Continuous monitoring of the Earth's radiation field at the top of the atmosphere (TOA) is essential for understanding natural and anthropogenic-induced changes to the earth's climate and for improving predictive performances of the global climate models (GCMs). To achieve this important science goal, high quality continuous long term broadband radiation data sets are needed. Currently, the most accurate long term time series of tropical broadband radiation measurements between November 1984 and September 1999 are the NASA ERBE/ERBS nonscanner data set. While excellent agreement is found between this data set and the various shorter broadband radiation sets (i.e., Nimbus 7 nonscanner, ERBE scanner, CERES scanners, and ScaRAB scanners), considerable disagreements are noticed between the ERBE/ERBS nonscanner data set and the broadband longwave radiation data set derived from narrowband radiation instruments (i.e., AVHRR and HIRS time series). Large differences are found between tropical mean LW fluxes in the AVHRR, HIRS, and ERBS broadband records. Concerns for the narrowband AVHRR and HIRS data sets include narrowband to broadband conversion models as well as varying diurnal sampling. Similar comparisons are performed to access the performance of current global climate models. Tropical mean time series of TOA longwave, shortwave, and net radiation from AMIP model runs of five different GCMs and NCEP Reanalysis project are compared with those from the ERBE/ERBS nonscanner time series. The results show noticeable deficiencies in the current GCMs for capturing both the magnitude and the phase of the observed tropical variability. For example, while some GCMs show skills in capturing many regional aspects of the 1998 ENSO event, none of the models were able to reproduce all of the observed large magnitude changes (from ERBE/ERBS and CERES/TRMM data) that occurred over the whole tropics during this ENSO period.