Biogeosciences Analyzing the major drivers of NEE in a Mediterranean alpine shrubland

Two years of continuous measurements of net ecosystem exchange (NEE) using the eddy covariance technique were made over a Mediterranean alpine shrubland. This ecosystem was found to be a net source of CO 2 (+ 52± 7 g C m−2 and + 48± 7 g C m−2 for 2007 and 2008) during the two-year study period. To understand the reasons underlying this net release of CO 2 into the atmosphere, we analysed the drivers of seasonal variability in NEE over these two years. We observed that the soil water availability – driven by the precipitation pattern – and the photosynthetic photon flux density (PPFD) are the key factors for understanding both the carbon sequestration potential and the duration of the photosynthetic period during the growing season. Finally, the effects of the self-heating correction to CO 2 and H2O fluxes measured with the open-path infrared gas analyser were evaluated. Applying the correction turned the annual CO2 budget in 2007 from a sink ( − 135± 7 g C m−2) to a source (+ 52± 7 g C m−2). The magnitude of this change is larger than reported previously and is shown to be due to the low air density and cold temperatures at this high elevation study site. Correspondence to: A. S. Kowalski (andyk@ugr.es)


Introduction
The concentration of carbon dioxide and other greenhouse gases in the atmosphere has been increasing since the industrial revolution, and is believed to be causing global surface temperatures to rise (Forster et al., 2007).Global warming is a serious concern as it seems to be driving global climate change (IPCC, 2007).The consequences of climate change are far from being completely quantified, and of particular concern is its impact on terrestrial ecosystems.
In recent decades, the eddy covariance technique (Baldocchi, 2003) has emerged as one of the most reliable techniques for quantifying mass and energy exchange.The eddy covariance technique is currently applied worldwide covering many different terrestrial ecosystem types with the aim of understanding the positive or negative feedbacks that climate change will have on Earth's surface.These sites have been grouped into different continental networks assessing net ecosystem exchange (NEE) such as EUROFLUX (Aubinet et al., 2000), AmeriFlux (Pryor and Barthelmie, 1999), and ChinaFLUX (Yu et al., 2006), forming the global FLUXNET (Baldocchi et al., 2001).Nonetheless, gaps in knowledge regarding some ecosystems still lead to uncertainties in global assessments.
In this regard little is known about high altitude sites (Schimel et al., 2002).The mountain and alpine regions of the world cover 10% of the Earth's terrestrial surface (Spehn and Körner, 2005).Understanding the patterns and drivers of NEE in these environments is very important for accurate determinations of the global carbon budget and for creating new Published by Copernicus Publications on behalf of the European Geosciences Union.B. R. Reverter et al.: Analyzing major drivers of NEE in a Mediterranean alpine shrubland perspectives in policymaking (Mountain Research Initiative; Becker and Bugmann, 2001).However, there is scarce data available to parameterize or validate models on alpine shrublands.
The effect of climate change on the carbon balance of high altitude ecosystems is a pressing subject of debate (Becker and Bugmann, 2001).As global temperatures rise, an earlier snow melt results in longer growing seasons which may lead to increased carbon assimilation (Lafleur and Humphreys, 2007).However, this increase in the growing season may result in reduced carbon sequestration, as a result of water limitations (Hu et al., 2009).On the other hand, permafrost melt in the tundra, a consequence of rising temperatures, is enhancing soil microbial respiration to a larger degree than photosynthetic carbon assimilation (Oechel et al., 1993).Understanding the complexity of biosphere-atmosphere interactions and the drivers of seasonal changes in NEE in coldlimited, high altitude ecosystems is far from resolved, particularly for high-altitude shrublands, where eddy covariance measurements are still lacking.
Another factor adding uncertainty to NEE estimates of cold ecosystems is related to the anomalous downward CO 2 fluxes observed during off-seasons (Amiro et al., 2006a, b;Grelle and Burba, 2007;Ono et al., 2008;Lafleur and Humphreys, 2007) or during snow covered periods when assessed by open-path eddy covariance (Skinner, 2007;Lafleur and Humphreys, 2007).Small photosynthesis rates (< 0.1 µmol m −2 s −1 ) have been observed over snowcovered vegetation (Starr and Oberbauer, 2003), if the snowpack is not too thick.However, the contribution of soil respiration through the snow cracks (soil temperatures are still high enough below the snowpack) should overcome any small uptake by photosynthesis (Musselman et al., 2005;Monson et al., 2006a).The recently-published Burba correction (Burba et al., 2008) rectifies such apparent net CO 2 uptake.The relevance of the Burba correction for the accurate determination of the carbon balance in cold ecosystems remains controversial.It has been experimentally observed to be negligible at some sites (Haslwanter et al., 2009), but not elsewhere where caution has been urged in its use (Blanken et al., 2009).
In this study we present two years (2007 and 2008) of CO 2 fluxes over a Mediterranean alpine shrubland.The general goals of this paper are: a) to identify the main variables driving the annual behaviour of NEE, b) to examine, focusing mainly on 2007, the explanations for seasonal and daily changes in NEE and c) to assess the role of the Burba correction for accurate measurements of the carbon balance.

