Turbulence scales for eddy covariance quality control over a tropical dry forest in complex terrain

The assessment of the potential for carbon sequestration in tropical dry forests has lagged behind the work done in other tropical environments; the particularities of the carbon dynamics in these ecosystems are largely absent from models of global primary productivity and effects of climate change on vegetation. In Mexico, the largest portions of preserved tropical dry forests have been confined to conservation areas on hilly terrain where productive activities like agriculture are difficult. The location of these remnants of original tropical dry forest in complex terrain presents an operational challenge to directly assess CO2 exchange between the forest and the atmosphere by means of the eddy covariance (EC) technique. We installed and maintained an EC system over an intact tropical dry forest in the Chamela-Cuixmala Biosphere Reserve in the lowlands of the Pacific Coast of Mexico, to assess the suitability of this technique to measure CO2 flux, sensible heat and latent heat flux over moderately complex topography in a remote location. The quality of the datasets produced was evaluated through the degree of energy balance closure achieved. Modelling the incoming and outgoing radiation to take into account the slope and aspect of the underlying terrain raised the energy balance closure from 71% to 75%. We also compared the effect of applying three different filters to discriminate conditions of well-developed turbulence: the friction velocity u*, the standard deviation of the vertical wind σw, and two dimensionless indices based on a modified turbulence intensity scale uTKE. u* and σw discriminated fluxes coming from topographically complex areas, according to a flux footprint model, but the filtered subset of data did not showed higher energy balance closure compared to the unfiltered dataset. The uTKE-based filters were more effective screening out periods dominated by horizontal or vertical advective transport instead of turbulence exchange, and raised the slope of the energy balance regression in the filtered data between 3%-17% - depending on the season-,compared to the unfiltered dataset. Together, these procedures of quality assurance and control produced a flux dataset of quality comparable to those obtained in more ideal surfaces.