Studies on pollen resources by stingless bees frequently suffer from low sample size due to difficulties concerning the acquisition of harvested pollen. Here we describe a funnel-trap that allows non-invasive and automated sampling of pollen-rich garbage pellets that are expelled from colonies by workers bees. Single garbage pellets of Trigona collina from Sabah, Malaysia, contained between 7 and 11 different morphotypes of pollen and the similarity of the pollen composition of pellets expelled by a given colony on a given day was very high (quantitative Steinhaus index: 71 to 90% ). The turn-over of pollen types in samples taken at consecutive points in time was relatively low over periods of three weeks (52 to 75% similarity) and variable over periods of four to six months (13.6 to 58.5% similarity). The comparison of pollen in corbicular loads and garbage pellets indicates that garbage pollen is derived from both feces of pollen-consuming workers and larval feces (meconia). The slow turn-over of pollen in garbage suggests that sampling at relatively long intervals (4-6 months) will be sufficient for a crude assessment of long-term pollen resources of stingless bee colonies.