Site description
The area of study is situated within an endorheic basin (2300 m a.s.l.) belonging to the Sierra Nevada mountain range, in the southeast of Spain (37 • 05 N 2 • 57 W), 39 km from the Mediterranean coast.The basin is an ancient cirque stemmed by a moraine and subsequently remodeled in the glacial age (Sánchez et al., 1988).Mean annual precipitation is 800 mm and the mean annual temperature 5.5 • C (Aguilar et al., 1986).The Sierra Nevada mountain range does not belong to the permafrost zone, however, certain processes occur that are also quite relevant to tundra: fluted and polygonal soils as a consequence of frosting and defrosting of the soil, and pipkrake processes (Gil-de-Carrasco et al., 1997).Furthermore, Prieto-Fernández (1975) describes vegetation species found in Sierra Nevada that are typical of arctic tundra.The site is usually snow covered from December to March.The subsequent snow melt causes water to flow across the basin and accumulate at the bottom creating a small lagoon that usually remains for approximately one month during the onset of the growing season.The flux tower is situated on a gentle slope separated from the lagoon by a fetch exceeding 150 m.
Festuca indigesta and Cytisus purgans are the dominant plant species, reaching to 20 cm height and homogeneously covering 45% of the ground.A remaining 35% corresponds to litter and 20% is bare soil.The presence of other plant species is intermittent throughout the year, and very sparse.Measurements of leaf area index (LAI), assessed with destructive methods, provided an average value of 1.23 m 2 /m 2 which is approximately constant over the seasons.
The soil is composed of sand (56%), silt (30%) and clay (14%) with a bulk density of 1110 kg m −3 , determined in situ using soil cores.The permanent wilting point was estimated at a volumetric water content of 2.6%.Soil organic matter ranged from 3.9% to 4.6% in the area around the flux tower.Soil analysis were conducted following recommendations from the American Society of Agronomy and Soil Science Society of America (Klute, 1986;Sparks, 1996).

