B52B-01
WRF-VPRM Modeling System And Its Role In Predicting Regional Carbon Budget
One of the key questions in climate science is to estimate a carbon budget of a given region. Atmospheric measurements of CO2 from global networks, mostly consisting of remote sites are used in combination with inverse modeling to estimate exchange fluxes of carbon between land and ocean biosphere and the atmosphere. Enhanced spatial resolution of such estimates is targeted with increasing density of the network, with a significant fraction of observations made in the continental boundary layer. These continental measurement sites, close to variable sources and sinks of CO2, are often located in meteorologically complex areas: terrain induced mesoscale phenomena such as sea-land, (lake, river, forest, etc.) breezes and mountain-valley circulations make the representation in global scale coarse resolution atmospheric models that are used in the inversions quite difficult. We setup a modeling system which combines a mesoscale meteorological model, the Weather Research and Forecasting (WRF) model with a diagnostic biospheric model, the Vegetation Photosynthesis and Respiration (VPRM). VRPM uses EVI and LSWI vegetation indices from MODIS satellite. In addition VPRM uses four parameters for each vegetation class, also temperature and radiation to produce biospheric CO2 fluxes. The WRF-VPRM modeling system was designed to realistically simulate atmospheric CO2 concentration fields at mesoscales, starting at 2km. Here we present our simulation results for different domains – SW France and SE Germany, where we have continuous measurement sites. The first domain contains ocean, land and mountains in the south and the east, while the second domain is the Ochsenkopf, a ~1 km tall hill in northern Bavaria. This gives us an opportunity to study the different kind of local mesoscale circulations and their influence on CO2 distribution. The study shows that in order to interpret local concentration measurements one has to perform high-resolution simulations which resolve local effects. The coastal station (Biscarosse tower in SW France) detects the remarkable variations in CO2 concentration due to sea-land breeze. This mesoscale effect leads to average CO2 concentration over land near the coast being higher than further inland (3D rectifier effect). The different kind of meteorological data obtained by ground and aircraft measurements allows us to investigate transport model deficiencies in detail and improve our models for using these fields in inversions. For performing inversions we're coupling Stochastic Time-Inverted Lagrangian Transport Model (STILT) to WRF-VPRM. The limited number of parameters in VPRM allows optimizing them against CO2 mixing ratio measurements in atmospheric inversion by using adjoint transport model - STILT. This provides a test for upscaling methods from flux towers to regional scales. The WRF-VPRM system provides the basis for a model-data-fusion system that simultaneously assimilates fluxes and concentrations to estimate high resolution regional scale biospheric CO2 fluxes over long-time periods.
B52B-02
Regional Carbon Dioxide Simulations Using Coupled Stochastic Time-Inverted Lagrangian Transport, Weather Forecast and Research, and Vegetation Photosynthesis and Respiration Models
Transport errors are known to be a large source of uncertainty in the inverse ("top-down") carbon flux estimates on global, continental, and regional scales. With a view to reducing these errors, we have configured the Stochastic Time-Inverted Lagrangian Transport (STILT) model to be driven by meteorological fields from the Weather Forecast and Research (WRF) model. Simulations of tower-based and airborne CO2 measurements have been performed using the STILT/WRF model coupled to the Vegetation Photosynthesis and Respiration Model (VPRM). The use of WRF meteorology leads to superior model performance compared with standard meteorological products and the combined STILT/WRF/VPRM model produces promising simulations of both ground-based and airborne CO2 data collected during spring and summer of 2004, especially in daytime conditions. Nevertheless, persistent simulation biases are evident as well, particularly with regard to the nighttime CO2 build-up, the morning transition to convective conditions, and the overall high bias in the airborne simulations. A limited model inter- comparison study, comparing STILT simulations driven by the WRF model against those driven by another mesoscale atmospheric model, the Brazilian developments on the Regional Atmospheric Modeling System (BRAMS) model, has revealed that, while model transport ensembles can ameliorate the most obvious simulation deficiencies, the overall improvement is not large. This is reflected in the fact that the model-to-model differences are smaller than between an individual transport model and observations, pointing to systematic errors in the simulated transport and the yet-to-be-quantified contribution from errors in the biosphere model. Overall, the STILT/WRF/VPRM offers a powerful tool for continental and regional scale carbon flux estimates. However, the best hope of reducing transport uncertainties hampering these estimates lies with a model ensemble approach involving an interchange of both transport and biosphere models.
