| Abstrakt: |
We explore the potential of using a non‐parametric statistical method called Alternating Conditional Expectations, ACE, to quantify functional relationships in biogeosciences. Here, ACE is used to quantify the non‐linear and multi‐faceted responses of greenhouse gas fluxes to a set of biophysical forcings, when the shapes of those response surfaces are unknown. We evaluated the statistical method over two contrasting ecosystems and two contrasting time steps. One case involved quantifying the biophysical controls of water vapor and carbon dioxide (CO2) fluxes over a semi‐arid oak savanna using daily integrated fluxes. The other case evaluated the responses of CO2 and methane (CH4) flux measurements to a set of biophysical forcings at a restored tidal wetland using thirty‐minute averages. The statistical model, based on 4 independent variables, explained up over 90% of the variation in daily integrated flux densities of water vapor and net carbon dioxide exchange at the savanna site. This fit was defined by distinct non‐linear responses to such drivers as gross primary production, photosynthetically active radiation, air temperature, vapor pressure deficit and soil moisture. At the tidal wetland site, we evaluated net carbon dioxide and methane fluxes with short‐term measurements to capture the influence of rising and falling tides and seasonality in biological activity. The statistical model defined the shape of the forcing of fluxes due to the roles of carbon exudates, water table depth, oxygen level in the water column, temperature and vegetation status. The statistical fits of the greenhouse gas fluxes were less precise than the savanna case. The fetch varies on a run‐to‐run basis as it is comprised of a heterogeneous mosaic of open water and vegetation. Furthermore, it is difficult to monitor the environmental conditions of the archaea and bacteria in the sediments that produce methane and carbon dioxide. Plain Language Summary: The breathing of the biosphere can be assessed by measuring the fluxes of carbon dioxide, methane and water vapor. Attribution of the biophysical controls of these fluxes is often complex as these forcings can be non‐linear or conditional. We introduce an alternative statistical method that is better suited to disentangle the complex and non‐linear interaction of biophysical forcings on the trace gas emissions. It is called Alternating Conditional Expectations. In this paper, we demonstrate the efficacy of this method with test cases over a savanna woodland and a tidal wetland ecosystem. The method provides new insights on how these trace gas fluxes are modulated. Key Points: Alternating Conditional Expectations can define the shapes of non‐linear biophysical forcings on trace gas fluxesSimple statistical models can explain over 80%–90% of the variance in fluxesMethod is robust for test cases over ecosystems in dry and wet conditions [ABSTRACT FROM AUTHOR] |
Plný text