Carbon Exchange Variability over Amazon Basin Using Coupled Hydrometeorological-Mixed Layer PBL-CO2 Assimilation Modeling System Forced by Satellite-Derived Surface Radiation & Precipitation

With the aid of a 3-part modeling system forced by various satellite-remote sensed atmospheric inputs controlled by cloudiness, this study: (1) describes the space-time variations of surface net CO2 flux exchange over the large scale Amazon basin, (2) determines the key factors controlling these variations, and ultimately (3) determines the optimal spatial configuration of a network of tower observing posts, which if deployed, could capture in area-wide averages the main variant properties of the Amazon basin's surface net carbon flux on an absolute basis. The philosophy guiding this research is that whereas a sufficiently detailed model can do very well in capturing the space-time gradients of carbon flux exchange and thus the relative source-sink properties of the Amazonian landscape, current modeling knowledge does not allow an adequate model determination of absolute source-sink properties. Direct observations are needed to obtain meaningful absolute accuracies of the source-sink properties, properties that are highly sensitive to environmental and bio-physiological factors that effectively produce a heterogeneous fabric of source and sink magnitudes across the basin at any given instant of time. However, for the Amazon basin, and as a general rule of thumb in carbon budget monitoring over a large expanse, there seems to be never enough observation posts to eliminate the systematic bias problem -- nor are those that do exist sited according to a network strategy that optimizes their collective ability to eliminate such a bias problem. A hydrometeorological model coupled to a mixed layer (ML) model of the planetary boundary layer (PBL) then equipped with a set of three CO2 assimilation models, and finally forced by high resolution satellite-retrieved incoming surface radiation fluxes and rainfall, forms a detailed carbon flux exchange modeling system linked to satellite inputs, that achieves the desired research objectives. The forcing of the model by remotely sensed solar/infrared radiative flux and rainrate variables, which exert dominant influences on the surface carbon budget and whose variant properties are determined by the position and diurnal timing of cloudiness, is an essential element of the modeling system. This is because one of the greatest shortcomings in prognostic modeling is the ability to reproduce real clouds, particularly convective clouds, in the right place at the right time. In understanding how environmental and bio-physiological factors exert their respective controls on carbon flux exchange variability, the underlying variables are classified into four categories: (1) meteorological factors; (2) radiation factors; (3) water cycle factors; and (4) bio-physiological factors. The three different CO2 assimilation models are investigated to achieve optimal performance insofar as obtaining validated surface carbon, heat, and moisture fluxes in the framework of the Florida State University (FSU) hydrometeorological model -- developed over the last decade by E.A. Smith & H.J. Cooper. Of the three carbon models examined, the one selected for the final net CO2 flux calculations was developed by G. Bonan, beginning with his Ph.D. dissertation research and now included in a land surface model (LSM) facility at NCAR. This carbon scheme contains a respiration component consistent with its photosynthetic component and physically couples the CO2, sensible, and latent heat fluxes through stomatal resistance. Test calculations of net ecosystem productivity (NEP) show that it is essential to model canopy-boundary layer interactions in order to reproduce observed morning CO2 effluxes measured at various forest sites located within Brazil and operated as part of the Large Scale Biosphere-Atmosphere Research Programme for Amazônia (LBA) -- specifically the LBA tower sites at Manaus and Jaru. This is because under typical conditions of a stable nocturnal PBL, the forest canopy remains stagnant, allowing carbon concentrations to become elevated until after sunrise when PBL stability flips and CO2 is rapidly vented into the atmosphere. In the PBL model developed for the study, CO2 concentrations and the concomitant fluxes are determined for five layers in and above the forest canopy following the progression of the ML during the daytime and the nocturnal boundary layer at night, which are treated as separate components of the diurnal PBL. It is important to point out that canopy heat capacity must be accounted for in the hydrometeorological modeling (an oft-overlooked factor in LSM modeling), to prevent sensible heat fluxes from being systematically overestimated. Values of observed canopy heat storage (needed in the development of the heat capacity scheme) are found using observed differences between net incoming radiation and sensible-latent heat fluxes, or observed total residual energy. Calibration and validation of CO2, sensible, and latent fluxes at the three LBA tower sites are accomplished using modeled total residual energy at the forest sites and modified photosynthesis parameters at the pasture site. Application of the modeling system over the large-scale Amazon basin shows that while vegetation type is the most important factor controlling area-wide CO2 fluxes, incoming surface solar radiation and ambient temperature (both directly responsive to the cloud field) are the primary factors producing spatial and temporal variability of CO2 fluxes at a given location. Modeled CO2 fluxes show mean monthly uptake values in the range of 1-3 mmol m-2 s-1. Due to the superimposed annual and daily march of the solar elevation angle, diurnal progressions of large coherent expanses of CO2 efflux over forest areas, progressing from SE to NW in December and from NE to SW in June, are an essential variational mode in the surface carbon budget. Inspection of area-wide modeled fluxes near the tower sites reveals that the systematic use of ECMWF-analyzed winds and temperatures in forcing the modeling system creates instances of spurious nocturnal stability that produce larger morning efflux magnitudes than observations suggest. Finally, CO2 fluxes at some 20,000 forest grid points within the Amazônia region and for eight months of model output, are analyzed to determine the optimal sampling configuration vis-à-vis capturing in area-wide averages, the space-time variability of net CO2 flux. These results lead to the conclusion that flux observations from five strategically placed towers, measuring in conjunction with the existing three LBA towers at Manaus and Jaru, would be sufficient in baselining area-wide net CO2 fluxes needed for an understanding of carbon sequestration within the Amazon basin on an absolute scale.