SAMPLING ERROR IN CLIMATE PROPERTIES DERIVED FROM SATELLITE MEASUREMENTS - CONSEQUENCES OF UNDERSAMPLED DIURNAL VARIABILITY

The diurnal cycle present in many climate properties is undersampled i n asynoptic data, which, through aliasing, introduces a bias into time -mean behavior derived from satellite measurements. This source of sys tematic error is investigated in high-resolution Global Cloud Imagery (GCI), which provides a proxy, with realistic space-time variability, for several climate properties to be observed from space. The GCI, whi ch resolves mesoscale and diurnal variability on a global basis, is sa mpled asynoptically according to orbital and viewing characteristics f rom one and multiple platforms. Sampling error is then evaluated by co mparing the resulting time-mean behavior against the true time-mean be havior in the GCI. The bias from undersampled diurnal variability is m ost serious in polar-orbiting measurements from an individual platform . However, it emerges even in precessing measurements, which drift thr ough local time, because diurnal variability is still sampled too slow ly to be truly resolved in such observations. A ''mean diurnal cycle'' can be constructed by averaging precessing measurements, provided tha t the ensemble of observations at individual local times is large enou gh (e.g., that observations are averaged over a long enough duration). The pattern of time-mean error closely resembles the pattern of error in the mean diurnal cycle. Time-mean behavior can therefore be determ ined only about as accurately as can the mean diurnal cycle. Determini ng accurate time-mean properties often requires averaging measurements from an individual platform over several months, which cannot be perf ormed without contaminating mean behavior with seasonal variations. Th e sampling limitations from an individual orbiting platform are allevi ated by sampling from multiple platforms, which provide observations f requently enough in space and time to determine accurate monthly mean properties.