Data from DOE Atmospheric Radiation Measurement Program Allows Evaluation of Surface Models

Robock, A., Rutgers University

Surface Properties

Cloud Modeling

Robock, A., Luo, L., Wood, E. F., Wen, F., Mitchell, K. E., Houser, P. R., Schaake, J. C., Lohmann, D., Cosgrove, B., Sheffield, J., Duan, Q., Higgins, R. W., Pinker, R. T., Tarpley, J. D., Basara, J. D., Crawford, K. C., Evaluation of the North American Land Data Assimilation System over the Southern Great Plains during the warm season, J. Geophys. Res., 108(D22), 8846, doi:10.1029/2002JD003245, 2003

An example of the model discrepancies is shown in a comparison of monthly mean diurnal cycle data from July 1999 at the ARM Southern Great Plains site. Data are for net radiation, latent heat flux (LE), sensible heat flux (H) and ground heat flux (G).

The North American Land Data Assimilation System (NLDAS) is a multi-institutional project focusing on providing accurate values of land surface conditions by combining, or "assimilating," observed atmospheric data into model simulations of these conditions. DOE's Atmospheric Radiation Measurement (ARM) Program is contributing to this project—part of the National Oceanic and Atmospheric Administration's Global Energy and Water Cycle Experiment Continental Scale International Project (GCIP)—by providing observational data sets from its highly instrumented Southern Great Plains (SGP) site. Research results published in the Journal of Geophysical Research describe an evaluation of land surface models using observations. Their data are focused on warm season precipitation, one of the most difficult variables to forecast accurately.

Four state-of-the-art land surface models are used in NLDAS: Noah, Mosaic, variable infiltration capacity (VIC), and Sacramento. Observed data came from 24 ground based instrument stations throughout Oklahoma and Kansas within ARM's SGP site, as well as from 72 environmental monitoring stations from the Oklahoma mesoscale network, or "Mesonet." These stations measure atmospheric temperature, humidity and winds, along with soil temperature and moisture at varying depths, precipitation, and radiative fluxes in short- and long-wavelengths, both incoming and outbound. In comparing these variables in the models against the observed measurements, the models were inconsistent (some overestimated measurements while others underestimated them) in almost every category except for soil temperature. The authors trace many of the discrepancies to soil-related information. Soil parameters vary continuously and any one land surface point in a model may contain a variety of soil types. Additionally, the amount of water in soil at any time and its availability to plants has a major effect on surface energy fluxes and hence, on local atmospheric conditions.

In analyzing their results, the authors found that changes to a subset of model parameters to more "realistic" values did not guarantee improvement in model performance as a whole. Thus, they concluded that model parameters are better calibrated as a complete set. This latter approach requires multiple well-instrumented sites in order to have sufficient data over a broad range of conditions. The ARM Program, along with the Oklahoma Mesonet, provides essential data to scientific collaborators with the GCIP, allowing them to better evaluate the effectiveness of modifications to computer model parameterizations. Once proven for weather forecasting, these improvements can also be applied to climate simulations.