VIC / GOES Radiation Evaluation

A recently published paper by Roads et al (2003) describing the GCIP Water and Energy Balance Synthesis (WEBS) showed that the Variable Infiltration Capacity model (VIC) model under-simulates “observations” of net solar radiation by an average of 50 w/m in the summer months of 19961999 over the Mississippi basin. The net solar radiation “observations” used by Roads et al (2003) are a GOES-based satellite data set produced by the University of Maryland. The “VIC” values were in fact produced by Maurer et al (2002) as VIC model forcings (hence are unrelated to the VIC model) using the algorithm of Thornton and Running (1999), which is based on the daily temperature range (we nonetheless refer herein to the Thornton and Running values as “VIC” as do Roads et al). This note diagnoses downward shortwave and net shortwave radiation values over the Mississippi basin and at the six SURFRAD locations across the U.S. Averaged over the Mississippi in June, July, and August of 1998 we found a 40 to 50 w/m net radiation deficit between VIC and GOES, as did Roads et al (2003). However, for downward shortwave radiation, the difference on average was about 20 w/m; the balance of the difference was attributed to a large (factor of 2) difference in albedo between that used by GOES and VIC. A comparison of GOES albedos with SURFRAD observations, and values used in the LDAS versions of the NOAH, and Mosaic models, indicates that the GOES albedo is low relative to both observations and other models. Both VIC and GOES downward solar radiation values were compared to SURFRAD measurements, which showed that VIC was closer to the observed daily averages (especially for summer 1998), while GOES better reproduced daily peaks. Introduction / Approach Roads et al (2003) compare various predictions of water and energy budget terms over the Mississippi River basin produced by coupled land-atmosphere models, and uncoupled runs of the VIC model, the latter described by Maurer et al (2002). This note focuses on differences noted by Roads et al (2003) between VIC solar radiation (actually predicted downward solar radiation based on the temperature range-based algorithm of Thornton and Running (1999), a satellitebased (GOES) product from the University of Maryland, and direct observations at SURFRAD sites, all during the time period 1996-1999. For the VIC and GOES products, the domain is the Mississippi basin. Figure 1 shows the study domain as well as the locations of the six U. S. SURFRAD stations, of which five lie within the Mississippi basin. Figure 1: Mississippi River Basin as represented by VIC 1/8 degree grid with locations of SURFRAD stations The basis of the analysis conducted here is Figure 7.4 of Roads et al (2003), reproduced here as Figure 2 (taken from http://ecpc.ucsd.edu/gcip/webs.htm), which compares net solar radiation from the GOES product (from Rachel Pinker and colleagues at the University of Maryland). The Pinker et al product is labeled in Roads et al (2003) as “observations’. The apparent low bias of VIC is important, as the Maurer et al (2002) data are for a 50-year simulation over the period 1950-2000, most of which predates the availability of satellite solar radiation data. If the source of biases in the VIC forcing data can be identified, it might be possible to adjust them to improve model simulations over the entire 50