EC and meteorological and soil measurements
Carbon dioxide and water vapor densities along with the 3 components of wind speed were measured at 20 Hz using an open-path infrared gas analyzer (IRGA, Li-7500, Lincoln, NE, USA) tilted ca. 12 • from the vertical (toward the north) and a sonic anemometer (USA-1, METEK, Elmshorn, Germany), mounted on a tower at 2.25 m above ground.Data were saved as 5-s averages, variances and covariances by a logger (METEK, Elmshorn, Germany) and subsequently converted to half-hour bases following Reynolds' rules.We applied 2-D coordinate rotations (McMillen, 1988;Kowalski et al., 1997), density corrections (Webb et al., 1980) Richardson et al., 2006).Resulting standard deviations were used to add artificial noise following a Laplace distribution.Every R 2 was greater than 0.83. the Burba correction (Burba et al., 2008;Järvi et al., 2009) as well as quality control checks with home-made MATLAB routines.The IRGA was calibrated bimonthly (except when weather limited site access) with gas standards of N 2 for zero and 479.5 ppm as a span CO 2 reference.Air temperature and humidity were measured by a thermohygrometer (STH-5031, Geonica, Madrid, Spain) at 1.5 m above the ground.Incident and reflected photosynthetic photon flux densities (PPFD) were measured by two quantum sensors (Li-190, Li-cor, Lincoln, NE, USA), also at 1.5 m.Net radiation was measured with a net radiometer (CN1-R, Middleton Solar, Brunswick, Australia) at 1 m above the surface.Rainfall was measured by a tipping bucket (0.2 mm) rain gauge (PLUVIOM 52203, RM Young, Traverse city, MI, USA).Three temperature and soil water content (EC-20, ECH2O, Decagon Devices, Pullman, WA, USA) sensors were also installed at 4 cm depth under both bare soil and plant cover having a sensitivity of 0.1 • C and 0.2%, respectively.These measurements were made at 1 Hz, but then averaged to 10 min and stored in data loggers (Meteodata 3000c, Geonica).

Data quality control, gap filling and statistical analysis
The 5-s data that did not satisfy a de-spiking routine were eliminated.Half-hours statistics were not computed when eliminated data exceeded 25% of the total.The subsequent processed half-hour statistics were also excluded from further analysis when rain or condensation (including frequent fog or dew) caused poor performance of the open-path IRGA.Nighttime data below the u * threshold of 0.2 m s −1 were also rejected for not fulfilling the turbulence measurement premise (Carrara et al., 2003;Wohlfahrt et al., 2005).This threshold was determined by plotting the nighttime CO 2 flux versus u * , and determining a value above which there was no dependence as described by Goulden et al. (1996).There were 16% and 20% of gaps in the 2007 and 2008 datasets due to electronic malfunction, calibration, power outages and poor performance of the anemometer and Li-7500 during inclement meteorological conditions.Nights with low turbulence resulted in 11% and 9% of data removed from 2007 and 2008 dataset.Finally, 6% and 3% were also rejected respectively from the two years due to spikes on half-hour timescales, yielding a total gap fraction of 33% and 32%.To obtain a continuous dataset for assessing the annual NEE, data were gap-filled following Falge et al. (2001) and Reichstein et al. (2005; http://gaia.agraria.unitus.it/database/eddyproc/index.html).

Uncertainty analysis
An uncertainty analysis established confidence intervals for CO 2 and water vapor balances.Random uncertainty and errors introduced by the gap-filling process were jointly calculated by realizing Monte Carlo simulations (Richardson and Hollinger, 2007).First, normally distributed random gaps were created in continuous annual datasets.Second, artificial noise was added to the remaining data following a Laplace distribution with a site-specific standard deviation that scaled with the magnitude of the CO 2 and water vapor fluxes (see Table 1; Richardson et al., 2006).Finally, these synthetic datasets were gap-filled so that different annual sums were obtained.Twice the standard deviation of such sums was taken as our annual error.

Respiration fitting
Nighttime CO 2 fluxes were fitted to the model following Raich and Schlesinger (1992):

Meteorological conditions in 2007 and 2008
Both years showed very similar patterns of air and soil temperature and PPFD (Fig. 1a, b and c).In fact, annual mean air temperature was 5.   Our precipitation measures are likely underestimated during winter, since rain gauges are not designed to measure solid precipitation.Furthermore, thirty-seven days of rain-gauge measurements were missing during 2008, due to instrument malfunction and inclement weather (25 mm were measured during this missing period at a station 12 km distant).However, soil water content (SWC) at the end of this 37-day period (DOY 129) was comparable for both years (12.6% for 2007, and 13.6% for 2008), suggesting that precipitation, at least during the last days of this period, was comparable.