B52B-03
Improving Regional Estimates of Carbon Exchange through a New Method of Analyzing Aircraft Flux Data
Regional carbon flux measurements are routinely performed using top down or bottom-up methodologies. Merging, comparing, or cross-validating these methods may require more data and further description of variance in reported data. While some surfaces are homogeneous at both local and regional scales, local scale ( ~ 1 km) heterogeneous or patchwork landscapes are becoming ever more common. Individual flux towers may accurately reflect the flux from a single patch, or a unique assemblage of several patches. However, the ability of a single tower (or even several) to quantitatively represent the distribution of fluxes from a variety of land covers and management practices is limited. Flux aircraft have often been used to represent regional fluxes. Local- scale heterogeneities may limit the usefulness of aircraft data for regional scaling. Traditional flux calculations use 3-5km averaging lengths that are often too large to capture such heterogeneities. The flux fragment method is a newly developed method for aggregating fluxes within an aircraft transect not by contiguous length (that often convolutes separate surface covers) but by highly resolved land classification. The result of this method provides a land use specific flux over a studied region (~10-100km). This can then be scaled by land use and compared to top-down regional (~1000km) estimates. Moreover, the flux fragment method provides an estimate of the spatial variance of fluxes per land use. This variance can then be used to place the results of a single flux tower within a larger context. Such variance descriptions are imperative when comparing top-down and bottom-up approaches to determine if differences between methods are significant or fall within the range of expected uncertainty from each approach. This flux fragment method was developed and the utility verified within the maize and soybean ecosystem of mid-continental North America.
B52B-04
Modeling Carbon Fluxes Between the Arctic Atmosphere, Ocean and Land Ecosystems
Recent studies suggest that Arctic ecosystems may be more vulnerable to future climate change than most other parts of the globe. However changes in its carbon-cycle and their feedbacks on climate are not yet fully understood. Using the MIT/MBL GLS (Global Land System) model coupled to a two-layer ocean model and a box- atmosphere model, we have studied carbon dioxide (CO2) fluxes between the major Arctic ecosystem components (atmosphere-ocean and atmosphere-land) and identified the dominant mechanisms that determine these fluxes for the Arctic region (52 N° and above). In most scenarios, we find that the Arctic represents a regional net sink of atmospheric CO2. The size of this Arctic sink depends on assumptions about lateral inputs and biological export of dissolved inorganic carbon in the Arctic Ocean, as well as on assumptions associated with the CO2 fertilization effect and warming temperatures in the terrestrial Arctic. Horizontal transport of atmospheric CO2 from extra-Arctic regions is a major control on local atmospheric CO2 concentrations in the Arctic. This suggests that anthropogenic processes in mid-latitudes may partly influence Arctic carbon dynamics. In addition, model runs with two hypothetical land cover change scenarios (conversion of Arctic land area to all tundra or all boreal) demonstrate that changes in land cover associated with potential future climate change may modify the land carbon fluxes by altering soil wetness and available soil nitrogen.
B52B-05
The ORCA West Coast Regional Project - Use of Top-Down Modeling in a Regional Carbon Budget Approach to Estimate Gross Carbon Fluxes for Oregon-California
The ORCA project aims at determining the regional carbon balance of Oregon and northern California, with a special focus on the effect of disturbance history and climate variability on carbon sources and sinks. ORCA provides a regional test of the overall NACP strategy by demonstrating bottom-up and model-data fusion approaches to derive carbon balances at subregional to regional scales. The top-down modeling component of ORCA focuses on coupling simple process models for gross primary productivity (GPP), autotrophic (RA) and heterotrophic respiration (RH) with atmospheric transport modeling to interpret measured carbon dioxide concentration data. This approach builds on high-resolution remote sensing products (e.g. Landsat, MODIS) as well as re-analysis meteorological datasets (EDAS-40, DayMet) to drive the models. We couple BRAMS (Brazilian Regional Atmospheric Modeling System) mesoscale modeling to the STILT (Stochastic Time-Inverted Lagrangian Transport) footprint model to identify sources and sinks for carbon that influence observed carbon dioxide concentrations at measurement sites in the ORCA domain. This approach of linking surface fluxes to time series of atmospheric data allows one to extract information from the latter for process model optimization, with the objective of obtaining regionally representative parameter settings for different combinations of ecoregion and land cover type. We present a proof-of-concept of the ORCA top-down modeling approach by comparing modeled to measured carbon dioxide concentration data. This comparison also addresses the uncertainties introduced by different components of this approach. To demonstrate the effect of the improved process model parameterization, the results are compared with regional fluxes derived by the ORCA bottom-up modeling component.
