higher than ambient during the day and lower than ambient at night, due to radiative heat transfer -- especially in light winds. These cases sometimes are apparent in the temperature profiles or in comparisons with tc from a nearby sonic anemometer. We have filtered these data by removing data with |Tc(z)-tc(z)|>0.5C during the period when the fans were logged as having problems. (Tc is the acoustic virtual temperature computed from T, RH, and P.) Note that this is a pretty liberal criteria which probably retains some bad data, though it also could reject good data when the sonic data are bad.

Specific problems:

Roof-top test results:

TestID	S/N	50Y		FLOSS Location	Roof:dT	Bath:dT	Roof:dRH Bath:dRH        Notes
				beg     end     degC    degC     %        %
a	001	T3510005	-	0.5	+0.059	+0.10	-1.6	-3.0		bell broken
b	206	R4910001	0.5	1	-REF-	+0.01	-REF-	 0.0		bell broken
c	104	T1140031	-	2	-0.006	+0.07	-1.0	-2.0		bell broken
d	701	T3510004	-	5	-0.112	-0.05	-1.2	-1.5		bell broken
e	702	T2620019	5	10	-0.060	-0.07	-0.1     0.0
f	007	T0820012	15	same	+0.011	+0.03v	-0.3    -0.5
g	704	09N001		20	same	+0.024	+0.00	-0.4	-1.5		new fan
h	008	U5020029	25	same	+0.051	 0.00	 0.0    -0.5
i	003	T1130007	30	same	-0.008	+0.03	+0.2     0.0
j	101	P1110024	sage	same	-		-			no inlet
k	502	R4910005	lake	same	-		-			no inlet
l	703	P1110027	1	-	+0.007	-0.02	-0.6    -0.5
m	004	U5020030	2	-	-0.012	 0.00	-2.3	-3.5		bell broken during tear-down
n	204	P4920040	10	-	-0.034	+0.01v	+1.0     0.0

i	003	T1130007	30	same						bad until end
j	206	R4910001	0.5	1	-REF-	+0.01	-REF-    0.0
k	101	P1110024	sage	same	+0.015	+0.01	-0.2     0.0
l	502	R4910005	lake	same	-0.025	-0.01	-1.0    -1.5

-	005	P1110023	spare	spare	-	+0.01	-	-1.0


There were a lot of radiometers, since this experiment tried to repeat (and extend) the longwave-radiative flux divergence study first done during CASES99, in addition to measuring all 4 components to know the surface energy balance at each site.

All upward-looking sensors were set up for forced ventillation, though it was discovered that electrical noise from the fans caused noise in the Tsoil measurements (same problem as in IHOP02). Thus, the fans were unplugged 12/4?-12/18 at Lake and Sage. (The fans were left on at the main tower site.) While they were unplugged, snow could (and did at least sometime) accumulate to cover the domes. This problem was "fixed" on 12/13-18 by both reprogramming the loggers to delay before sampling the Tsoil probes and by adding a capacitor in parallel with the fans to reduce the amount of electrical noise. After these fixes, Tsoil looked okay at Lake and Sage, though the tower sensors still show a slight amount of noise.

Net radiometers were used at all 3 sites. They were the only radiometers at Lake and Sage from 10/11 - 11/18. However, during this period, a logger program problem prevented data from being acquired until 11/18-19. All Rnet values should be normal (albeit, sometimes with ice cover) after then.

Upward and downward pyranometers were deployed at all 3 sites starting 11/18-20 using Kipp and Zonen radiometers. All appear to have operated normally.

Upward and downward pyrgeometers were deployed at all 3 sites starting 11/18-20 using Epply sensors. In addition, downward-looking pyrgeometers were deployed over two other surface types (mostly bare ground (12/17) and greasewood (1/7)) near the main tower to extend the lower sampling. An upward and a downward-looking sensor were mounted at 4m over this bare patch (also 12/17) as a second level in the flux-divergence array. A pair (for redundancy) of upward and downward-looking sensors were mounted at 25m (12/17, though trolley not extended until 12/18) as the third level of the flux divergence array.

Note that the "bare" site was approximately north of the main tower and thus was shaded by it from direct sun about noon every day. This wasn't considered a problem since the long-wave flux divergence sub-experiment only needs data at night.

All of these pyrgeometers were carefully calibrated in our oil bath, so that accuracies of <1 W/m2 could be obtained. However, Tdome-Tcase from the 25m and 4m.in.bare pyrgeometers were odd after setup. A wiring problem in the logger was discovered and fixed on 1/27. This cable swap also caused the wrong calibration coefficients to be applied to these signals. Our S-functions dat.Tdome and dat.Tcase have been changed to correct this problem. The 4m.in.bare "problem" appears to have corrected itself by January.

