News & Updates
July 27 - August 7, 2015
JCSDA News and Newsletters
Evaluation and Improvement of Land Surface States and Parameters To Increase Assimilation of Surface-Sensitive Channels and Improve Operational Forecast Skill
From the Spring 2015 issue of the JCSDA Quarterly
Land surface models (LSMs) exist within a wide spectrum of complexity. Current NOAA/NCEP/EMC LSMs, such as the Noah LSM, use a bulk surface treatment, meaning the vegetation, snow, and soil surface are treated as a combined unit with one surface temperature. Bulk LSMs have been effective at providing accurate lower boundary condition fluxes of heat and water to the atmosphere in operational settings.
Recent LSM developments such as the Noah-MP LSM consider a more process-based approach, with multi-layer snow packs and explicit vegetation canopies that have dynamic growth. These new LSMs can more accurately simulate situations when surface heterogeneities exist (e.g., canopy overlying snow), provide more detailed information about individual land surface processes (e.g., multiple surface temperatures), and may increase the assimilation of atmospheric and land surface observations to enhance model performance.
To satisfy the increasing demand of predicting the earth system (hydrology, water quality and resources, agriculture, etc.), it is likely that U.S. operational centers will transition from the bulk approach to the processbased approach, with explicit canopies and dynamic vegetation modules, and expand the utility of the LSM beyond its traditional purpose of providing atmospheric boundary conditions. These LSM structural changes provide opportunities to assimilate more satellite land data, such as albedo and leaf area index (LAI).
The changes will also present a challenge to the land data assimilation community, however, as LSMs move beyond using satellite data directly (e.g., prescribing albedo or vegetation as is done in current operational models) to using satellite-observed land surface states. These new models also contain many more unobservable parameters that can be estimated using satellite land products, especially those from relatively high frequency global sensors, such as MODIS and VIIRS.
As part of our JCSDA-funded project, we are developing a framework for land data assimilation systems (e.g., NASA LIS) to effectively use available satellite land data products (e.g., MODIS BRDF/albedo and LAI) to update model states, and to estimate model parameters. As an example, we have extracted the Noah-MP canopy two-stream radiative transfer model to create a forward model for estimation of model radiation-relevant parameters.
The forward model takes inputs of LAI (either from a prescribed climatology or satellite observations), solar zenith angle, and canopy leaf/stem reflectivity and transmissivity (in broadband VIS and NIR) and outputs surface albedo comparable to the MODIS surface albedo product (MCD43C1). To estimate canopy parameters, we use the forward model and minimize a cost function of surface-absorbed solar radiation.
Figure 1. Diurnal cycle of albedo from MODIS (black), Noah-MP with estimated parameters (red) and default Noah-MP (blue).
Initial results of the parameter estimation system have shown promise in improving canopy radiation-relevant parameters. For example, the system was tested for a 10°x10° section of the U.S. Midwest containing the states of Minnesota, Iowa, and Wisconsin at 0.05° spatial resolution and dominated by cropland and forest. Figure 1 shows the June diurnal cycle of surface albedo in the cropland pixels from the MODIS product, default Noah-MP model, and Noah-MP model with estimated parameters termed as "optimal." The "optimal" clearly reproduces the MODIS albedo with higher fidelity than the default Noah-MP. These simulations also yielded a 30-40 percent reduction in temperature bias over the domain. To date, the system has only been used in hindcast mode for parameter estimation. In the time remaining in our project, we plan to extend the capabilities to near real-time parameter adjustment.
Michael Barlage (NCAR)