CHAMBER FLUX MEASUREMENTS
CO2 and H2O fluxes were measured using a Licor 6400 photosynthesis system (Li-Cor Inc., Lincoln, Nebraska, USA) connected to a 1m x 1m plexiglass chamber in canopies dominated either by Salix pulchra or Betula nana shrub species. The height of the chamber varied depending on the height of the canopy being measured; chamber bases were constructed of PVC pipe to accomodate canopies with heights up to 125 cm. In addition to the plexiglass chamber, we also constructed a plexiglass "sleeve" that could extend the height of the rigid portion of the chamber by 0.25m.

To set up each chamber, a location was chosen where the base would be level enough to ensure a complete seal with the plexiglass chamber and shrub branches could be moved either in or out of the chamber without creating large gaps in the canopy inside the chamber. Branches were included within the chamber if they were rooted within the chamber and excluded otherwise. Once the base was in place, we drove hollow PVC pipe legs into the permafrost and inserted an aluminum frame with foam campermount tape along the top edge for the plexiglass chamber and/or sleeve to rest upon, creating an airtight seal. The aluminum frame had taped to it semi-transparent, plastic skirt which extended to the ground (+30cm). We sealed the skirt to the tundra by weighting the skirt with heavy chains, pushing them firmly into the moss layer where possible and adding additional plastic materials as needed to ensure a good seal. We screwed the LiCor custom chamber head attachment over the holes drilled into the plexiglass chamber, again sealing with a rubber gasket. The air in the chamber was mixed using 4-8 small fans (depending on chamber height) powered by a 12v battery.

At each plot we took measurements to create two light curves: one under direct light and one under diffuse light conditions. In order to determine the fraction of diffuse light, we used a DeltaT Beam Fraction Sensor (BF3, Delta-T Devices Ltd, Burwell Cambridge, UK) which quantifies the total irradianc and total diffuse light from which the diffuse light fraction (diffuse light/total light) can be calculated. For each day of flux measurements, the BF3 logged an instantaneous reading every 30-60 seconds set up on a leveled tripod at approximately 2 m above the ground. For the purpose of correlating the diffuse light fraction with each flux measurement, the LiCor 6400 and BF3 sensor were synchronized to read the same time (+/- 1 sec) at the start of each day.

Different light levels for both diffuse and direct light curves were achieved by taking measurements under a variety of conditions: ambient light (no manipulation), successive shading levels (covering the chamber with 1-5 fine mesh net cloths), and intercepting direct light with photographic diffuser panels, as well as reflecting light into the chamber to increase the amount of diffuse light with white photographic panels. When the diffuser panels were used, they were carefully positioned to intercept all direct light that would otherwise enter the chamber. Whenwhite reflector panels were used, they were positioned on the side of the chamber opposite the sun and angled towards the chamber so as to increase the amount of diffuse light entering the chamber (these were used in conjuction with the diffuser panels). For these 'artificial' diffuse light measurements, we did not diffuse the BF3 sensor, thus the diffuse fraction calculations during these flux measurements do not represent the light conditions in the chamber. After field tests of using the diffuser and reflector panels, we determined that the panels effectively block all direct light, and thus we assume the diffuse light fraction is greater than 0.7 for these measurements. At each light level a flux measurement lasted 45 - 60 secs in total, with CO2 and H2O concentrations in the chamber recorded by the LiCor 6400 every 2 secs. After each measurement we lifted the chamber until CO2 and H2O concentrations had stabilized at ambient levels. We made an effort to obtain a wide range of flux measurements for light levels between 0-1600, and used whatever chamber light treatments were needed to achieve that based on the ambient light conditions.

In addition to light measurements, we made at least three measurements in the dark for each day we took flux measurements. These were achieved by covering the chamber in an opaque tarpaulin cloth. These measurements represent the ecosystem respiration.

After each light curve we determined chamber volume by taking depth measurements from the top of the chamber base to the ground. We measured the chamber base depth with 36 measurements made at regular 20cm intervals determined by placing a 1m x1m plastic frame with a 20cm x 20cm string grid on top of the base. The volume determined by these depth measurements (chamber surface area*average depth) was added to the volume of the plexiglass chamber (and sleeve, as needed) . The surface area of the inside of the 1 m x 1 m plexiglass chamber was 0.8836m2.

For measurements with a linear change in CO2 (r2>0.97), NEP is calculated from the last 45 seconds of the measurement. [The first 12 seconds of data were always discarded as the LiCor 6400 bases its flux calculations on 10sec averages.] When abnormalities in CO2 slope were observed due to leaks, very small changes in CO2, or changes in light levels, certain portions of the measurement were discaded. CO2 slope was always taken from contiguous data points (i.e. points were never removed from the middle of the measurement period). Often at low light levels near the light compensation point, very small changes in CO2 resulted in measurements within the detection limits of the LiCor 6400, and large segments of the measurement had to be discarded to use a representative, linear segment of the data. All other variablels (H2O flux, pressure, chamber air temperature, CO2 concentration, PAR, PAR range) are calculated over the same time window as NEP. These variables were computed as the average over the course of the measurement.

