Improving the Mechanistic Representation of Arctic Carbon Dynamics in the Community Land Model Using Data Assimilation of Leaf Area Index
Xueli Huo, The University of Arizona, huoxl90@email.arizona.edu (Presenter)
Andrew Fox, Joint Center for Satellite Data Assimilation, afox@ucar.edu
Tim Hoar, National Center for Atmospheric Research, thoar@ucar.edu
Jeffrey Anderson, National Center for Atmospheric Research, jla@ucar.edu
William Kolby Smith, The University of Arizona, wksmith@arizona.edu
Hamid Dashti, The University of Arizona, hamiddashti@arizona.edu
Charles Devine, The University of Arizona, cjdevine@email.arizona.edu
Siyu Wang, Beijing Normal University, sywangbnu@hotmail.com
David Moore, The University of Arizona, davidjpmoore@arizona.edu
More and more evidence shows that the global carbon cycle is responding to climate change. In our NASA Arctic Boreal Vulnerability Experiment Phase 2 project we are addressing the question of “How are the magnitudes, fates, and land-atmosphere exchanges of carbon pools responding to environmental change, and what are the biogeochemical mechanisms?”. We will improve the mechanistic representation of biochemical processes within the Community Earth System Model (CESM). This model can help us understand what is causing changes in the global carbon cycle, and evaluate whether warming in the arctic has and is likely to cause amplifying feedbacks to climate. However, the model is typically biased and the accuracy that represents the real world needs to be improved. Data assimilation (DA) is proven to be an effective and efficient way to correct model bias and improve model accuracy, and we use DA to ensure that the model agrees with historical LAI in the region.
Here, we develop and implement a sequential state data assimilation technique to reduce errors and biases in modeled LAI. We employ the Ensemble Adjustment Kalman Filter (EAKF) within the Data Assimilation Research Testbed (DART), which is coupled with the Community Land Model version 5 (CLM5). Building on previous work, we have been optimizing the settings of DA including localization, inflation, observation quality control, observation error and which variables to include in the state vector and update.
We used the bi-weekly GIMMS LAI3g data as LAI observations and conducted a decade-long data assimilation of LAI into CLM 5 globally from 2000 to 2011. Then we compared the modeled LAI, gross primary productivity (GPP), and latent heat flux (LE) with those from the corresponding decade-long free run globally. By comparison, we assessed the impact of assimilating LAI into CLM on improving modeled global carbon, water and energy flux exchange between land and atmosphere. Particularly, We highlight the impact this has had in the Arctic, a region where the model is known to exhibit a very large overestimation in both productivity and LAI. The results show that after assimilation the modeled annual LAI in the Arctic deceases by ~22% and in some locations the reduction is over 50%. The assimilation of LAI also corrects the overestimation of annual GPP in the Arctic with an average reduction about 18% and maximum reduction over 30% locally. The water cycle closely coupled with the carbon cycle through photosynthesis and transpiration is also impacted, and the annual latent heat flux in this domain is reduced by 8% compared with the free run.
Our data assimilation is very flexible, and we have begun to investigate using multiple types of observations simultaneously in addition to LAI. This includes vegetation optical depth derived biomass, soil moisture and solar induced fluorescence. We are transitioning our work to a high resolution simulation in the ABoVE domain, and will also focus on site-level simulations at data rich locations within the domain.
Presentation: iPoster
Associated Project(s):
Presentation Type: Poster
Science Theme: Carbon Dynamics