2002 Annual

Monday, 14 January 2002: 10:15 AM
Evaluation of streamflow and snowpack simulations in the land surface models of the Land Data Assimilation System (LDAS) Project
Dag Lohmann, NOAA/NWS/NCEP/EMC, Camp Springs, MD; and K. E. Mitchell, P. R. Houser, J. C. Schaake, E. F. Wood, D. Tarpley, R. W. Higgins, R. T. Pinker, A. Robock, D. P. Lettenmaier, B. Cosgrove, Q. Duan, J. Sheffield, and L. Luo
Traditional coupled land-atmosphere 4-D data assimilation systems (4DDA) often yield significant errors and drift in a) soil moisture and temperature and b) surface energy and water fluxes owing to substantial biases in precipitation, surface radiation and air temperature in the attendant surface forcing fields of the parent atmospheric models. The GCIP regional PILPS-2C experiment and the ISLSCP Global Soil Wetness Project demonstrated the viability of executing physically-based, distributed, uncoupled, land-surface models over large spatial domains, provided that moderately dense observations of precipitation were available. Hence, as a land-surface alternative to coupled 4DDA, we have undertaken the development, execution, and evaluation of an uncoupled Land Data Assimilation System (LDAS) -- a REALTIME, hourly, distributed, uncoupled, land-surface simulation system on a U.S. national domain at 0.125-degree resolution. This LDAS partnership is using four distinct land-surface models (NOAH, MOSAIC, VIC, SAC-SWA), running in tandem on a common 1/8-th degree CONUS grid and driven by common surface forcing anchored by model-independent, observation-based precipitation and solar insolation fields. Also, a common streamflow routing model is applied to each LSM's gridded runoff.

The goals of the LDAS project are to 1) improve LSM physics by sharing methodologies and data sources, 2) identify causes of the spread in magnitudes of surface water fluxes and surface water storage typically seen in LSM intercomparisons, 3) compare land states of the uncoupled LDAS with traditional coupled 4DDA, 4) demonstrate how to assimilate land-state related satellite retrievals (e.g. snowpack, skin temperature) and 5) provide land-state initial conditions (e.g. soil moisture and snowpack) for a) retrospective land-memory predictability studies and b) realtime coupled model predictions of weather and seasonal climate.

In this paper, we intercompare LDAS results of snowpack simulations over CONUS and streamflow simulations (over about 25 small and medium-sized basins located primarily in the eastern half of the U.S.) from the land-surface models participating in LDAS. We will present some results from both the realtime simulations during the period of Apr 99 - present, as well as retrospective results for the period Sep 96 - Sep 99. In the latter, the common forcing was re-processed in order to improve forcing quality, e.g. by taking advantage of observations not available in realtime, such as cooperative-observer gage observations of precipitation.

As companions to these LDAS streamflow and snowpack intercomparisons, we will highlight the different approaches in the participating land-surface models in the treatments of a) surface infiltration and surface drainage, b) complexity of snowpack physics, and c) water storage. In some instances, we augment the results of the snowpack intercomparisons with impact studies that utilize reruns wherein the physics of certain snowpack processes such as sublimation and albedo treatments are changed.

Supplementary URL: