A Lagrangian view is adopted for establishing the spatiotemporal cloud statistics and the scale dependent radiative properties using satellite data. Individual clouds are identified using a newly developed scheme. We sort all clouds by cloud type, cloud area, and number of clouds in each area bin, as well as their radiative properties. For seven different cloud types our analyses provide radiative properties, such as albedo and cloud top temperature, as a function of the cloud spatial scale. All clouds are marked by local time, and large clouds are tracked over time. These analyses provide diurnal variability, lifetimes, and evolution of cloud systems as a function of their spatial scales. These scale dependent cloud properties can be objectively used in guiding the development and evaluation of cloud parameterization in global climate models (GCMs). Particularly, we show how our Lagrangian approach can be used to establish the relative importance of resolvable and fully parameterized clouds to the total cloudy area and to the total amount of reflected visible irradiance. Focus in this 1 month satellite study is on the convective-stratiform cloud systems over the western and central tropical Pacific Ocean, including the so-called warm pool. We adopt the hourly Japanese geostationary satellite (GMS) window channel radiances in the visible and IR window region for cloud classification and characterization. To study the radiative contributions of different clouds in the area, we computed the bidirectional model (BDM) for the Visible and Infrared Spin Scan Radiometer instrument aboard GMS, which we show to agree well with the BDM of the Earth Radiation Budget instrument aboard the Nimbus 7 satellite. An iterative two-stage cloud detection scheme was developed to identify individual clouds. Furthermore, a tracking algorithm was developed to study the time evolution of mesoscale convective systems (MCS). It operates on area and orientationally equivalent ellipsoidal representations of these MCS. We show that the temporal statistics of these convective anvil clouds show good agreement with those reported in the literature. Our data indicate that for the convective-stratiform systems in the tropical Pacific, 95% of the radiatively important clouds (containing a core with an effective brightness temperature <219 K) are of scales resolvable by a GCM of about 50 km¿50 km. On the other hand, a GCM of 250 km¿250 km will only be able to resolve 50% of the radiatively important clouds. This, however, does not mean that the processes responsible for the formation and maintenance of these systems are also resolvable. The low clouds that are unattached to convective-stratiform systems are mostly unresolvable by available GCMs.¿ 1997 American Geophysical Union