Each flux measurement was placed in one of four categories: direct, diffuse, intermediate, or respiration. These categories are defined as follows:
Direct (sunny measurements): diffuse light fraction < 0.40
Diffuse (cloudy or diffused-light): diffuse light fraction > 0.70
Intermediate (partial sun/cloud): 0.40 < diffuse light fraction < 0.70
Respiration (dark): measurements taken in complete darkness; used for all light curve calculations

NDVI UNISPEC DATA
We measured NDVI on each flux a single channel Unipec spectral analysis system (PP Systems Inc, Amesbury, MA, USA). The unispec spectral analyser measures reflected light intensity in 256 portions of the visible spectrum from ~300nm to ~1100nm. A foreoptic cable transmits light reflected from the target to the instrument, a measurement scan lasts for ~10ms. Nine scans were measured in a regular grid for each of the flux plots. The end of the fibre optic was kept approximately 30cm vertically above the top of the canopy. The NDVI values were calculated using the equation below and then the NDVI values averaged together for each of the nine scans. Each scan was corrected for both incident radiation as well as sensor error. The incident radiation was accounted fo with a "reference scan" taken by holding the foreoptic cable vertically above a white, reference standard under the same light conditions as subsequent measurements. The scans were also corrected for with a "dark scan" taken with the sensor exposed to complete darkness (covering the sensor input with a black cloth) to account for the intrinsic error in the sensor itself. The program Multispec5.1.5.exe was used to compile and integrate the Unispec reflectance spectra from the raw target spectra.

LAI SUNSCAN DATA
The methods used to collect leaf area index (LAI) estimates at many heights within each canopy were the same used to collect PAR (see "SunScan_PAR_Data"). We measured the LAI using a DeltaT SunScan wand in conjuction with the BF3 sensor (Delta-T Devices Ltd, Burwell Cambridge, UK). The SunScan wand compares indident light readings measured on the BF3 sensor with the 64-PAR readings on the light wand to calculate LAI along the length of the 1m-long wand.

LAI was measured by inserting the SunScan wand as near to the ground as possible--typically ~5cm from the ground as the wand rested on top of moss--and then measured vertically every 15 cm with the last measurement being above the canopy. Measurements were taken from the side of the chamber or point frame opposite the sun at three locations under both direct (ambient) and diffuse light conditions . In many cases the replicates at each height were differentiated by row (1-3, or occasionaly 3-8 which correspond to the point frame pins). Typically diffuse light conditions were achieved by shading both the BF3 sensor and the shrub canopy with photographic diffuser panels. On occasion, measurements were taken during cloudy light conditions where the diffuse light fraction was greater than 0.7 and no diffuser panel was needed; on these occasions the direct and diffuse light estimates may have been taken in slightly different locations as they were taken at different times and the precise position of the SunScan wand could not be replicated exactly.

The SunScan wand measures PAR along 64 points along a 1-m horizontal profile. When sampling PAR within the shrub canopy, the data are from the raw output from each PAR sensor. When sampling LAI, there is an internal calculation performed though the Delta-T software that compares the reading above the canopy to the reading within the canopy, and takes into account the percent of absorbed PAR (assumed to be 0.85), and the ellipsoidal leaf angle distribution parameter (ELADP*) (assumed to be 1.0).

While the LAI data from each height in the chamber flux canopies are available, the data here are only from the lowest, ground-level measurements (~5cm) to represent the canopy LAI.

CALCULATIONS:
Fluxes are calculated from the slope of chamber CO2 (umol mol-1) [or H2O (mmol mol-1)] concentration against time.

NEP = rho x vol x dCdry/dt rho = P_av x 1000
SA R T

where NEP = net CO2 flux [umol m-2 s-1]
rho = air density [mol/m3]
P_av = pressure [kPa]
R = ideal gas consant 8.314 [J mol K-1]
T = temperature [K] = Temp_av [0c] + 273.
vol = chamber volume {m3}
dCdry/dt = slope of chamber water-dilution-corrected CO2 conc against time [umol mol-1 s-1]
SA = chamber surface area [m2] = 0.8836

For net fluxes, a negative flux represents CO2 uptake by vegetation, a positive flux represents CO2 loss from the ecosystem to the atmosphere.

RE = NEP during dark measurement
GEP = RE - NEP

NDVI = (RII-RI)
(RI + RII)

where RI = average reflectance from 570nm to 680nm
RII = average reflectance from 725nm to 1000nm.