Winter/spring dynamics and the onset of the first growing season
Although little differences were apparent in annual temperature or precipitation, we did observe important differences at seasonal and weekly scales, which developed into large annual NEE differences.For instance, some snow fall events (Fig. 1d), assessed via the reflected/incident PPFD ratio, occurred at different times of the year, and nearly three times as much rain fell during summer in 2007 (42 mm) as in 2008 (15 mm; Fig. 2b).Differences in the time courses of precipitation and temperature appear to have affected the onset of the growing season as reflected in the temporal trends in NEE for both years (Fig. 2c).The low values of NEE between the days 70 and 77 seem to indicate an early onset of the growing season in 2007.However, a sudden rainfall (Fig. 2b) on day 78, enhancing soil respiration, and subsequent considerable snow cover that began day 85, prevented growth development (see Fig. 3).Indeed, favourable synoptic weather conditions in early spring (days 125 to 150) appear to be fundamental, marking the onset of the growing season and the transition from source to sink of CO 2 .
The two years also presented different patterns in snow fall and snow cover to which we attribute differences in NEE during the days 77 to 110.During this time, air temperature and PPFD were much higher in 2007 than in 2008 (Fig. 1a and  c), probably due to decreased cloudiness in the 2007 early spring.Although the transition from source to sink of CO 2 happens between days 125-150 in both years, the strength of the CO 2 sink is higher in 2007 than in 2008.This indicates that air temperature and PPFD are key variables driving the onset of the growing season.Water availability is seldom a limiting factor during the alpine Mediterranean spring (SWC was much higher than the permanent wilting point, around 15.4% in 2007 and 15.7% in 2008).In fact, there was more rainfall and therefore higher SWC in 2008, as compared to 2007 (Fig. 2a and b) in this period, which supports the notion that temperature and PPFD, and not precipitation, control the onset of the growing season.Maximum daily CO 2 uptake in 2007 was − 2.2 g C m −2 , and took place around mid June (DOY 167), whereas maximum daily CO 2 uptake in 2008 was lower (− 2 g C m −2 ) and happened approximately one month later (Fig. 2c).

Summer dynamics and the end of the first growing season
Maximum ecosystem carbon assimilation occurred during mid June (DOY 167) and mid July (DOY 200) in 2007 and 2008 respectively.During this period, the ecosystem was characterized by mild air temperatures and high PPFD, which would favour the photosynthetic period to continue.However, after that time the lack of water resources (Fig. 2a  and b), which arises from the typical summer drought in the Mediterranean, starts to limit photosynthetic activity, and strongly constrains NEE (Fig. 2c).

Autumn dynamics and the onset of the second growing season
The lack of water resources limits biological activity by the mid and late summers (from DOY 163 and 200 on) of 2007 and 2008, respectively.Even so, some photosynthesis con-  tinues, as hydrological resources from the soil water stock are still available.This trend continues until the last days of summer when the first rains come after a long dry period.These autumn rains were accompanied by soil remoistening (Fig. 2a and b) that activated heterotrophic respiration, causing large, sudden releases of carbon (Huxman et al., 2004).These new water inputs ease plant stress and foster tissue repair and rehydration in the days following rainfall (Sala et al., 1982).Such plant recovery, which may last up to 7 days (Sala et al., 1982), pushes NEE negative again, starting a second growing season (around day 250 of Fig. 2c).This phenomenon is characteristic of Mediterranean climates (Wohlfahrt et al., 2008a) where the first rains after a long drought period usually arrive either in the last days of the summer or in the first days of autumn, and summer-like meteorological conditions continue afterwards for a while (the so-called Veranillo de San Miguel).The extended sunny period that favoured the second growing season of 2007, right after the first summer/autumn rains, did not take place during 2008.