FLOSSII Pyrgeometers

Variable Name Station S/N B * (lab) B * (roof) f(%) * (roof) Notes
Rlw.in 1 (grass site) 26416 2.9 2.4 0.9  
Rlw.in 2 (sage) 26417 2.6 2.1 1.3  
Rlw.in 3 (lake) 29137 1.7 - - Faulty case temperature (Tcase). Removed Dec 4.
Rlw.out 1 (grass site) 29260 3.4 2.9 1.1  
Rlw.out 2 (sage) 27907 3.5 2.5 0.2  
Rlw.out 3 (lake) 29136 2.3 1.8 1.9  
Rlw.in.4m 0 (bare site) 31981 3.6 2.1 1.1 Installed Dec 17
Rlw.out.4m 0 (bare site) 31980 3.9 1.9 1.2 Installed Dec 17
Rlw.out.bare 0 (bare) 31976 3.0 2.5 0.5 Installed Dec 17
Rlw.out.grease 0 (greasewood) 31978 3.5 3.0 0.9 Connected Jan 7
Rlw.in.a.25m 0 (tower) 31975 2.9 2.4 0.7 Installed Dec 17
Rlw.in.b.25m 0 (tower) 31979 3.2 2.7 0.8 Installed Dec 17.
Removed Jan 7, to check questionable case and dome temps.
Reinstalled Jan 27, 15:30 MST
Rlw.out.a.25m 0 (tower) 31974 3.1 2.6 0.5 Installed Dec 17.
Removed Jan 7, to check questionable case and dome temps.
Reinstalled Jan 27, 15:30 MST
Rlw.out.b.25m 0 (tower) 31977 2.4 1.9 1.2 Installed Dec 17

* The above B and f values are used in the calculation of the long wave radiation (Rlw) from the pyrgeometer thermopile value (Rpile) and the case and dome temperatures:

Rlw = Rpile + SB * ( Tcase^4 + B * ( Tcase^4 - Tdome^4 ) ) + f * Rsw

where SB is the Stefan-Boltzmann constant, 5.67e-08 W/m^2/degK^4.

At the end of the project, we ran a side-by-side comparison of all of the pyrgeometers on the roof of FLAB. During this comparison, we made a set of shading tests to evaluate the response to short-wave radiation. Several of the short-wave tests did not make sense (show an inverse top-hat response when shaded) using the above (laboratory-derived) B values. It was possible to obtain top-hat response for all sensors by making the B values smaller. However, using the set of B and f values thus obtained, there still are differences of up to 5 (typically 2) W/m2. (We note that even larger differences were seen during periods when the sensors weren't cleaned daily.) The QC-corrected data use the coefficients from the rooftop tests. We will leave any additional corrections to the user.


There were lots of soil measurements for FLOSSII because we wanted to sample the heterogeneous surface. Each of the 3 main sites had 3 subplots with a complete set of sensors (see diagram). Each subplot had a 3-level temperature profile (2.5, 5, and 10cm), 2-level moisture profile (5 and 10cm), and 1 heat flux measurement (10cm). All of these sites were installed 10/4-5 to give them time to settle in. (They were "set" by pouring water over the site after installation.) In addition, each of the main sites had one thermal properties probe installed 12/16-18 to determine heat capacity and thermal conductivity. Many of the sensors to implement this spatial sampling have been provided by the NCAR Hydrometeorological Land and Water Initiative (see here for more information). Note that the installation in October was in normal soil, but the December installations were in frozen ground. It appears that the TP01 sensors installed in the frozen ground did not couple well to the soil until a thaw that occured 14 Jan.

The soil temperature measurements were good, with the exception of the fan noise problem noted above in the radiometer section. Also, Tsoil.a.2.5cm at Sage had extreme variations. This could have been due to the sensor being exposed to air. However, subplot "a" was directly underneath a sage bush, so the dirt around this sensor could have been heated by the plant material (which is black). We decided to leave this sensor alone and its data have been retained in the final data set.

The soil heat flux probes were mostly okay, with the following notes:

All of the bad data due to these cable and sensor problems have been deleted.

The soil moisture probes (Ech2o) all reported values, with the following notes:

However, an e-mail exchange that we had with the manufacturer after the project explained that these probes (and others that depend on the electrical properties of the soil, including our CS615 and TRIME probes) do not detect frozen water. Thus, MOST OF OUR SOIL MOISTURE MEASUREMENTS ARE NOT VALID. The data have been edited to remove all Qsoil measurements when the corresponding Tsoil is below zero. This removes all but the last few weeks (spring thaw) of data at the Sage and Lake sites. (The grass cover at Grass insulated the soil so it remained frozen through the end of the project.) We made a few gravimetric measurements at each subplot which also should be valid. A plot shows all of the good soil moisture measurements. In this plot, rows correspond to the grass, sage, and lake sites (from top to bottom) and subplots a, b, and c (left to right); Ech2o measurements are lines (blue=10cm; magenta=5cm) and gravimetric samples are dots, colored to correspond to the Ech2o locations when appropriate. Obviously, not much data remain. It should be possible to correlate the TP01 heat capacity values (see below) with soil moisture, though these measurements were not made in the same subplots. (The TP01 values are shown in the above plot as green lines.) This task will be left to the user.

