USGS - science for a changing world

South Florida Information Access (SOFIA)


publications > paper > controls on mangrove forest-atmosphere carbon dioxide exchanges in western Everglades National Park > research methods

Controls on mangrove forest-atmosphere carbon dioxide exchanges in western Everglades National Park

Abstract
Introduction
> Methods
Results
Discussion
Summary & Conclusions
References
Figures & Tables

2. Research Methods

2.1. Site Description and Measurements

aerial photo of the study site including a visualization of available fetch from all wind directions
Figure 1. Aerial photograph of the study site including a visualization of available fetch from all wind directions. NEE values were determined to be valid if >50% of the cumulative flux originated within the fetch boundaries contained by the tidal channels bordering the site. [larger image]

[5] The study site (25.3646°N, 81.0779°W) is located within an extensive riverine and fringing mangrove forest close to the mouth of the Shark River in western Everglades National Park (Figure 1). The dominant tree species at the site are red (R. mangle), black (A. germinans), and white (L. racemosa) mangroves reaching heights of 15-20 m [Ewe et al., 2006]. The forest understory is sparse and composed of seedlings and juvenile mangroves with an average height less than 4 m. The region experiences semidiurnal tides and is inundated twice during most 24 h periods. High tides can reach up to 0.5 m above the sediment surface [Krauss et al., 2006]. However, the sediment surface can be exposed for several days at a time during the annual minima in the solar tidal cycle which corresponds to the periods of low discharge through Shark River, generally in February, March, and April. The sediment surface at the site is ~0.2 m above mean sea level. Peat thickness beneath the forest increases toward the Gulf of Mexico in this region and at our site reaches 5 to 6 m [Spackman et al., 1966].

graph of monthly rainfall totals during January 2004 through August 2005
Figure 2. Monthly rainfall totals during January 2004 through August 2005. Values represent the average of rainfall totals from the Gulf Coast (GI) Everglades National Park station and the Shark Slough (SH2) United States Geological Survey station. Stations GI and SH2 are 5 km and 13 km E-NE of the study site. [larger image]

[6] Approximately 60% of the annual rainfall in the Everglades falls during the May-October wet season [Duever et al., 1994]. Seasonal rainfall patterns (Figure 2) are strongly influenced by the passage of tropical cyclones, usually between June and October, and by the infrequent passage of cold fronts during the winter months. With the onset of the wet season, total daily irradiance (Figure 3a) becomes variable due to frequent afternoon convective thunderstorms. Minimum daytime air temperatures (TA) in the Everglades rarely fall below 10°C between December and February (Figure 3b). From March through November, the daily maximum TA in the region is generally above 27°C [Duever et al., 1994]. During 2004-2005, the minimum daytime TA ranged from 10 to 15°C during the winter dry season, while the May-October wet season values were consistently above 25°C and less variable (Figure 3b). Soil surface water salinity at the site varies with tidal cycles and rainfall patterns. On daily time scales, salinity values increase from 1 to 12 ppt with incoming tides (Figure 3c). However, the annual minimum salinity values (2-18 ppt) during 2000-2010 occur when water levels are at their highest during the peak of the annual tidal cycle and freshwater discharges. Salinity values are highest (30-35 ppt) at the end of the dry season in May and early June. Annual minimum water levels occur during the early dry season in February and March when troughs in the lunar monthly tidal component combine with minimal fresh water flow through Shark River. Water levels are relatively high (>0.3 m) during this period (Figure 3d) due to increased fresh water discharge into Shark River. The peak of the annual tidal cycle also occurs during the wet season [Stumpf and Haines, 1998].

graphs of incident solar irradiance, average daily air temperature, average daily surface water salinity, and average and maximum daily water level
Figure 3. (a) Incident solar irradiance, (b) average daily (24 h) air temperature measured at 27 m above the ground, (c) average daily surface water salinity, and (d) average and maximum daily water level during January 2004 through August 2005. The 15 day centered moving averages of all daily values are included. [larger image]

[7] A 30 m flux tower and 250 m boardwalk from the banks of Shark River were constructed in June 2003. The tower base is 1.5 m above the surface and is supported by a square grid of central tiers (9 cm by 9 cm by 3.7 m long) driven 3 m into the sediment. Crossbeams to peripheral tiers provide additional stability and prevent the structure from sinking into the peat. Guy wires are anchored on smaller platforms with similar tiered construction. An elevated and waterproof wooden hut at the tower base houses twelve 6 V (260 A hr each) rechargeable batteries. All electronics are housed in a waterproof box elevated 2 m above the sediment. Five 120 W solar panels maintain battery charge. The tower (Universal Manufacturing, Clinton Twp., Michigan) is composed of 22" wide by 10' tall triangular aluminum sections. From the tower site, specific and uniform fetch distances (Figure 1) are determined from river boundaries and are as follows (where 0° is N, increasing in the clockwise direction): 300 m from 0° to 70°, 1500 m from 70° to 120°, 1000 m from 120° to 135°, 800 m from 135° to 180°, 1300 m from 180° to 270°, 250 m to 300 m from 270° to 360°.