Diurnal and monthly differences between 2007 and 2008
Some diurnal differences are encountered between the two years.Figure 3a  )

Hour of day (h)
May07 Jun07 Jul07 Aug07   that CO 2 uptake follows PPFD throughout each year, except during precipitation or water stress, which alter the relationship between PPFD and NEE.The same figure also displays asymmetry in NEE around noon.A higher daily uptake, peaking around 10 h, is found in 2007 and between 9 h and 12 h in 2008.The means for months of higher photosynthetic activity are displayed in Fig. 4, which further shows such asymmetry: maximum uptake before noon and subsequent NEE reduction in the afternoon.This afternoon depletion may also be observed in the daily relationship between NEE and PPFD, such that NEE was more negative for a given level of PPFD in the morning than in the afternoon (Fig. 5), representing hysteresis.This could be due to a series of interacting factors, which may be viewed as either environmen- tal or endogenous (that is, resulting from plant behaviour).
Temperature is typically higher in the afternoon than in the morning for a given level of PPFD, which could stimulate soil respiration.Figure 6a displays different nocturnal respiration trends as a function of temperature for different classes of SWC.We found positive exponential trends for every SWC class (R 2 = 0.56 for SWC > 18%, R 2 = 0.80 for 12% < SWC < 18% and R 2 = 0.51 for 8% < SWC < 12%) except for the lowest values (SWC < 8%), for which warming, such as that in the afternoon, does not increase soil respiration.When the light-response curve (Fig. 6b) is plotted for such a dataset (SWC < 8%) a clear hysteresis cycle is observed.Thus, afternoon NEE reduction appears not to be Biogeosciences, 7, 2601-2611, 2010 www.biogeosciences.net/7/2601/2010/due exclusively to an enhancement of soil respiration in the afternoon.
There are several factors that can potentially interact to shape the observed hysteresis in the NEE-PPFD relationship.First, as the Bowen ratio and VPD increase at midday, stomatal conductance drops to avoid runaway cavitation (Tyree and Sperry, 1988), at the expense of diminishing carbon uptake.Water stored within the stem xylem diminishes during the day such that, after midday stomatal closure, plants are not fully capable of recovering the level of stomatal conductance they had in the morning (Sperry et al., 2008;Hölttä et al., 2009).Second, the accumulation of sugars within a leaf may lead to a feedback inhibition of rubisco activity (Lambers et al., 1998).Third, daily patterns of carbon assimilation have been shown to undergo circadian regulation, such that plants are "programmed" to have a higher carbon uptake in the morning than in the afternoon, independent of environmental factors (Webb, 2003;Dougthy et al., 2006;Resco et al., 2009).Canopy conductance, computed using the Penman-Monteith equation (Jones, 1992), shows a slight decrease during the afternoon (data not shown), corresponding to the decrease in carbon uptake evident in the light response curve of Figs. 5 and 6b (hysteresis).

Annual differences between 2007 and 2008
Differences in the length and strength of the growing season and in the timing of the rain and snow fall seem to be responsible for general NEE disparities between the two years.The study site is a net source of CO 2 during both years.Annual emissions of CO 2 were + 52± 7 g C m −2 and + 48 ± 7 g C m −2 for 2007 and 2008, respectively.These differences are due mainly to disparities in carbon uptake during the growing season.The growing season of 2007 was shorter and more skewed than that of 2008 (Fig. 2c).It lasted longer in 2008 and presented a less steep slope, but ultimately with less area and therefore less CO 2 uptake.Rain pulse events following long dry periods took place in both years bringing about large emissions, but in different amounts depending on their timing.Both years presented large releases during the first 125 days contributing to the annual source of carbon.Each of three events of snowfall and thawing during these 125 first days of 2007 (Fig. 1d) may have contributed to net release (Monson et al., 2006a).Snow inhibits the photosynthetic activity covering the vegetation leaves, but not bacterial processes since the snow cover prevents soil temperature from falling excessively (Monson et al., 2006b).During snowcovered periods respiration is expected to dominate NEE, and therefore the timing of snows-thaws may also account for annual differences between the two years.
Our Mediterranean alpine site was a carbon source for both 2007 and 2008.It may be that high-altitude ecosystems are undergoing a CO 2 uptake depletion in the same manner as tundra sites (Oechel et al., 1993).The study site may have been a carbon sink in previous years, but the rise of the mean

