Town Hall on Ensemble Methods for Land Data Assimilation

Tuesday, February 28th at 9:00 – 10:30 (UTC +0)
Tuesday, February 28th at 23:00 – 0:30 (UTC +0)
Wednesday, March 1st at 4:00 – 5:30 (UTC +0)
Registration: Closed

Organizers: Prashant Kumar (1), Shunji Kotsuki (2), Andy Fox (3), Tristan Quaife (4), Hannah Liddy (5)
(1) Space Applications Centre, ISRO, India; (2) Center for Environmental Remote Sensing, Chiba University, Japan; (3) NASA GSFC, USA; (4) University of Reading, UK; (5) Columbia University/NASA GISS, USA
† Organized by the AIMES Land Data Assimilation Working Group

Overview: As a solution to nonlinear problems in land surface modeling, ensemble data assimilation techniques have been successfully implemented by the land, hydrological, biogeochemical modeling communities to improve land surface water and carbon cycle predictions. While these methodological advances have resulted in innovative development and reduced the computational burden of analysis, there remain challenges to the technique that need to be addressed by the land DA community. In this Town Hall, we invite participants to explore the benefits and challenges of assimilating land observations using ensemble based techniques.

In particular, we ask participants to consider:

  • How can we optimally treat bounded variables?
  • What approaches can be used when distributions are highly non-gaussian?
  • What developments beyond the traditional ensemble Kalman filter, including rank histogram, smoothers and particle filters, might be most useful for land DA?
  • What inflation methods are suitable for land models to maintain ensemble spread adequately?
  • Can we couple neural networks and ensemble data assimilation, such as for surrogating models and observation operators and mitigating forecast bias?
  • How can we apply techniques such as localization to avoid the impact of spurious correlations in space and/or time?
  • Can you assimilate un-recognized measurements in land DA?
  • What are the possibilities of developing low cost ensembles using machine learning?
  • To achieve balance in atmospheric and land surface parameters, what are the needs of coupled land surface DA systems?

Through posters and lightning talks, attendees can share their methodologies, address challenges, and meet others working in this area. The goal of this town hall is to provide a venue to address the technical challenges associated with the development of ensemble data assimilation techniques and to identify if there are additional needs that can be addressed through opportunities to convene the land DA community.

Featured Posters:

  • Plans and progress for new surface analyses at DWD Gernot Geppert*(1) and Martin Lange (1), (1) Deutscher Wetterdienst (DWD), Germany
  • Investigating appropriate inflation methods for soil moisture data assimilation Daiya Shiojiri* (1) and Shunji Kotsuki (1), (1) Center for Environmental Remote Sensing, Chiba University, Japan
  • Estimating the global precipitation from gauge observations using the Local Ensemble Transform Kalman Filter Yuka Muto*(1) and Shunji Kotsuki (1), (1) Chiba University, Japan
  • Seeking an optimal observation location for data assimilation by sparse sensor placement Mao Ouyang*(1) and Shunji Kotsuki (1), (1) Chiba University, Japan
  • Setup of a land data assimilation system to assimilate aboveground biomass and MODIS leaf area index observations into the Community Land Model in the Arctic and Boreal region Xueli Huo*(1); Andrew Fox (2,3); Hamid Dashti (4); Charles Devine (1); Timothy Hoar (5); Brett Raczka (5); William Gallery (1); William Kolby Smith (1); David Moore (1), (1) The University of Arizona, USA; (2) GMAO, NASA GSFC, USA; (3) GESTAR-II, Morgan State University, USA; (4) University of Wisconsin-Madison, USA; (5) National Center for Atmospheric Research (NCAR), USA
  • A Quantile Conserving Ensemble Filtering Framework: Next Generation Nonlinear and Non-Gaussian Data Assimilation Capabilities for DART Jeffrey Anderson*(1); (1) NCAR, CISL, Data Assimilation Research Section, USA