[8] Environmental variables were measured above the canopy at 1 s intervals, averaged over 30 min on two data loggers (model CR23X, Campbell Scientific, Logan, Utah), and uploaded to a laptop for storage. These measurements include net radiation (model CNR 1, Kipp and Zonen, Bohemia, New York) and incoming and reflected PAR (model LI-190SB, LI-COR, Inc., Lincoln, Nebraska). Measurements also include air temperature (TA) and humidity (model HMP45C, Campbell Scientific, Inc., Logan, Utah) and wind speed and direction (model 05103 RM Young, Traverse City, Michigan) measured at 27 m. Aspirated and shielded thermometers (model 107 temperature probes, Campbell Scientific, Inc.) measure air temperature at 20 m, 15 m, 11 m, 6 m, and 1.5 m above the ground. Heat flux plates (model HFT 3.1, Campbell Scientific, Inc.) record soil heat fluxes, and soil thermocouples (model 105T, Campbell Scientific, Inc.) measure soil temperature (Ts) at -5 cm, -10 cm, -20 cm, and -50 cm. Further details on tower measurements are provided by Barr [2005]. Hydrologic data were continuously monitored and recorded every 15 min at a station 30 m south of Shark River and 150 m west of the flux tower. Measurements included specific conductivity and temperature (model 600R water quality sampling sonde, YSI Inc., Yellow Springs, Ohio) of surface well water and water level (model Waterlog H-333 shaft encoder, Design Analysis Associates, Logan, Utah).

[9] The eddy covariance (EC) system is mounted at 27 m. The EC consists of a three-dimensional sonic anemometer (model RS-50, Gill Co., Lymington, England) and thermistor and an open path infrared CO2 and water vapor (H2O) gas analyzer (model LI-7500, LI-COR, Inc., Lincoln, Nebraska). High-frequency (10 Hz) measurements are stored and processed with custom software to derive half-hourly CO2, latent and sensible heat, and momentum exchanges between the forest and the overlying atmosphere. High-frequency data processing consists of spike removal [Vickers and Mahrt, 1997], a two-dimensional coordinate rotation of the wind field, a time lag correction of CO2 concentration to maximize covariance with vertical wind speed variation, buoyancy corrections of sonic air temperatures [Schotanus et al., 1983], and conversion of the turbulent flux into the total constituent flux [Webb et al., 1980], which accounts for the positive vertical mass flow resulting from positive buoyancy of less dense air parcels. Storage of CO2 in the air column below the EC system was estimated based on the half-hourly rate of change of CO2 concentrations at the infrared analyzer level [Morgenstern et al., 2004; Humphreys et al., 2005]. This storage term was added to the fluxes derived from the EC system to determine NEE. The algorithms used to calculate NEE were independently verified using AmeriFlux "gold file" data sets (http://public.ornl.gov/ameriflux/standards-gold.shtml).

2.2. Missing Data

[10] Missing or invalid EC fluxes are commonly referred to as "gaps." Gaps occur when gas concentrations are out of range (as occurs during precipitation events), when turbulence is weak or intermittent, or when there is insufficient fetch. The CO2 fluxes during these gap periods need to be included to determine annual NEE cycles [Falge et al., 2001]. At the study site, short-duration gaps (≤ 4.5 h) occurred primarily at night but also happened as the result of thunderstorms and breaks in the power supply. Nighttime flux data were discarded during periods of weak turbulence [Goulden et al., 1996; Lee et al., 1999] when the friction velocity (u*) was less than 0.21 m s-1. This u* threshold was calculated by first dividing nighttime NEE values into 20 u* classes for each bimonthly period and then defining a u* value above which NEE became invariant or, for those bimonthly periods where no clear relationship between NEE and u* was apparent, we chose a u* value which corresponded to an NEE value ≥ 85% of the maximum bimonthly NEE. The global u* threshold of 0.21 m s-1 applied in the data analysis is the median value of all bimonthly u* threshold values, which varied between 0.15 m s-1 and 0.30 m s-1. During three bimonthly periods, the u* threshold was >0.25 m s-1. However, the differences in fluxes calculated during these periods using a u* threshold of 0.21 m s-1 versus greater values up to 0.3 m s-1 were not significant. Therefore, the global u* threshold of 0.21 m s-1 was applied to all bimonthly periods.

