Skip to main content
Dryad logo

Barro Colorado Island - eddy covariance flux data (2012-2017)

Citation

Detto, Matteo (2022), Barro Colorado Island - eddy covariance flux data (2012-2017), Dryad, Dataset, https://doi.org/10.5061/dryad.3tx95x6j5

Abstract

This dataset contains CO2/H2O eddy covariance fluxes and other microclimatic observations observed on Barro Colorado Island (Panama) from 2012 to 2017. Data were acquired and processed using standard routines (see Methods). Data are organized in a spreadsheet with each variable in a column.

Methods

Turbulent fluxes and meteorological variables were measured from a 41 m Eddy Covariance (EC) flux tower located on the central plateau of the island at about 140 m ASL from July 2012 to August 2017. The eddy covariance system included a sonic anemometer (CSAT3, Campbell Scientific) and an open-path infrared CO2/H2O gas analyzer (LI-7500A, LiCOR Bioscience). Hi-frequency (10Hz) measurements were acquired by a datalogger (CR1000, Campbell Scientific) and stored on a local PC. Additional environmental variables used in this study were recorded at 5-min intervals and include rainfall (Tipping Bucket Rain Gauges TB4, Hydrological Services), temperature and relative humidity (HC2S3, Rotronic), solar radiation (CMP11, Kipp&Zonen), upwelling and downwelling shortwave and longwave radiations (CNR1, Kipp&Zonen), direct and diffuse PAR (BF5, Delta-T Devices), and canopy temperature obtained from five infrared thermometers (SI-131, Apogee Instruments) pointing toward the crowns of the four canopy trees closest to the tower. Three time domain reflectometers (CS616, Campbell Scientific) were installed vertically into the soil around the tower and calibrated with gravimetric soil moisture samples taken in different moisture regimes in close proximity to the probes.

Data were processed with a custom program using a standard routine described in (Detto et al. 2010). Fluxes were computed on a 30-min moving averaging window with a 5-min time step. This redundancy allowed better identification of gaps, especially gaps due to frequent rains, non-stationary and not ideal turbulent conditions, and increased temporal resolution. QA/QC criteria follow guidelines for removing erroneous values (Mauder et al. 2013) and exclude periods of non-ideal turbulent conditions, during and immediately after rain, outliers of several scalar statistics, poor energy budget closure and light response, non-stationary conditions, and known sensor malfunctioning.

GPP was derived from net ecosystem exchange (NEE) daytime values by adding the corresponding mean daily ecosystem respiration (RECO). Two approaches were implemented. In the first, RECO equaled the intercept of the light response curve (Lasslop et al. 2010). In the second, RECO equaled mean nighttime fluxes. Both methods were performed on a ±15-day moving window using a relatively restricted filter (friction velocity >0.4 m s-1) to minimize flux underestimation. Because of relatively small variation in air temperature, lack of temperature dependence of soil respiration (Rubio & Detto 2017), and not detectable temperature dependence of nighttime fluxes, the temperature was not included as a covariate for estimating RECO. The two methods provided similar results, but the first method yielded higher RECO values. However, when compared to soil chamber measurements available for the same period in the tower's footprint (Rubio & Detto 2017), these fluxes appeared underestimated, considering that RECO includes also above ground respiration which can contribute up to 40-50% of total respiration (Malhi et al. 2011). The energy balance closure was comparable to closures in most other Fluxnet sites (Wilson et al. 2002) with an imbalance of about 25%, and no appreciable differences between wet and dry seasons (Fig. S10).

In order to compute daily and monthly time-integrated values of GPP, evapotranspiration (ET), and sensible heat (H), gaps in the time series were filled using an Artificial Neural Network (Papale & Valentini 2003) with hydro-meteorological inputs as predictors (soil moisture, various radiation components, temperature, VPD, and air pressure). The data were randomly divided into a training set (70%), a validation set (15%), and a test set (15%). A two-layer feed-forward network with 10 sigmoid hidden neurons and linear output neurons was trained using the Levenberg-Marquardt algorithm until the mean square error (MSE) of the validation set stop improving (Hagan & Menhaj 1994). Performance, in terms of MSE, was evaluated using the test set at the end of the training. This procedure was repeated 100 times to produce 100 estimates of GPP and ET. Training multiple times generates different results due to different initial conditions and random sampling of the training, validation and test sets. The ensemble was obtained as a weighted average from the 100 Artificial Neural Network predictions using the reciprocal of MSE of the test set as weights.

 Detto, M., Baldocchi, D. & Katul, G.G.G. (2010). Scaling Properties of Biologically Active Scalar Concentration Fluctuations in the Atmospheric Surface Layer over a Managed Peatland. Boundary-Layer Meteorol., 136, 407–430.

Hagan, M.T. & Menhaj, M.. (1994). Training feedforward networks with the Marquardt algorithm. IEEE Trans. Neural Networks, 5, 989–993.

Lasslop, G., Reichstein, M., Detto, M., Richardson, A.D. & Baldocchi, D.D. (2010). Comment on Vickers et al.: Self-correlation between assimilation and respiration resulting from flux partitioning of eddy-covariance CO2fluxes. Agric. For. Meteorol., 150, 312–314.

Malhi, Y., Doughty, C. & Galbraith, D. (2011). The allocation of ecosystem net primary productivity in tropical forests. Philos. Trans. R. Soc. B Biol. Sci., 366, 3225–3245.

Mauder, M., Cuntz, M., Drüe, C., Graf, A., Rebmann, C., Schmid, H.P., et al. (2013). A strategy for quality and uncertainty assessment of long-term eddy-covariance measurements. Agric. For. Meteorol., 169, 122–135.

Papale, D. & Valentini, R. (2003). A new assessment of European forests carbon exchanges by eddy fluxes and artificial neural network spatialization. Glob. Chang. Biol., 9, 525–535.

Rubio, V.E. & Detto, M. (2017). Spatiotemporal variability of soil respiration in a seasonal tropical forest. Ecol. Evol., 7.

Wilson, K., Meyers, T., Goldstein, A., Baldocchi, D., Falge, E., Tenhunen, J., et al. (2002). Energy balance closure at FLUXNET sites. Agric. For. Meteorol., 113, 223–243.

Funding