In addition, we deployed a soil thermal property probe (Hukseflux TP01) at each site. This is a new measurement for ISFF, but should provide a better estimate of soil heat capacity than we have made for earlier studies (using mass-weighted bulk thermal properties and crude estimates of soil composition). Also, these probes measure the soil thermal conductivity, which is needed to correct our soil heat flux measurements. Unfortunately, we did not obtain these probes until the soil was frozen and we also had some wiring errors (though little data were lost because of the wiring since the probes took a long time to settle). Thus, a complete data set is not available until 1/27. It appears that these probes needed to settle in the frozen ground -- they were installed with a slight (<1mm) air gap with the soil. The data appear to show that none of the probes were really settled until a soil warming period at the end of January (see plot). Additionally, the signals from the probes at lake and sage had significant noise, caused by the voltage controller of the PAM station. We fixed this noise with a voltage regulator on 2/22, though the data before appear to usable.


These worked as expected. Main site's PTB100 had a second-order temperature effect, causing errors on the order of 4mb in the cold temperatures. (See logbook comment 44.) This has been corrected in the data using the fit: Perr(mb) = 3.4 - 0.126*Tbox.cosmos(C), where these coefficients were calculated using the data from the Sage site, adjusted by -0.1mb to account for the 1m height difference. Hoses between the pressure ports and sensor were not connected at Lake and Sage until 10/10. Data from this period have not been removed since a slightly bad pressure should be more useful than no pressure.


There was a weighing precipitation gauge and an ultrasonic snow depth gauge at the main tower site (near grass radiation stand). Both were installed on the 11/21 visit. The weighing gauge did not report values between 11/21 and 12/15 for an unknown reason. It appeared to work later. Its response was tested on 1/7 and it was only one "tip" off (43 reported; 44 expected), so the data should be okay. These tips have been removed from the data.

The snow depth gauge was noisy since it was deployed, similiar to its behavior during FLOSS01. Thus, the data probably are useless, even though the mean value is nearly correct. We moved it on 1/6 to a location which shouldn't see interference from the radiometer stand legs, but the noise didn't change. Nevertheless, these data are included in the final data set since they appear to be a reasonable indication of when it was snowing (when the signal has large variations and often goes to zero)!

As mentioned above, the krypton hygrometers appear to be a good indicator of fog!

Data System Effects

The October and November data were gathered exclusively through the GOES satellites. Our recent experience with data taken this way is that about 10% of the data have errors. We have eliminated the data when checksums were not correct.

FLOSS2 Field Logbook

A computer-readable field logbook of comments noted by NCAR and other personnel is available in read-only html form and updated every day.


Main Tower
Oct 1, 2002 Nov 20, 2002 Nov 21, 2002 Dec 4, 2002 Jan 27, 2003 IOP: Feb 17 - Mar 7
Ice Detectors

Sage Site
Oct 10, 2002 Nov 20, 2002 Jan 7, 2003 Jan 27, 2003 IOP: Feb 17 - Mar 7 TCAM test: Mar 26-27

Lake Site
Oct 10, 2002 Nov 20, 2002 Dec 4, 2002 Jan 27, 2003 IOP: Feb 17 - Mar 7
Univ. of Wyoming KingAir Wildlife Misc

Photos from Tower

An Axis 2420 network camera is installed on the tower at about 28 meters, looking west. Photos of VGA resolution (704x480) were saved in JPEG form every 15 minutes from 05:00 to 20:00 local time (MST=GMT-7). Contact ISFF for these data. Photos at 08:00, 12:00 and 16:00 local time each day are available online.

Note that the exposures of these images vary considerably -- even between the 15-minute shots. We highly recommend examining the entire set of images for any serious use of these images.

Daily QC Plots

Preliminary time series, profile and spectra plots are available for review of data quality. The time series plots are of the 5-minute statistics, covering 48 hours, centered on local noon of each day. Hourly profiles from the 30 meter tower are plotted from 5-minute or 30-minute means, as labeled. Spectra plots are made from 30-minute periods of high-rate sensor data, starting at the stated time.

FLOSSII Data Access

The NCAR FLOSSII data is available for download in the following forms: [an error occurred while processing this directive]
This page was prepared by Steven Oncley, NCAR Research Technology Facility