[11] Flux data were discarded when the flux footprint [Schuepp et al., 1990; Schmid, 2002] extended beyond the forest fetch. The fetch exceeded the footprint most frequently (66% of the time) during the nighttime when winds originated from the NW to NE. Longer gaps (>4.5 h) in the data set were generally caused by instrument or data acquisition malfunction and on rare occasions lasted for several days. The combined duration of gaps comprised 61%, 28%, and 46% of the total nighttime, daytime, and combined data sets, respectively. These values are comparable to average nighttime gap duration of 65% reported for 10 forested sites in Europe [Moffat et al., 2007].

2.3. Gap Filling and Error Analysis

[12] Several strategies are available to gap fill eddy covariance CO2 fluxes [see Moffat et al., 2007; Gu et al., 2005; Falge et al., 2001]. We chose a mean diurnal variation (MDV) method to fill short gaps and look-up tables (LUT) for longer gaps. The MDV utilizes a 14 day moving window centered on the day of the gap, and the missing values are filled with the mean fluxes within this window occurring during the same half-hourly period as the gap. For longer gaps, separate daytime and nighttime LUTs were developed for each 2 month interval beginning on 1 January 2004. Nighttime TA was better correlated with CO2 fluxes than TS and was therefore chosen as the independent variable in the LUT. For each 2 month interval, half-hourly nighttime CO2 fluxes were partitioned into 20 TA bins, each containing the same number of values. For daytime half-hourly CO2 fluxes, a two-dimensional LUT was constructed using 16-PAR and 3-TA bin categories. Falge et al. [2001] provide additional details on appropriate LUT dimensions when gap-filling EC fluxes.

[13] Potential error and bias in the fluxes, introduced by the MDV and the LUT, were estimated by randomly creating and then refilling a set of artificial gaps [Moffat et al., 2007] overlapping with valid data periods. On a monthly basis, the sum of the artificial daytime and the nighttime gaps were constructed to have the same duration as actual gaps. The root-mean-square error (RMSE) and the bias error (BE) were determined by comparing the fluxes estimated from 100 simulations of randomly generated, then filled, artificial gaps to the concurrent valid observations (Table 1). Small BE values (-0.021 ± 0.054 µmol (CO2) m-2s-1) suggest this method imputed minimal bias in annual NEP estimates. The RMSE (3.56 ± 0.058 µmol (CO2) m-2s-1) was within the range of those reported for six forested sites in Europe [Moffat et al., 2007].

Table 1. Summary of Annual NEP and Errors Associated With the Gap-Filling Technique
Parameter Average ± Standard Deviation
Bias error, BE (µmol CO2 m-2 s-1) -0.021 ± 0.054
RMSE (µmol CO2 m-2 s-1) 3.56 ± 0.058
NEP 2004 (g C m-2 year-1) 1170 ± 127
NEPa 2005 (g C m-2 year-1) 832 ± 97
aThrough the end of August 2005.

[14] Confidence intervals on monthly NEP were computed using the results of the gap filling procedure. In this method, all of the half-hourly NEE values for each month, including valid and gap-filled data, were integrated to give an estimate of monthly NEP for each of the 100 gap-filling simulations. The 5th and 95th percentiles of monthly NEP for each month were derived from the results. The average percent relative error (RE) was then calculated as the difference between the 95th and the 5th percentiles of gap-filled NEP divided by 2. In some cases, particularly during summer nighttime periods, the duration of actual gaps represented more than 50% of the total duration of nighttime periods within a 1 month interval. For these months, the amount of valid data points was considered insufficient to accurately calculate the RE associated with the gap-filling procedure. To account for this, a relationship between RE and the percentage of artificial gaps was derived by imposing a range of artificial gap fractions up to 80% of the valid data periods for those months with actual gaps <10%. A power function was then fit to the complete set of RE values associated with the fraction of data gaps (ƒgap) for each month using least squares regression:

RE = 7.66 ƒ 0.366 . (1)
gap

Monthly NEP confidence intervals were calculated as the product of NEP values derived from valid data points and RE based on the fraction of gaps (ƒgap) in that month using (1).