The Burba correction
Although some have claimed the Burba correction to be negligible (Haslwanter et al., 2009) or even problematic (Wohlfahrt et al., 2008b), we found it to be very important for our experimental site.Figure 7 shows daily NEE with and without the correction during the first weeks of 2007.
The shaded areas indicate periods of snow cover.It is evident that the final Burba-corrected CO 2 flux is larger than the uncorrected CO 2 flux.During the first week-long snow period, an apparent net uptake of 0.38 g C m −2 is corrected to yield an emission of 3.8 g C m −2 , which seems much more reasonable as the average air temperature during these 7 days was just 3.4 • C and the ecosystem was completely snow covered.A second snowfall took place on day 23, where noncorrected values estimated six days of implausible uptake, which disappears once the correction is applied.The Burba correction eliminates apparent uptake, and also introduces a positive increment that strongly affects longterm integrations.In fact, it is so important as to reverse conclusions regarding the annual CO 2 source/sink behaviour.Figure 8a displays the cumulative NEE throughout the year 2007, yielding − 135 ± 7 g C m −2 without correction versus a source of 52 ± 7 g C m −2 when applying the Burba considerations.For the year 2008 applying the Burba correction converted an annual CO 2 sink of −100 ± 7 g C m −2 to a source of 48 ± 7 g C m −2 (data not shown).Thus there is an increment in estimated annual carbon loss of 187 ± 14 g C year −1 in 2007 and 148 ± 14 g C year −1 in 2008 when the correction is applied.This annual magnitude of the Burba   correction, higher than those previously reported (Burba et al., 2008), is due to the fact the site is at 2300 m elevation.
For such an elevation, the 30% reduction in air density (and thereby heat capacity), means that instrument heating causes higher perturbations in temperature and density effects.Furthermore, cool or even cold summer nights at such elevation cause the Burba correction to be non-negligible for all seasons.
Figure 8b shows cumulative evapotranspiration (ET) for the year 2007.In contrast to the case for cumulative NEE, the effect of applying the correction changes to a lesser extent.The corrected data yield an annual ET of 386 ± 6 mm, versus a non-corrected value of 346 ± 5 mm.These small differences in ET are similarly reflected at daily timescales (typically ranging between 1 × 10 −5 and 2 × 10 −4 g m −2 s −1 ).Because ET/ET is 100 times lower than F CO2 /F CO2 , the importance of the Burba correction for ET on an annual basis is quite small compared to NEE, consistent with the findings of Burba et al. (2008).Therefore, the importance of the correction in ET on an annual basis is quite small if compared to NEE.

Conclusions
Seasonal variability in NEE at this high-altitude ecosystem seems to be mainly driven by the interaction between PPFD and precipitation/snow and to some extent by one or more environmental or endogenous factors on daily timescales.When snow covers the site, respiration dominates over photosynthetic processes.During spring when snow melts, moistening soil and roots, increases in temperature and PPFD drive the beginning of the growing season.At this point fair weather conditions are fundamental for development of the growing season, not only for high uptake rates but also for enduring uptake.The lack of water resources in the soil likely halts the growing season, thus reducing the uptake period.From then on, water limitation is the major driver of carbon fluxes (Granier et al., 2007) until the first rains arrive after a dry summer.Once the soil is remoistened, large respiration rates occur followed by a second uptake season during late summer or early fall if meteorological conditions permit.Hysteresis was found in the light response curves that we attributed to endogenous or environmental factors or to both of them (see Sect. 3.5).
Although some have found a negligible effect when applying the Burba correction, it may have profound impacts on annual budgets due to daily systematic error propagation (Moncrieff et al., 1996).Such bias may not be evident for ecosystems with great CO 2 exchange potential, but large relative changes are observed here.The Burba correction strongly affects annual NEE of CO 2 , converting conclusions about the system from sink to source, but influencing the annual ET budget to a far lesser extent.Apparently, this correction is necessary for achieving credible carbon exchange measurements in very cold conditions, particularly for ecosystems with near-neutral capacity as carbon sources/sinks.