B52B-06
Regional Ecosystem Carbon Exchange in the Southern Great Plains: Measurements, Modeling, and Scaling
The extremely heterogeneous landscape of the ARM (Atmospheric Radiation Measurement) Climate Research Facility (ACRF) in the U.S. Southern Great Plains is representative of the southern boundary of the NACP Midwest intensive experiment. The area is largely agricultural with vegetation cover type and status that vary on sub- kilometer scales. In this study we developed, applied, and tested a "bottom- up" approach to inferring terrestrial C exchanges at fine scales (down to 250 m). Measurements at the ACRF include a 60 m tower instrumented with eddy covariance (ECOR) systems at several heights, about 20 permanent ECOR towers, several portable ECOR systems, many atmospheric and cloud sensing systems, and regular balloon sonde and aircraft measurements. We applied the land-surface model ISOLSM (with recent modifications to the plant physiological submodel) forced with OK and KS Mesonet climate datasets and MODIS vegetation indices. A method to infer vegetation cover type using satellite data and archetypal LAI annual profiles was developed and successfully tested against USDA census data for the region. The model's net CO2 exchange estimates were calibrated and tested using eddy correlation data from the dominant surface covers. Three years spanning a substantial precipitation gradient (2003 - 2005) were then simulated. Large differences in annual regional CO2 exchanges were predicted corresponding to expected system responses to available moisture. Spatial scaling analysis from 250 m to 100 km indicated that homogenizing LAI and vegetation cover can impact annual NEE substantially, including changing the region from a predicted net CO2 source to a net sink. Further, differences in NEE associated with spatial scaling differed between years, indicating that accurate bottom-up NEE estimates in this heterogeneous region require fine-scale analysis approaches.
B52B-07
Spatial Estimates of GPP Using LiDAR- and Quickbird-Derived fPAR
Regional- to global-scale gross primary production (GPP) is commonly estimated with light-use efficiency models, which are largely dependent on remotely sensed estimates of the fraction of photosynthetically active radiation absorbed by vegetation (fPAR). Methodologies to quantify spatial variability of fPAR and improve GPP estimates have not been established for mixed forests and heterogeneous landscapes in the Great Lakes Region, and are needed to estimate photosynthetic sinks for the Mid-Continent Regional Intensive Campaign. In this study, hemispheric photos were collected during the 2006 growing season to estimate fPAR, plant area index (PAI), leaf inclination angle, and clumping factors in >130 lowland and upland stands within the footprint of a 400 m eddy covariance flux tower near Park Falls, Wisconsin, USA. Airborne LiDAR and Quickbird imagery were acquired during leaf-on and leaf-off periods to make predictions of canopy structure, and resulting PAI/fPAR estimates were compared with litterfall measurements and products derived from the Moderate Resolution Spectroradiometer (MODIS). GPP was modeled with the MODIS MOD17A2 algorithm, using fine-resolution land cover and fPAR inputs that were spatially aggregated into units ranging from 30 m to 1 km square. Uncertainties and errors associated with fPAR methods and spatial resolutions are discussed based on agreement with flux tower observations.
B52B-08
Using crop specific phenology models coupled with the Simple Biosphere Model (SiB) for improved prediction of Land-Atmosphere Exchanges Across the Midcontinental Region of North America
Land-atmosphere exchanges in the Mid-Continent Intensive (MCI) region are dominated by agricultural management of cropland ecosystems. An accurate representation of crop dynamics (and hence crop phenology and physiology) in the existing model/s is important in predicting carbon and other exchanges. The current study was carried out in an attempt to develop, implement, and evaluate a numerical modeling system to estimate time-varying exchanges of carbon, water, and energy across the 5-state region of the Mid-Continent Intensive (MCI) experiment of the North American Carbon Program (NACP). This is to be achieved by using a coupled modeling system (i.e. SiB-RAMS; The Regional Atmospheric Modeling System (RAMS), a mesoscale meteorological (non-hydrostatic) model developed at Colorado State University, coupled to the Simple Biosphere Model (SiB)) tested against the observations at a variety of spatial scales. Since the original SiB had only minimal representation of cropland dynamics, we developed crop-specific phenology models to be coupled with SiB to improve the fluxes from the mid western region which is dominated by the managed agroecosystems. Coupled SiB-phenology models replaced the need for the use of remotely sensed NDVI information, on which SiB was originally relying for deriving Leaf Area Index (LAI) and the fraction of Photosynthetically Active Radiation (fPAR) for estimating carbon dynamics. Also, the use of phenology models remarkably improved the representation of LAI, Net Ecosystem Exchange, etc., by SiB, on validation against the site-specific observed values at different temporal scales.