2.4. Light and Temperature Responses

[15] Daytime NEE responses to PAR were determined separately for "high" (TA > 28°C) and "low" air temperatures (TA < 21°C). High and low air temperature included 16.2% and 21.3% of the daytime flux data set, respectively. Daytime NEE values were further grouped by a clearness index (Kt), defined as S/Se, where S is incoming solar irradiance (W m-2) and Se is extraterrestrial irradiance at the top of the atmosphere on a plane parallel to the Earth's surface:

Se = Ssc(1 + 0.033 cos(360 td /365))cosθ. (2)

Ssc is the solar constant (1370 W m-2), td is day of year, and θ is the solar zenith angle [Spitters et al., 1986; Gu et al., 2002]. Lower (Kt ≤ 0.65) and higher (Kt > 0.65) values of Kt represent cloudy skies and clear skies, respectively. The threshold value of Kt (0.65) was selected as the median of daytime values during the study period. Daytime NEE values were distributed evenly into cloudy and clear sky bin categories across the high and low temperature ranges. To reduce the scatter and variability of half-hourly NEE versus PAR, NEE values were bin averaged across 30 intervals of PAR. A form of the Michaelis-Menten equation (3) was fit to the bin-averaged NEE data using nonlinear least squares regression:

NEE = - a' PAR + Rd. (3)

(1-(PAR/2000) + (a' PAR/GEP2000))

The variable a' represents the ecosystem quantum yield (µmol (CO2) per (µmol (photons)). GEP2000 is the gross ecosystem photosynthesis (µmol (CO2) m-2s-1) defined as the sum of daytime NEE and the ecosystem respiration rate Rd (µmol (CO2) m-2s-1) when PAR equals 2000 µmol (photons) m-2s-1.

2.5. Salinity Effects

[16] Direct salinity effects on NEE in mangroves are difficult to quantify because both short- and long-term fluctuations in salinity are also typically accompanied by changes in tidal cycles, temperature, and solar irradiance, all of which influence canopy-scale CO2 fluxes. We examined the potential effects of salinity on ecosystem functioning by comparing the relationships between daytime NEE, PAR, and TA at salinity values above and below the daytime annual median of 29 ppt. PAR and daytime TA data were divided into 20 and 15 bins, respectively, for each period characterized by "high" (>29 ppt) or "low" (≤29 ppt) salinity. Bin ranges were selected such that NEE values were equally distributed along PAR and TA dimensions. Contours of equal NEE values were then constructed across the two dimensions of the PAR-TA matrix (Sigma Plot Version 11, Systat Software, Inc., San Jose, California). We also investigated salinity effects on GEP [Lopez-Hoffman et al., 2006; Theuri et al., 1999; Ball and Pidsley, 1995; Suárez and Medina, 2006]. Daily total GEP was normalized by daily total PAR (here termed the light use efficiency (LUE)). We investigated the covariance between LUE and daily average salinity at PAR > 600 µmol (photons) m-2s-1:

LUE= ΣPAR>600 GEP . (4)

ΣPAR>600 PAR

2.6. Ecosystem Respiration and Tidal Effects

[17] Equation (3) was used to estimate average daytime plant and soil respiration rates separately during high and low tides. In this method, a moving 7 day window of half-hourly PAR and NEE data was centered on each day in the record. Nonlinear regression was used to calculate daily Rd separately for high- (water level > 0.2 m) and low-tide (water level ≤ 0.2 m) periods. At least 30 valid NEE values were required within each tidal cycle during the 7 day window to calculate a high- and low-tide Rd for each day. An Arrhenius-type relationship [Lloyd and Taylor, 1994] was used to model daytime Rd as a function of air temperature:

Rd = Rd20 exp [Ea/R (1/293K - 1/TK)] . (5)

Rd20 (µmol (CO2) m-2s-1) is the ecosystem respiration rate at 20°C, Ea (in J mol-1) is the apparent activation energy, R is the universal ideal gas constant (J mol-1 K-1), and TK is the average absolute air temperature during the 7 day moving window. The base respiration rate at 20°C was included rather than the more commonly used 10°C since daytime temperature values of 10°C are rare at the study site. Half-hourly GEP values were calculated as the sum of -NEE and Rd, with the results assigned to either the high or the low-tide category. High- and low-tide GEP values were summed and used to determine daily LUE in (4). Nighttime Rd was modeled as a function of temperature using (5) for high- and low-tide periods. Three or more consecutive half-hourly Rd values within each tidal cycle were required for inclusion in the analysis.

2.7. Seasonal and Annual NEP

[18] Half-hourly, gap-filled NEE values were converted to carbon equivalents and summed over 24 h periods to produce daily total net ecosystem production NEP (g C m-2 d-1). Daytime and nighttime components of NEP were calculated separately. Monthly sums of daily NEP illustrate seasonal changes in mangrove carbon assimilation in relation to climatic and physical drivers such as salinity and water levels.

< Introduction | Results >



| Disclaimer | Privacy Statement | Accessibility |

U.S. Department of the Interior, U.S. Geological Survey
This page is: http://sofia.usgs.gov/publications/papers/mang_exchange/methods.html
Comments and suggestions? Contact: Heather Henkel - Webmaster
Last updated: 04 September, 2013 @ 02:04 PM(TJE)