Fig. 1 .
Fig. 1.Average daily environmental conditions in the alpine Mediterranean shrubland located in Sierra Nevada mountain range (southeast of Spain): (a) air temperature, (b) soil temperature, (c) ratio of reflected to incident photosynthetic photon flux density, (d) absorbed photosynthetic photon flux density.

Figure 2 .
Figure 2. Average daily environmental conditions in the Mediterranean alpine shrubland located in the Sierra Nevada mountain range (southeast Spain): (a) soil water content, (b) precipitation and (c) net ecosystem exchange flux.The Burba correction was applied.

Fig. 2 .
Fig. 2. Average daily environmental conditions in the Mediterranean alpine shrubland located in the Sierra Nevada mountain range (southeast Spain): (a) soil water content, (b) precipitation and (c) net ecosystem exchange flux.The Burba correction was applied.

21Figure 3 .
Figure 3. Annual and daily evolution of NEE (µmol m -2 s -1 ) by the alpine Mediterranean shrubland located in the Sierra Nevada mountain range during (a) 2007 and (b) 2008.

Fig. 3 .
Fig. 3. Annual and daily evolution of NEE (µmol m −2 s −1 ) by the alpine Mediterranean shrubland located in the Sierra Nevada mountain range during (a) 2007 and (b) 2008.
and b represents the daily evolution of NEE throughout the year 2007 and 2008, respectively.We found www.biogeosciences.net/7

Figure 4a .
Figure 4a.Mean diurnal NEE flux during May, June, July and August 2007 in the alpine Mediterranean shrubland located in Sierra Nevada mountain range.

Fig. 4a .Figure 4b .
Fig. 4a.Mean diurnal NEE flux during May, June, July and August 2007 in the alpine Mediterranean shrubland located in Sierra Nevada mountain range.

Fig. 4b .
Fig. 4b.Mean diurnal NEE flux during May, June, July and August 2008 in the alpine Mediterranean shrubland located in Sierra Nevada mountain range.

Fig. 5a .•Fig. 5b .
Fig. 5a.Mean hourly NEE flux vs. PPFD during May, June, July and August 2007.Hysteresis was found for every month, with the temporal trend proceeding counterclockwise.

Fig. 6b .Figure 7 .
Fig. 6b.Mean hourly NEE flux vs. PPFD for SWC < 8% in 2007.Clear hysteresis was found for lowest SWC values.NEE flux for the lowest values of SWC does not present any increasing relationship with temperature (Fig.6a), however, it does present a clear hysteresis cycle indicating that afternoon soil temperature enhancement (and soil respiration with it) is not the consequence of hysteresis.

Fig. 7 .
Fig. 7.The NEE flux during the first 45 days of 2007.Black line with squared points is the NEE flux with no correction.Black line with round points is the same NEE but corrected.Shades areas show periods when the ecosystem was snow-covered.Some snowed days have an apparent downward CO2 flux when not correction was applied.The NEE flux corrected from Burba is always above the non-corrected one.

Figure 8a .Figure 8b .
Figure 8a.Cumulative net ecosystem exchange (g C m -2 ) through the year 2007 (dashed line without and black line with the Burba correction applied).

Fig. 8a .Figure 8a .Figure 8b .
Fig. 8a.Cumulative net ecosystem exchange (g C m −2 ) through the year 2007 (dashed line without and black line with the Burba correction applied).

Fig. 8b .
Fig. 8b.Cumulative evapotranspiration (mm) throughout the year 2007 (dashed line without and black line with the Burba correction applied).

Table 1 .
Empirically determined linear relations used to estimate random errors (σ ) in fluxes as a function of the flux magnitude for the Laguna Seca field site (see