Sunday, 27 November 2016

Data Assimilation


[image:]  Propagating probability distributions of stand variablesResearch Issue

Data sampling routinely involves some degree of uncertainty in the measurements, often called noise, which is unpredictable. Likewise, models used to estimate system processes also contain uncertainty, which can come in many forms, not the least of which is their imprecise formulation based on incomplete understanding of the phenomenon being modeled. The question is how best to estimate the state of the system under consideration through time, along with any unknown model parameters, from these (possibly nonlinear) noisy model predictions and incoming measurements.

Our Research

Data assimilation seeks to merge models with data sequentially, as the data arrives, to estimate the unknown state of the system probabilistically, thus taking measurement and model uncertainty into account. We have been adapting sophisticated statistical techniques used in nonlinear systems estimation for work on forest estimation problems. These problems include, but are not limited to, estimating the carbon flux in forests using eddy covariance techniques, and propagating probability distributions of forest stand variables for use in decision making for management.

Expected Outcomes

Computer software, research papers, and management guides for practitioners.

Research Results

Moffat, Antje M.; Papale, Dario; Reichstein, Markus; Barr, Alan G.; Beckstein, Clemens; Braswell, Bobby H.; Churkina, Galina; Desai, Ankur; Falge, Eva; Gove, Jeffrey H.; Heimann, Martin; Hollinger, David Y.; Hui, Dafeng; Jarvis, Andrew J.; Kattge, Jens; Noormets, Asko; Richardson, Andrew D.; Stauch, Vanessa J. 2007. Comprehensive comparison of gap filling techniques for eddy covariance net carbon fluxes. Agricultural and Forest Meteorology. 147:209-232.
Gove, J.H.; Hollinger, D.Y. 2006. Application of a dual unscented Kalman filter for simultaneous state and parameter estimation in problems of surface-atmosphere exchange. Journal of Geophysical Research. 111. D08S07, doi:10.1029/2005JD006021. 21 p.
Gove, J.H., and D. R. Houston. 1996. Monitoring the growth of American beech affected by beech bark disease in Maine using the Kalman filter. Environmental and Ecological Statistics 3(2):167-187.

Research Participants

Principal Investigator

  • J.H. Gove, US Forest Service- Northern Research Station Research Forester 
  • D. Y. Hollinger, US Forest Service- Northern Research Station Plant Physiologist
  • H.T. Valentine, US Forest Service- Northern Research Station Research Forester

For further details log on website :
http://www.nrs.fs.fed.us/sustaining_forests/conserve_enhance/timber/butternut/

No comments:

Post a Comment