89th American Meteorological Society Annual Meeting

Monday, 12 January 2009: 2:00 PM
Analysis, Validation and Application of the NCEP Multi-model NLDAS Products for Drought Monitoring and Prediction
Room 127BC (Phoenix Convention Center)
Youlong Xia, NOAA/NWS/NCEP, Camp Springs, MD; and K. E. Mitchell, E. F. Wood, L. Luo, J. Sheffield, D. P. Lettenmaier, A. W. Wood, B. A. Cosgrove, C. J. Alonge, J. Meng, H. Wei, M. Ek, P. Restrepo, J. C. Schaake, K. Mo, and R. T. Pinker
The NCEP Environmental Modeling Center (EMC) collaborated with its CPPA (Climate Prediction Program of the Americas) partners to develop a North American Land Data Assimilation System (NLDAS, http://www.emc.ncep.noaa.gov/mmb/nldas). The multi-institution and multi-model NLDAS system includes both an analysis/monitoring mode and an ensemble seasonal prediction mode.

The monitoring mode consists of a retrospective 29-year (1979-2007) historical execution and a realtime daily update execution using four land surface models (Noah, Mosaic, SAC, and VIC) on a common 1/8th degree grid using commonly hourly land surface forcing. The non-precipitation surface forcing is derived from NCEP's retrospective and realtime North American Regional Reanalysis System (NARR). The precipitation forcing is anchored to a daily gauge-only precipitation analysis over CONUS that applies a Parameter-elevation Regressions on Independent Slopes Model (PRISM) correction. This daily precipitation analysis is then temporally disaggregated to hourly precipitation amounts. The NARR-based surface downward solar radiation is bias-corrected using seven years (1997-2004) of GOES satellite-derived solar radiation retrievals.

The 29-year NLDAS retrospective is used to derive the climatology of each of the four land models. Then current realtime land states (soil moisture, snowpack) and water fluxes (evaporation, total runoff, routing streamflow) of each of the four land models from daily update executions are depicted as anomalies and percentiles with respect to their own model climatology. One key application of the realtime updates is drought monitoring over CONUS, shown at the "NLDAS Drought" tab of the NLDAS web site given above and at the CPC drought web site at http://www.cpc.ncep.noaa.gov/products/Drought/. Additionally, the realtime NLDAS land states provide initial conditions for the NLDAS prediction mode described below.

The uncoupled ensemble seasonal prediction mode utilizes the following three independent approaches for generating downscaled ensemble seasonal forecasts of surface forcing: (1) Ensemble Streamflow Prediction, (2) CPC Official Seasonal Climate Outlook, and (3) NCEP CFS ensemble dynamical model prediction. For each of these three approaches, twenty ensemble members of forcing realizations are generated using a Bayesian merging algorithm developed by Princeton University. The three forcing methods are then used to drive a chosen land surface model (currently VIC) in seasonal prediction model over thirteen large river basins that together span the CONUS domain. One to nine month ensemble seasonal prediction products such as air temperature, precipitation, soil moisture, snowpack, total runoff, evaporation and streamflow are derived for each forcing approach. The anomalies and percentiles of the predicted products for each forcing approach may be used for the purpose of US drought prediction. The above seasonal prediction approach, currently applied to the VIC land model, will be expanded to include all four land models cited earlier.

The NLDAS products of the four land models are compared and validated using USGS observed streamflow, observations of surface energy fluxes from flux stations, Oklahoma and Illinois soil moisture observations, and satellite-based snow cover analyses. The advantages and disadvantages of different land surface models are emphasized. This will be beneficial for the land model development community. Additionally, improvement of the land surface models will reduce the uncertainty and errors in NLDAS products for both monitoring and prediction mode and increase the accuracy and reliability of NLDAS products.

Supplementary URL: