1 Use of in situ and airborne multiangle data to assess 1 MODIS - and Landsat - based estimates of surface albedo 2 3 4

23 The quantification of uncertainty of global surface albedo data and products is a critical part 24 of producing complete, physically consistent, and decadal land property data records for studying 25 ecosystem change. A current challenge in validating satellite retrievals of surface albedo is the 26 ability to overcome the spatial scaling errors that can contribute on the order of 20% disagree27 ment between satellite and field-measured values. Here, we present the results from an uncertain28 ty analysis of MODerate Resolution Imaging Spectroradiometer (MODIS) and Landsat albedo 29 retrievals, based on collocated comparisons with tower and airborne multiangular measurements 30 collected at the Atmospheric Radiation Measurement Program’s (ARM) Cloud and Radiation 31 Testbed (CART) site during the 2007 Cloud and Land Surface Interaction Campaign (CLAS32 IC’07). Using standard error propagation techniques, airborne measurements obtained by 33 NASA’s Cloud Absorption Radiometer (CAR) were used to quantify the uncertainties associated 34 with MODIS and Landsat albedos across a broad range of mixed vegetation and structural types. 35 Initial focus was on evaluating inter-sensor consistency through assessments of temporal stabili36 ty, as well as examining the overall performance of satellite-derived albedos obtained at all diur37 nal solar zenith angles. In general, the accuracy of the MODIS and Landsat albedos remained 38 under a 10% margin of error in the SW (0.3 5.0 μm) domain. However, results reveal a high 39 degree of variability in the RMSE (root mean square error) and bias of albedos in both the visible 40 (0.3 0.7 μm) and near-infrared (0.3 5.0 μm) broadband channels; where, in some cases, re41 trieval uncertainties were found to be in excess of 20%. For the period of CLASIC’07, the prima42 ry factors that contributed to uncertainties in the satellite-derived albedo values include: (1) the 43 assumption of temporal stability in the retrieval of 500 m MODIS BRDF values over extended 44 periods of cloud-contaminated observations; and (2) the assumption of spatial and structural un45 iformity at the Landsat (30 m) pixel scale. 46